Games Analysis |OT| Time To Argue About Pixels And Frames!

[No-1HoloLens-Fan]

@Louay

Do you really think XSX can’t handle 2000Mhz or 1900MHz with it’s cooling ? lol no it can but they decided not do it because the fan will get louder. there is no nonsense about this. also this is not about PS5. i’m comparing XSX to RDNA 2 GPUs for example RX 6800 XT 72CUs wider and faster 2,015 MHz.

I can see what you are trying to say.

People are just completely missing you point.

[Zappy5]

@Louay

Do you really think XSX can’t handle 2000Mhz or 1900MHz with it’s cooling ? lol no it can but they decided not do it because the fan will get louder. there is no nonsense about this. also this is not about PS5. i’m comparing XSX to RDNA 2 GPUs for example RX 6800 XT 72CUs wider and faster 2,015 MHz.

I’m not being funny but you are just saying numbers with absolutely no basis for them. When they engineer these things they calculate a strict power and heat and cost budget. They don’t just say ‘hey what happens if we whack the clocks up’. And given the excellence we’ve seen from Xbox in terms of engineering I think we can safely and confidently assume they’ve produce a machine with everything at the peak it can possibly be without compromising on heat issues. There is a pretty simple calculation for fan size and box size to heat produced.

And for consoles you bet they leave themselves some headroom because they don’t want mass failures on their hands. And unlike PC people aren’t opening these things up to clean them out or reapply thermal paste every 6 months and people expect them to last ten years. They work out their budget and stick to it.

[LifeForms]

@LucasTaves

A 5700XT has around just a 10% performance uplift over the 5700, it may go higher on some games but in general a lot closer to 10 than 15-20 range. Though that was for older games. Benchmarks for 5700 became harder to come by after 5700XT launched because everyone defaulted to that one.

TBH the ps5 result in here impressed me. Wasn’t expecting the settings to be that high (and previously Alex mentioned they were using the same settings as X, but that doesn’t seem to be the case), and the performance to be that high comparatively.

Hope they do more comparisons like that. It doesn’t seem it’s reaching that level across other games but it didn’t seem it was that high in here at first either.

Looks here the 5700xt performs better then PS5 considering it outperforms a rtx2080.

[Zappy5]

@LucasTaves

A 5700XT has around just a 10% performance uplift over the 5700, it may go higher on some games but in general a lot closer to 10 than 15-20 range. Though that was for older games. Benchmarks for 5700 became harder to come by after 5700XT launched because everyone defaulted to that one.

TBH the ps5 result in here impressed me. Wasn’t expecting the settings to be that high (and previously Alex mentioned they were using the same settings as X, but that doesn’t seem to be the case), and the performance to be that high comparatively.

Hope they do more comparisons like that. It doesn’t seem it’s reaching that level across other games but it didn’t seem it was that high in here at first either.

Why isn’t it the case that the settings used are the same on both consoles? I’ve not seen anything to suggest that isn’t the case.

[LifeForms]

@No-1HoloLens-Fan

I can see what you are trying to say.

People are just completely missing you point.

Thats because its a completely random point, nothing to do with the discussions at hand. Might as well start comparing mac + cheese recipes.

[LucasTaves]

@LifeForms

Looks here the 5700xt performs better then PS5 considering it outperforms a rtx2080.

Hmm, interesting. It seems the game is an outlier then, the 5700XT doesn’t usually outperforms a 2080 from what I’ve seen.

[LucasTaves]

@Zappy5

Why isn’t it the case that the settings used are the same on both consoles? I’ve not seen anything to suggest that isn’t the case.

I said X not SX

Previously Alex said both SX and Ps5 were using the same settings as X, but apparently they are above (not only that but watching some footage it seems the lods being more aggressive on X)

[LifeForms]

@LucasTaves

Hmm, interesting. It seems the game is an outlier then, the 5700XT doesn’t usually outperforms a 2080 from what I’ve seen.

Yep, ACV performs terrible on Nvidia cards, some are claiming conspiracy because its amd sponsored.

