Digital Foundry Article Technical Discussion [2020]

Yes, and that makes the PS5 gpu ballpark 2070 performance (or the named 2060 super) in non RT performance, or atleast thats their opinion. Dunno why were even questioning this though, as we already knew or suspected these figures. Il get the PS5 for its games i cant get elsewhere, not for its hardware specs etc. It's a 399/499 box afterall, a very good price anyway.

 

lol it does? hmm I was always under the impression that the 5700XT was competing fairly on par with the 2070. Which is quite an achievement considering that the Turing INT32 performance is not included in the flop count but certainly reflective on benchmarks.

 

Great thread fom Yuri O Donell

 

I think DF's opinion will be very close to the truth

 

5700xt beats a 2070 quite handily and in the higher end, more optimized titles its often on par and sometimes beating a 2070 super.

 

Is that preview image suppose to be an RT reflection? It looks photoshopped. Or is that an RT based artifact where objects too close to the reflective surface aren’t accurately rendered in the reflection?

 

I may have misunderstood (likely, given its 1am on a Saturday night here, and drink time!), but @function seemed to be referring to Ray Tracing performance which of course no RDNA1 GPU can begin to compete in with an RTX GPU.

 

Nice video @Dictator had time to finally step through the whole thing. Great primer and some of the best explanations there are.
I think in the 4 step process represented in your video, that's pretty much what's going to have to happen for consoles. There's not a long of evidence currently that would suggest of there being enough power to do ML denoising on the CUs to make it worth while.

I think Nvidia, my understanding is that they setup the cards to put the pressure on step 4, which is denoising. So you have faster times on steps 1-3, and a very long step 4, which is then handled by ML on tensor cores

 

so again, in this df video there was strong suggetstion that ps5 gpu is on pair with 8gb vram 8tf rdna1 and it's simply no true and I don't know why Alex making such a suggestion as other info was done very profound and good, more caution for Alex that is my advice because his videos are usally really good so no reason to fuel consoles/pc wars bs with his lack of attention of comparision or not saying implictly that he mention only rt performance

 

INT32 should never be counted towards FLOPs, because, INTs aren't FLOPs.

 

Regarding the 5700 Thing I put in the Video, that is a just a circa equivalent that I wrote in there - where you could also put in the 5700 xt if you wanted. GPUs that over of a large grouping of titles perform similar to eachother. In that case I chose the RTX 2060 Super for the Video because I confered with the Team about what GPU to use and we thought the 2070 S Was too close to the 2080 and we do not think the PS 5 GPU will be that high up there necessarily (but it is just a guess). It is more of an informed guess based on what we saw of XSX GPU and our expectations for the scaling between the the PS 5 GPU and XSX GPU using our GPU ladder.
https://www.eurogamer.net/articles/...27-gpu-power-ladder-all-graphics-cards-tested

Somewhere floating around a 5700-5700xt, 2060S, 2070. We cannot be more precise without having an unlocked PS5 game to test with (hopefully hitman 3 when that comes out as we have great contact with IO and they would tell us the exact GPU settings for that game I think). That is why in the Video I said the 2060S is likely a great comparative point for the PS5 GPU - not the same or exact same, but something where we will be using it in comparisons against the PS5 when it comes out.

 

And where on that ladder do you expect the XsX to fall?

 

People are getting hung-up over trivial aspects, rather than the messaging itself. That console gamers are getting RT gaming as well... within reason.

 

insomniac dev said that the RT reflections are even better now than in thoses shots, so we still have to wait a bit more to really judge the quality and PS5 performance in RT with that game.

 

True. But I expect it to be better RT filtering and sharpening techniques, rather than higher RT precision/resolution reflection sampling.

 

thx for explanation but you probably should just use 2070 (not super) and on table put rx5700xt (not rx5700), all this cards are ideed close but better not to fuel console wars, about hitman performance I really doubt ps5 backward compatibilty will be on pair wtih xsx as it uses directx and playstation lower level api and probably will be harder to use ps5 power on ps4 titles without any code change edit: my mistake, you mention hitman 3 not 2, yeah if you know console settings it could be good indicator

 

Here I am talking about Hitman 3 not 2, so a native made for PS5 game that will probably have an unlocked framerate and resolution Option where I think 4K native and 60 fps target could have some areas with GPU related dips where we could then see where other GPUs fair against it in those scenarios. So not a back compat game.
I honestly do not think using the 2060 S fuels console warring as if you Look at the GPU ladder it is basically an RTX 2070 which is why I mention that in the Video. The 2070 is not even produced anymore and by the time the equivalent 3xxx series card that comes out that Lands at the same point on the GPU ladder, i will use and compare that probably in its stead.

 

To anyone worrying about whether the PS5 is closer to 2070 or 2070 Super performance: have a hearty meal and then a hefty wank and ask yourself again if you're really that bothered.

I'm not.

*Burp*

 

understand availability argument but still don't get putting rx5700 (not xt) on description

 

Hi, does Digital Foundry have comparison between cross-platform games of PS5 and Xb Sx? Thanks.