AI Server Boom: Vietnamese Retailer Builds Powerful Rigs with Multiple RTX 5090 GPUs

Seven RTX 5090 GPUs power AI server worth over $30,000 -- over 4000W of power and 224GB of memory in a single frame

Professionals are willing to pay top dollar for RTX 5090-powered AI servers. It's no secret that Nvidia's RTX 5090 is a powerhouse for AI workloads, given Blackwell's architectural improvements and support for the lower precision data formats. A Vietnamese shop, Nguyencongpc (via I_Leak_VN), recently shared a few photos of an AI rig they set up for a customer, featuring seven RTX 5090s hooked up to multiple kilowatt PSUs in an open-air GPU frame. Despite the average RTX 5090 costing nearly $4,000, professionals find their performance worth the investment, even though Nvidia is not positioning these cards as workstation solutions. The elusive RTX 5090 has been off the shelves and out of reach for most of its lifespan unless you're willing to pay through the nose. This is mainly because Nvidia's profit margins are significantly higher on their server-focused Blackwell B100/B200/B300 GPUs than on typical GeForce products. For Nvidia, allocating the limited number of wafers it procures from TSMC toward these AI accelerators is simply more profitable. The AI system resembles mining rigs from the pandemic era. The open-air GPU frame carries seven Gigabyte RTX 5090 Gaming OC units connected using what we can see are PCIe riser cables, powered by multiple Super Flower Leadex 2000W PSUs. The system can easily be valued at over $30,000, considering these GPUs go for $3,500-$4,000 on a good day. This single frame offers a massive 224GB memory pool for AI training and inference. Of course, this isn't unified memory, so developers must rely on techniques like model parallelism to distribute workloads efficiently across multiple GPUs. On that note, Nvidia's forthcoming Blackwell workstation cards are outfitted with up to 96GB of VRAM and cost between $8,000 and $9,000. Even though the RTX Pro 6000 Blackwell offers more memory, multiple RTX 5090s provide a performance advantage for AI tasks where raw compute is essential, especially on a limited budget. However, when price is not a concern and you wish to scale AI more effectively, more VRAM per individual GPU becomes essential to handle complex models with hundreds of billions of parameters. With the advent of AI, Nvidia now differentiates its GPU offerings by memory capacities, which results in a steep price curve if you're looking for higher memory configurations. There have been workarounds by modders, where blower-style RTX 4090s with 48GB of memory are common in the Chinese community. It might take months before prices normalize since practically every new GPU flies off the shelves instantly and is never seen at MSRP. You might want to check out the used market for last-generation GPUs, which can usually be found near their launch MSRPs or even lower.

Barebone air-cooled AI server with 7 RTX 5090 GPUs goes on sale in Vietnam prompting fears of huge price rise

Crypto mining in the early 2020s caused prices of GPUs to spike, the same is likely to happen with AI During the early 2020s, cryptomining was a driving force behind a lot of GPU shortages and inflated prices, as high-end graphics cards were snapped up in bulk to mine cryptocurrencies, leading to record pricing across multiple generations of GeForce GPUs. Nvidia attempted to address the issue by introducing mining limiters, but miners, naturally, found ways to bypass these, and retail prices remained well above the MSRP for long periods. History could well be about to repeat itself, but this time based around artificial intelligence demand rather than cryptocurrency. Videocardz (via I_Leak_VN) reports that a Vietnamese retailer called Nguyencongpc has started selling barebone AI servers equipped with seven GeForce RTX 5090 graphics cards. The setup, cooled by air and powered by a 2000W PSU, is designed to handle large AI workloads, including support for large language models. The RTX 5090, with its 32GB of GDDR7 memory, offers a blend of high performance and relatively low cost compared to Nvidia's dedicated in-demand AI accelerators. When several are used in tandem, as is the case with the server Nguyencongpc is selling, they can match or exceed the memory capacity needed for enterprise-level AI applications - without the high price tag of the RTX Pro or H-series products. While you've got to admire the ingenuity on display here, the knock on will inevitably be mounting supply pressures. The RTX 5090 is already facing stock issues, and its price has climbed past $3000 in many markets, with some listings close to $4000. This mirrors the RTX 4090's trajectory, which driven by cryptomining, rarely dropped below MSRP and remained in high demand for non-gaming uses. With modders reportedly exploring ways to increase onboard memory even further, the RTX 5090 could follow the same path as its predecessors - designed for games, but monopolized by entirely different industries. Nvidia has yet to take steps to prevent system integrators or retailers from redirecting gaming GPUs toward AI markets and if it doesn't, gamers could well once again find themselves unable to purchase the latest graphics hardware at fair prices.

