Nvidia is gearing up to launch the successor of its Turing GPUs. Codenamed Ampere, the GPUs are all set to be revealed in the upcoming few weeks. Nvidia had recently announced that its annual trade show, GPU Technology Conference 2020 will be not be streamed owing to coronavirus concerns. While announcements will be made via news releases, it’ll be interesting to see if Nvidia actually makes major announcements. Nonetheless, the latest leak suggests some intriguing information about the Ampere GPUs.
According to Twitter user and leaker, KittyCorgi, the latest GPUs from Nvidia won’t be based on 7nm architecture. Instead, Nvidia will be using Samsung’s 10nm node. This is pretty surprising given the fact it was hinted that Nvidia will be using the 7nm architecture for its Ampere GPUs. Another interesting fact revealed by the leaker is that RTX will be available for everyone. That is, even the entry-level equivalents of GTX 1650s will get RTX.
As far as the performance improvements are concerned, much is not to be expected as per the leaks. The Ampere lineup will supposedly improve RTX performance while the rasterization performance might not see any significant gains as such. The CUDA core won’t be doubled, but this may happen in the next-gen, i.e, Hopper. FP32 units, however, will be doubled.
If this leak is indeed true, it’ll be interesting to see how the GPUs go head to head against AMD’s RDNA2 architecture. With AMD gearing up to launch its refined 7nm GPUs, the gap might close even further.