NVIDIA's GeForce RTX 5090 D V2: A New China-Specific GPU Launching in August

NVIDIA's new GeForce RTX 5090 D v2: China-ready GPU should launch in August with 24GB GDDR7

NVIDIA's new custom made-for-China AI GPU is reportedly called the GeForce RTX 5090 D V2, previously rumored as the RTX 5090DD, coming in August. As an Amazon Associate, we earn from qualifying purchases. TweakTown may also earn commissions from other affiliate partners at no extra cost to you. NVIDIA's new GeForce RTX 5090 D v2 is reportedly the name of its new China-specific GPU, according to a new post on X by leaker "MEGAsizeGPU". In another post, HKEPC reports that NVIDIA is preparing its new GeForce RTX 5090 D v2 for an August launch, something we heard about in previous reports. This means that gamers in China will only have another month before they can secure themselves the new (tweaked) flagship GeForce RTX 5090 D v2 graphics card. Inside, the NVIDIA GeForce RTX 5090 D v2 has the same GPU core count as the RTX 5090D with 21760 CUDA cores and the tweaked GB202-240-K-A1 GPU. The VRAM has been scaled down to 24GB (down from 32GB on the RTX 5090) on a 384-bit memory bus clocked at 28Gbps and the same 575W TDP but leaks tease "and there's a surprise". NVIDIA launched its cut-down GeForce RTX 5090 back in January 2025, tweaked to meet US export controls, its AI TOPS power was sliced down, but the GPU retained its same GB202 GPU with 21760 CUDA cores and 32GB of GDDR7 memory. This new RTX 5090 D v2 has the VRAM cut down from 32GB to 24GB, a big difference, but retains all of its gaming power. The original RTX 5090D featured the GB202-250 GPU, the RTX 5090 uses the GB202-300, while the new RTX 5090DD uses a newer GB202-240 GPU. We should expect the new RTX 5090 D v2 to use a newer PCB design -- PG145 SKU 40 -- which is required because of the changes to the VRAM and GPU layout.

NVIDIA GeForce RTX 5090 D V2 Coming To China, Cut-Down VRAM, To Be Fully Compliant With US Regulations

NVIDIA's next export-compliant GPU will be a re-release of the GeForce RTX 5090 under the new "GeForce RTX 5090 D V2" naming scheme. Tariffs and Export regulations between the US and China have become a headache for gamers, especially those on Desktop PC platforms. While prices in the US reach a time high, Chinese gamers are restricted from access to high-end graphics cards. This all started back in 2023 and has continued throughout 2024 and now 2025. NVIDIA knew about this and has come up with export-compliant GPU offerings, both for HPC and gaming segments. But due to the ever-changing nature of export controls, you never know when the supply of these products can end, only to wait for months for the same or a revised offering. That's the story of the Chinese variant of the GeForce RTX 5090. NVIDIA devised a GeForce RTX 5090 D from the start this time to comply with US regulations, but that didn't go well. A few months later, a new set of regulations was passed, which meant that the 5090 D could no longer be sold in China. That's why reports of a second China-exclusive variant of the 5090 came about. Now, we have more details on what this new card is going to be called, and it looks like NVIDIA won't be using the "GeForce RTX 5090 DD" naming as was previously rumored. According to MEGAsizeGPU, the new graphics card will be called the GeForce RTX 5090 D V2. It almost feels like NVIDIA is running out of ideas to name these revised variants. HKEPC further follows up with their report, and states that the NVIDIA GeForce RTX 5090 D V2 is now scheduled for a launch in August, which means that Chinese gamers will have to wait for one more month before they can get their hands on the new flagship. According to previous reports, the GPU will feature the GB202-240 die versus the GB202-250 on the 5090 D and GB202-300 on the standard 5090. The card will also feature a newer PCB design, the PG145 SKU 40, which is due to changes within the VRAM and GPU layout. The NVIDIA GeForce RTX 5090 D V2 will feature the same 21,760 cores, but a 24 GB GDDR7 memory configuration across a 384-bit interface. That's a VRAM cut of 25%. The GPU will also feature a lower TBP, and we don't have more info regarding the clock speeds, but we can expect gaming performance to remain similar so far that the VRAM doesn't become a bottleneck. The pricing is also expected to remain the same for China, despite the vast reduction in VRAM capability, so that's a bummer for Chinese audiences who previously got 32 GB and now would only get up to 24 GB. This will also affect entry-level AI setups and businesses that were using the large memory capacity to run decently sized LLMs.