7 x GeForce RTX 5090s power custom AI servers in Vietnam: 28 x 8-pin power cables, 4000W power

TL;DR: Vietnamese shop Nguyencongpc is building custom AI servers using NVIDIA GeForce RTX 5090 graphics cards, with seven cards consuming 4000W of power. This setup could lead to GPU shortages for gamers, similar to past cryptocurrency mining issues. The RTX 5090 features 32GB of GDDR7 memory, beneficial for AI workloads. Vietnamese shop Nguyencongpc have been building custom AI servers based on NVIDIA GeForce RTX 5090 graphics cards, with 7 x RTX 5090 cards used with 28 power cables in total, and 4000W of power consumption for the custom AI servers. If you thought cryptocurrency mining back a few years ago caused issues with GPU shortages, it seems we have to fear custom AI servers using consumer-focused GeForce RTX 50 series gaming GPUs creating shortages for PC gamers. At the time, NVIDIA gimped crypto mining performance, but crypto miners continued to find ways around that. However, NVIDIA's new GeForce RTX 5090 graphics card ships with a huge 32GB of ultra-fast GDDR7 memory, with large pools of memory favored by AI workloads, so a Vietnamese shop has used 7 of them in a custom AI server for various AI workloads, and 7 x 32GB of RAM is a big deal for AI users. The new AI servers are using dual 2000W power supplies with a total of 28 power cables (4 x 8-pin PCIe power cables converted to the 16-pin power connector on the RTX 5090) and with 7 x RTX 5090s, that's a total of 28 power cables and 4000W of power flowing to the graphics cards. The other thing is, we've seen upgraded GeForce RTX 4090 graphics cards with 48GB and even a huge 96GB of VRAM... upgraded from their stock 24GB of GDDR6X memory. This is for AI workloads, and once we get modders working on RTX 5090 upgrades from 32GB to higher pools of VRAM... well, AI servers like this will become even more popular.

Vietnamese Retailer Builds AI Servers With Several GeForce RTX 5090 GPUs Amid Shortages

A Vietnamese retailer builds AI servers with RTX 5090s to leverage higher memory capacity for AI workloads. Builds machines using various configurations. The GeForce RTX 5090 was launched at the end of January, and it has been over two months, and yet we don't have sufficient stocks of the GPU. While this is common across all the released RTX 50 series GPUs, the RTX 5090 is particularly more affected due to its high demand and lower supply. The RTX 5090 is the only gaming GPU to feature 32 GB of memory, which makes the 5090 a special one even for non-gaming tasks. As higher memory is highly beneficial for AI tasks, a Vietnamese retailer called "Nguyencongpc" has started to build AI servers using as many as seven RTX 5090 GPUs. In the pics, you can see seven Gigabyte RTX 5090 models installed on the server, amounting to almost 4000W of total power consumption. Similarly, there are rigs built using two MSI RTX 5090 GPUs that are aimed at less intensive AI tasks. We don't know the exact specifications of servers with seven RTX 5090s, but the ones with the MSI Gaming Trio RTX 5090 use the AMD platform, featuring the ROG Crosshair X870E Hero motherboard. It appears the retailer has plenty of cards not only to sell to gamers but also for dedicated AI servers for small businesses. The price of the dual RTX 5090 rig must be around $10,000 and around or over $30,000 for the AI server with seven GPUs. This is because an individual RTX 5090 generally costs $2500+ for most editions at the moment, even though the card was launched for $2000. The Gaming TRIO RTX 5090 is a premium edition, so it might be worth around $3000. The current state of the GPU market and the emergence of such AI servers remind us of the early 2018 era when GPU mining was at its peak, and you would hardly find a gaming GPU. The same repeated in 2021, making it extremely difficult to get your hands on an RTX 30 series GPU. Hopefully, this AI 'mining' servers trend won't pick up the pace, or it will be hard to find the RTX 5090 GPUs. For such purposes, there are already professional GPUs available with higher VRAM capacity, and we have already seen RTX 4090 with 48 GB memory to do such jobs, but the card is hardly good at gaming.