• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

DF: Control PS5 Vs Xbox Series X Raytracing Benchmark

ethomaz

Banned
You sure as shit seem very invested in this for an adult man. Jesus christ.

The PS5 is the best console ever and even pcs are shit compared to it. There, be happy. Also, DF are in MS pocket and they do this comparisons on purpose so Phil can be happy and the shit engineers MS has.

And while you were on it, since you're bringing information from beyond3d you could at least post the comparison Dictator made of the infamous corridor with the green light, that you and you buddies posted across two threads as some kind of gotcha.
Truth > what you thing about me.

BTW I have no ideia what are you talking about because I never even mentioned that corridor (I don’t even know where it is in the game lol).

But you can post it if you think it is worth lol
 
Last edited:

rsouzadk

Member
You sure as shit seem very invested in this for an adult man. Jesus christ.

The PS5 is the best console ever and even pcs are shit compared to it. There, be happy. Also, DF are in MS pocket and they do this comparisons on purpose so Phil can be happy and the shit engineers MS has.

And while you were on it, since you're bringing information from beyond3d you could at least post the comparison Dictator made of the infamous corridor with the green light, that you and you buddies posted across two threads as some kind of gotcha.
He wont. You should know that by now. It is just a "bench" and people are going crazy about it.
 
Last edited:
Pretty much why comparing this to the gameplay is pretty interesting. The GPU does have that power but it doesn't necessarily translate to the gameplay. A very interesting study.
Yes Sir it is, but certain individuals are trying to create something out of nothing, I feel sorry for Alex and the DF team. The usual suspects are up to their usual shenanigans to discredit them, shameful.
 

01011001

Banned
I am actually wondering if it would be a good idea to forbid threads like these... the sheer stupidity in them is mindboggling to say the least. low IQ take after low IQ take from both sides in each of these stupid threads. facts are not accepted as such, the "losing side" always looks for stupid small details to try to discredit the source... it is ridiculous
either that or make fanboying inside threads like these bannable... this shit is not fun or has any value
 
Last edited:
I am actually wondering if it would be a good idea to forbid threads like these... the sheer stupidity in them is mindboggling to say the least. low IQ take after low IQ take from both sides in each of these stupid threads.
either that or make fanboying inside threads like these bannable... this shit is not fun or has any value
They should remain open, but the fanboys should be hit with the ban hammer 🔨. Threads like this are great for those who want to discuss, but horrible for fanboy fodder.
 
Last edited:
Yes Sir it is, but certain individuals are trying to create something out of nothing, I feel sorry for Alex and the DF team. The usual suspects are up to their usual shenanigans to discredit them, shameful.

I wouldn't call all members of Digital Foundry unbiased. But I don't think they are manipulating their results to put Xbox ahead.

What bugs me the most is instead of trying to explain why the PS5 might be ahead in something they merely say that they don't know.
 
I am actually wondering if it would be a good idea to forbid threads like these... the sheer stupidity in them is mindboggling to say the least. low IQ take after low IQ take from both sides in each of these stupid threads. facts are not accepted as such, the "losing side" always looks for stupid small details to try to discredit the source... it is ridiculous
either that or make fanboying inside threads like these bannable... this shit is not fun or has any value
They seem to have a particular dislike to Alex and its borderline harassment as far as I'm concerned. What many don't realise is that we only know exact settings and information about different versions because developers talk and trust DF. They don't talk to NXGAMER, VG tech etc, DF give us the full picture and for that reason alone they should be commended.
 

ethomaz

Banned
I wouldn't call all members of Digital Foundry unbiased. But I don't think they are manipulating their results to put Xbox ahead.

What bugs me the most is instead of trying to explain why the PS5 might be ahead in something they merely say that they don't know.
VGTech is not saying DF is making up results but that the tools are lacking.

That is something Batalhia already confirmed when he said that he had to manual count the framerate in Vanalla due the tools having issues with screen tearing.
 
So he's saying that some next gen games will perform better on the PS5?

I'm wondering what kind of games those will be.
Well he is right. We already have at least one next gen game: NBA 2K21 has next-gen assets specific to the new consoles. DF about it:

but with the arrival of NBA 2K21 on the new wave of consoles, we're seeing something genuinely fascinating here, with a proper generational leap in fidelity across the board...It's really just an academic difference, but it is there, and PS5 does ultimately have the edge here.

Using the same res and effects It runs at least 10% better on PS5.
 

