Yesterday, we briefly told you that chips dedicated to artificial intelligence were banned from export to China. Initially, the ban concerned the A100 and H100, then the A800 and H800, boards designed precisely for export. This morning, via TechpowerUp, we learn that this restriction concerns a larger number of GPUs, including the RTX 4090, which is expected to be affected!
The RTX 4090 and other cards affected by these restrictions!
From NVIDIA, we learn this morning that the chip export restrictions are actually broader than expected. In fact, while the A800 and H800 are affected, so too are the L40S and L40 from the Chameleon… And also the RTX 4090, even though it’s part of the company’s all-purpose segment.
Indirectly, exports of systems incorporating these references have been suspended. As a result, China is unlikely to receive the famous HGX and DGX from the Chameleon, for example.
This decision does not only affect NVIDIA. It also affects products from AMD, Intel and other companies. These include Applied Material, Lam and KLA semiconductor.
What’s more, the US government is closing the loopholes by monitoring the destination countries of shipments. As a result, regions such as Macau and Hong Kong are unlikely to be supplied. The fear is that foreign companies will be tempted to ship back to China.
Finally, we also learn that chips intended for datacenters will be subject to restrictions if they exceed a certain power limit, or a certain power density. This limit is to be calculated in flops/mm².
We remind you that the official motives of the American government are to slow down the modernization of the Chinese army. To this end, the US is attacking the artificial intelligence sector in order to curb China’s development in this area.
