NVIDIA's GeForce RTX 50 Series: A New Era of AI-Powered Gaming GPUs

COLORFUL iGAME GeForce RTX 5080 Ultra W Review - Overclocked and Ready for 4K

COLORFUL offers a range of different-looking GeForce RTX cards. With the arrival of the new Blackwell-powered GeForce RTX 50 Series, we enjoyed checking out the new COLORFUL iGAME GeForce RTX 5080 Ultra W OC model for review. Now, one of the first things you'll notice about this particular RTX 5080 card is that it looks unlike any other GeForce RTX 5080. According to COLORFUL, the physical look was inspired by hip-hop and street art culture. It doesn't stop there, as it's also a design tailor-made for white PC builds, where on top of the white and pink colors, you've got some holographic flourishes and some bright neon purple lighting when you plug it in and turn it on. A quick look at the COLORFUL iGAME GeForce RTX 5080 Ultra W When picking up a new GPU, a lot of people aren't too phased about the look. You might fall into this camp or look for a clean, minimal, all-black affair, something with a little bit of RGB, something with a lot of RGB, or something as vibrant as what COLORFUL's cooked up here. As someone who does like the all-white PC aesthetic that has become popular in recent years, with its neon colors and graffiti art, the COLORFUL iGAME GeForce RTX 5080 Ultra W OC is something I'd consider picking up and slotting into a white rig. Or building an entire system around it. Of course, when it comes to reviewing a new GPU, gaming performance and cooling efficiency are more important than a funky physical design. However, seeing something a little different in a sea of black-and-grey industrial sci-fi-looking GPUs is always great. As the name suggests, the iGAME GeForce RTX 5080 Ultra W is an OC model. As we've benchmarked a few different RTX 5080s, we can safely say that it's a GPU with enough headroom to get an additional 5% or even 10% extra performance over the reference or Founder Edition spec, depending on the game or workload. For this review, we ran our benchmarks using the GPU's "Turbo" OC mode, which you can activate via a button on the exhaust. What does this mean for its overall gaming performance? Well, when it comes to 4K gaming across a wide range of titles with and without ray tracing, the COLORFUL iGAME GeForce RTX 5080 Ultra W OC is 22% faster than the GeForce RTX 4080 and an impressive 88% faster than the GeForce RTX 3080. Throw in DLSS 4, and you have one of the best high-end 4K PC gaming options. Below is a summary of NVIDIA's GeForce RTX 50 Series and RTX Blackwell architecture, applicable to all models. NVIDIA describes 'Neural Rendering,' which includes all previous versions of DLSS and the brand-new DLSS 4, as the 'next era for computer graphics.' They're not alone; the Lead System Architect for the PlayStation 5 Pro console, Mark Cerny, recently said that ray-tracing is the future of games and that AI will play an integral role in making that happen. DOOM: The Dark Ages developer id Software shared a similar sentiment, adding that the arrival of DLSS was an 'inflection point' for PC game visuals and performance and on par with the arrival of dedicated GPUs and programmable shaders. With the arrival of the Blackwell generation and the GeForce RTX 50 Series, AI is now being used to accelerate programmable shaders with the brand-new RTX Neural Shaders. Yes, these are actual neural networks that use live game data, and the power of Tensor Cores to do everything from compress textures, render lifelike materials with a level of detail impossible to match using traditional rendering methods, and even use AI to partially trace rays and then infer "an infinite amount of rays and bounces for a more accurate representation of indirect lighting in the game scene." RTX Mega Geometry is incredible in its own right; it essentially increases a scene's geometry detail and complexity (triangles or polygons) by up to 100x. 100 times the detail, it's hard to wrap your head around - but the added benefit in a game like Alan Wake 2 is dramatically improving the performance of the game's Full Ray Tracing or Path Tracing mode. With DLSS 4 and RTX Neural Shaders, NVIDIA's GeForce RTX 50 Series and RTX Blackwell architecture (which includes the same AI optimizations as data center Blackwell) can be viewed as the turning point for PC gaming - the moment when AI becomes integral to everything from designing a game to programming and then finally rendering it on a 4K display to play. DLSS 4 includes more goodies than NVIDIA's highly touted new Multi Frame Generation technology, but let's start there. DLSS 3's version of Frame Generation has evolved with DLSS 4, powered by Blackwell hardware and software, and an innovative use of AI to generate frames 40% faster while using 30% less VRAM. Switching to a new model also means that Frame Generation and Multi-Frame Generation could soon come to GeForce RTX 20, 30, and RTX 40 Series owners. DLSS 4 benefits all GeForce RTX gamers. However, with the 5th Generation of Tensor Cores in the GeForce RTX 50 Series delivering 2.5X more AI performance, NVIDIA's latest GeForce RTX 50 Series GPUs can execute five complex AI models - covering Super Resolution, Ray Reconstruction, and Multi Frame Generation in a couple of milliseconds. Part of the reason it happens so quickly is the addition of hardware Flip Metering, which shifts frame pacing to the Blackwell display engine - the result is frame rates of up to 4K 240 FPS and higher without stuttering issues. With up to 15 of every 16 pixels generated by AI, the result is up to 8X the performance when compared to native rendering or rasterized performance. DLSS Super Resolution and Ray Reconstruction are also switching to a new 'Transformer' model, with over double the parameters and four times the compute requirement. This is one of the most exciting aspects of the GeForce RTX 50 Series, as it pushes DLSS into a new realm of image quality and performance. The best part is that it will work on all GeForce RTX GPUs; however, there will be a performance hit compared to running it on an RTX 50 Series GPU. Already available in games, DLSS 4's Transformer model is another DLSS 2.0-like moment for the technology, and the results speak for themselves. Even better, DLSS 4 is being integrated into the NVIDIA App with a new 'DLSS Override' feature that allows users to experience the latest tech without waiting for a path or game update. DLSS 4 is built to be backward compatible, with 75 games and apps supported. It doesn't stop there, as the new AI Management Processor (AMP) allows AI models to share the GPU with graphics workloads. As a result, expect to see digital humans in games alongside AI assistants like NVIDIA's Project G-Assist become more prevalent in the coming years. This filters down to the creator side, with AI assistants for streamers, who will also benefit from the GeForce RTX 50 Series' expanded creator features. RTX Blackwell introduces 4:2:2 chroma-sampled video encoding and decoding. The ninth-generation NVENC encoder also improves AV1 and HEVC quality. The flagship GeForce RTX 5090 supports up to three encoders and two decoders to deliver a 50% gen-over-gen improvement in speed compared to the GeForce RTX 4090. The GeForce RTX 5080 adds a second decoder compared to the GeForce RTX 4080. The RTX Blackwell is a game changer for creators and editors, especially with the new low-voltage and cutting-edge GDDR7 memory that dramatically improves memory bandwidth and speed. Here's a look at the specs for the flagship GeForce RTX 50 Series GPUs, the GeForce RTX 5080 and GeForce RTX 5090, compared to the previous Ada generation. With the GeForce RTX 50 Series generation, outside of the flagship GeForce RTX 5090, with the same or similar process node, the actual specs only see a modest upgrade over the GeForce RTX 40 Series. With the GeForce RTX 5080, you're looking at 11% more CUDA, Tensor, and RT Cores than the GeForce RTX 4080. Although Blackwell features architectural upgrades for new AI rendering and other improvements, the typical generation uplift of 30+ percent for overall or raw performance is missing. Ultimately, this means that for the first time, an 80-class GeForce RTX graphics card isn't outperforming the previous generation's flagship - in this case, the GeForce RTX 4090. The GeForce RTX 5080 is more than 11% faster than the GeForce RTX 4080, which means that performance and efficiency improvements can be gained outside of moving to a smaller process node. However, it does feel like we're getting closer to the point where we'll no longer see the significant gains you get moving from 12nm to 8nm or 8nm to 4nm. The good news is that NVIDIA hasn't just discovered AI; with the Blackwell architecture and new AI-enhanced tools and technologies, it feels like they've hit a new level of performance and quality. DLSS 4's new Multi Frame Generation feature might be exclusive to the GeForce RTX 50 Series. Still, the significant updates to the existing Frame Generation model, DLSS Super Resolution, and Ray Reconstruction's new Transformer model are transformative to image fidelity and overall performance. These features are available to all GeForce RTX gamers, with GPUs like the GeForce RTX 5080 introducing more advanced Tensor Core hardware to ensure these more complex DLSS models run faster. In addition, Blackwell is primed and ready for RTX Neural Rendering, a suite of new and impressive AI technologies designed to boost ray-tracing performance and image quality. Only a single piece of it is available to test - RTX Mega Geometry has been added to Alan Wake 2. This will be one part of the new GeForce RTX 50 Series that will take some time to filter down into the games we play. As an OC model, the COLORFUL iGAME GeForce RTX 5080 Ultra W OC offers a Turbo mode that takes the reference boost clock speed of 2617 MHz and pushes it to 2655 MHz. A modest upgrade, however, with COLORFUL's iGAME Center software or a third-party app (or even the NVIDIA App), can be pushed even higher. Inside the funky while and hip-hop-inspired exterior, the COLORFUL iGAME GeForce RTX 5080 Ultra W OC houses an effective cooling system that has no issue keeping temperatures in check. Although it doesn't quite sit on the large side for a GeForce RTX 5080, the three-slot thickness and 1840-gram weight are nothing to sneeze at. With three fans, where the middle fan spins on a different axis from the outer fans, and a decent-sized heatsink, there's ample cooling headroom for overclocking. A lot of the features are what you'd expect to find on a high-end GPU like this: dual ball bearings for long-lasting and quiet fans, an auto-stop feature that has the fans stop once the temperatures drop below a certain threshold, high-quality heat pips, and fins for heat dissipation, a metal backplate with vents for airflow, and software support for monitoring and tweaking settings - wither everything from lighting to fans speeds and curves accounted for. As the introduction mentions, the COLORFUL iGAME GeForce RTX 5080 Ultra W OC stands out in a sea of custom GeForce RTX 5080 designs. The blend of graffiti art, hot neon pink, holographic flourishes, and an RGB strip running down the side make it a looker once installed. Unfortunately, the only thing missing is a custom white PCB. However, COLORFUL packs a white 8-pin to 16-pin power adapter, which is a nice touch. It's a stylish design, perfect for all-white builds or those looking for something much more vibrant, and a thermal solution that works well to cool the GPU even after you press the 'Turbo' button next to the video outputs. PC gaming not only covers a wide range of genres and styles, from indie games with simple 2D graphics to massive 3D worlds lit by cutting-edge real-time ray tracing technology. With that, the needs and requirements of each gamer vary. High refresh rates and latency reduction become more important than flashy visuals or playing at the highest resolution possible for those who live and breathe fast-paced competitive games. For those who want to live in a cinematic world and become a key player in an expansive narrative, ray-tracing, and high-fidelity visuals are a stepping stone toward immersion. Our chosen benchmarks cover various games, engines, APIs, and technologies. For the GeForce RTX 5090, all tests are run at 4K and 1440p and include results for performance-boosting Super Resolution technologies like NVIDIA DLSS 4 - including Frame Generation and the new Multi Frame Generation. In many ways, DLSS numbers are more important in 2025 than native rendering - a title with ray tracing isn't meant to be played without Super Resolution. Also, DLSS technologies like Ray Reconstruction and the new RTX Mega Geometry dramatically improve visual fidelity and detail compared to native rendering. However, our benchmark results are still sorted using 'raw performance' or native rendering. Here's the breakdown of games, settings, and what's being tested. The COLORFUL iGAME GeForce RTX 5080 Ultra W OC is the sort of GPU that you need to pair with a 4K or Ultrawide display to get the most out of it. Only at these ultra-high resolutions can you get a clearer picture of the performance uplift it provides over previous generation 80-class GPUs like the GeForce RTX 4080 and the GeForce RTX 3080. Compared to the GeForce RTX 4080, you're looking at 22% faster 4K gaming on average, with that number dropping to 17% when looking at the GeForce RTX 4080 SUPER - a refresh that is only a year old. Looking back further to the GeForce RTX 3080, the new GeForce RTX 5080 is a whopping 88% faster - a massive improvement. As the third most powerful GPU for 4K gaming, the GeForce RTX 4090 is 16% faster than the COLORFUL iGAME GeForce RTX 5080 Ultra W OC, while the new flagship GeForce RTX 5090 is 47% faster. Looking at the flagship Radeon RX 7900 XTX, which could be AMD's most powerful option for another generation depending on how the Radeon RX 9070 XT pans out, the iGAME GeForce RTX 5080 Ultra W is a sizable 29% faster. For a more nuanced look at game-to-game performance, go through the individual benchmarks below, as the results can vary wildly from title to title. Call of Duty: Black Ops 6 runs 17% faster on the GeForce RTX 5080 compared to the GeForce RTX 4080; however, it's still 5% slower than the Radeon RX 7900 XTX - the only 4K win for AMD's flagship in our testing. Looking at two Unreal Engine 5 games, Black Myth: Wukong and competitive shooter Marvel Rival, both run 18% faster on the RTX 5080 than the RTX 4080. On the other end of the spectrum, Counter-Strike 2 runs 29% faster on the RTX 5080. Drop the resolution to 1440p, and the COLORFUL iGAME GeForce RTX 5080 Ultra W's lead over the GeForce RTX 4080 shrinks to 14%, with its lead over other GPUs also dropping - Radeon RX 7900 XTX (16%) and GeForce RTX 3080 (65%). At this resolution, we see the GPU run into more CPU and game engine bottlenecks as the GeForce RTX 5080's performance also sees its performance get within 10% of the GeForce RTX 4090. That said, when it comes to games with lots of ray-tracing like Cyberpunk 2077, Dragon Age: The Veilguard, and Hogwarts Legacy - dropping the resolution down to 1440p with DLSS enabled does deliver smooth triple-digit FPS gaming. 3DMark offers a suite of synthetic benchmarks built to test GPUs in various scenarios. 3DMark Steel Nomad is a cutting-edge DirectX 12 benchmark with newer, modern rendering techniques designed to push GPUs to their limit. The 'Light' version tests at 1440p, while the main Steel Nomad benchmark tests pure native 4K rendering. Port Royal is a benchmark focusing exclusively on real-time ray tracing for lighting effects like reflections, shadows, and more. 3DMark Steel Nomad is the successor to Time Spy and has been designed to test more modern non-RT rendering techniques, such as those used in Unreal Engine 5. In the 1440p-based 3DMark Steel Nomad Light synthetic benchmark, the COLORFUL iGAME GeForce RTX 5080 Ultra W's score is 20% higher than the GeForce RTX 4080's score - which is indicative of the in-game results we're seeing. Bump things up to the 4K-based 3DMark Steel Nomad, and that lead grows to 24%, a result we only see in a handful of games. Switching to the ray-tracing 3DMark Port Royal synthetic benchmark, the lead grows again to 26%; however, when it comes to the ray-tracing workloads in the games included in our benchmark suite, the figure we're seeing is closer to 20%. When it comes to super-intensive Path Tracing, cards like the GeForce RTX 4090 and the new GeForce RTX 5090 benefit from their increased VRAM and more RT Cores. DLSS 4 and Multi Frame Generation are impressive bits of technology, thanks mainly to the overall improvements to performance and latency on the Frame Generation side and the new 'Transformer' model for Super Resolution and Ray Reconstruction. We used the DLSS 'Quality' mode preset for these benchmarks, often delivering better-than-native image quality. Multi Frame Generation is an excellent and promising new technology designed to make the most out of high refresh-rate displays. As our GPU test bench includes testing on a 4K 240 Hz display, it's rare to see the game get close to taking full advantage of its capabilities. Multi Frame Generation, alongside NVIDIA Reflex for latency reduction, DLSS Super Resolution for upscaling, and RTX Blackwell 'Flip Metering' hardware, delivers a silky smooth experience as long as the baseline performance is around 70-80 FPS. As we conducted our tests using the DLSS Quality preset and the new Transformer model, dropping this to Balanced in 4K improves the overall experience in Cyberpunk 2077 and Hogwarts Legacy. Frame Generation (2X) and Multi Frame Generation (3X, 4X) require tinkering and testing to get all of the settings right - which is unfortunate for newcomers or those inexperienced with adjusting graphics settings in a PC game - but the effort is worth it. Even with some noticeable artifacts, the overall smoothness and improved motion clarity are hard to ignore. A game like Dragon Age: The Veilguard feels more immersive at 4K 240 FPS than it does running 4K 65 FPS - and compared to native, the latency is lower when you take the 90+ DLSS Quality input. Frame Generation and Multi Frame Generation will be a big part of PC gaming's future, especially when it comes to Full Ray Tracing or Path Tracing. Path Tracing, or Full Ray Tracing, arrived with the GeForce RTX 40 Series and DLSS 3 and is leveling up with the GeForce RTX 50 Series and DLSS 4. It's only possible thanks to AI technologies like DLSS Super Resolution, Ray Reconstruction, and RTX Neural Shader technology like RTX Mega Geometry. It's designed specifically for these technologies, and we're only including native or rasterized performance to highlight just how intensive it is on a GPU as powerful as the GeForce RTX 5080. In fact, outside of the massive increase in performance, these games also look notably worse without DLSS 4. Path Tracing is more than looking at an RTX Off and RTX On comparison clip or screenshot and spotting the difference; it's about being immersed in a jaw-dropping digital world. The magic of realistic lighting doesn't come from a single scene or moment but a collection of moments that add up to deliver a seamless world to explore and get lost in. It's cutting-edge technology and perfect for cinematic titles that push the visual envelope. Alan Wake 2 with DLSS 4, the new Transformer model, and RTX Mega Geometry are sights to behold, as are the lush forests and environments of Black Myth: Wukong. The critically acclaimed Indiana Jones and the Great Circle has yet to get its big DLSS 4 update. It perfectly exemplifies what Full Ray Tracing can bring to a game's visuals and cinematic immersion. This game looks excellent running on an Xbox console or a mainstream GPU like the GeForce RTX 4060. However, jump over to the COLORFUL iGAME GeForce RTX 5080 Ultra W, enable DLSS and Frame Generation, and you've got a genuinely stunning title running at 100+ FPS that looks an entire generation or two ahead of the same game running in 4K on the Xbox Series X. With its 360W power rating, the COLORFUL iGAME GeForce RTX 5080 Ultra W draws more than 300W during heavy gaming workloads, sometimes going all the way up to 350W or so when overclocked. However, the GeForce RTX 5080's power management and scaling are impressive, so the average power usage is well below its capacity when gaming in 4K in a range of titles - especially when you enable DLSS 4. A max temperature of 65 degrees was recorded during our stress test, with audible but minimal noise output; the COLORFUL iGAME GeForce RTX 5080 Ultra W's cooling and thermal performance is impressive. Enough to warrant an additional manual overclock for more performance - even though you probably won't need it. Yes, we would have loved to have seen the GeForce RTX 5080 launch with GeForce RTX 4090 performance levels for $999. Its more modest 22% uplift over the GeForce RTX 4080 for 4K gaming presents better value for those upgrading from lesser hardware or something like the GeForce RTX 3080 or RTX 3090. Upgrading every GPU generation doesn't make sense for most people, even if the gen-on-gen uplift is 30-35%, but thanks to the GeForce RTX 40 Series' massive gen-on-gen gains, the GeForce RTX 5080 in COLORFUL iGAME GeForce RTX 5080 Ultra W form is 88% faster than the GeForce RTX 3080. Ultimately, these raw performance numbers have become more academic than helpful for daily gamers as technologies like DLSS continue improving and provide tools for enhancing performance and fidelity. When it comes to the GeForce RTX 50 Series, DLSS 4, and the promise of RTX Neural Rendering, it is already giving us a taste of what the AI-enhanced future of gaming will look like. DLSS and real-time ray-tracing have been around for a couple of GPU generations now; however, with the GeForce RTX 50 Series, it feels like we're entering the era where AI is not only becoming essential for boosting performance but also image fidelity, smoothness, and latency, and even rendering more CGI-like detail on the screen. Regarding custom GeForce RTX 5080 models, the COLORFUL iGAME GeForce RTX 5080 Ultra W is an excellent option if you're cool with the stylish and cool physical design. The thermal performance is great, with enough room to overclock and boost performance, and we're told the price will stick to the $999 MSRP set by NVIDIA.

