Random_Matt said:
Actually compatible with the 3000 series, they decided not to implement it. |
Well supposedly, the reason they didn't is because it was too laggy on Ampere/Turing as the instruction sets on Ampere wasn't fast enough to compute. The tensor core was the biggest improvement this gen going from 320 Tensor ops from Ampere (3090Ti) to 1400 in Ada (Eventual 4090Ti). So I don't think they are totally lying. I am sure if you go down the stack low enough with like a 4050, that could have similar tensor performance as 3080/3090. But they can't just support a few GPUs, they would have to support all the way down to a 3050 which then would get bottlenecked.
So while I don't like it, I can see the reasoning if its legit. The good thing is that DLSS 2 will be supported in every game that has DLSS 3 so Ampere/Turing users will continue to get DLSS support. They are also going to continue to improve DLSS 2 as it directly affects DLSS 3 as well and any improvements to DLSS 2 will apply to Ampere/Turing users.
PC Specs: CPU: 7800X3D || GPU: Strix 4090 || RAM: 32GB DDR5 6000 || Main SSD: WD 2TB SN850