Nvidia will add more restrictions to RTX 4090 D
In response to the United States government’s new export controls on cutting-edge artificial intelligence (AI) chips, Nvidia plans to launch a China-specific version of the GeForce RTX 4090 D, equipped with AD102-250, to replace the GeForce RTX 4090, which has been included in the restricted list. This new model is expected to arrive early next year. Reports indicate that the RTX 4090 D will have a slightly higher base frequency than the RTX 4090’s 2235 MHz, reaching 2280 MHz, while maintaining the same boost frequency of 2520 MHz.
According to Benchlife, although the exact specifications, such as the number of CUDA cores, are still unclear, the RTX 4090 D will have a lower total power consumption than the existing RTX 4090, dropping from 450W to 425W. Moreover, Nvidia is adding more restrictions to the RTX 4090 D, such as disabling overclocking, which implies limitations set through firmware or other means. Reportedly, graphics card manufacturers are set to receive samples for testing this week, potentially leading to the leakage of the RTX 4090 D’s specific specifications online.
The RTX 4090 D must comply with a Total Processing Power (TPP) limit of 4800, whereas the current RTX 4090, whether in FP8 or FP16, has a TPP of 5286, exceeding the limit by about 10%. Under normal circumstances, Nvidia would need to reduce the number of CUDA cores in the RTX 4090 D, which is currently less than the RTX 4090’s 16384. With the overclocking limitation in place, it seems that Nvidia is opting to minimize the reduction in CUDA cores and not provide any leeway for manufacturers.
Rumors suggest that the RTX 4090 D will still be equipped with 24GB of GDDR6X memory, matching the specifications of the RTX 4090. Additionally, both models are expected to have the same price point of $1599.