MSI GeForce RTX 5090 SUPRIM LIQUID SOC Review - 4K Gaming Unleashed

When you think about a liquid-cooled GPU, you often think about custom water blocks and closed-loop systems with tubes, colored liquids, and plenty of RGB lighting. Or a PC modding project where you take a high-end GPU like the GeForce RTX 5090 or GeForce RTX 4090, pull it apart, and then rebuild it with a water block, fittings, and radiator. The exception to the dual or triple fan with heatsinks and vapor chambers rule. In the CPU space, the off-the-shelf AIO liquid-cooled solution is common; in the GPU space, you're limited to rare variants like the MSI GeForce RTX 5090 SUPRIM LIQUID SOC. A quick look at the MSI GeForce RTX 5090 SUPRIM LIQUID SOC The GeForce RTX 5090 is already what you'd consider a halo or luxury item for someone who wants the absolute best gaming, content creation, or even at-home AI performance. This is MSI's most premium GeForce RTX 5090, carrying an eye-watering price tag of around $2500 - a $500 premium over the MSRP. The SUPRIM LIQUID is a hybrid design in that the water block also has a fan, one of MSI's new STORMFORCE fans that you can find on the MSI GeForce RTX 5090 SUPRIM SOC we recently reviewed. It also ships with a large 360mm radiator - larger than previous SUPRIM LIQUID models from MSI, probably because it is an OC model rated at 600W. One of the first questions that come to mind when thinking about a water-cooled GPU, especially one as power-hungry as the GeForce RTX 5090, is whether or not it works in the sense that it's a more effective or just-as-effective solution as a traditional GPU cooler for a high-end model. The answer to that is a resounding yes. The MSI GeForce RTX 5090 SUPRIM LIQUID SOC is not only the most powerful gaming GPU we've ever encountered, but even during gaming sessions with lots of ray tracing, the GPU temperature barely cracked 60 degrees Celsius - a remarkable achievement for a 600W GPU. And when it comes to gaming with the GeForce RTX 5090, there's just nothing quite like it. As the successor to the GeForce RTX 4090, a GPU that is still an absolute beast in 2025, you're looking at 4K gaming performance that is 30% faster on average. Throw in DLSS 4's impressive updates and the new Multi Frame Generation, and you're able to play Alan Wake 2 with Full Ray Tracing on a 4K 240 Hz display like the MSI MAG 321UPX QD-OLED in our test suite, maxing out that the resolution and refresh rate. Yes, 4K 240 FPS. Below is a summary of NVIDIA's GeForce RTX 50 Series and RTX Blackwell architecture, applicable to all models. NVIDIA describes 'Neural Rendering,' which includes all previous versions of DLSS and the brand-new DLSS 4, as the 'next era for computer graphics.' They're not alone; the Lead System Architect for the PlayStation 5 Pro console, Mark Cerny, recently said that ray-tracing is the future of games and that AI will play an integral role in making that happen. DOOM: The Dark Ages developer id Software shared a similar sentiment, adding that the arrival of DLSS was an 'inflection point' for PC game visuals and performance and on par with the arrival of dedicated GPUs and programmable shaders. With the arrival of the Blackwell generation and the GeForce RTX 50 Series, AI is now being used to accelerate programmable shaders with the brand-new RTX Neural Shaders. Yes, these are actual neural networks that use live game data, and the power of Tensor Cores to do everything from compress textures, render lifelike materials with a level of detail impossible to match using traditional rendering methods, and even use AI to partially trace rays and then infer "an infinite amount of rays and bounces for a more accurate representation of indirect lighting in the game scene." RTX Mega Geometry is incredible in its own right; it essentially increases a scene's geometry detail and complexity (triangles or polygons) by up to 100x. 100 times the detail, it's hard to wrap your head around - but the added benefit in a game like Alan Wake 2 is dramatically improving the performance of the game's Full Ray Tracing or Path Tracing mode. With DLSS 4 and RTX Neural Shaders, NVIDIA's GeForce RTX 50 Series and RTX Blackwell architecture (which includes the same AI optimizations as data center Blackwell) can be viewed as the turning point for PC gaming - the moment when AI becomes integral to everything from designing a game to programming and then finally rendering it on a 4K display to play. DLSS 4 includes more goodies than NVIDIA's highly touted new Multi Frame Generation technology, but let's start there. DLSS 3's version of Frame Generation has evolved with DLSS 4, powered by Blackwell hardware and software, and an innovative use of AI to generate frames 40% faster while using 30% less VRAM. Switching to a new model also means that Frame Generation and Multi-Frame Generation could soon come to GeForce RTX 20, 30, and RTX 40 Series owners. DLSS 4 benefits all GeForce RTX gamers. However, with the 5th Generation of Tensor Cores in the GeForce RTX 50 Series delivering 2.5X more AI performance, NVIDIA's latest GeForce RTX 50 Series GPUs can execute five complex AI models - covering Super Resolution, Ray Reconstruction, and Multi Frame Generation in a couple of milliseconds. Part of the reason it happens so quickly is the addition of hardware Flip Metering, which shifts frame pacing to the Blackwell display engine - the result is frame rates of up to 4K 240 FPS and higher without stuttering issues. With up to 15 of every 16 pixels generated by AI, the result is up to 8X the performance when compared to native rendering or rasterized performance. DLSS Super Resolution and Ray Reconstruction are also switching to a new 'Transformer' model, with over double the parameters and four times the compute requirement. This is one of the most exciting aspects of the GeForce RTX 50 Series, as it pushes DLSS into a new realm of image quality and performance. The best part is that it will work on all GeForce RTX GPUs; however, there will be a performance hit compared to running it on an RTX 50 Series GPU. Already available in games, DLSS 4's Transformer model is another DLSS 2.0-like moment for the technology, and the results speak for themselves. Even better, DLSS 4 is being integrated into the NVIDIA App with a new 'DLSS Override' feature that allows users to experience the latest tech without waiting for a path or game update. DLSS 4 is built to be backward compatible, with 75 games and apps supported. It doesn't stop there, as the new AI Management Processor (AMP) allows AI models to share the GPU with graphics workloads. As a result, expect to see digital humans in games alongside AI assistants like NVIDIA's Project G-Assist become more prevalent in the coming years. This filters down to the creator side, with AI assistants for streamers, who will also benefit from the GeForce RTX 50 Series' expanded creator features. RTX Blackwell introduces 4:2:2 chroma-sampled video encoding and decoding. The ninth-generation NVENC encoder also improves AV1 and HEVC quality. The flagship GeForce RTX 5090 supports up to three encoders and two decoders to deliver a 50% gen-over-gen improvement in speed compared to the GeForce RTX 4090. For creators and editors, the RTX Blackwell is a game changer. The new low-voltage and cutting-edge GDDR7 memory dramatically improves memory bandwidth and speed. Here's a look at the specs for the flagship GeForce RTX 50 Series GPUs, the GeForce RTX 5090 and GeForce RTX 5080, compared to the previous Ada generation. The RTX Blackwell generation is an anomaly because it's the first generation of GeForce RTX hardware to be built on a process node similar to the outgoing generation. Although the custom TSMC 4N process has been modified or optimized for the new GeForce RTX 50 Series, without a process shrink to 3nm or even 2nm, this means the raw performance uplift compared to the GeForce RTX 40 Series is a little underwhelming. As seen in our review of the GeForce RTX 5090 Founders Edition model, when looking at 4K gaming, you can expect around 27% more performance - which is decent but not mind-blowing. However, the GPU inside the GeForce RTX 5090 is an absolute beast - it's the most densely packed and powerful gaming silicon ever produced. 92.2 billion transistors on a 750mm die, with 21760 CUDA Cores and 32GB of GDDR7 memory on a 512-bit bus with an overall bandwidth of 1792 GB/sec. The fifth-gen Tensor cores can deliver double the AI performance of the GeForce RTX 4090 - thanks to native FP4 support. With 33% more silicon for rendering and ray-tracing, the power rating has shot up to 575W, or 600W, when using the MSI GeForce RTX 5090 SUPRIM LIQUID SOC's overclocked mode. This is essentially the limit of a single 16-pin power connector, and you'll want at least a 1000W PSU to drive this thing - we recommend 1200W so you've got some headroom. The good news is that, on average, the GeForce RTX 5090 doesn't draw 600W when gaming, with the average 4K power draw of the MSI GeForce RTX 5090 SUPRIM LIQUID SOC sitting much lower. The SUPRIM LIQUID SOC model ships with an out-of-the-box Boost Clock speed of 2565 MHz - 158 MHz higher than the reference design. In our testing, we noticed that in some games, this led to a notable increase in overall performance, including the higher 1% Low FPS performance for a smoother and more stable gaming experience. MSI GeForce RTX 5090 SUPRIM LIQUID SOC is the sort of GPU you build an entire system around, as the large 360mm radiator needs to be installed in a case alongside your CPU's AIO or air-cooler and other system fans. The benefit not only comes from having an RTX 5090 that runs significantly more remarkable than the Founders Edition model (check out our thermal performance summary for more information), but the actual water block is smaller and lighter - offering a much cleaner-looking system. Again, this is a GPU that you build an entire system around. As a hybrid design with some passive cooling thanks to a single STORMFORCE fan sitting on the GPU casing, MSI has been cautious with the design - which we're chalking up to the 600W power rating. The MSI GeForce RTX 5090 SUPRIM LIQUID SOC GPU is 2.5 slots thick and longer than the GeForce RTX 4090 SUPRIM LIQUID design. It is still one of the smallest GeForce RTX 5090 models on the market once you remove the space required for the radiator, so once fully installed, it's a sleek and stylish look that still maintains the SUPRIM aesthetic and upgraded look with RGB for the SUPRIM label and logo. The water block design employed by MSI is patented with a high-performance pump designed to keep the GPU cool with the hybrid cooling system. With clay-based thermal pads, a thickened layer of copper inside the PCB, and dual BIOS 'Gaming' and 'Silent' modes, this is easily one of the most impressively designed GeForce RTX 5090 GPUS. The proof is in the performance; the MSI GeForce RTX 5090 SUPRIM LIQUID SOC is faster than the Founders model and can keep the overall GPU temperature at around 55 degrees under load. For a 600W GPU, this is a remarkable result. PC gaming not only covers a wide range of genres and styles, from indie games with simple 2D graphics to massive 3D worlds lit by cutting-edge real-time ray tracing technology. With that, the needs and requirements of each gamer vary. High refresh rates and latency reduction become more important than flashy visuals or playing at the highest resolution possible for those who live and breathe fast-paced competitive games. For those who want to live in a cinematic world and become a key player in an expansive narrative, ray-tracing and high-fidelity visuals are a stepping stone toward immersion. Our chosen benchmarks cover various games, engines, APIs, and technologies. For the GeForce RTX 5090, all tests are run at 4K and 1440p and include results for performance-boosting Super Resolution technologies like NVIDIA DLSS 4 - including Frame Generation and the new Multi Frame Generation. In many ways, DLSS numbers are more important in 2025 than native rendering - a title with ray tracing isn't meant to be played without Super