In reaction to the U.S. export restrictions on AI hardware, NVIDIA has introduced the RTX 6000D Blackwell professional graphics card, a model tailored exclusively for the Chinese market, on the Geekbench benchmarking platform. This card is powered by the GB202 core, which incorporates 156 SM (Streaming Multiprocessors), translating to 19,968 CUDA cores. This represents a significant reduction of approximately 17% compared to the 24,064 cores found in the standard RTX PRO 6000.
When it comes to memory, the RTX 6000D is outfitted with 84GB, a decrease from the standard version's 96GB. Additionally, the memory bandwidth has been scaled down to 448Bit. The core clock speed of the RTX 6000D is capped at 2430MHz, which is notably lower than the standard version's clock speed that exceeds 2600MHz, leading to a consequent decline in overall computational performance.
In the OpenCL benchmark test, the RTX 6000D achieved a score of 390,656 points, a marked drop from the standard version's performance range of 450,000 to 500,000 points.