ToTTenTranz

Banned
What bugs me the most is instead of trying to explain why the PS5 might be ahead in something they merely say that they don't know.

Perhaps because they actually don't know.
Microsoft has been open about their hardware implementation, especially so towards DF with whom they partnered with to announce the consoles' specs to the world.

On Sony's side they may simply not know as much about the hardware. They don't know how much effective memory bandwidth is being gained by the cache scrubbers or the custom "14C" GDDR6 with probably tighter latency, if there is some kind of Infinity Cache, how much L3 is there in the CPU and whether or not it's using a single CCX with 8 Zen2 cores and unified L3 (effectively making it "Zen 2.5"?), like RedGamingTech has been suggesting for months.
Microsoft has released a ton of slide decks explaining in some detail all their features and what to gain from them, but Sony is a complete mystery in comparison.

From the high-level specs numbers they have (the same as we do), Sony console should come out behind in almost all cases, and the fact that it isn't is leaving them scratching their heads.


The only seemingly "exclusive" PS5 info I've seen a DF member say or write was Dictator/Alex saying something along the lines of "it's to be expected that the Xbox GPU has more advanced features, because the PS5 had to be completed earlier due to its original 2019 release date".
I'm assuming this is something he heard from developers and not just a wild guess, though.
 
Last edited:

MonarchJT

Banned
VGTech is not saying DF is making up results but that the tools are lacking.

That is something Batalhia already confirmed when he said that he had to manual count the framerate in Vanalla due the tools having issues with screen tearing.
vgtech (even if you are not sure yet) has not found such different results ... they also give the xsx as more performing but probably with some frames (according to them duplicate) less. Anyway I didn't like the post written on beyond3d especially if they still don't even have the certainty to prove what they say. They would have done well to notify DF privately and then if certain of what was happening to release the news and DF I'm sure would have clarified.
 
Last edited:

TGO

Hype Train conductor. Works harder than it steams.
After watching several videos on the control benchmark, i must confess that chick is one ugly looking Broad. "lollipop_disappointed:
I don't think she's that bad, the piss filter doesn't help much and the game model don't do the actress justice
But I wouldn't say she ugly
controlultimateeditior4k88.jpg

controlultimateeditiovgjys.jpg

controlultimateeditiogqjjn.jpg
 

Dibils2k

Member
Why would anybody care about a pixel being off or a fps being higher or lower.

Are people just trying to find things to discredit DF or something for the sake of it? because it doesn't favor there box? pathethic.
yep they do it early on everytime, at the peak of interest then later when its proven to be just BS they quiet down until the next H2H and we start the cycle again
 
you are contradicting yourself and jumping to conclusions based on PR of tech instead of reading how it work and how its being compared, you talk about games not designed around XBSX designed to take advantage of its capabilities that is correct but then you talk about hitman 3 as a correct example of using the capabilities of XBSX and PS5 when its a game that works on PS4 and Xbox one also is a PSVR game on PS4, its game that targets last gen and even PSVR, its not a good example of using PS5 and not even XBSX to its capabilities

about SFS, first it doesnt improve RAM, it minimizes the data it requires to get to ram, your ram stays the same and because of that other graphic buffers are not going to work twice as fast, also it doesnt achieve 2.5x more data streaming on top of other partial resident textures solutions, in fact it uses already PRT as part of the process
It boosts effective RAM capacity by significantly cutting down on the amount of texture data that needs to be stored in VRAM at any given moment. If the data requirements for scenes are significantly lowered to get the same visual result, then the effective capacity has improved no matter how you slice it.
 
It boosts effective RAM capacity by significantly cutting down on the amount of texture data that needs to be stored in VRAM at any given moment. If the data requirements for scenes are significantly lowered to get the same visual result, then the effective capacity has improved no matter how you slice it.

the data required for a visual result on a scene is more than just texture buffers(there is also geometry and other buffers for many effects, shadows, light, etc), even if you can update textures very quickly the rest is the same and the updates are not going to be much faster the difference may be for space available(but that is also true if you dont require too much textures) but even that depends how much you need and how much you can use, that is why its wrong to say it improves RAM capacity that is simply not true, saying it improves RAM involves other things, you dont get more speed or more space in RAM you lower your texture requirements which is not bad
 