Resolution. Also, DLSS technologies like Ray Reconstruction and the new RTX Mega Geometry dramatically improve visual fidelity and detail compared to native rendering. However, our benchmark results are still sorted using 'raw performance' or native rendering. Here's the breakdown of games, settings, and what's being tested. As seen with the traditional air-cooled GeForce RTX 5090 SUPRIM SOC, we reviewed the SUPRIM LIQUID is 30% faster than the GeForce RTX 4090 for 4K gaming while also being around 3-5% faster than the GeForce RTX 5090 Founders Edition model in specific gaming workloads. Even more impressive is that the 1% low performance is 32% faster than the GeForce RTX 4090 for 4K gaming. You're also getting a smoother and more stable gaming experience with MSI's premium liquid-cooled variant. Now, you might be thinking that this is an impressive result, but it is not mind-blowing because the GeForce RTX 4090 provided one of those rare generational uplifts that was precisely that - mind-blowing. Well, as we've now had time to put the Ampere-powered GeForce RTX 3090 to the test - the MSI GeForce RTX 5090 SUPRIM LIQUID SOC is, on average, a whopping 122% faster for 4K gaming. Yeah, just let that figure sink in. If you're looking to upgrade from the RTX 3090 to the RTX 5090, you can expect some truly incredible performance improvements. Based on the individual game results, we see some evidence of certain games running into a CPU or game engine bottleneck when gaming in 4K on the MSI GeForce RTX 5090 SUPRIM LIQUID SOC. Dragon Age: The Veilguard runs 37.5% faster than the GeForce RTX 4090 with 38% faster 1% Low performance - an excellent result for MSI's premium GeForce RTX 5090. On the other hand, Warhammer 40K: Space Marine 2 is only 16% faster, with 10% better than 1% Low performance, which makes the GeForce RTX 5090 the only GPU running into any bottleneck with this game. There are a few more examples like this where some games are running 35-40% faster, with others only running 15-20% faster than the GeForce RTX 4090. With DLSS Super Resolution enabled using the Quality preset, plus the new and impressive DLSS 4 Transformer model in titles where that's available, the average 4K gaming performance of the MSI GeForce RTX 5090 SUPRIM LIQUID SOC sits at 163 FPS. However, DLSS for the RTX 5090 means that because the internal resolution is 1440p, the average performance gains over the GeForce RTX 4090 drops to around 25% from 30%. Power consumption also drops when you enable DLSS 4, so it's still a net positive because you're getting better image quality, more performance, and more efficient GPU. The MSI GeForce RTX 5090 SUPRIM LIQUID SOC is the fastest GPU on the planet for all gaming, including 1440p. However, dropping the resolution from 4K means the GeForce RTX 5090 can hit the brakes, draw less power, and essentially switch on cruise control. This means that some games will only run 10% faster than on the GeForce RTX 4090, while others will still see that 25%+ performance uplift. Seeing Counter-Strike 2 hit 421 FPS with Very High settings on a stress test map is jaw-dropping. With a display that can hit this refresh rate, you're looking at unprecedented motion clarity in one of the most popular competitive shooters. Likewise, with the MSI GeForce RTX 5090 SUPRIM LIQUID SOC, you're looking at the only GPU (alongside other RTX 5090s, of course) that can run Cyberpunk 2077 with Ultra quality ray-tracing natively at 100+ FPS with detail settings cranked at 1440p. However, as this is a title with DLSS 4 support that improves image quality, 108 FPS jumps up to 135FPS - and that's without Frame Generation or the new Multi Frame Generation. Ultimately, the GeForce RTX 5090 is meant for 4K gaming on a high refresh-rate display, so pairing it with a 1440p display would be a mistake. 3DMark offers a suite of synthetic benchmarks built to test GPUs in various scenarios. 3DMark Steel Nomad is a cutting-edge DirectX 12 benchmark with newer, modern rendering techniques designed to push GPUs to their limit. The 'Light' version tests at 1440p, while the main Steel Nomad benchmark tests pure native 4K rendering. Port Royal is a benchmark focusing exclusively on real-time ray tracing for lighting effects like reflections, shadows, and more. The 3DMark synthetic benchmark results for the MSI GeForce RTX 5090 SUPRIM LIQUID SOC showcase a much higher difference in overall performance than our in-game tests. The 3DMark Steel Nomad score is a whopping 59% higher than the GeForce RTX 4090's score, and the ray-tracing-based Port Royal score is 48% higher. We're not sure why this is; it could come down to these workloads being more indicative of workloads where the RTX 5090 comes into its own. There are titles where we see a 40% uplift with specific settings at 4K, so perhaps we'll see this improvement with new games coming in 2025. Maybe a few driver updates from NVIDIA will unlock some additional performance in the coming months. Either way, with the new RTX Blackwell architecture, including optimizations for new RTX Neural Rendering technologies, there's more to come regarding performance and image fidelity. DLSS 4 and Multi Frame Generation are impressive bits of technology, thanks mainly to the overall improvements to performance and latency on the Frame Generation side and the new 'Transformer' model for Super Resolution and Ray Reconstruction. We used the DLSS 'Quality' mode preset for these benchmarks, often delivering better-than-native image quality. With the GeForce RTX 50 Series, NVIDIA's new Multi Frame Generation technology is currently limited to the new cards because it is powered by the latest hardware and AI optimizations that deliver up to 3 additional AI-generated frames with only a minor increase in latency. For the MSI GeForce RTX 5090 SUPRIM LIQUID SOC, with Multi Frame Generation, you can run Cyberpunk 2077, Dragon Age, and Hogwarts Legacy in 4K with Ultra quality ray tracing at 4K 240 Hz with plenty of headroom. This immediately ramps up the level of immersion you feel when playing these visually stunning titles; the presentation is not only smoother, but motion clarity is dramatically improved. With NVIDIA Reflex and DLSS 4 Super Resolution and Ray Reconstruction (supported on all GeForce RTX GPUs), you're also getting a more detailed image and higher-quality ray tracing effects, such as more detailed reflections, shadows, and object detail. The improved Frame Generation and the new Multi Frame Generation might not be suitable for all games and scenarios. Still, when enabled in single-player games like Cyberpunk 2077, Dragon Age, and Hogwarts Legacy can be a game changer. Sure, the 360 FPS in Dragon Age has more latency than running the game natively at 360 FPS, but that's not something you notice. The overall latency is excellent, so the trade-off is worth it. DLSS 4's Multi Frame Generation isn't perfect; you see some artifacts, but it's a technology built for visually intensive games and maximizing your display's refresh rate. Path Tracing, or Full Ray Tracing, arrived with the GeForce RTX 40 Series and DLSS 3 and is leveling up with the GeForce RTX 50 Series and DLSS 4. It's only possible thanks to AI technologies like DLSS Super Resolution, Ray Reconstruction, Frame Generation, and RTX Neural Shader technology like RTX Mega Geometry. It's designed specifically for these technologies, and we're only including native or rasterized performance to highlight just how intensive it is on a GPU as powerful as the GeForce RTX 5090. In fact, outside of the massive increase in performance, these games also look notably worse without DLSS 4. Alan Wake 2 with Full Ray Tracing is probably the best game to showcase the potential of the GeForce RTX 5090 because the build we used for benchmarking included the full DLSS 4 suite of AI-powered technologies alongside the new RTX Mega Geometry technology - which developer Remedy uses to add a new 'Ultra' quality preset for its Path Tracing. With DLSS 4 without Frame Generation, the MSI GeForce RTX 5090 SUPRIM LIQUID SOC is 40% faster than the GeForce RTX 4090; with Multi Frame Generation, it's over 2X faster. Playing Alan Wake 2 with this level of visual fidelity at 4K 240 Hz is one of those 'seeing is believing' moments - it's a glimpse at what gaming might look like on the PlayStation 7 or 8, let alone PlayStation 6. Path Tracing is more than a step or two above the ray-tracing we see in titles like Resident Evil 4 or F1 24's suite of in-race effects. Unless AMD's RDNA 4 delivers a monumental increase in RT performance and FSR image quality, it'll stay exclusive to the GeForce RTX line-up - with only GPUs as powerful as the GeForce RTX 4080 and RTX 5080 able to provide a playable experience in 4K. With Indiana Jones set to get a DLSS 4 upgrade soon and id Software's DOOM: The Dark Ages set to support Full Ray Tracing with DLSS 4 and Multi Frame Generation support, we're going to see more and more games offer this next-gen level of image fidelity and immersion as we head into 2025 and beyond. The MSI GeForce RTX 5090 SUPRIM LIQUID SOC's 600W TGP is unprecedented for a GPU and a sizable step up from the 450W GeForce RTX 4090. In a way, it feels like heading into uncharted waters, which makes the thermal performance we see here pretty incredible. MSI's SUPRIM LIQUID hybrid water-cooled solution only hit a GPU temperature of 55 degrees Celsius under load, which goes to show you that it's not only possible to cool a GPU as power-hungry as the GeForce RTX 5090 but in a premium build like this - you're looking at lower temperatures than most mainstream low-power gaming cards. The good news is that the MSI GeForce RTX 5090 SUPRIM LIQUID SOC includes a 'Silent' BIOS that drops the power limit to 575W. Regardless of the mode, it will not consistently draw 600W when gaming. The 485W average power usage we saw across all 14 games running in 4K that we benchmarked is slightly lower than the air-cooled SUPRIM SOC we reviewed. And when you add in the lower temps, the GeForce RTX 5090 is tailor-made for liquid cooling setups. When we reviewed the GeForce RTX 5090 Founders Edition, we couldn't test as many GPUs as we wanted for comparison, with the main two being the GeForce RTX 5090 and the Radeon RX 7900 XTX. As far as OC models go, the MSI GeForce RTX 5090 SUPRIM LIQUID SOC is an absolute beast, with 4K gaming performance that is 30% faster than the GeForce RTX 4090 and 94% faster than the Radeon RX 7900 XTX. Yes, we're including ray-tracing results because the technology is everywhere now - including PS5 and Xbox Series consoles. However, with the GeForce RTX 5080 out in the wild, the MSI GeForce RTX 5090 SUPRIM LIQUID SOC is 56% faster for 4K gaming. The GeForce RTX 5080 is a great 4K gaming GPU, so you can see why the MSRP has risen to $2000 for the RTX 5090. Also, we've had some additional time in the past week to test the flagship Ampere-generation GeForce RTX 3090, and here, the MSI GeForce RTX 5090 SUPRIM LIQUID SOC is a massive 122% faster for 4K gaming. Now, most people don't upgrade or build a new rig every single GPU generation, so for those still rocking a $1500 GeForce RTX 3090 (the price of the 3090 in 2020), the GeForce RTX 5090 is a giant leap forward for gaming performance. Of course, MSI's premium liquid-cooled model's $2500 price tag makes it a product that very few gamers will rush out to purchase; there's more to RTX Blackwell than running Alan Wake 2 at 4K 240 FPS with Full Ray Tracing and DLSS 4. After going hands-on with several demos at CES, we're pretty confident that RTX Blackwell's Neural Rendering capabilities will usher in a new era of real-time graphics - it's just a matter of how long before we see in more than a handful of games. For content creators, there's a significant improvement to video production, with 4:2:2 chroma-sampled video encoding, enhanced AV1 and HEVC quality, DisplayPort 2.1, and more. AI enthusiasts will also be eying the 3352 AI TOPS of performance and 32GB of GDDR7 memory for training and inferencing. The GeForce RTX 5090 is the full package, and in MSI SUPRIM LIQUID form, it is something to behold.

