Digital Foundry: Control PS5 vs Xbox Series X Ray Tracing 'Benchmark' - Unlocked FPS In Photo Mode!

[plagiarize]

Then again, the PS5 version of the game is about 45% smaller in size than the XSX version.

I do agree on the loading thing though, I don't understand what is happening there. There simply is absolutely no reason that the PS5 and the XSX should load a game in the exact same amount of time. Like NONE. Absolutely none. There I no technical basis for it to even remotely occur. Even if everything that had to the with IO processing on the PS5 was 30% slower than the XSX, from just raw IO transfer throughput the PS5 should still end up loading faster.

Click to shrink...

It's not that simple. You've got *more* compressed files on the PS5, possibly using different compression methods. Unless we know how much of the loading time is data throughput and how much is decompression we can't draw a conclusion like that. The PS5 version is twice as compressed. It could be spending a lot more of its load time decompressing than those files than the Series X, and it could just so happen to work out at about the same exact total time.

Whatever the case, both load so much faster than previous gen versions, that in practice both load plenty quickly for my liking. Maybe they could have further sped the PS5 version up with more work, but they figured it was already more than fast enough. Maybe they put more priority on getting the file sizes as low as possible while matching Series X load times.

And if so, that's fine by me too, cause storage expansion options are a bigger problem on PS5 at this exact moment of time (though hopefully not for long).

 

[digitalrelic]

TBH i expect it to be the opposite, a bigger team will be able to squeeze out a lot more out of more powerful hardware. That is why it would be nice to know why the XSX is doing 30% better here. Like somebody said above, the game is probably CPU limited.

Click to shrink...

It's averaging 16% better here. Which is literally exactly right in line with what the raw hardware between the XSX and PS5 tells us.

 

[Pheonix]

I have a hard time believing it's due to the SSD, but rather maybe better/more mature compression methods, but the huge difference in file sizes fascinates me. I'd love to know if other games have a difference in size.

Click to shrink...

He does kinda have a point. It's at least plausible, though I fear it's also invalid.

I could see how the PS5 has more compressed data that takes more time to decompress and ends up with you having similar load times with the XSX but benefitting from smaller overall file size. But then you look at games where they both have identical file sizes but still have mostly similar load times. And those here the PS5 only has like a 10-20% loading advantage. Which is like saying one loads in 8 seconds and the other in 10 seconds.

If they have so much overhead, a 60fps mode with better effects without RT would be great!

Click to shrink...

Not for the devs. We are seeing framerates dancing from 32 - 55% during photomode, be rest assured that during gameplay it would generally dip even lower than that and be more sporadic based on whats happening on screen. For what its worth, the 60fps mode probably averages above 85fps internally with drops to as low as 55fps lol. Devs love

If the loading times are bound by the CPU, and the CPUs are near identical, it could be the reason you're looking for.

Click to shrink...

But how though?

That's why i said even if everything that deals with IO processing is worse (you can add the same) on the PS5.

Everything, in this case, would include, data calls, decompression, and even mapping data. For them to still be identical would mean that the actual time spent moving data is minuscule to non-existent. And thats impossible.

Forgive me for getting slightly technical, let's say the PS5 moves data at 5GB/s. That's 5GB every 1000ms. Vs 2.4GB every 1000ms. Now let's throw in some bottlenecks and say for whatever reason the PS5 loses 300ms and the XSX loses only 100ms. The PS5 should still come out faster. The only way it doesn't is if devs go in there and enforce a hard restriction on the PS5. Basically, limit its throughput speed to match the XSX.

 

[digitalrelic]

Here we go.....

Sigh. How many more games we need to have to see that both consoles are practically the same in all metrics that matter? Which is good. End the war, warriors.

Click to shrink...

Agreed. Both consoles are very similar, with just enough difference in their hardware approach to make first-party exclusives really interesting. Third party games? It'll always be really minimal differences that 95% of consumers won't notice or care about.

 

[Pheonix]

It's not that simple. You've got *more* compressed files on the PS5, possibly using different compression methods. Unless we know how much of the loading time is data throughput and how much is decompression we can't draw a conclusion like that. The PS5 version is twice as compressed. It could be spending a lot more of its load time decompressing than those files than the Series X, and it could just so happen to work out at about the same exact total time.

