Now that the Ada Lovelace architecture whitepaper is out (annoyingly with some important details missing, but hey ho...), we can now see exactly how the new 4090 and 4080s compare, on paper, to some of the previous models:
The most important change, bar the big increase in the number of CUDA cores, has to be the massive increase in L2 cache -- and we can thank AMD for that for showing just how important cache is on a graphics card. This should go quite some way to alleviating the narrowing of the memory bus but the high resolution benchmarks will still be interesting to see.
The most important change, bar the big increase in the number of CUDA cores, has to be the massive increase in L2 cache -- and we can thank AMD for that for showing just how important cache is on a graphics card. This should go quite some way to alleviating the narrowing of the memory bus but the high resolution benchmarks will still be interesting to see.