INNO3D GeForce RTX 5080 X3 Review - SFF Ready for 4K

When it comes to gaming GPUs, the general rule is that the higher up the stack you go in terms of configuration and performance, the bigger the component gets. In the GeForce RTX 5080 realm, it's not uncommon to find cards with complex cooling solutions, three large fans, and enough weight and thickness that a GPU brace is required for installing it in the traditional horizontal orientation. Compact high-end enthusiast 4K gaming GPUs are rare, which makes the new INNO3D GeForce RTX 5080 X3 so impressive - it's an SFF-Ready card that won't have any problem fitting into a small gaming build. A quick look at the INNO3D GeForce RTX 5080 X3 Considerably lighter and more compact than even NVIDIA's GeForce RTX 5080 Founders Edition model, INNO3D's design is perfect for putting together a compact 4K gaming rig - perhaps something that could sit underneath a large 4K OLED TV next to a console like the PlayStation 5. That's the idea that first popped into my head when connecting the INNO3D GeForce RTX 5080 X3 up after testing a number of larger GeForce RTX 50 Series GPUs. Of course, going smaller means that, as an MSRP model, there's no out-of-the-box overclocking. There's undoubtedly a little wiggle room to push performance a little higher; however, that will necessitate a more aggressive fan curve than the almost silent profile you get. The INNO3D GeForce RTX 5080 X3 is every bit the 4K gaming GPU as other models, delivering excellent but not mind-blowing performance at this resolution. That comes down to the modest gen-on-gen uplift NVIDIA's latest 80-class GPU offers over the GeForce RTX 4080, with INNO3D's MSRP reference spec design performing around 19% faster than the RTX 4080 and 15% faster than the RTX 4080 SUPER. This comes from our 14-game average looking at raw 4K gaming performance across a wide range of titles. However, it still has an excellent performance and offers greater overall value to gamers at half the price of the flagship GeForce RTX 5090. With the arrival of DLSS 4, NVIDIA's suite of AI-powered tools for boosting performance and image fidelity has been leveled up for the GeForce RTX 50 Series. The new Transformer model for DLSS Super Resolution and Ray Reconstruction is a game changer. Multi Frame Generation delivers on its promise of smooth and responsive high refresh-rate gameplay. And with RTX Neural Rendering on the horizon, games will not only run faster on a GPU like the INNO3D GeForce RTX 5080 X3 but look better, too. As a compact GPU perfect for an SFF build, DLSS improves the overall efficiency for a cooler and quieter PC gaming experience. Below is a summary of NVIDIA's GeForce RTX 50 Series and RTX Blackwell architecture, applicable to all models. NVIDIA describes 'Neural Rendering,' which includes all previous versions of DLSS and the brand-new DLSS 4, as the 'next era for computer graphics.' They're not alone; the Lead System Architect for the PlayStation 5 Pro console, Mark Cerny, recently said that ray-tracing is the future of games and that AI will play an integral role in making that happen. DOOM: The Dark Ages developer id Software shared a similar sentiment, adding that the arrival of DLSS was an 'inflection point' for PC game visuals and performance and on par with the arrival of dedicated GPUs and programmable shaders. With the arrival of the Blackwell generation and the GeForce RTX 50 Series, AI is now being used to accelerate programmable shaders with the brand-new RTX Neural Shaders. Yes, these are actual neural networks that use live game data, and the power of Tensor Cores to do everything from compress textures, render lifelike materials with a level of detail impossible to match using traditional rendering methods, and even use AI to partially trace rays and then infer "an infinite amount of rays and bounces for a more accurate representation of indirect lighting in the game scene." RTX Mega Geometry is incredible in its own right; it essentially increases a scene's geometry detail and complexity (triangles or polygons) by up to 100x. 100 times the detail, it's hard to wrap your head around - but the added benefit in a game like Alan Wake 2 is dramatically improving the performance of the game's Full Ray Tracing or Path Tracing mode. With DLSS 4 and RTX Neural Shaders, NVIDIA's GeForce RTX 50 Series and RTX Blackwell architecture (which includes the same AI optimizations as data center Blackwell) can be viewed as the turning point for PC gaming - the moment when AI becomes integral to everything from designing a game to programming and then finally rendering it on a 4K display to play. DLSS 4 includes more goodies than NVIDIA's highly touted new Multi Frame Generation technology, but let's start there. DLSS 3's version of Frame Generation has evolved with DLSS 4, powered by Blackwell hardware and software, and an innovative use of AI to generate frames 40% faster while using 30% less VRAM. Switching to a new model also means that Frame Generation and Multi-Frame Generation could soon come to GeForce RTX 20, 30, and RTX 40 Series owners. DLSS 4 benefits all GeForce RTX gamers. However, with the 5th Generation of Tensor Cores in the GeForce RTX 50 Series delivering 2.5X more AI performance, NVIDIA's latest GeForce RTX 50 Series GPUs can execute five complex AI models - covering Super Resolution, Ray Reconstruction, and Multi Frame Generation in a couple of milliseconds. Part of the reason it happens so quickly is the addition of hardware Flip Metering, which shifts frame pacing to the Blackwell display engine - the result is frame rates of up to 4K 240 FPS and higher without stuttering issues. With up to 15 of every 16 pixels generated by AI, the result is up to 8X the performance when compared to native rendering or rasterized performance. DLSS Super Resolution and Ray Reconstruction are also switching to a new 'Transformer' model, with over double the parameters and four times the compute requirement. This is one of the most exciting aspects of the GeForce RTX 50 Series, as it pushes DLSS into a new realm of image quality and performance. The best part is that it will work on all GeForce RTX GPUs; however, there will be a performance hit compared to running it on an RTX 50 Series GPU. Already available in games, DLSS 4's Transformer model is another DLSS 2.0-like moment for the technology, and the results speak for themselves. Even better, DLSS 4 is being integrated into the NVIDIA App with a new 'DLSS Override' feature that allows users to experience the latest tech without waiting for a path or game update. DLSS 4 is built to be backward compatible, with 75 games and apps supported. It doesn't stop there, as the new AI Management Processor (AMP) allows AI models to share the GPU with graphics workloads. As a result, expect to see digital humans in games alongside AI assistants like NVIDIA's Project G-Assist become more prevalent in the coming years. This filters down to the creator side, with AI assistants for streamers, who will also benefit from the GeForce RTX 50 Series' expanded creator features. RTX Blackwell introduces 4:2:2 chroma-sampled video encoding and decoding. The ninth-generation NVENC encoder also improves AV1 and HEVC quality. The flagship GeForce RTX 5090 supports up to three encoders and two decoders to deliver a 50% gen-over-gen improvement in speed compared to the GeForce RTX 4090. The GeForce RTX 5080 adds a second decoder compared to the GeForce RTX 4080. The RTX Blackwell is a game changer for creators and editors, especially with the new low-voltage and cutting-edge GDDR7 memory that dramatically improves memory bandwidth and speed. Here's a look at the specs for the flagship GeForce RTX 50 Series GPUs, the GeForce RTX 5080 and GeForce RTX 5090, compared to the previous Ada generation. The 90-class flagship, halo, or Titan GPU in the GeForce line-up is still relatively new - debuting with the GeForce RTX 3090 in 2020. As we saw with the GeForce RTX 40 Series, the difference between the RTX 4090 and the RTX 4080, spec-wise, grew considerably, and that trend continues with the new GeForce RTX 5080 and RTX 5090. On paper, the GeForce RTX 5080 looks like half of the GPU the GeForce RTX 5090 is, with 50% fewer CUDA Cores, Tensor Cores, RT Cores, and even memory capacity - going from 32GB of GDDR7 on the RTX 5090 down to 16GB of GDDR7 on the RTX 5080. And with the same or similar custom TSMC process node as the GeForce RTX 40 Series, the RTX Blackwell-powered GeForce RTX 5080 only features 11% more CUDA Cores, Tensor Cores, and RT Cores than its previous-generation counterpart. The good news is that the GeForce RTX 5080 arrives at half the price of the flagship GeForce RTX 5090, with an MSRP that is around 20% lower than the GeForce RTX 4080 and on par with 2024's GeForce RTX 4080 SUPER refresh. And even though we haven't got that process node shrink this time (a rarity in the GPU space), the RTX Blackwell architecture includes several design changes built around accelerating AI performance for things like DLSS 4 and the new RTX Neural Rendering technologies set to debut this year. The GeForce RTX 5080 includes 16GB of faster GDDR7 memory, which delivers a memory bandwidth of 960 GB/sec - 35% faster than the GeForce RTX 4080's GDDR6X memory. With the introduction of FP4, you've gotten up to double the AI performance of the previous generation. Throw in faster clock speed, and the GeForce RTX 5080 delivers a decent but not mind-blowing performance uplift over the RTX 4080. However, the RTX 4080 brought with it a pretty significant performance increase over the RTX 3080 - for those that upgrade every other generation (which is more common than upgrading every single GPU generation), the INNO3D GeForce RTX 5080 X3 is an excellent choice. 