NVIDIA is reportedly working on a refreshed GeForce RTX 4090 graphics card that has been customized for the Chinese market, with the upcoming GeForce RTX 4090 D... the "D" stands for Dragon because it's the Year of the Dragon in 2024 for the Chinese New Year.
We've heard that NVIDIA will be using a tweaked AD102-250 GPU for the RTX 4090 D, which will feature less CUDA cores than the AD102-300 GPU that powers the RTX 4090. The new RTX 4090 D will also have a lowered 425W TDP down from the RTX 4090 and its 450W TDP, but new information is here about the GPU and power consumption: the RTX 4090 D will NOT support overclocking.
NVIDIA's new GeForce RTX 4090 D being locked away from overclocking makes sense, given that it needs to not go over the US government and its constantly updated GPU restrictions in China, so these limitations need to happen. CUDA core counts, GPU clock speeds, TDP, and overclocking support for the RTX 4090 D will be what sets it apart from the RTX 4090.
- Read more: China gets the D: NVIDIA GeForce RTX 4090 D complies with US export restrictions
- Read more: More details emerge on the GeForce RTX 4090D for China, GPU clocks
- Read more: NVIDIA's GeForce RTX 4090D for Chinese market uses cut-down AD102-250 GPU
- Read more: RTX 4090 dismantled in China, sold as Al GPU instead
NVIDIA won't be changing the memory configuration on the RTX 4090 D, where it should still have 24GB of GDDR6X memory... an important part of any AI accelerator is lots of high-speed video memory. We should expect changes to stay with the GPU side of things -- clock speeds, CUDA cores, TDP, etc -- which is what will make the RTX 4090 D different to the RTX 4090 for China. Let's see how long this card lasts on the market in the country before it's scooped up by everyone that's thirsty for AI accelerators in China.