Whatever the case, both load so much faster than previous gen versions, that in practice both load plenty quickly for my liking. Maybe they could have further sped the PS5 version up with more work, but they figured it was already more than fast enough. Maybe they put more priority on getting the file sizes as low as possible while matching Series X load times.

And if so, that's fine by me too, cause storage expansion options are a bigger problem on PS5 at this exact moment of time (though hopefully not for long).

Click to shrink...

Oh, I agree...I alluded to this in my previous post which I got distracted away from posting lol.

But I also pointed out why this also seems strange to me too. This is the first game we are seeing where the PS5 is almost half the size as the XSX version. If being about 40% smaller equated to the same loading speed, shouldn't that mean when it's identical in size it should at least load about 40% faster?

 

[VanWinkle]

It's not that simple. You've got *more* compressed files on the PS5, possibly using different compression methods. Unless we know how much of the loading time is data throughput and how much is decompression we can't draw a conclusion like that. The PS5 version is twice as compressed. It could be spending a lot more of its load time decompressing than those files than the Series X, and it could just so happen to work out at about the same exact total time.

Whatever the case, both load so much faster than previous gen versions, that in practice both load plenty quickly for my liking. Maybe they could have further sped the PS5 version up with more work, but they figured it was already more than fast enough. Maybe they put more priority on getting the file sizes as low as possible while matching Series X load times.

And if so, that's fine by me too, cause storage expansion options are a bigger problem on PS5 at this exact moment of time (though hopefully not for long).

Click to shrink...

That's a very fair point. Doesn't explain why so many other multiplatform games load about the same time between the two systems, though.

If given the choice, I would be extremely happy with PS5 games being half the size and the same loading times as Series X (as it is in Control), rather than the same file size and 2x faster loading.

 

[arsene_P5]

while static and not having action where the cpu would also play a bigger role. cpu is miniscule difference. this explains why xsx was behind in various comparisons. faster clocks better tools make the difference when consoles are close.....

Click to shrink...

Faster clocks isn't making the difference (in every scenario). There are GPU tasks which are more reliant on more CU's.

Cool, I like to play games in photo mode! Seriously, why?

Click to shrink...

Because it's the closest comparison yet to benchmark the GPU we got. Most of the other games are locked 30/60/120 and then you don't see the overhead. Or the games aren't representing like for like scenario's, which is impossible to do during gameplay anyways. So Photomode it is.

Wasn't AMD architecture lacking actual cores for this?

Click to shrink...

Yes, there are no cores on AMD hardware (RDNA2 or previous architecture)

There simply is absolutely no reason that the PS5 and the XSX should load a game in the exact same amount of time. Like NONE. Absolutely none.

Click to shrink...

Not for this game, but there are reasons and one of them is networking. Bungie already talked about the potential bottleneck. Want a example?

• Game XY loads data of the SSD in 2 seconds on PS5
• Game XY loads data of the SSD in 4 seconds on Xbox
• Game XY needs 5-6 seconds for network and stuff.

How long will it load on those consoles? You be the judge!

 

[Laver]

But how though?

That's why i said even if everything that deals with IO processing is worse (you can add the same) on the PS5.

Everything, in this case, would include, data calls, decompression, and even mapping data. For them to still be identical would mean that the actual time spent moving data is minuscule to non-existent. And thats impossible.

Forgive me for getting slightly technical, let's say the PS5 moves data at 5GB/s. That's 5GB every 1000ms. Vs 2.4GB every 1000ms. Now let's throw in some bottlenecks and say for whatever reason the PS5 loses 300ms and the XSX loses only 100ms. The PS5 should still come out faster. The only way it doesn't is if devs go in there and enforce a hard restriction on the PS5. Basically, limit its throughput speed to match the XSX.

Click to shrink...

I don't know how the loading process actually works, but if the game needs to load eg. 7GB of assets to render the first frame, but the first 500MB are the most crucial data that CPU will use to set up the engine and all the parameters, then the CPU will get that data in 0.2 seconds on XSX and 0.1 seconds on PS5, then will take 8 seconds to initialize the engine, loading the less important data in the meantime, resulting in the 8.1 second loading time on PS5 and 8.2 second loading time on XSX, ie. virtually identical.