84% faster for 4K gaming and 62% for 1440p gaming. As an MSRP model, the INNO3D GeForce RTX 5080 X3 doesn't ship with additional overclocking, so the overall performance is on par with NVIDIA's reference Founders Edition model. With physical dimensions of 300 x 116 x 41mm and roughly 1080 grams, the INNO3D GeForce RTX 5080 X3 is less than half the size and weight of some GeForce RTX 5080 models we've reviewed. With its compact two-slot thickness and SFF-Ready form factor, it's a GPU perfect for compact builds or those wanting a card that takes up as much room as a GeForce RTX 4060 or RTX 4070. With a more compact build, the three 88mm fans are much smaller than the fans you'll find on large models - however, the difference here is size-related primarily and not quality. With a copper base for the GPU and an aluminum base for the memory, six heat pipes, and generous fins with a metal backplate and vent system, the INNO3D GeForce RTX 5080 X3 has been designed to cool the GeForce RTX 5080 effectively, and its increased power limit of 360W. The default thermal profile is tuned for silent performance, so it's not the sort of GPU you'll hear working overtime to keep the temperatures in check. In practice, this means it runs warmer than NVIDIA's Founders Edition and other premium OC models. Still, it never gets to the point where you absolutely need to increase fan speeds or worry about thermal throttling. The design is also clean and minimal, free from RGB lighting, and has a two-tone black and grey look that is robust and great. INNO3D also offers a White variant, so you're covered if you're putting together a white SFF build. It's a design that we loved when it came to the RTX 40 SUPER Series launch a year ago, and the refinements and changes here are all for the better. However, the 'INNO3D' branding on the fans is an obvious sticker - a criticism we'd love to see corrected in future models. PC gaming not only covers a wide range of genres and styles, from indie games with simple 2D graphics to massive 3D worlds lit by cutting-edge real-time ray tracing technology. With that, the needs and requirements of each gamer vary. High refresh rates and latency reduction become more important than flashy visuals or playing at the highest resolution possible for those who live and breathe fast-paced competitive games. For those who want to live in a cinematic world and become a key player in an expansive narrative, ray-tracing, and high-fidelity visuals are a stepping stone toward immersion. Our chosen benchmarks cover various games, engines, APIs, and technologies. For the GeForce RTX 5090, all tests are run at 4K and 1440p and include results for performance-boosting Super Resolution technologies like NVIDIA DLSS 4 - including Frame Generation and the new Multi Frame Generation. In many ways, DLSS numbers are more important in 2025 than native rendering - a title with ray tracing isn't meant to be played without Super Resolution. Also, DLSS technologies like Ray Reconstruction and the new RTX Mega Geometry dramatically improve visual fidelity and detail compared to native rendering. However, our benchmark results are still sorted using 'raw performance' or native rendering. Here's the breakdown of games, settings, and what's being tested. The INNO3D GeForce RTX 5080 X3's 4K gaming performance is excellent, with an average frame rate of 92 FPS - which increases to 114 FPS when looking at DLSS performance using the 'Quality' Super Resolution preset. This is enough to make it roughly 19% faster than the GeForce RTX 4080 and 84% faster than the GeForce RTX 3080. It's also 26% faster than AMD's flagship Radeon RX 7900 XTX. However, NVIDIA's new GeForce RTX 5090 is still around 50% faster for 4K gaming. When it comes to desktop GPUs for 4K gaming, the GeForce RTX 5080 comes in third regarding performance but first regarding value for money. Results vary depending on the game, the genre, and whether or not it includes ray-tracing or DLSS Super Resolution. Call of Duty: Black Ops 6 is a title that favors Radeon GPUs, which makes it the only game where the INNO3D GeForce RTX 5080 X3 falls short of Radeon RX 7900 XTX's performance. However, take a look at Cyberpunk 2077 with the Ultra quality ray tracing preset, and the GeForce RTX 5080 is 21% faster than the GeForce RTX 4080 and a whopping 75% faster than the Radeon RX 7900 XTX. Look at the RTX 5080 versus the RTX 4080, F1 24 runs 14% faster, Hogwarts Legacy runs 20% faster, and Dragon Age: The Veilguard runs 25% faster. Competitive gaming on the INNO3D GeForce RTX 5080 X3 is also god-tier in 4K, with Counter-Strike 2 running 27% faster on the RTX 5080 than the RTX 4080. Looking at two of the biggest games of 2024, Warhammer 40K: Space Marine 2 runs 27% faster in 4K, while Black Myth: Wukong runs 17.5% faster. Dropping the resolution to 1440p, you're looking at diminishing returns for the INNO3D GeForce RTX 5080 X3, GeForce RTX 4090, and GeForce RTX 5090. Thanks to CPU bottlenecks and other limitations, when you start gaming with a GPU like this, you're better off pairing it with a 4K display or an Ultrawide - which has a pixel count between 1440p and 4K. On the plus side, the GeForce RTX 5080 becomes super efficient at this resolution, with hardly any gaming workloads drawing more than 280W. At 1440p, the GeForce RTX 5080's performance uplift over the GeForce RTX 4080 isn't all that prominent or even noticeable in games like Warhammer 40K: Space Marine 2, Call of Duty: Black Ops 6, and DOOM Eternal. In ray-tracing titles or visually demanding Unreal Engine 5 games, a definite improvement comes from gaming on the GeForce RTX 5080. However, the generational improvement drops from around 19% to 12%. 3DMark offers a suite of synthetic benchmarks built to test GPUs in various scenarios. 3DMark Steel Nomad is a cutting-edge DirectX 12 benchmark with newer, modern rendering techniques designed to push GPUs to their limit. The 'Light' version tests at 1440p, while the main Steel Nomad benchmark tests pure native 4K rendering. Port Royal is a benchmark focusing exclusively on real-time ray tracing for lighting effects like reflections, shadows, and more. 3DMark Steel Nomad is the official Time Spy replacement, the latest synthetic gaming benchmark designed to push modern GPUs to their limit with cutting-edge effects and rendering techniques in DirectX, Vulkan, and engines like Unreal Engine 5. Steel Nomad Light is a 1440p benchmark, and here, the INNO3D GeForce RTX 5080 X3 delivers a score that aligns with our 4K gaming results. The 4K-based Steel Nomad synthetic benchmark score for the INNO3D GeForce RTX 5080 X3 sees it outperform the GeForce RTX 4080 and RTX 4080 SUPER by 25% and 22%, respectively - a result that is more than a few percent higher than our in-game benchmarks. However, it's a result that can be replicated on an overclocked GeForce RTX 5080. Switching over to 3DMark Port Royal, a synthetic ray-tracing benchmark, we also see a slightly higher generational uplift than real-world RT gaming. However, when it comes to titles with cutting-edge Path Tracing like Alan Wake 2 - which we were able to test with both DLSS 4 and RTX Mega Geometry - the Port Royal results for the GeForce RTX 5080 versus the GeForce RTX 4080 and RTX 4080 SUPER line up perfectly. DLSS 4 and Multi Frame Generation are impressive bits of technology, thanks mainly to the overall improvements to performance and latency on the Frame Generation side and the new 'Transformer' model for Super Resolution and Ray Reconstruction. We used the DLSS 'Quality' mode preset for these benchmarks, often delivering better-than-native image quality. When you enable the entire suite of DLSS 4 technologies in a game running on the GeForce RTX 5080 in 4K or 1440p, which includes Super Resolution, Ray Reconstruction (in RT titles), and Multi Frame Generation, performance, smoothness, and even image quality is leveled up. It's not the same as rendering a game natively at a particular resolution. Still, it is the future of PC and console gaming - as it also brings efficiency improvements and the ability to target higher resolutions and refresh rates. The new Transformer model for DLSS 4 Super Resolution is a game changer for image fidelity. It makes the new DLSS 'Performance' and 'Balanced' presets look as good as or better than the previous model's DLSS 'Quality' preset. It's undeniably an improvement, and even though some issues exist (it's still in Beta), it's a big reason to pick up a new GeForce RTX GPU over anything else. Multi-frame generation, exclusive to the GeForce RTX 50 Series, is also a selling point, but it requires knowledge to get the best experience. Paired with the right game and an input frame rate of at least 75 FPS, the result is like watching a magic trick unfold before your eyes. When tuned to match your display's refresh rate, it strikes the right balance between performance, smoothness, and responsiveness. Path Tracing, or Full Ray Tracing, arrived with the GeForce RTX 40 Series and DLSS 3 and is leveling up with the GeForce RTX 50 Series and DLSS 4. It's only possible thanks to AI technologies like DLSS Super Resolution, Ray Reconstruction, and RTX Neural Shader technology like RTX Mega Geometry. It's designed specifically for these technologies, and we're only including native or rasterized performance to highlight just how intensive it is on a GPU as powerful as the GeForce RTX 5080. In fact, outside of the massive increase in performance, these games also look notably worse without DLSS 4. Path Tracing or Full Ray Tracing is the future of visual fidelity in cinematic or AAA titles. To be playable in 2025 and beyond, it requires DLSS 4, frame generation, and the new RTX neural rendering suite of technologies. Alan Wake 2 is an excellent example of where the technology is headed in the next year or two, as it includes the brand-new RTX Mega Geometry technology for ray-tracing, which on its own improves performance by around 10-15% on all GeForce RTX GPUs. Throw in DLSS 4 and Multi Frame Generation, and you've got an unforgettable third-person psychological horror experience on the GeForce RTX 5080 that looks significantly better than how it looks running on the PlayStation 5 Pro. Indiana Jones and the Great Circle is another one of those night-and-day moments for image quality, where a great-looking first-person game becomes jaw-dropping once you turn on Full Ray Tracing. This game is getting its DLSS 4 update, which will also introduce RTX Hair technology (another piece of the RTX Neural Rendering suite), which will notably improve its visuals yet again. With 2X Frame Generation, it runs at 130+ FPS on the INNO3D GeForce RTX 5080 X3 to deliver a smooth, immersive experience. Running on id Tech technology, it's an early look at what we can expect from the highly anticipated DOOM: The Dark Ages - which is launching with DLSS 4 support, Full Ray Tracing, and RTX Neural Rendering. As the smallest and most compact GeForce RTX 5080 we've reviewed, the INNO3D GeForce RTX 5080 X3 also runs the hottest, with temperatures under load hitting 74 degrees Celsius. However, this is in line with the materials INNO3D sent us before our testing and review, which noted that as the GPU has been designed for near-silent performance, the GPU temperature will hit around 74 degrees during heavy gaming workloads. Interestingly, the idle power draw of this model compared to the Founders Edition design is higher, a trend we've noticed across various partner cards. With a modest generational uplift compared to the previous generation, the GeForce RTX 5080 is the first 80-class GeForce card to fall short of the last-generation flagship. NVIDIA hinted at this during the big GeForce RTX 50 Series reveal at CES 2025, and it's a little disappointing. But, when you've got this level of 4K gaming performance, alongside the arrival of the brilliant DLSS 4 and Multi Frame Generation, it's hard to stay disappointed after you fire up Cyberpunk 2077, Alan Wake 2, or even Counter-Strike 2. And for those with a GeForce RTX 3080 looking to upgrade, the INNO3D GeForce RTX 5080 X3 delivers - with monumental 4K gains that deliver on that next-gen promise. The GeForce RTX 5080 and the RTX 5090 also arrive when you can feel a change or shift in what gaming performance means. It took a few years, but NVIDIA DLSS has ushered in a new era of AI-enhanced gaming across PCs, laptops, and consoles like the PS5 Pro and Nintendo Switch 2. With NVIDIA leading the charge, it also means that the desktop PC still remains the absolute high-water mark for both image fidelity and performance. Games look better and run faster on the GeForce RTX 5080 than any card from AMD's Radeon line-up. Of course, this doesn't cover every game, but DLSS is everywhere, and DLSS 4 is the new gold standard for Super Resolution and Frame Generation. Although it's not the fastest model on the market, the compact SFF-Ready INNO3D GeForce RTX 5080 X3 is a brilliant choice for those looking to put together the ultimate small gaming rig that could slot in right next to that PS5 or Xbox Series console. From hardware to software, this is next-level 4K gaming.