Last edited:
the data required for a visual result on a scene is more than just texture buffers(there is also geometry and other buffers for many effects, shadows, light, etc), even if you can update textures very quickly the rest is the same and the updates are not going to be much faster the difference may be for space available(but that is also true if you dont require too much textures) but even that depends how much you need and how much you can use, that is why its wrong to say it improves RAM capacity that is simply not true, saying it improves RAM involves other things, you dont get more speed or more space in RAM you lower your texture requirements which is not bad

The amount of texture data you require in physical RAM is significantly cut down due to Sampler Feedback Streaming. It provides the textures you need at the very instant you need them, meaning less need to stage a lot of texture data in RAM. Textures are one of the biggest consumers of VRAM. There's a reason I'm using the word 'effective' RAM capacity because it's a fact that it will improve effective RAM capacity on Series X, just as infinity cache in a very different fashion greatly improves effective memory bandwidth on RDNA2 PC cards.

I'm not saying SFS is Infinity Cache. By using SFS you have a far greater chance of Series X's GPU never needing to dip over into slower standard memory due to the memory savings developers get from using Sampler Feedback Streaming. Sampler Feedback Streaming is going to be a game changer in the games industry and for GPUs in general. It's the next big step or evolution in texture streaming, and Xbox Series X has support for it right now. That's a big damn deal.
 

Neo_game

Member
You all keep doing this not so subtle dance where you whisper in dark corners that the Series X hardware is fundamentally bottlenecked and will never achieve the potential of its spec. You all know full well these types of things are nothing more than early development hiccups that will be completely ironed out with time, right? I'll be fair, though, I'd totally be taking my early shots wherever I could get them too, but this is all building towards a fundamental, inescapable reality.

words point GIF

:messenger_tears_of_joy: (wanted an excuse to use this gif)

There is no innate flaw in the hardware design of Series X that secretly makes PS5 the more powerful console. There is nothing that youtube "engineers" like moores law and redtech have seen that Microsoft and AMD have somehow overlooked or miscalculated. No amount of Cerny magic can change one simple reality: Not only is PS5 the weaker console, it also isn't even close enough as to be considered "more or less equal in performance" to Series X as some hope is the ultimate outcome when all cards from both platforms are fully on the table in each of their most impressive future games. Series X is just flat out stronger. The reason performance between the two consoles - save for a few edge cases - is looking most often like a wash is largely because Series X is playing with one arm tied behind its back.

How so? Well, between games clearly being designed more around PS5 first and then merely ported over to Series X with, for the most part, the exact same configuration and data organization - which often won't properly address the difference in GPU performance across 10GB of GPU Optimal memory which runs at 560GB/s and 3.5GB of standard memory which runs at 336GB/s for Series X's asymmetric memory design - this ends up contributing to an occurrence where Series X's much larger GPU ends up running into many more situations where all those additional shader ALUs - 1024 more to be exact compared to the PS5 - end up being bandwidth starved.

This is how you get performance stutter that isn't present on PS5, or how you get bizarre performance drops when simple text is being shown on screen. Unless people actually want to believe Xbox Series X can't handle large text appearing on screen in Control without big dips in performance when weaker PCs and consoles manage the task just fine?

On the PS5 side it comes down to nothing more than simplicity and ease. PS5 has a more straightforward unified memory setup where all available game memory sees the same 448GB/s. Asymmetric memory is the biggest factor so far in Series X not showing the gap. It's a memory bandwidth thing, and this is why Sampler Feedback Streaming is so important and vital to the Series X design. It increases effective RAM capacity by 2.5x. So that 10GB of GPU optimal memory will seem more like 25GB. Even scaling it back some, and saying only 5GB of the 10GB (since not all 10GB will be for textures) that still works out to 12.5GB of effective memory, still leaving the other 5GB of GPU optimal, thus pushing it to an effective 17.5GB of memory that will all operate at the 560GB/s memory speed. In this entire thing I've left out and haven't even factored in the 2.5GB of slower memory. And before some of you start telling me these numbers/memory savings are fantastical and aren't at all realistic, I refer you all to your kraken/oodle PS5 SSD compression/decompression posts/threads. If we can/should take those numbers seriously, we can take the far less fanciful SFS numbers provided by Microsoft seriously as well.

