NVIDIA has quietly replaced the China-bound RTX 5090D on its product pages with a new GeForce RTX 5090D v2 and set the suggested retail price at ¥16,499, the same sticker as the outgoing D model. The new RTX 5090D v2 keeps the GB202 family's compute muscle intact, with 21,760 CUDA cores, Blackwell architecture and a 575 W TGP, but trims the memory to 24 GB of GDDR7 on a 384-bit bus instead of the previous 32 GB on 512 bits. That change appears aimed at preserving a high-end gaming SKU for Chinese buyers while adapting memory and routing to regional supply and compliance realities. Retail listings and product pages from partners show the V2 already replacing prior entries, which suggests NVIDIA will, or has already, phased out the original D variant in that market.

According to EXPerview , hands-on testing suggest a split picture for buyers. In pure gaming, the RTX 5090D and RTX 5090D v2 cards are often virtually tied, with frame rate differences usually within a percent or two, so most current 4K titles will not noticeably feel the missing 8 GB. For AI workloads and model inference, the gap is more apparent, with single-digit to low-double-digit performance drops where extra memory matters. Partners are preparing custom builds with larger coolers and multi-fan designs that will likely sell above the stated price, given tight supply and demand. For gamers who want local warranty and service in China, the RTX 5090D v2 is now the top-tier choice. For teams and shops that rely on the fuller-memory D models for large-scale AI work, procurement strategies may shift, or premiums will have to be paid to secure the remaining 32 GB inventory.