4 ways the Nvidia GeForce RTX 5090 is superior to the RTX 4090

Struggling to decide between the Nvidia GeForce RTX 5090 and its predecessor? If you're already running with the RTX 4090 and are considering the upgrade? I'd hold off unless you truly require the entire suite offered by DLSS 4 and even more frames at higher resolutions. It's more than good enough for most PC builds and the RTX 5090 takes everything to the next level. While reviews are impressed by the performance of the RTX 5090 and its AI-powered Deep Learning Super Sampling (DLSS) and frame generation technologies, it's not the smoking gun Nvidia hoped it would be. Still, it's the undisputed champion of GPUs, and here are three reasons why it's better than the outgoing RTX 4090. 4 The RTX 5090 has a new architecture Meet Blackwell At the heart of the new Nvidia GeForce RTX 50 series, including the RTX 5090, is the Blackwell architecture. Replacing Ada Lovelace, this new platform promises some major improvements over RTX 40 series GPUs. We've got newer GDDR7 memory with a bandwidth of up to 36 Gb/s. This is a notable improvement over previous-gen cards and allows the entire GPU stack to perform better at higher resolutions. What makes the RTX 5090 perfect for high-fidelity gaming is the Multi Frame Generation with DLSS 4 and its 4,000 AI TOPS, 38 ray tracing TFLOPS, and 125 Shader TFLOPS. What all this means for gaming is a substantial leap in overall performance without packing the GPU with many more cores and other heat-generating parts. Still, we've got newer system memory with more bandwidth, newer generation cores across the board, and some other notable enhancements that the RTX 4090 simply cannot contend with. Related Nvidia's GeForce RTX 50 series promises RTX 4090 performance for $549 It's like an RTX 4090 only much more powerful and with considerably more RAM. Posts 7 3 The RTX 5090 has more memory More VRAM is always better So, what does all this extra memory allow the RTX 5090 to do? GDDR7 is a substantial improvement over GDDR6X that's featured on the RTX 4090. There's also much more of it, totaling 32GB. The RTX 4090 shipped with "just" 24GB, though we've been calling on Nvidia to include more RAM with its GPUs, especially in the mid-range segment. At 30Gbps, we're seeing a colossal increase in memory bandwidth at 1,792 GB/s, which allows the GPU to access and store data in memory at faster rates. Alongside the newer (and more numerous) CUDA cores, the RTX 5090 is positioned to become the most powerful GPU on the market. 2 The RTX 5090 has more cores I heard you seek performance? Source: Nvidia Nvidia's GeForce RTX 4090 has 16,384 CUDA cores, which are used to handle much of the work when playing games and the RTX 5090 upped this number to a whopping 21,760. That alone will provide a much-needed lift to make 8K gaming more of a reality, but these are joined by newer-gen RT and AI cores. Nvidia shared some slides during its CES 2025 announcement to show just how much of a jump this GPU is compared to its predecessor. With all Nvidia frame-gen features enabled, you could look to experience a two-times frame rate improvement in games with ray tracing. 1 The RTX 5090 has full access to DLSS 4 Generate all the frames! This generation is all about DLSS, something Nvidia has been perfecting since its debut in 2018, and with RTX 50 we're now up to DLSS 4. The sheer performance bump offered by Blackwell GPUs will allow technologies such as DLSS to perform better when used in conjunction with ray tracing and other more advanced graphics. As well as rendering more frames, the latest DLSS version will also attempt to create new frames before the GPU even requires them. Think of it as the GPU AI predicting the future, generating an additional two extra frames for each one that traditional DLSS would offer. Software will play a major role with the RTX 50 launch and I can see Nvidia working to further fine-tune the technologies included with this GPU to offer the best performance possible. Some of the DLSS Software will play a large part here and Nvidia will likely work to continue improving DLSS and its AI tech to better utilize the improved hardware. Related 11 best games to play on your RTX 5000 series Whether you go for the flagship RTX 5090 or the midrange, these are the best games to show off the power of your new GPU Posts 4 But it's not all good news for the RTX 5090 Nvidia's GeForce RTX 4090 is an impressive GPU and one that can still hold its own against its successor even with weaker specifications. It won't hit quite as many frames-per-second as the RTX 5090, but it's a good match for 4K monitors and could be the better value option if you can locate one at a reduced price, at least compared to the new 50 series. Graphics cards don't appear to be getting more affordable and so we have to make do with the current GPUs installed inside our systems. The RTX 4090 is still a capable card that will be more than a match for gaming for many years.

Nvidia's RTX 5080 doesn't dethrone the RTX 4090

Table of Contents Table of Contents Pricing and availability Specifications Performance Thermals and sizing Begrudgingly, the 5080 is the "better" option When you come at the king, you best not miss. Fortunately, it's almost a given that Nvidia's XX80 GPU would beat the XX90 graphics card that preceded it. That's just how it tends to go. But we've been hearing rumblings of the RTX 5080 underperforming expectations since its original unveiling and Nvidia's praise sweetened with such a heavy dose of AI. Our RTX 5080 review didn't do much to change that tale. So, how does the RTX 5080 compare to the RTX 4090? Can it compete? Does it offer a value upgrade for 30-series and 20-series owners looking for the best graphics card they can afford? Let's find out. Recommended Videos Pricing and availability The RTX 4090 debuted in October 2022 with an at-the-time record price for a top-consumer graphics card of $1,600. That price bounced around a lot during GPU shortages but ultimately remained relatively constant over the past few years. Stock of the card has all but disappeared at the time of writing, but used models have skyrocketed in price after the mediocre improvements of the RTX 5090. The 4090 is now priced between $1,500 and $2,500 even on second-hand markets. Get your weekly teardown of the tech behind PC gaming ReSpec Subscribe Check your inbox! Privacy Policy The RTX 5080 is on sale now, but stock is extremely limited. Its suggested retail price is $1,000, but third-party cards are already up to $1,500, and shortages and scalpers may drive the price up further in the short term. Specifications Nvidia RTX 5080 Nvidia RTX 4090 CUDA Cores 10,752 16,384 RT Cores 84, 4th generation 128, 3rd generation Tensor Cores 336, 5th generation 512, 4th generation Boost clock 2.6GHz 2.5GHz Memory size 16GB GDDR7 24GB GDDR6X Memory bus 256-bit 384-bit Memory speed 30Gbps 21Gbps Memory bandwidth 960GBps 1,008GBps TBP 360W 450W The RTX 5080 is, on paper, a weaker card than the RTX 4090, with around 50% fewer CUDA cores, fewer RT and Tensor cores (albeit from a later generation), 50% less memory, and a narrower memory bus. The RTX 5080 does have a higher boost clock and much faster memory, but we'd be hoping that the new Blackwell architecture does more of the heavy lifting of bridging the gap between these two cards. The RTX 5080 is more efficient than the last-gen king, too, demanding just 360W, whereas the RTX 4090 could easily pull 450W when pushed hard. Specs rarely tell the whole story, though, and they certainly don't cover the new DLSS features that the RTX 5080 supports. Its next-generation Tensor cores let it utilize multi-frame generation, which has the potential to artificially boost frame rates in compatible games by quite some margin. Performance Real-world performance testing is always where the GPU rubber meets the gaming road, and unfortunately despite all its potential, the RTX 5080 really doesn't impress. Its average performance across our suite of games at 4K and 1440p showed it as a very fast card, and 10-15% faster than the RTX 4080 Super that it's replacing, but falling far short of the RTX 4090 when we don't make use of the new multi-frame generation feature. In 1440p the gap between the 5080 and 4090 is lessened, but the 4090 still holds sway, particularly in slightly older games. Although we didn't run into any VRAM issues during testing, it's worth noting that super-demanding games like Alan Wake 2 have been shown to fill as much as 15GB of VRAM when running at full specifications, so we're already seeing the 5080 potentially miss out on future gaming features because it just doesn't have the VRAM room as the 4090 does. Ray tracing performance does see the 5080 almost catch up with the 4090, and pull further ahead of the 4080 Super (as well as obliterating AMD's alternatives), but the additional RT cores in the 4090 keep it more than relevant as the second-best RT graphics card out there -- albeit at distance from the 5090. The frame generation feature is the only real standout for the RTX 5080, and arguably its only big selling point for any 4080 or 4080 Super owners considering an upgrade. In compatible games, it can double or triple frame rates when enabled, though not all gamers will be willing to make the visual fidelity sacrifices that such a boost to FPS demands. The fewer tensor cores on the RTX 5080 (versus the 5090) means that its performance hit from enabling multi-frame gen is more pronounced, too, so you're more likely to see the artifacts and other issues frame generation can cause. Marvel Rivals - DLSS 4 Gameplay Overall, the RTX 5080 is a better card than the 4080 Super, but it can't quite eclipse the 4090 without its AI-augmented frame generation. Thermals and sizing At this level of graphics card with these kinds of power demands, you know they are going to run hard and maybe hot, but with the coolers on these cards, they don't actually get too toasty. The RTX 4090's cooler is gigantic and still works really well at chilling down that toasty GPU, but the RTX 5080's is far slimmer (dual slot!), making it a card that's easier to fit into smaller form factor builds. That's not nothing, but it's a very minor selling point for an otherwise underwhelming GPU. Begrudgingly, the 5080 is the "better" option I wish there were enough RTX 4090s out there for everyone to buy at an affordable price because then I'd probably just recommend you buy that instead. It's still incredibly fast, and though it does demand more power, it has boatloads of VRAM with no signs of saturation on the horizon. But I can't. Because the RTX 4090 isn't on sale anymore, and the second-hand options are massively overpriced. So, where does that leave gamers with around $1,000 burning a hole in their pocket, looking for a GPU upgrade: The RTX 5080 -- if you can find it at a good price. If you can't, one of these might be better. It's still an improvement over anything besides the 4090 and 5090 (and does get close to the former), and multi-frame generation might not be the game changer Nvidia wants it to be, but in the right setting, it makes a huge difference. There's a reason we've all been using Lossless Scaling so much. The RTX 5080 isn't the big, new-gen card we'd hoped it would be, and it is absolutely not worth it if you're already gaming on a 4080 or 4080 Super. For anyone else though, eh? It's OK. It might be worth waiting to see whether the RX 9070 XT from AMD gets close at a fairer price, but that's still a couple of months away. At the very least, don't pay scalper prices for the 5080. It isn't worth that.