It's possible I'm entirely wrong here, but such hypothetical scenario would explain what we are observing.

 

[BubblegumNemesis]

It's averaging 16% better here. Which is literally exactly right in line with what the raw hardware between the XSX and PS5 tells us.

Click to shrink...

It doesn't really work like that though. Not about averages.

 

[digitalrelic]

It doesn't really work like that though. Not about averages.

Click to shrink...

Using the average may not be the perfect metric to go by, but using peak numbers (30%) absolutely isn't the way to go.

 

[gothi]

It's not that simple. You've got *more* compressed files on the PS5, possibly using different compression methods. Unless we know how much of the loading time is data throughput and how much is decompression we can't draw a conclusion like that. The PS5 version is twice as compressed. It could be spending a lot more of its load time decompressing than those files than the Series X, and it could just so happen to work out at about the same exact total time.

Whatever the case, both load so much faster than previous gen versions, that in practice both load plenty quickly for my liking. Maybe they could have further sped the PS5 version up with more work, but they figured it was already more than fast enough. Maybe they put more priority on getting the file sizes as low as possible while matching Series X load times.

And if so, that's fine by me too, cause storage expansion options are a bigger problem on PS5 at this exact moment of time (though hopefully not for long).

Click to shrink...

He does kinda have a point. It's at least plausible, though I fear it's also invalid.

I could see how the PS5 has more compressed data that takes more time to decompress and ends up with you having similar load times with the XSX but benefitting from smaller overall file size. But then you look at games where they both have identical file sizes but still have mostly similar load times. And those here the PS5 only has like a 10-20% loading advantage. Which is like saying one loads in 8 seconds and the other in 10 seconds.


Not for the devs. We are seeing framerates dancing from 32 - 55% during photomode, be rest assured that during gameplay it would generally dip even lower than that and be more sporadic based on whats happening on screen. For what its worth, the 60fps mode probably averages above 85fps internally with drops to as low as 55fps lol. Devs love

But how though?

That's why i said even if everything that deals with IO processing is worse (you can add the same) on the PS5.

Everything, in this case, would include, data calls, decompression, and even mapping data. For them to still be identical would mean that the actual time spent moving data is minuscule to non-existent. And thats impossible.

Forgive me for getting slightly technical, let's say the PS5 moves data at 5GB/s. That's 5GB every 1000ms. Vs 2.4GB every 1000ms. Now let's throw in some bottlenecks and say for whatever reason the PS5 loses 300ms and the XSX loses only 100ms. The PS5 should still come out faster. The only way it doesn't is if devs go in there and enforce a hard restriction on the PS5. Basically, limit its throughput speed to match the XSX.

Click to shrink...

The PS5 has hardware accelerated decompression using Kraken and Oodle plays into that to the point the tools developers say it makes the retrieval of data on the PS5 even faster than the raw SSD speed indicates. The game being smaller on PS5 doesn't mean loading will be slower, if we are to believe Sony and Rad Game Tools it should make it even faster!

But loading isn't about raw SSD speed or compression blocks alone. The data has to go somewhere, memory speeds are orders of magnitude faster than the SSD. Both the PS5 and Xbox have fast memory, Xbox happens to be faster. Does that explain why the Xbox and PS5 appear to have similar I/O speeds in games? Probably not.

There's a great quote from one of the Rad Game Tools guys (that I can't find right now) basically stating that devs aren't going to be drag racing SSD's, they're about to find out how inefficient all their file I/O code and built-in helpers are.

 

[Pheonix]

Not for this game, but there are reasons and one of them is networking. Bungie already talked about the potential bottleneck. Want a example?

• Game XY loads data of the SSD in 2 seconds on PS5
• Game XY loads data of the SSD in 4 seconds on Xbox
• Game XY needs 5-6 seconds for network and stuff.

How long will it load on those consoles? You be the judge!

Click to shrink...

Fair enough, but that only explains a scenario that we are looking at a network tied game. What about games like Assasins Creed? Or hitman? Or dirt 5?