Some may call the asymmetric memory design a flaw in Series X, but it's far from it. It was a clever decision to get the power they wanted at the price they wanted. It was either that or 20GB of RAM and price yourself out of the game. The other alternative was going with even faster GDDR6 RAM, another move that could price them out of competitiveness. This is something developers have to design for, but it isn't all on developers. Microsoft has to make it easier, more error free. They need to get more developers actually using Sampler Feedback Streaming in their titles. This is literally the reason Microsoft said that Xbox Velocity Architecture is the soul of Xbox Series X. It's how the console will meet its full potential.

Hf1hogb.jpg




So when I see these early game comparisons and all this talk of PS5 "punching above its weight," another way I've come to look at it is that Series X's basic memory setup seems to be getting ignored in cases despite the very real performance drawbacks of doing so, and it's only because of the built in power advantage that it still ends up easily matching the PS5 experience, or only barely losing till patch arrives, or even having the edge over PS5. But we've already seen cases where Series X clearly outdoes PS5. Hitman 3 doesn't happen by accident or pure chance. It still represents the largest gap between both consoles thus far.


I do not think there is any porting from PS5 to the SX. It is from SS, may be SS will be the limitimg factor for SX thoughtout the gen ? Some cross gen games are scaled from last gen consoles. For SS it from from last gen S and for SX it from X. Moreover remdey are PC devs and have more games on Xbox than Sony consoles. So if this game is ported from PC they should not have any trouble. You would think they are quite well versed with DX and Microsoft tools. As a matter of fact, I think their next game Crossfire X is timed or console exclusive to Xbox. As far as memory and BW constraints. I think there is advantage for SX if games are using 10gb or less above that, it is not ideal and probably PS5 has an advantage with unifrom bus width. But we do not know how much RAM these games are using.
 

Md Ray

Member
didn't he call that the corridor of doom for a reason as even on pc its under heavy load?
That's the point. It's the heaviest area in the game and a perfect way to gauge perf for worst case scenario and decide whether to cap the frame-rate or not in order to avoid large inconsistencies.
 
The amount of texture data you require in physical RAM is significantly cut down due to Sampler Feedback Streaming. It provides the textures you need at the very instant you need them, meaning less need to stage a lot of texture data in RAM. Textures are one of the biggest consumers of VRAM. There's a reason I'm using the word 'effective' RAM capacity because it's a fact that it will improve effective RAM capacity on Series X, just as infinity cache in a very different fashion greatly improves effective memory bandwidth on RDNA2 PC cards.

I'm not saying SFS is Infinity Cache. By using SFS you have a far greater chance of Series X's GPU never needing to dip over into slower standard memory due to the memory savings developers get from using Sampler Feedback Streaming. Sampler Feedback Streaming is going to be a game changer in the games industry and for GPUs in general. It's the next big step or evolution in texture streaming, and Xbox Series X has support for it right now. That's a big damn deal.

again its wrong, Ill give you an example, PS2 has 4 MB of eDRAM, it uses typically 2 MB for texture buffer but it can update this buffer many times during frame, let say it updates 10 times per frame so would you say it has 22 MB of VRAM? if your answer was yes you are wrong because if you require more VRAM or more bandwidth for another particular effect or for redrawing(something PS2 uses a lot)then you dont have as much texture avilable as before or as if you had 20 MB more, just because you can update data fast doesnt mean you have more RAM they are two different things


I think you exaggerate with SFS as PRT is its base and do most of its work and its benefits(there are many tiers so I am talking generally), SFS helps with a special filter that ensures there is a minimal mip map available so dont have to search and cause a dip, I think PS5 have the upper hand in this regard
 
again its wrong, Ill give you an example, PS2 has 4 MB of eDRAM, it uses typically 2 MB for texture buffer but it can update this buffer many times during frame, let say it updates 10 times per frame so would you say it has 22 MB of VRAM? if your answer was yes you are wrong because if you require more VRAM or more bandwidth for another particular effect or for redrawing(something PS2 uses a lot)then you dont have as much texture avilable as before or as if you had 20 MB more, just because you can update data fast doesnt mean you have more RAM they are two different things


I think you exaggerate with SFS as PRT is its base and do most of its work and its benefits(there are many tiers so I am talking generally), SFS helps with a special filter that ensures there is a minimal mip map available so dont have to search and cause a dip, I think PS5 have the upper hand in this regard
We had that conversation in regards to the 32MB EDRAM in the xbone back in 2013 😭.
 
I did have a theory that it was something to do with texture steaming. Although I thought the textures were just loading faster due to the PS5s I/O.
What's interesting is that the stuff not loaded caused discrepencies in the performance (it may load faster on the PS5 as well)... That game has this bug on my PC too, sometimes an item on the wall will remain in N64 mode for quite a while, often it loads only after I restart a section many times because I'm no good at it.
 

