Digital Foundry Article Technical Discussion [2020]

This is interesting this what I expected from Xbox Series X and not PS5 and better than a 5700XT. Not bad at all.

Around RTX 2080 is not bad.

The DRS option is not good at all on PC, why limit the option.

 

Around RTX 2080 performance is just for this game? I don't think we can generalize across all PS5 /XBSX games at this early stage.

 

No surprises here!

 

as expected, rtx2060 super and radeon rx5700 (not xt) was easly defeated by ps5 so theory of lacking bandwidth and putting ps5 gpu on pair with rx5700 was wrong as I expected

 

So if a game favors Nvidia's apis/drivers, we should automatically make assumptions that PS5/XBSX hardware is less equivalent than Nvidia's GPU based products?

 

I do think PSman raises a valid point here that Valhalla heavily favours AMD hardware. Heck, Alex also pointed that out in the video. Looking at the following benchmarks the AMD GPU's are performing well above their normal positions with the 5700XT - usually around 2070-2070S level performing closer to a 2080Ti. In this game based on Alex's video the 5700XT would probably be at least as fast as the PS5. Personally I was expecting the PS5 to match or exceed the 2080Ti in this title.

https://www.techspot.com/review/2146-amd-radeon-6800/

It was compared to a 5700 non-XT in the video. It's likely the 5700XT is around PS5 performance or even a little better in this title.

 

https://www.guru3d.com/articles_pag..._graphics_performance_benchmark_review,6.html

5700 and 5700XT only differ several percent for this game.

PS5 is better than 5700XT in AC Vahhalla.

 

So the question becomes, which games should be compared on gauging performance towards making comparable comparisons between PC GPUs and console hardware? Games where AMD or Nvidia haven't peppered their tech with or have associated deals with.

 

I don't think it matter so much, we now know that ps5 gpu is not slower than 5700xt and there will be always games that favorite more amd or nvidia architecture

 

Does Digital Foundry test other PC games such as COD?

Generally the performance gap between RTX2080s and 2070 is about 25%. With more and more optimization

for consoles we would see PS5 performance closer to RTX2080s as time goes on.

 

If you look at the recent games, you'll see that more games favor AMD cards. With "next-gen" kicking off, more games will perform better on AMD cards.
Turing cards will not age gracefully in rasterization workloads.

 

It will be interesting to follow the performance of games in the months and year to come. Games will all be released with Direct X 12. The three multiplatform games without triangle based RT performs very good on AMD GPU(AC Valhalla, Godfall and Dirt 5).

Some games will go with non based triangle raytracing too, it will be interesting to compare like on Unreal Engine 5 and Lumen.

 

That's kind of my point. Because PS5/XBSX hardware is AMD based and a particular game may favor AMD tech, doesn't necessarily mean we can't make certain observations on Nvidia GPU equivalencies without someone always constantly reminding us its an AMD favored game.

 

Yea I don't want to take anything away from PS5 performance here; so I'll start my post with: PS5 is a beast. That clutter option is ace.

Following though, I had originally pegged PS5 to perform about 5700XT (for many months before launch) but with even higher clocks (5700XT 50th anniversary was my common comparison GPU with an even higher clocked GPU than the standard 5700XT).
I guess it's still actually sitting probably around that range, but what's changed here is the general position of the 5700 series. It's just performing remarkably well when optimized for it. I wish I knew what changed on the coding side of things to make this alternation of typical positions between a 2080S and a 5700XT.

I think it's pretty clear no one really knows what's happening with XSX. Just going to wait for more data to see if this behaviour will stay consistent for the next 6 months or if it's going to start changing.

The killer for me was when Alex showcased the VRAM requirements being below 10GB for 1440p. I was just confused at that point in time. I no longer understand what's happening.

 

That's another data point but it doesn't align with either the Techspot results I posted further up, or more importantly, with Alex's own video which shows the 2070S performing above the 5700 (they aren't directly compared but look at both cards framerate at the PS5 low point of 51 fps). You simply can't make the claim "PS5 is better than 5700XT in AC Vahhalla" based on the available data. It may well be, but we have conflicting data points which suggest it also may not. Naturally I would expect the PS5 to perform better so if it doesn't, it would be interesting to understand why. Bandwidth may be the answer in that particular scene at least.

Surely the answer to this is "not one in isolation?" A cross section of games is going to be needed and I'm sure we'll get that over time from the awesome work at Digital Foundry. You may be absolutely right that the trend moving forwards will be next gen multiplatform games in general preferring AMD hardware, at least in non-RT workloads. And if that's the case then yes, it's fair to say that in modern titles the PS5 is in the 2080/S league. But for now, games like Valhalla, Dirt 5 and Godfall are outliers.

In any case, the one hopefully non-controversial conclusion I can take from this game is that the PS5 is performing roughly where we'd expect it to in relation to PC hardware and thus the parity with XSX appears to be an issue with the XSX underperforming rather than with the PS5 overperforming. At least in this title.

 

so the final anwser is/ we don't know....

yet

 

PS5 average performance leads 5700 by 25~30% in the video. And generally 5700XT leads 5700 by 15% or so. Therefore PS5 should lead 5700XT 10% or above.

By the way do you have any evidence PS5 performs under 5700XT?

 

One thing to note here, the infamous 1.0.4 patch led to a 6% hit on PS5 on the scene used to benchmark performance between PS5 and PC

 

These are first gen games too. If we look at the evidence of the last generation, the perfomance in relation to equivalent PC hardware will move favourably for the consoles over time. Considering VRAM allocations, etc.

 

It's more than an outlier, it's unprecedented if this is just optimization.

If we talk about AMD optimization, Decima engines should unfairly be optimized for AMD considering they only had PS4 to work with. All the titles have largely been optimized for GCN architectures because of the consoles. In doing so, we've never seen a family of GPUs jump several positions higher suddenly due to a release of a console based on the newer architecture a full year and 3 months later. With PS4 generation, the 5850 and 5870 didn't suddenly start performing out of price bracket; it's one thing where we see the arguments that there is secret sauce in PS5 therefore performing above expectation but it's another thing entirely when the older architecture of 5000 family of cards is also lifted in performance. I don't think we've seen that before there's no precedent we've had in forum discussion for this really so to speak, we are in uncharted territory.

The only thing I can link between the newest 6000 series of cards, with the 5000 series of cards and PS5 - the only commonality that all 3 share than the Nvidia cards don't, is that NGG / primitive shader pipeline that can apparently be done driver side. Or if not, then they explicitly used Primitive shaders. Mesh shaders are supposed to be the same thing but require explicit programming - the wins could possibly be entirely front end based. Of which they can take advantage of this during the cross generation transition. It may explain why XSX is doing so poorly here even compared to the 5700XT as it's performing closer to that 2070S level of expected performance (but I'm not holding my breath on this one)