Nvidia GeForce RTX 5080 vs 4090: The old flagship still rules (but you can't get it any more)

Nvidia GeForce RTX 5080 The RTX 5080 is the new xx80 tier using the Blackwell architecture, which brings GDDR7 to a consumer GPU for the first time. It's also $300 less than the MSRP of the outgoing RTX 4080, which is better pricing from Nvidia even if the GPU doesn't meaningfully improve on performance. Pros & Cons GDDR7 memory for lots of bandwidth Plenty of ray tracing power Lower MSRP compared to the card it replaces First xx80 card to not beat the outgoing flagship in some time Hard to find in stock 16GB VRAM isn't enough for its tier $1000 at Best Buy $1000 at Newegg Nvidia GeForce RTX 4090 Founders Edition The outgoing flagship GeForce RTX 4090 is still an absolute beast (if you can find one in stock). Unfortunately, Nvidia stopped production in 2024, so they're getting harder and harder to find. Pros & Cons 24GB of VRAM Plenty of CUDA cores Good ray tracing performance Not being produced any more Nvidia's GeForce RTX 50 series is here, and if you were lucky to get one of the best GPUs in the brief time they were available, you know you'll be playing DLSS-enhanced games as soon as you plug it in. Only the top two from the range are out, which is the usual release cadence for any GPU release. The RTX 5080 joins a long line of XX80 graphics cards from the company, and the general rule is that the new XX80 card is more powerful than the previous flagship. That's the RTX 4090 in this case, and it's only fitting if we put that long line of historical facts to the test. Related It's time to accept Nvidia's "fake frames" might be the new normal Like it or not, Nvidia's RTX 50 series shows that AI-generated frames are here to stay Posts 3 RTX 5080 vs RTX 4090: Price, specs & availability The new lieutenant squares up against the previous captain The Nvidia GeForce RTX 5080 launched with an MSRP of $1,000, the same price as the outgoing RTX 4080 Super. It's designed for 1440p and 4K gaming, especially because it can use DLSS frame generation to create up to three AI-generated frames per rasterized frame made from the traditional rendering process. The new Blackwell architecture powers it with 10,752 shader units and 84 ray tracing cores and the inclusion of a PCIe 5.0 interface for the first time on a consumer GPU. The Nvidia GeForce RTX 4090 was the flagship of the last generation, powered by the Ada Lovelace architecture. It had an MSRP of 1,600 when it launched in 2023, has 16,384 shader units and 128 ray tracing cores, and is out of production as of 2024, so you'll be lucky to find any new units anywhere near the MSRP. Nvidia GeForce RTX 5080Nvidia GeForce RTX 4090Shader Units10,75216,384Ray Accelerators/Cores84128Base Clock Speed2,295 MHz2,235MHzBoost Clock Speed2,617 MHz2,520MHzMemory Capacity16 GB GDDR724GB GDDR6XMemory Bus256-bit384-bitMemory Bandwidth960 GB/s1,008GB/sPower Draw360 W450WArchitectureBlackwellAda Lovelace Architecture Blackwell takes over from Ada Close The Nvidia GeForce RTX 5080 comes with a new architecture, Blackwell. This builds on the Ada Lovelace architecture used in the GTX 4090, with two key new features. While Ada had single frame generation, which uses the Tensor cores for AI-generating one frame per raster frame when enabled, the RTX 5080 can generate up to three frames per raster frame, drastically increasing frame rates at higher resolutions without incurring much penalty for latency or system resources. The other key feature, DLSS 4, is mostly backward compatible with the RTX 40 series, with some features also working on the RTX 30 and RTX 20 cards. It includes a new deep learning anti-aliasing (DLAA) model for fewer jaggies, enhanced DLSS Super Resolution for rendering at a lower resolution and upscaling to your monitor's native resolution, and DLSS ray reconstruction for better lighting effects with ray traced titles. Source: Remedy Entertainment For the first time, the RTX 5080 uses GDDR 7, which has 16GB of speedy memory and a 256-bit memory bus, which means 960GB/s of memory bandwidth for the new GPU. This is fast compared to the 24GB of GDDR6X in the RTX 4090, which manages 1,008GB/s out of its wider 384-bit memory bus, but the larger VRAM will make the older card better for AI models or particularly demanding games. Related Nvidia GeForce RTX 4090 review: The untouchable king of performance There is no other graphics card like the Nvidia GeForce RTX 4090. Its power is unmatched, as is its size and its power consumption. Posts Performance and power Disappointing showing from the RTX 5080 here, but it is cheaper There's no easy way to say this. Without more CUDA cores, the RTX 5080 is always going to lose against the RTX 4090. It's just a question of by how much. The saving grace here is that the difference is less than the % difference in either price or power usage. While we haven't got our samples yet to test, other outlets have been busy over the last two weeks, and the picture shows that the RTX 4090 is roughly 20% faster in pure raster performance than the RTX 5080. If you mostly play older games, the RTX 4090 is easily the one to go for, if you can find it in stock anywhere. When DLSS, particularly Multi Frame Generation, is involved, the two GPUs are at a dead heat. Plus, the RTX 4090 uses up to 450W of power, while the RTX 5080 uses up to 360W. That alone might be the decider, as not everyone has a powerful PSU or wants to use that much power every time they play a game. Related 3 reasons why the Nvidia GeForce RTX 5080 is better than an RTX 4080 Looking to see if the RTX 5080 is better than the RTX 4080? It is, but here's why. Posts RTX 5080 vs RTX 4090: The student doesn't quite surpass the master For those of you who want to play DLSS-enhanced games, there's no question that the GeForce RTX 5080 is the better option. It might not be as powerful in pure raster performance, and the 16GB of VRAM might run into issues as games get more resource-hungry, but it's $600 cheaper. It's also nearly 100W less power-hungry, has a better cooler, and, most importantly -- is being actively produced. Whether it's good value is very dependent on the games and workloads you want to throw at it, but it's a viable option for upgraders if you have an RTX 30 series or earlier. Nvidia GeForce RTX 5080 If you can find one in stock, the Nvidia RTX 5080 is a powerful GPU for ray traced games. It's also $300 cheaper than the RTX 4080 was, making it (slightly) better value. $1000 at Best Buy $1000 at Newegg If you run AI workloads or play older games, the RTX 4090 is going to be more suited to your needs. The snag is that Nvidia stopped production in 2024, to dry up stock before the RTX 50 series launch. If you can find one near or around MSRP, it's a cracking deal still, and will last you for many years to come until games need more than 24GB of VRAM. Nvidia GeForce RTX 4090 The outgoing flagship GeForce RTX 4090 is still an absolute beast (if you can find one in stock). Unfortunately, Nvidia stopped production in 2024, so they're getting harder and harder to find.

3 reasons why the Nvidia GeForce RTX 5080 is better than an RTX 4080

Nvidia announced the Blackwell-powered RTX 50 series of graphics cards at CES 2025. This new generation of GPUs brings numerous benefits, including the latest iteration of DLSS (with multi-frame generation), improved CUDA and Tensor cores, more advanced memory, and beefier RT cores. Nvidia introduced Blackwell as an architecture for generational AI and the company's entire CES presentation was about AI. You'd be forgiven for believing Nvidia was an AI-only company if you're new to the GPU space. The RTX 5080 is a mid-range GPU, sitting below the flagship (and incredibly expensive) RTX 5090. But how does it compare to the outgoing RTX 4080? 3 Enhanced visuals with DLSS 4 DLSS 4 isn't exclusive to RTX 50 series GPUs, but you will not enjoy all the benefits it offers without upgrading to one. Nvidia will want to sell some RTX 50 series cards so it was inevitable that some features would be exclusive to this new generation of AI-powered GPUs. What's unique for RTX 50 series cards on the Blackwell architecture is DLSS Multi Frame Generation. This differs from normal DLSS frame generation in that the AI cores are used to take the one extra frame generated through DLSS and crank it up to three. So instead of two frames, you get four. Does this equate to better performance? Not exactly. If you're struggling with low frame rates, DLSS 4 with Multi Frame Generation will not magically improve your FPS to more than a hundred, but it will smooth out the feed to the monitor, especially if you have a high-refresh-rate panel. All other improvements to DLSS, including memory optimizations, will be transferred to select older Nvidia GeForce RTX GPUs. Related 3 reasons DLSS is still the most revolutionary technology in PC gaming DLSS has come a long way in six years, but I'm still excited about what the future holds Posts 1 2 Better price to performance The Nvidia GeForce RTX 5080 launched with an MSRP of $999, the same as the RTX 4080. AIB partners will likely charge slightly more for aftermarket cards, and stores could increase prices further due to demand. Because of the same price but higher reported performance, the RTX 5080 is a better deal. During its CES 2025 briefing, Nvidia promised RTX 4090 performance with the RTX 5070 through AI development instead of raw performance and rasterization. We can see this in the specifications comparison between the RTX 5080 and RTX 4080 with only slight bumps in certain areas but a considerable gain in reported performance. According to Nvidia, the RTX 5080 offers up to two times the performance across a range of tests, including some heavier games and rendering suites. Cyberpunk 2077, Alan Wake, and Black Myth: Wukong all saw a two-times increase in performance over the RTX 4080, which is likely due to the AI boost with DLSS 4 and other advancements made through the company's Tensor cores. The slight improvement to the specifications would carry some of the weight, but these generational gains are powered by Nvidia's AI tech. We'll have the RTX 5080 in our hands soon enough for testing to see how it fares against the RTX 4080. It should be considerably better when comparing full game tests with DLSS and other graphical settings cranked up high, but how much so will depend on the CPU used, the game, and other parts of your PC. Related Best GPUs in 2025: Our top graphics card picks Picking the right graphics card can be difficult given the sheer number of options on the market. Here are the best graphics cards to consider. Posts 1 More powerful architecture Close Nvidia used Ada Lovelace for the RTX 40 series of graphics cards and the RTX 5080 runs on Blackwell, which launched at CES in 2025. Nvidia didn't push system upgrades as hard with the RTX 5080 and other Blackwell GPUs, instead focusing on the enhancements made to its AI capabilities. For better or for worse, Nvidia has transformed from a simple GPU brand to one with heavy development in artificial intelligence and machine learning. This effort brought us DLSS, an incredible piece of kit that can boost in-game performance when enabling advanced graphics such as ray tracing. Nvidia GeForce RTX 5080Nvidia GeForce RTX 4080Shader Units10,7529,728Ray Accelerators/Cores8476Base Clock Speed2,295 MHz2,205 MHzBoost Clock Speed2,617 MHz2,505 MHzMemory Clock Speed1,875 MHz1,400 MHzMemory Capacity16 GB GDDR716 GB GDDR6XMemory Bus256-bit256-bitMemory Bandwidth960 GB/s716.8 GB/sPower Draw360 W320 WArchitectureBlackwellAda LovelaceProcess5 nm4 nmAI Accelerators/Cores336304 The RTX 5080 has some notable improvements over the RTX 4080. In terms of specifications, we're looking at an additional 1,024 CUDA cores (shader units), 12 ray tracing cores, 32 Tensor cores (AI accelerators), GDDR7 RAM (compared to GDDR6X), higher memory bandwidth, and faster clock speeds. All this is based on the new Blackwell architecture with a slightly higher power draw of 360W vs. 320W for the RTX 4080. Though we are seeing GDDR7 RAM included, it's still 16GB, which is identical to the outgoing RTX 4080. It's a tricky upgrade path Upgrading from an RTX 40 series to an RTX 50 series graphics card is a difficult sell. An RTX 4080 isn't a terrible GPU. It may be viewed as a better value upgrade for other PC gamers with the introduction of the newer generation since some will want to offload their existing GPUs to make way for an upgrade. Should you buy either of these graphics cards for a fresh PC build, I'd go with the latest and greatest, the RTX 5080. It's the same price as the RTX 4080, though good luck locating stock and listings that aren't charging considerably more.

NVIDIA GeForce RTX 5080 FE review: the perfect balance between power and price?