I don't know how the loading process actually works, but if the game needs to load eg. 7GB of assets to render the first frame, but the first 500MB are the most crucial data that CPU will use to set up the engine and all the parameters, then the CPU will get that data in 0.2 seconds on XSX and 0.1 seconds on PS5, then will take 8 seconds to initialize the engine, loading the less important data in the meantime, resulting in the 8.1 second loading time on PS5 and 8.2 second loading time on XSX, ie. virtually identical.

It's possible I'm entirely wrong here, but such hypothetical scenario would explain what we are observing.

Click to shrink...

You aren't wrong at all, and that's mostly how it works. Mostly. Game critical data is loaded into RAM, and for the most part, it always stays in RAM. That's what the engine needs to set up its runtime environment on the platform. From that point onwards, its only pulling in relevant to what is happening (or about to happen). Think along the lines of fast traveling to anew place or loading a levelfor the first time and dying in the same level and restarting. In one case, an argument can be made that all the relevant data to initialize the level and load its assets needs to go into RAM, in the other? most of that data should already be in RAM.

It doesn't really work like that though. Not about averages.

Click to shrink...

But that's exactly how benchmarking works. It's literally about averages. Finding and focusing on the average performant value of any given hardware is 1000 times more meaningful data than isolating best-case or worst-case scenarios.

If I drop in rez to 1080p 1% o the time but average a resolution of 1800p 90% of the time and you NEVER drop to 1080p but average a resolution of 1440p 90% of the time... My image would always appear to be better than yours. Even though you may have higher bounds and never reach my lows. That is why averages are important.

 

[AndrewDean84]

This kind of test is why I subscribe to DF on YouTube.

 

[BubblegumNemesis]

Using the average may not be the perfect metric to go by, but using peak numbers (30%) absolutely isn't the way to go.

Click to shrink...

No peak numbers are also not the way to go, I was just wondering why they are so high given the hardware we know about.

 

[Raide]

Well, that's interesting. I wish more titles had unlocked options for framerate, since it's a good way to really push these new systems.

 

[Terbinator]

Does seem kind of crazy to not have a limit remover included even if it is hid from the end user.

If the system is setup for VRR then let that baby run. You don't even need to have it ad an option.

 

[Pheonix]

This kind of test is why I subscribe to DF on YouTube.

Click to shrink...

Yup, they are awesome.

Though this was kinda gifted to them by Remedy lol. Not only did Remedy use identical settings and resolutions for both, they even unlocked the rez in photo mode. Almost as if they were making an unofficial benchmarking tool for the consoles lol.

 

[arsene_P5]

No peak numbers are also not the way to go, I was just wondering why they are so high given the hardware we know about.

Click to shrink...

Let's say the average of the hardware is 16% for this argument, then you still get a smaller difference, if for example the CPU is the limitation or the demanding tasks in the scene benefit more from GPU frequency. You'll get a bigger diffence, if the tasks benefit from a wider GPU (more CU) and/or more RAM bandwith. Point is results will vary, depending on the demands of a particular scene and each consoles has their own benefits.

 

[Gero]

MS did a great job with the series consoles.

 

[Prine]

This video shows the consoles have the overhead to do a 1080p/60 FPS RTS mode if Remedy puts it in with optimizations.





I don't think this is true in this case.

Going by the video, the XSX has an advantage of 3% to 11% to 20% depending on a scene.

Click to shrink...

God damn! Beastly.

 

[Arklite]

Always offer an unlocked framerate option, the benefits in controller response outweigh the frame judder, IMO.
A bit moot with Control having a locked 60fps mode but obviously a welcome benefit to RT mode.

 

[jett]

Interesting stuff. Seems the framerate difference lines up almost exactly to the teraflop difference.

 

[HonestAbe]

DirectML isn't an AI upscaler. it's the foundation of which one can be built on

Click to shrink...

yes, that's why i said software/API but it was still using NVIDIA hardware.

I think MS will be coming up with some ML solution/DLSS "close enough equivalent" though.

 

[Pheonix]

God damn! Beastly.

Click to shrink...

Sigh... knew that post would start warring...

We could also say "God Damn! The PS5 is $100 cheaper!", or start looking for games where the PS5 actually performed better.

Point is, this video, and data like this, isn't meant to be used the way some of you are trying to use it.

 

[Flappy Pannus]

Cool, I like to play games in photo mode! Seriously, why?

Click to shrink...

