NVIDIA is gearing up to release the GeForce RTX 5090 D V2, a China-exclusive graphics card, on August 12, 2025, tailored to comply with U.S. export restrictions while delivering high-end gaming performance. This new GPU, featuring the GB202-240 chip and 24GB of GDDR7 memory, marks NVIDIA’s latest effort to maintain its foothold in the Chinese market amidst tightening regulations.
The RTX 5090 D V2 follows the RTX 5090 D, which was launched in January 2025 to meet U.S. export controls limiting high-performance GPUs in China. While the original RTX 5090 D retained the same 21,760 CUDA cores and 32GB of GDDR7 memory as the global RTX 5090, it featured reduced AI performance (2,375 AI TOPS compared to 3,352) to comply with regulations. However, new restrictions halted RTX 5090 D deliveries, prompting NVIDIA to develop the V2 variant with further tweaks to ensure compliance.
The Release date of the GeForce RTX 5090 D V2 with 24GB memory has been confirmed. If there are no major surprises, this compliant GeForce RTX 50 series graphics card will be Released on time on August 12.
~ Benchlife (Translated)
The RTX 5090 D V2 uses the GB202-240 GPU, a step down from the GB202-250 in the RTX 5090 D and GB202-300 in the standard RTX 5090. The most significant change is the reduction in memory, from 32GB to 24GB of GDDR7 across a 384-bit memory bus, resulting in a 25% decrease in memory bandwidth compared to the 512-bit interface of its predecessor. Despite this, the card maintains the same 21,760 CUDA cores and a 575W TDP, ensuring gaming performance remains robust.
| Name | RTX 5090 D V2 |
|---|---|
| GPU | GB202-240 |
| CUDA Cores | 21760 |
| Memory | 24 GB GDDR7 |
| Memory Bus | 384-bit |
| Process Node | TSMC 4N |
| Bandwidth | 1.34 TB/s |
| TBP | 575W |
| Release Date | Coming August |
The V2 uses a new PCB design (PG145 SKU 40) to accommodate the adjusted VRAM and GPU layout. While the memory reduction may impact AI workloads and synthetic benchmarks, gaming performance is expected to remain comparable to the RTX 5090 D, as 24GB of VRAM is sufficient for most gaming scenarios before bottlenecks occur. The card also retains the same clock speeds as the RTX 5090 D, and overclocking capabilities are expected to be similar, though NVIDIA has implemented restrictions to limit AI and crypto-mining performance, including a 3-second performance cut-off for such workloads.
The launch comes at a critical time for NVIDIA in China, a key market where the company faces no direct high-end competition from AMD. The RTX 5090 D faced supply challenges due to export bans, with stock nearly depleted by May 2025. The V2 aims to fill this gap, but its launch has been clouded by uncertainty. Recent reports suggested NVIDIA might cancel the V2 if U.S. restrictions on GPUs like the H20 AI chip were eased, allowing resumed sales of the RTX 5090 D. Although NVIDIA has not officially announced the MSRP for the RTX 5090 D V2, the August 12 launch remains confirmed as of now.