Fredrik

Member
50% of the refresh rate is 30fps.
So with VSync on when the hardware can’t hold 60fps it will stutter to 30fps.

There is a game on PS4 that suffered this issue because it was using VSync... when the framerate dropped below 60fps it automatically dropped to 30fps due VSync... the framerate come back to 60fps it become 60fps all the time with VSync... it never showed the framerate between 31 and 59... just 60 or 30.
I don’t know what that PS4 game was doing, maybe showing frame times? A functional fps counter shows the average frames per second and uncapped vsynced 59fps is nothing strange at all, just means that one frame was doubled the last second.
 
More proof about my theory that Microsoft only check the pure number and don't check how the single parts run together. Seems to me like they out in a powerful GPU which will never in a game reach its full potential because it is bottle-necked by the CPU.

How about Microsoft make a new slogan: The most powerful console ever built... in photo mode!
 

SlimySnake

Flashless at the Golden Globes
More proof about my theory that Microsoft only check the pure number and don't check how the single parts run together. Seems to me like they out in a powerful GPU which will never in a game reach its full potential because it is bottle-necked by the CPU.

How about Microsoft make a new slogan: The most powerful console ever built... in photo mode!
well, i dont know if its a theory. Phil went on record several times saying he wanted to win the power crown. It was very important to him.

As for whether or not, he ended up creating a bottleneck somewhere by focusing on a large gpu chip, that remains to be seen. Some last gen games are not performing well so far, but some are doing just fine. lets wait and see. I think Battlefield 6 will be the first true test for both machines.
 

muteZX

Banned
in a very simplistic way ..

PS5 has faster rasterization /lets say about +18%/ because of ROPs and co. running at higher speed and there is a much faster SSD IOP.
XSX has faster raytracing /lets say about +18%/ because of higher amount of RT/CU units and partly much faster RAM.

The Mendoza map showed XSX performance "problems" with alpha at native 4K and XSX would have to reduce the resolution to the PS5 level, so the games would run exactly the same.
 

Caio

Member
PS5 has a worse GPU than Series X, this is a GPU test PS5 cannot win in this scenario, just accept it and move on. There is nothing interesting going on lol

Don't be so defensive, nobody is attacking your Box, mate relax. I was just asking. I'm simply waiting for VGTech face-off.
I do not need to accept anything unless it comes from a reliable source, and don't ever tell me what I should accept or not if you cannot even understand what people say in their posts.
 

MonarchJT

Banned
again its wrong, Ill give you an example, PS2 has 4 MB of eDRAM, it uses typically 2 MB for texture buffer but it can update this buffer many times during frame, let say it updates 10 times per frame so would you say it has 22 MB of VRAM? if your answer was yes you are wrong because if you require more VRAM or more bandwidth for another particular effect or for redrawing(something PS2 uses a lot)then you dont have as much texture avilable as before or as if you had 20 MB more, just because you can update data fast doesnt mean you have more RAM they are two different things


I think you exaggerate with SFS as PRT is its base and do most of its work and its benefits(there are many tiers so I am talking generally), SFS helps with a special filter that ensures there is a minimal mip map available so dont have to search and cause a dip, I think PS5 have the upper hand in this regard
i think that the xsx have basically more bw....Xbox architect did bet on 10 gb of fast ram to be enough fr the next gen, to which you can add 3 GB (im excluding the +or- 3gb that both console will use for the OS) ranging at 336gb/s that are not that slow! ...to put it into perspective the ps4 had 176gb/s, the ps4 pro 218 gb/s and the Xbox One x which was meant for 4k have 326 GB /s.
And again is not for nothing the rtx 3070 has less RAM and less bw than the faster ram in the series X , 8gb@448gb/s vs 10gb@560gb/s. If you add to all this the possibility to add 3gb of ram as needed for the frame buffer and to limit the bandwidth consumption through SFS (reduce till 1/3 the data needed for texturesl) , Mesh Shader (reduce the data decreasing the draw calls made by the CPU and allowing the culling of the triangles not visible in the scene) DirectML Super Resolution (hopefully given that they have used the console hw for machine learning that they will release a customized albeit simplified version of the Nvidia dlss). I think it's not bad as a solution
 
Last edited:
Top Bottom