Nvidia Ampere Discussion [2020-05-14]
- Thread starter Man from Atlantis
- Start date
-
- Tags
- nvidia
Rootax
Veteran
Yeah, I'm hyped. Don't really know why, as I won't change gpu soon. But, it feels (or it's only me), than nVidia will bring the big guns this time. I don't expect much in classic rasterisation (maybe 30% vs 20xx ?), but it's their second shoot at RT, I really hope a big increase, like 50-75%.
And then I will be curious of amd reaction, since they can't have a too little too late moment again imo. It will be fun.
And then I will be curious of amd reaction, since they can't have a too little too late moment again imo. It will be fun.
Rootax
Veteran
It will depend a lot of the competition.
Competition will need to at least match feature set.
84 vs 82 is so marginally more (+2.x %) and TGP seems high aleady, so they probably cannot clock it meaningfully higher either.
Big Navi probably has more than (Texture) Sampler Feedback Tier 0.9, HDMI 2.1 and Raytracing are basically confirmed (through consoles) as well and all the other checkboxes are ticked by Navi already. Or what am I missing here?
Man from Atlantis
Veteran
DavidGraham
Veteran
Performance in Ray Tracing, the details given in Xbox Series X puts RDNA 2 to be less performant than Turing, Ampere could be significantly higher than both. Also DLSS 2, also performance in Mesh Shaders is a mystery.
Rootax
Veteran
a DLSS like solution ?
Imagine Ampere performing below Turing. Has that ever happened? (2003 5700 ultra?)
That's a software check box, not hardware.
DavidGraham
Veteran
Software and Hardware in NVIDIA's case.
I think it's the tensor cores helping out there.
Jawed
Legend
We know that these newest process nodes have fairly high variability in clocks/power - the silicon lottery only gets worse as time goes by.
84SM + best silicon + ~8% better bandwidth?
If you're interested in a 3090, are you not going to wait to find out what the 3090Ti does? Or is that going to be the Titan, which gamers just ignore because $4000?
I'm really curious as to what they end up releasing. There is so much speculation. I'm sure it'll make for fascinating reading when the entire history and product is revealed. As a consumer I'm more interesting in seeing how RT slots into the product, especially in lower end devices. If RT is to reach a semblance of general acceptance it needs to be possible even in the mid-range.
I just hope the performance boost is better, and the value proposition, than RTX had. I'm still disappointed in the rather tepid performance improvement it offered over the previous gen. RT as a feature wasn't worth it then, and I have a hard time believing it will be fully worth it now. But we shall see. Exciting times ahead!
I just hope the performance boost is better, and the value proposition, than RTX had. I'm still disappointed in the rather tepid performance improvement it offered over the previous gen. RT as a feature wasn't worth it then, and I have a hard time believing it will be fully worth it now. But we shall see. Exciting times ahead!
No confirmation yet.
It's about how performant any competing ml upscaling that amd can do is.
If RDNA2 and ampere is in the same ballpark in general performance but ampere can render at 1080/1440p and achieve equivalent image quality, then that would be a huge performance gain.
I know not all games use dlss, but it is enough and growing to make a reasonable difference, even if only for a headline statement.