I am not surprised that this is the usual practice of Nvidia to be honest. And with competition getting intense of late, they may likely double down on this path. While it’s true some of these reviews don’t talk about the other features on Nvidia cards, but rasterization is bread and butter for a gamer and his/her graphic card. Rest are mostly just cherries on top. DLSS and NVENC are likely the other 2 useful features, but again, DLSS is not supported in every game and not every gamer will stream their gameplay. And if so, then it falls back to the card’s rasterization capability.
not to mention it will be extremely dangerous for GPU market if these stuff were a high priority for consumers.
You guys must be quite young to the DIY PC scene I guess.
In the past, heck even now, there are games that run faster on nVidia cards, there are games the run faster on AMD / ATi cards. Seems to happen less now but in the past you could have games that are 25% or 50% faster on one particular card brand.
Check out wikipedia regarding shader / tmu / *** count. Notice the ratio is different? Yea and people will be quick to point out that nVidia and AMD's shader/tmu/*** have different speeds. Point is, every card is designed differently and have pros and cons. AMD cards are really good for mining and have many FLOPS due to high shader count. Now NVDA decided to do the same.
You guys know nvidia's shader count pretty much doubled in RTX 3000 series compared to RTX 2000? And more shaders means more flops!!! Like Xbox and PS5, lots of FLOPS!!! Even though the more seasoned DIY-ers all know FLOPS mean jack after 1060 vs RX480.
If you go way back, look at what is the shader count in the past compared to the *** count. Geforce 8800GTX shader count is 128, TMU of 64 and *** of 16!!! What's the ratio now, 8704:272:96??? *** count is now just 1% of shaders?
Not to mention this series is the first unified shaders. Before that we would have separate vertex and pixel shaders. Plus all the new hardware features like T&L and other DX8 / DX9 stuff. Yea I'm going very way back.
Point is, different cards can be made very differently, and result in different strengths. A high "brute force" card of an earlier era would actually be slower than a lower-end newer card with new features. Don't believe you can check the reviews.
"not to mention it will be extremely dangerous for GPU market if these stuff were a high priority for consumers."
Indeed it is. There were many people who said DX9 hardware features were useless because games looked just as good in DX8 while running faster in DX8. (True story.) They said the same about DX10. And about DX11. People shunned the idea of more shaders over ***. Then the era changed and shaders became everything.
Indeed if these stuff were a high priority for consumers, there will be no progress and we will still be stuck with DX7 with no unified shaders.
This is now just a repeat of what has happened in the past, except now we have these RT cores also. And the ratio in RTX 2000 is too low so they increased it in RTX 3000.
I think more important now tho, is that we also have influencers now.
Oh... and BTW, since DLSS is mentioned, did u guys know MSAA also didn't work properly for many titles in the past? We pretty much just used AA, or what we now retroactively call SSAA or supersampling AA.
Also does anyone ever feel stupid that in the past we used supersampling because we could render more pixels than our screen resolution, now games are resorting to doing subsampling (i.e. that render resolution % slider) because we couldn't render as many pixels as our screen resolution. *However*, there is an article on can higher resolution replace AA, and the short answer is, no. Because AA does more things intelligently than just double the render resolution and halving it. Yet again another demonstration of technique over brute force.
