NVIDIA's new GeForce RTX 4080 SUPER refresh GPU is currently in the development oven at Team Green, with some new ponderings about its specifications -- and how it can get around the new US sanctions against China -- which have seen the AD102-based GeForce RTX 4090 banned, and the price of the GPU skyrocketing in China.
The upcoming refreshed GeForce RTX 4080 SUPER has been rumored with 20GB of GDDR6X, but a new video from Moore's Law is Dead suggests that we'll see some changes to the refreshed SUPER card that will skirt around the US sanctions in China... and it totally makes sense for NVIDIA to do this.
The Biden administration pushed in new sanctions on China that have outright banned the GeForce RTX 4090 from being sold in the country, with prices jumping within hours. Moore's Law is Dead is reporting from his sources that NVIDIA will not be moving any GeForce RTX 4090 graphics card stock out of Asia and into any other country in any big numbers because shipping containers are so expensive as we head into the holiday season.
- Read more: NVIDIA GeForce RTX 4080 SUPER: 20GB GDDR6X on 320-bit memory bus
- Read more: NVIDIA GeForce RTX 4080 SUPER, RTX 4070 Ti SUPER, RTX 4070 SUPER teased
So the GeForce RTX 4090 will still be sitting in China, unable to be sold -- legally, at least -- leaving the new GeForce RTX 4080 SUPER up in the air for one of the biggest markets in the world for NVIDIA. The last rumor had the GeForce RTX 4080 SUPER using the AD102 GPU... but with the RTX 4090 banned in China -- powered by AD102 -- an AD103-based GeForce RTX 4080 SUPER makes sense.
NVIDIA could use a fully-enabled AD103 GPU with the GeForce RTX 4080 SUPER, which would get it around the sanctions and onto retailer shelves across the now-sanctioned countries... and it's not just China, either.
Moore's Law is Dead is predicting somewhere between 9856 and 10240 CUDA cores for the RTX 4080 SUPER, compared to the 9728 CUDA cores on the RTX 4080. GPU boost clocks would shift to between 2.55GHz and 2.8GHz, up from 2.51GHz on the RTX 4080, while base GPU clocks would climb to 2.31GHz up to 2.5GHz, from the stock 2.21GHz on the RTX 4080.
- Read more: US restricting sales of NVIDIA's flagship GeForce RTX 4090 into China
- Read more: US government restricts NVIDIA H800, A800 AI GPUs to China, new sanctions
- Read more: US restricts NVIDIA from exporting H100, A100 chips to China, Russia
- Read more: NVIDIA's new cut-down A800 GPU created because of US export controls to China
- Read more: Chinese economist: if US ramps sanctions, China needs to seize TSMC
16GB of GDDR6X would still be used, but I would think NVIDIA would step up to the faster 24Gbps GDDR6X memory from Micron (over the 21Gbps GDDR6X used on current RTX 40 series GPUs). TGP would sit around 350-400W, keeping it below the RTX 4090 and above the RTX 4080 in terms of specs, TGP, and performance.