[Staffy]

@javycane

I think its naive on both fronts to think the Series X would obliterate everything from the gate or that the PS5 would not be a quality machine and lagging behind.

Both are massive jumps in quality over the previous gen and might fine devices. Its been 1 month and it feels like a lifetime with all this pixel counting.

I can’t deny that I definitely thought it was a given that XSX would come out on top. Have small edges in terms of resolution and framerate and that’s about it. But that hasn’t been the case at all yet.

What I wonder is, if MS had stuck with XDK, would they have been leading in these analysis videos? The BC analysis videos are telling where resolution is clearly in favor of XSX.

Also, had all the tools been ready for developers for more than a year, chances are we may not have seen this happen. It sounds logical that very late tools are the reason we are seeing this, honestly it could have been way worse.

But let’s hope that’s the case, let’s hope there isn’t some other underlying problem that will turn out to be a long term issue for devs.

Take that anonymous developer that Tom Warren talked about. He said even basic things (he gave examples) proved to be challenging with GDK. Is that normal?

[No-1HoloLens-Fan]

@LifeForms

Thats because its a completely random point, nothing to do with the discussions at hand. Might as well start comparing mac + cheese recipes.

The topic at the hand is very large. It’s gonna involve hardware and software discussion & comparision.

[LucasTaves]

@Staffy

…n terms of resolution and framerate and that’s about it. But that hasn’t been the case at all yet. What I wonder is, if MS had stuck with XDK, would they have been leading in these analysis videos? The BC analysis videos are telling where resolution is clearly in favor of XSX. Also, had all the tools been ready for developers for more than a year, chances are we may not have seen this happen. It sounds logical that very late tools are the reason we are seeing this, honestly it could have been way worse. But let’s hope that’s the case, let’s hope there isn’t some other underlying problem that will turn out to be a long term issue for devs. Take that anonymous developer that Tom Warren talked about. He said even basic things (he gave examples) proved to be challenging with GDK. Is that normal?

It’s an interesting discussion. XDK with Gen9Aware apis seems to be exceeding expectations (around twice the X performance even in gcn mode), so in theory adding the apis for the new features and enabling rdna2 features on XDK would lead to a very performant and mature dev environment.

But I can see why MS would prefer to sunset the XDK sooner rather than later, and making some features exclusive to GDK would be an effective way to enforce that

[Zappy5]

@LucasTaves

I said X not SX

Previously Alex said both SX and Ps5 were using the same settings as X, but apparently they are above (not only that but watching some footage it seems the lods being more aggressive on X)

Ah ok yes fair enough!

[Fiendflag]

@Zappy5

I think it’s very clear that the Series GDK has less than performant APIs. It’s the only explanation possible here…if the issue had anything to do with CU counts or clock speeds you’d see that reflected in the latest Digital Foundry performance video comparing PS5 and PC yet you don’t see that - everything scales relative to its teraflop performance…

Sorry I was meant to say APU not API. The most likely scenario is that the dev tools are the biggest problem I agree.

About everything scaling to it’s teraflop performance, it seems to be true for PS5 but not for SX though…that’s why I said it’s like the game is not using the extra CU’s. It’s like the SX performance is utilizing the same CU’s as the PS5 version but due to the lower clock speed we see the lower resolution is some spots.

Also to be honest I am not in any way an expert on this kind of stuff, I just want to see what you guys think about this theory. I am probably full of BS so yeah tread lightly!

[Zappy5]

@Fiendflag

… Sorry I was meant to say APU not API. The most likely scenario is that the dev tools are the biggest problem I agree. About everything scaling to it’s teraflop performance, it seems to be true for PS5 but not for SX though…that’s why I said it’s like the game is not using the extra CU’s. It’s like the SX performance is utilizing the same CU’s as the PS5 version but due to the lower clock speed we see the lower resolution is some spots. Also to be honest I am not in any way an expert on this kind of stuff, I just want to see what you guys think about this theory. I am probably full of BS so yeah tread lightly!