DF should probably make a video about this, explaining it in more detail or something

 

[Edge]

I talked about the open FPS in the photo mode before here a few days ago and even thought, since it's the photo mode, that the graphical quality is even a bit higher in it as in gameplay.
When you turn the camera to a certain position where the head of her is very close and then you activate photo mode, you notice something changes and makes it look better.

So realistically outside of that photo mode with open fps, it should even run a bit better as in this benchmark, no?



By the way, someone like Digital Foundry who is in talk with devs, messages them and is responsible for some fixed quality issues in the past, should push more for the change of having games future proofed.
I know they did that before but recently there's not one single word about it anymore.
We simply need more simple options in games like this, like keep the quality mode, keep the performance mode, but also give us a simple fps toggle. That single thing would make this game future proof. Who knows if they will ever touch it again once this release is out of the release window and a simple open fps option would almost make sure that the ray tracing graphics mode would be locked 60 on a future PS5 Pro and Series XX or Z, or whatever this could be called.
This should be more pushed, there's no reason to not have this in every single game.
And if they don't want to confuse casual folks, fuck it, then hide it and let us geeks unlock it with the Konami code or some shit.

 

[Jiraiya]

Sigh... knew that post would start warring...

We could also say "God Damn! The PS5 is $100 cheaper!", or start looking for games where the PS5 actually performed better.

Point is, this video, and data like this, isn't meant to be used the way some of you are trying to use it.

Click to shrink...

Calling the series x beastly isn't warring....it's a powerful piece of hardware.

 

[isahn]

interesting results, I wonder how a more CPU intensive scenario could affect the performance profile of both consoles.

 

[Pancracio17]

Fair enough, but that only explains a scenario that we are looking at a network tied game. What about games like Assasins Creed? Or hitman? Or dirt 5?

Click to shrink...

At least Hitman is an always online game.

 

[DuvalDevil]

I'm kinda surprised how good the Series X runs in that video. Will be interesting what developers - especially 1st party - can squeeze out of that sexy box in the future.

 

[digitalrelic]

Sigh... knew that post would start warring...

We could also say "God Damn! The PS5 is $100 cheaper!", or start looking for games where the PS5 actually performed better.

Point is, this video, and data like this, isn't meant to be used the way some of you are trying to use it.

Click to shrink...

How is saying a console is beastly "warring"? It's only warring to those with a bias, looking for an argument.

 

[Jaypah]

Sigh... knew that post would start warring...

We could also say "God Damn! The PS5 is $100 cheaper!", or start looking for games where the PS5 actually performed better.

Point is, this video, and data like this, isn't meant to be used the way some of you are trying to use it.

Click to shrink...

To be fair the PS5 beating the XSX in various games has already been brought up. With more words than "beastly" too. I didn't find either case to be warring but 🤷🏿‍♂️

 

[VanWinkle]

Dictator do you have any theories why the difference in fps between the two consoles is smallest on the heaviest part of the test (the "corridor of doom")?

 

[Yautja_Warrior]

Makes sense, the Series X has a more powerful GPU.

Good old "Corridor of Doom" bringing both down 32-33fps, that corridor is a beast lol

 

[Pheonix]

Calling the series x beastly isn't warring....it's a powerful piece of hardware.

Click to shrink...

How is saying a console is beastly "warring"? It's only warring to those with a bias, looking for an argument.

Click to shrink...

To be fair the PS5 beating the XSX in various games has already been brought up. With more words than "beastly" too. I didn't find either case to be warring but 🤷🏿‍♂️

Click to shrink...

Let's not pretend like we don't know how these things go.

The issue isn't saying one is beastly or not. That true, hell I think for the money they both are.

It's handpicking screenshots like what was used in that post to make an argument for how there is some sort of great discrepancy between them. The same thing was done in some DF threads where the PS5 had some sort of advantage. Things like that does nothing but take away from what the whole point of the thread is really supposed to be about.

 

[BubblegumNemesis]

interesting results, I wonder how a more CPU intensive scenario could affect the performance profile of both consoles.

Click to shrink...

Well since both of them have almost the same CPU, if they hit a CPU bootleneck, the GPU performance won't matter that much.

So it looks like the pure GPU gap between XSX and PS5 is quite a bit bigger than 16%.

