According to reports, NVIDIA has shifted the supply of its “China-Only” Ampere A800 AI GPUs while supervising the US government’s export ban to unfriendly countries
After China was hit with a US ban, NVIDIA plans to make A800 AI GPUs available globally, giving businesses focused on AI another optionFor more details govindhtech.com
The A800 AI GPU is found in the dual-slot PCIe card known as the A800 40GB Active product from Nvidia For servers in China, the A800 AI GPU was also offered in the SXM form factor in addition to the PCIe card
Said to “bring the power of a supercomputer to your workstation and accelerate end-to-end data science workflows,” the A800 40GB Active GPU is billed as the “ultimate workstation development platform for AI
The Nvidia A800 40GB Active PCIe card, like the A100 40GB, contains 6,912 CUDA cores, 432 Tensor cores, 40GB of high-bandwidth HBM2 memory, and 240 watts
Nvidia stated in marketing materials that the A800 40GB Active GPU is 4.2 times faster for AI inference with the BERT Large model, 90% faster for AI training with the BERT Large model, 90% faster for the GTC benchmark
Along with other AI chips in its lineup, Nvidia claimed that the A800 40GB Active GPU is also included with a three-year subscription to Nvidia AI Enterprise
According to reports, NVIDIA has joined forces with US-based partners including PNY and system integrator Colfax International to increase sales of A800 AI GPUs, which were once solely meant for use in China