The game is using the extra CUs on Pc though…I’m certain it’s now down to new APIs just not being as performant. Like when MS went to Dx12 on Pc and for a while you saw games performing way better on DX11 still…

[Fiendflag]

@Zappy5

The game is using the extra CUs on Pc though…I’m certain it’s now down to new APIs just not being as performant. Like when MS went to Dx12 on Pc and for a while you saw games performing way better on DX11 still…

Yeah that’s why I said that the SX version might be a quick unoptimized port of the PS5 version (which probably was the lead platform at least on consoles due to Ubi having the dev tools earlier and all) which is optimized around 36 CU’s, also aren’t both PS5 and SX versions at the same exact graphics settings?

It’s hard to believe that only the dev tools are the reason for the SX version performing like a ~9.5-10.5tf machine. Anyway curious to see how the next wave of games perform, Cyberpunk will be an interesting one for sure.

[Staffy]

@LucasTaves

It’s an interesting discussion. XDK with Gen9Aware apis seems to be exceeding expectations (around twice the X performance even in gcn mode), so in theory adding the apis for the new features and enabling rdna2 features on XDK would lead to a very performant and mature dev environment.

But I can see why MS would prefer to sunset the XDK sooner rather than later, and making some features exclusive to GDK would be an effective way to enforce that

Yes, very understandable. That’s true.

I think all things considered, the tools being this late I think it could have been much worse. So either devs really just know their shit, the tools aren’t that bad or it’s both.

I think it can only improve from here.

[Louay]

(1) DF Developers: Ori And The Will of the Wisps - The Next-Gen Xbox Series X/S Challenge [Sponsored] - YouTube

[LucasTaves]

@Staffy

Yes, very understandable. That’s true.

I think all things considered, the tools being this late I think it could have been much worse. So either devs really just know their shit, the tools aren’t that bad or it’s both.

I think it can only improve from here.

There was an interview with Jason Ronald in the latest major Nelson podcast. It seems to be some factors at play

• Tools coming hot and not being as performant as they should until very late for launch games
• Some peculiarities around SX hardware that devs are not using well (perhaps the memory layout)
• Unexpected bugs and performance issues in the platform.

Jason said Ms is working both internally on their side to ensure the tools improve and also with developers to make sure they know how to use the hardware and patch their games as the performance issues are fixed.

But I imagine a developer that has had more experience with the hardware and the tools can get around them even in this state such as Coalition, can get around them and still deliver the expected performance like gears 5 going above ultra, even with some raytracing effects and still averaging 1800p.

But all things considered I think SX should be performing around 40-60% better than it is right now (considering both the bandwidth and the tflops), if that difference is solely from the devkits that’s actually not too different from what xbone had were going from the same dx11 as pc to a low level console specific version had gains at around 50% in the first year.

[Staffy]

@LucasTaves

…ve and also with developers to make sure they know how to use the hardware and patch their games as the performance issues are fixed. But I imagine a developer that has had more experience with the hardware and the tools can get around them even in this state such as Coalition, can get around them and still deliver the expected performance like gears 5 going above ultra, even with some raytracing effects and still averaging 1800p. But all things considered I think SX should be performing around 40-60% better than it is right now (considering both the bandwidth and the tflops), if that difference is solely from the devkits that’s actually not too different from what xbone had were going from the same dx11 as pc to a low level console specific version had gains at around 50% in the first year.

Starting to sound like these are things that we can’t even really blame on MS of devs, just unexpected unfortunate surprises basically. But it’s great to know that they are helping developers out where they can.

Good point about Gears 5. Some want to mention that it isn’t hitting native 4K, but they forget to mention as well that it’s doing things not even PC does. Xbox One is a good example, that made a real nice improvement where it started with 720p and then 900p and even after a while 1080p games. It should be fine eventually.

[LifeForms]

@No-1HoloLens-Fan

The topic at the hand is very large. It’s gonna involve hardware and software discussion & comparision.

It is, but not the topic of the posts he originally quoted were quite specific, if its a seperate point its makes sense to just have it as a stand alone post rather then a quote.