 

[Jiraiya]

Let's not pretend like we don't know how these things go.

The issue isn't saying one is beastly or not. That true, hell I think for the money they both are.

It's handpicking screenshots like what was used in that post to make an argument for how there is some sort of great discrepancy between them. The same thing was done in some DF threads where the PS5 had some sort of advantage. Things like that does nothing but take away from what the whole point of the thread is really supposed to be about.

Click to shrink...

This entire thread is about fps during photomode...seems appropriate.

I follow DF allot...and i don't remember another game having this particular benefit.

 

[arsene_P5]

Makes sense, the Series X has a more powerful GPU.

Good old "Corridor of Doom" bringing both down 32-33fps, that corridor is a beast lol

Click to shrink...

PS5 Pro and Series X2 - "but can they run Crysis... Ehm corridor!!!" :P

 

[Jaypah]

Let's not pretend like we don't know how these things go.

The issue isn't saying one is beastly or not. That true, hell I think for the money they both are.

It's handpicking screenshots like what was used in that post to make an argument for how there is some sort of great discrepancy between them. The same thing was done in some DF threads where the PS5 had some sort of advantage. Things like that does nothing but take away from what the whole point of the thread is really supposed to be about.

Click to shrink...

The post they quoted was more than a screenshot but, aight bruh. This is literally a waste of energy. Do your thing 👌🏿

 

[Yautja_Warrior]

PS5 Pro and Series X2 - "but can they run Crysis... Ehm corridor!!!" :P

Click to shrink...

Corridor is definitely the new Crysis lol

 

[Sho_Nuff82]

Which should never be the priority of any developer imo..VRR is a crutch for inconsistent performance. If a Dev is doing their job well it should not need VRR at all.

Click to shrink...

Why should a game be locked at 30 or 60 if there is overhead for more?

PC gamers aren't limited by console television refresh rates.

 

[Jaypah]

I've heard horrible things about this corridor, both from DF and folks on ERA. I haven't gotten there yet in my PC playthrough but considering the Windows Store version that's on Gamepass has a bug that doesn't enable DLSS I'm sure it won't be pretty.

 

[Gestault]

Why should a game be locked at 30 or 60 if there is overhead for more?

PC gamers aren't limited by console television refresh rates.

Click to shrink...

PC gamers who don't have a VRR setup between their GPU and monitor have the same limitation. It's why vsync settings exist in the first place.

 

[Scently]

I've heard horrible things about this corridor, both from DF and folks on ERA. I haven't gotten there yet in my PC playthrough but considering the Windows Store version that's on Gamepass has a bug that doesn't enable DLSS I'm sure it won't be pretty.

Click to shrink...

I think the PC GP version of the game now supports DLSS. Think Remedy tweeted about it recently.

 

[Gestault]

I think the PC GP version of the game now supports DLSS. Think Remedy tweeted about it recently.

Click to shrink...

Yeah, unless that bug is new-new, I think the DLSS issue was resolved in the Win10 version.

 

[Jaypah]

I think the PC GP version of the game now supports DLSS. Think Remedy tweeted about it recently.

Click to shrink...

Sweet, I haven't played in a few days. Thanks for the update!

 

[tapedeck]

My guess is as the Xbox GDK matures the performance difference most expected before these systems launched will start to become the norm..maybe 15-20% better on Series X for most 3rd party stuff, which of course is nothing Earth shattering..but an advantage nonetheless.

As others have said Im more surprised the loading times are so close, maybe down the line Sony 1st party stuff will load significantly faster than MS' 1st party..obv too early to tell. Either way Series X SSD is certainly no slouch.

 

[rokkerkory]

1080p60 with RT would be amazing next patch / upgrade!

 

[Deleted member 46804]

Sigh... knew that post would start warring...

We could also say "God Damn! The PS5 is $100 cheaper!", or start looking for games where the PS5 actually performed better.

Point is, this video, and data like this, isn't meant to be used the way some of you are trying to use it.

Click to shrink...

The point of the DF video is to put these consoles head to head in a like for like performance scenario. This is literally benchmarking the new consoles.

 

[KAMI-SAMA]

Wait this is for photomode? Lol, do people really put that much interest in photomode?