Why you can trust Creative Bloq Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test. You wait years for one, and then two come along at once. Yes, the NVIDIA GeForce RTX 5080 graphics card is here, and we've taken it for a test drive, too see how it will perform for power-hungry creatives. While the flagship card, the GeForce RTX 5090 FE (FE stands for Founders Edition), naturally comes with All The Power, I was intrigued to see how close the 5080 FE could run it, to see if it would be a more affordable top choice of graphics cards for graphic artists and 3D pros. And you know what? It was a closer race than I thought. The first thing I noticed as I unpacked the NVIDIA GeForce RTX 5080 FE from its fully recyclable packaging (which looked identical to the 5090 apart from the number at the top), was that it looks almost exactly the same as the 5090 card. The only noticeable difference is the model number etched into the topside (and that it's ever-so-slightly more prominent on the 5080, although that could just be my specific unit). A solid, weighty piece of hardware, slightly longer and narrower than the 40-series cards, it slots in very easily but leaves a little daylight in the back-panel notch on my tower. Thankfully, the only witnesses to that will be my wall and my overly curious cat. The power socket being inserted at an angle saves space, as the cable juts out less from the side of the card than it otherwise would, but inserting and removing the cable is a little bit more finicky as a result. Hardly a deal-breaker though... The GEFORCE RTX logo lights up too, and there's light protruding through from inside too, most prominent in the V-shaped corners near the centre as you look at it from above. The back, just like the 5090, has four ports, three DisplayPort and one HDMI. You'll be entirely unsurprised to find the DPs support up to 4K 480Hz or 8K 165Hz with DSC each, while the HDMI supports up to 4K 480Hz or 8K 120Hz with DSC, Gaming VRR, and HDR. The NVIDIA 5080 comes with the same feature and software offering as the 5090, sporting the same Blackwell core architecture, which has stepped up both AI-enhanced output and efficiency, while providing less of an uptick in pure VRAM terms. In fact, the 5080 comes with 16GB of the latter, the same number you'll find inside the 4080 version. So how is the new one better, I hear you howl in abject confusion at your favourite reviews editor? Well, I went to NVIDIA's Editors Day at CES 2025, and as I learned there, the AI so many of us hate to see used to create pictures of people with varying numbers of fingers and other appendages or dodgy boomer memes on Facebook, has been employed here to provide a dramatic uptick in power efficiency. For example, this is apparent in how it can shut down a majority of its cores more quickly during idle times during graphics rendering, which in turn means those cores are available to help with new tasks while the previous ones are still running, instead of those cores being spun up and doing nothing productive on said previous task. This means that with those AI-enhanced features activated, we can expect to see up to 40% higher performance from the same amount of VRAM during many 3D tasks, (and like I pointed out already in the 5090 review, multi-tasking in video rendering means that time-consuming job has become a lot faster than that). Yes, Mega Geometry, Neural rendering, DLSS 4 and a host of other tools will also be available on the new drivers for the 30- and 40-series of cards, but they don't have the added benefit of the Blackwell core architecture, if those not-so-marginal gains are important to you. NVIDIA's proprietary technology, CUDA cores, is central to the graphics performance of its GPUs. For many professionals in the 3D field, the number of CUDA cores on a card has been the primary factor in selecting a GPU. Additionally, the increasingly important Tensor Cores play a crucial role. Both CUDA cores and Tensor Cores are processing units, and having more of them generally improves performance. While CUDA cores excel at general graphics and 3D rendering, they are less efficient for neural processing, AI, and deep learning tasks compared to Tensor Cores. The GeForce 5080 Founders Edition (FE) features 1,801 Tensor Cores, which is over 30% more than the 1,321 found in the GeForce 4090. However, the 5080 FE has 10,752 CUDA cores, which is fewer than the 16,384 CUDA cores in the 4090. In all honesty, I expected a bigger dropoff in benchmark and real-world performance levels from the 5090 FE here. Unless you're working on the latest Pixar movie or a billion-dollar AAA game, the 5080 might just be enough for you for the next few years. PugetBench for Creators testing gave me near-identical numbers in the Photoshop test to what the 5090 delivered. I asked NVIDIA about this and the answer was that "there appear to be some scaling issues with the pre-launch build which means the results might not be completely representative of performance difference between the cards." This might already have been sorted out by the time you receive your card. I also used 4-2-2 camera video assets to run a concurrent render of 9 video streams, and again, the 5080 gave my 4070 Ti Super and its single render job a run for its money. I estimated about a 8.5x increase in speed here, which should vastly speed up work for many video editors. When it comes to Procyon Flux testing, we finally see a marked difference between the models. The 5080 FE took more than twice as long to generate images using the memory-demanding FP4 test than the 5090 FE did, but it was still impressively fast in its own right. There's also about a 15% reduction in performance from the 5090 to the 5080 in both the PugetBench DaVinci test and the Topaz Video AI upscaling test. Both figures still reside firmly in the 'staggering' category though. Also, H.264 encoding saw no dropoff in the 5080, with the lower-spec card churning through 128.98 frames per second, all-but identical to the 5090's 129.04 frames per second. And while the Geekbench 6 OpenCL testing numbers didn't match-up to the reality-bending figures churned out by the 5090, the 5080 FE produced a hugely respectable 272,788 points here, outscoring the 4090 in a test that's not yet been optimised for the AI-enhanced features of the Blackwell-equipped card. 3D processing is now drastically faster, thanks to innovations in the 50-series' feature offering. While software support is still being rolled out, it is clear that by the end of the year, nearly all major 3D and video software will have integrated the Blackwell suite of features. That means some of our standard benchmark tests are not fully compatible with the 50-series graphics cards. We will update this review with those tests as soon as we have the opportunity to run them. Gaming is also truly spectacular here, especially when I can make use of features like DLSS 4. I ran Cyberpunk 2077, EA FC 24 and more games on this card, and the step up from 4070 Ti Super I had previously is not only noticeable, it's staggering. Go to one of our gaming-focused sister sites for a full breakdown of the card for gamers if you're interested to know more about that. Also, extra point for releasing an impressively stable driver on the first go. At under a grand at launch, the NVIDIA GeForce RTX 5080 FE isn't exactly cheap, but while it's only half the price of the 5090 FE, it provides a heck of a lot more than half the performance. With DLSS 4 support, neural rendering, RTX Mega Geometry and a host of other innovations on board, 3D modellers, animators and perhaps most importantly game developers will be looking for the most bang for their buck when it comes to the 50-series. And the 5080 might be that card. Meanwhile, I'd actually recommend the 5080 over the 5090 for video editors regardless of the price benefit. Same for photo editors (although a 40-series card will remain more than enough power for most of you).

Review: ASUS TUF GeForce RTX 5080 Gaming OC edition

Introduction and Analysis In the ever-evolving realm of graphics technology, the ASUS TUF Gaming GeForce RTX 5080 16GB GDDR7 OC Edition emerges. With a blend of technology and design, this graphics card is positioned in the premium class of RTX 5080 products. The Heart of the beast: NVIDIA Blackwell Architecture. At the core of this formidable machine lies the NVIDIA Blackwell architecture. This design delivers unparalleled AI performance through its fifth-generation Tensor Cores and fourth-generation Ray Tracing Cores. The ASUS TUF Gaming series serves as an alternative to the ROG Astral and ROG Strix product families by providing a balanced approach to cooling and performance. While ROG Astral and Strix variants often include additional software features and specialized headers, TUF Gaming models focus on reliability and practical functionality. For instance, TUF cards maintain large heatsinks with triple-fan designs, offering efficient heat dissipation. They also include protective PCB coatings and use a vapor chamber on higher-tier models to manage temperatures. Though TUF Gaming products still feature factory overclocks and synchronized RGB elements, these are generally more moderate than their ROG counterparts. ASUS Clocks this card at 2700 MHz on the boost frequency and offers two BIOSes for performance and silent modes. Additionally, this GPU incorporates a 3.6-slot design to accommodate more robust cooling hardware. Cooling performance is supported by Axial-Tech fans and a large vapor chamber, both of which promote consistent thermal management. The card also includes a Dual BIOS switch, granting users the choice between a quieter operating mode or a higher-performance setting based on individual preferences. While advanced connections like FanConnect II may be absent compared to ROG Astral models, most gamers and professional users will find the TUF Gaming GeForce RTX 5080 sufficiently adaptable. The inclusion of military-grade components -- such as TUF chokes, MOSFETs, and capacitors -- reinforces its durability under intensive use. This graphics card is equipped with 84 streaming multiprocessors and a total of 10,752 CUDA cores, which is approximately half the core count of its higher-end counterpart, the RTX 5090. The RTX 5080 features a two-slot design, making it compatible with a wide range of PC builds. It is outfitted with 16 GB of GDDR7 memory operating on a 256-bit memory bus, achieving memory speeds of up to 30 Gbps. NVIDIA is set to release its newest GPUs -- the RTX 5090, RTX 5080, and RTX 5070 Ti -- between late January and early February. The RTX 5090 serves as the flagship model, targeting users who require substantial computational power for intensive tasks such as high-resolution gaming, 3D rendering, and data analysis. In contrast, the RTX 5080 is designed for users seeking a robust performance without the premium cost associated with the top-tier model. Both GPUs incorporate advanced features like DLSS4, NVIDIA's latest iteration of Deep Learning Super Sampling, which enhances image quality and performance by utilizing artificial intelligence to upscale lower-resolution images in real-time. The GeForce RTX 5080 is built on a GPU architecture that includes 45.6 billion transistors and occupies a die area of 378 mm², manufactured using TSMC's 4nm process technology. This advanced manufacturing process allows for higher efficiency and better performance per watt. NVIDIA continues its tradition of naming GPUs in honor of significant figures in science and technology. The RTX 5080, named "Blackwell," pays tribute to David Blackwell, a pioneering African-American mathematician and statistician known for his contributions to game theory, probability, and information theory. Blackwell's legacy in the National Academy of Sciences is immortalized through this GPU, reflecting NVIDIA's commitment to honoring innovators who have significantly impacted their fields. SpecificationRTX 5090RTX 5080RTX 5070 TiRTX 5070BlackwellGB202-300 GB203-400GB203-300-A1GB205-300-A1GPU SMs170 (192 Full)84 (84 Full)70 (84 Full)50 (50 SM Full)GPU Cores21,760 10,7528,9606,144Clock Speeds2010/24102300/26172300/24502160/2510L2 Cache98MB65MBMemory Capacity32 GB GDDR716 GB GDDR716 GB GDDR712 GB GDDR7Memory Bus512-bit 256-bit256-bit192-bitMemory Speed28 Gbps30 Gbps28 Gbps28 GbpsBandwidth1,792 GB/s960 GB/s896 GB/s672 GB/sTotal Board Power (TBP)575W360W300W250W Power Interface1x 12V-2x6 (16-Pin)1x 12V-2x6 (16-Pin)1x 12V-2x6 (16-Pin)1x 12VHPWR (16-Pin)Launch Date 2025January 30, 2025January 30, 2025February February Price$1999$999$749$549 Built on the Blackwell GB203 GPU, the RTX 5030 sports a large die size compared to its predecessor. It integrates the GB203 GPU with 84 (84 active) Streaming Multiprocessors, giving it 10752 shader cores. Enabled with 10752 KB L1 Data Cache/Shared Memory and 65536 KB L2 Cache Size the card has 112 ROPs and 336 Texture units. The GPU is tied towards 16GB DDR7 memory clocking in at 30 Gbps (960 GB/sec). Memory technology also sees a leap forward. The RTX 5000 series features 16 GB of GDDR7 VRAM with a 256-bit interface, running at healthy 30 Gbps. This setup offers a total bandwidth of up to 960 GB/s. A larger cache and new memory compression techniques enhance data flow and efficiency. These improvements support high-resolution gaming, real-time rendering, complex simulations, and more, promising smoother performance across various applications like 4K gaming and professional design. The GeForce RTX 5080 has a Total Board Power rating of 360 Watts. While this represents the maximum draw under full load, typical power consumption during gaming is expected to be somewhat lower, something we'll check out in this review. Next page please.

Nvidia's GeForce RTX 5080 Is a Solid High-End GPU

Excellent 1440p performance. Lower power requirement. Compact form factor. Nvidia's newest 50 Series GPUs are slowly trickling out, and as usual, the very top-end RTX 5090 (7/10, WIRED Recommends) is the graphics card that everyone is talking about. It boasts extreme 4K gaming, the latest in AI-powered gaming enhancements, and a power draw to match. I can't blame you for not wanting to spend $2,000 on a GPU; that's enough to build a midrange gaming PC on its own. At just $1,000 for the RTX 5080 Founders Edition, this still-expensive step-down card will be the model that more people seriously consider, even if it's still a splurge. It's a better choice from a performance perspective, meeting people where they already are in terms of monitor resolution, game choice, and existing power supplies. But how does it fare against the more expensive card, and how does it handle some of the more popular and evergreen games? Well enough to my eyes. If you're building your next high-end gaming PC and are looking for a high-end video card to match, this might be exactly what you're looking for -- if you can find one for sale. The form factor of the RTX 5080 is identical to its more powerful counterpart, with a true two slot design that should fit in most cases very comfortably. I really appreciate the size reduction overall, and I hope AIC cards follow suit. Where the RTX 5090 draws an immense 575 watts, the 5080 only asks for 360 watts with the same new power connector. Like the RTX 5090 FE, the RTX 5080 includes an adapter, and I imagine most partner cards will as well. That means a lower overall system power requirement, with Nvidia recommending just 850 watts for the Founders Edition. I expect this will be an easier requirement for existing rigs to meet without needing to buy a new 1,000-watt or higher PSU. Nvidia introduced a new version of its AI-powered enhancement tools for the RTX 50 Series. These notably add support for multi-frame generation, which uses AI to generate up to three frames between. If you're interested in learning more about the effects of using this tech on image quality, make sure to check out the RTX 5090 review. The short version is that multi-frame generation can produce minor artifacts, particularly in areas where two objects at different depths overlap, such as looking through a fence. These are hard to spot across a whole screen though, and the higher frame rate makes the gaming experience much smoother, so the frames are onscreen less time. I'll start by checking out performance in Cyberpunk 2077, one of the more demanding games that currently supports multi-frame generation.