Digital Foundry Article Technical Discussion [2024]

So there are separate discussions happening here wrt Space Marine: Firstly, is there even VRS? Secondly is it hardware or software VRS?

I should be clear, I’m looking for reasons why Series consoles are behaving out of the normal band that we expect. We haven’t confirmed anything yet, except I’m just looking at an older developer interview that indicates that they have full intention of putting these features into their engine.

Technically, you are correct. But I wouldn’t say VRS is a new feature then, since you can back port it as far back as DX9? Whereas they were explicitly talking about adding in the new features which would be hardware VRS.

I really don’t have a hard stance here on the performance aspects, I think we’ve long had a great back and forth discussion in b3d on whether custom software solutions are more performant than the equivalent hardware ones (and with good backing info that they are). Performance isn’t at debate for me, given what has been said. However, there isn’t a strong reason for series consoles to outperform their line of best fit here given the history of releases. More bandwidth, and more compute wouldn’t make up a 20% frame rate difference. So I’m looking for evidence of something else. PS5 and PC appear to be in line, so I’m looking at Xbox being the outlier here. But before we start talking performance of such features, we need to verify if it’s even present in Space Marine.

Thst means looking through PC and XSX looking for those shading differences. Seeing if there are options to remove software VRS and so forth.

This is intresting, i was thinking about this earlier. We havent really seen any xbox exclusive games that runs better on XSX then PC counterparts did we? Alex mentioned in one the DF vids that often PS5 is the main platform to develop on and PC XSX comes after. So this deviation is a interesting riddle. Did series version of the game got more love for some reason?
 
I suggested a few pages back that it could be the cut-down FPU units on the PS5 CPU causing the difference.
That could be the cause of this actually. If that game relies on AVX256 jobs it could significantly (meaning more than 5% or so) reduce GPU (and CPU ) clocks on PS5.

It's interesting we search for some answers here but why didn't we put the same curiosity in why PS5 outperform XSX for about the same amount in some others games (often in games using hardware RT)? For instance here PS5 outperforms XSX often by 25% in the fidelity RT mode of Hogwart Legacy. With XSX having 25% more bandwidth and 17% more FLOPS how is that even possible in the fidelity RT mode (high resolution, RT and low framerate mode)?
 
That could be the cause of this actually. If that game relies on AVX256 jobs it could significantly (meaning more than 5% or so) reduce GPU (and CPU ) clocks on PS5.

It's interesting we search for some answers here but why didn't we put the same curiosity in why PS5 outperform XSX for about the same amount in some others games (often in games using hardware RT)? For instance here PS5 outperforms XSX often by 25% in the fidelity RT mode of Hogwart Legacy. With XSX having 25% more bandwidth and 17% more FLOPS how is that even possible in the fidelity RT mode (high resolution, RT and low framerate mode)?
rt is cpu heavy and usualy ps5 has slight advantage vs xsx in cpu limited situation (lower api), space marine 2 is new exception
 
More info about this please. What exactly different?
It was a discussion point here before the consoles were released. @Locuza and a few others both here and on Twitter noticed from the die shot of the PS5 CPU that the FPU lookedhttps://chipsandcheese.com/2024/03/20/the-nerfed-fpu-in-ps5s-zen-2-cores/ much smaller than on the Series Consoles and Ryzen CPUs. The linked article for Chips and Cheese took a look at it and concluded that it wouldn't amount to much. In any case, I only mention it because of how CPU heavy this game is, and the considerable difference between the Seres X and the PS5.
 
rt is cpu heavy and usualy ps5 has slight advantage vs xsx in cpu limited situation (lower api), space marine 2 is new exception

PS5 has enjoyed an unexpected advantage so far this generation, something that no one expected when the specs were first released.

Wasn't it always somewhat expected that at some point Series-X would start to pull a head as the generation went on?
 
PS5 has enjoyed an unexpected advantage so far this generation, something that no one expected when the specs were first released.

Wasn't it always somewhat expected that at some point Series-X would start to pull a head as the generation went on?
in gpu limited situation should have ~20% advantage, big difference in cpu limited situation is rather surprising
 
in gpu limited situation should have ~20% advantage, big difference in cpu limited situation is rather surprising

Depending on how much the power use is affecting PS5, it might not be so surprising.

The high load could be causing PS5 to lose both CPU and GPU clocks, how's the game or algorithm proritising clocks on PS5 in these high load conditions?

So is it even the CPU? Is it GPU compute? Is it both?

It might be a mixture of multiple things adding up to a large difference.
 
rt is cpu heavy and usualy ps5 has slight advantage vs xsx in cpu limited situation (lower api), space marine 2 is new exception
Slight, yes, but we are talking 25% here. Shouldn't journalists interview developers to get to the bottom of this?

In compute limited games without HW RT (like those UE5 games) XSX do often performs 15-20% faster than PS5, as expected. Why would PS5 performs 25% faster in CPU limited scenarii with 10% less CPU available (as XBX CPU is clocked higher with more cores available)? And PS5 CPU clocks were supposed to be dynamic and often go lower than 3.5ghz too.

Well with this cpu for sure not in all titles as Space Marines 2 showed.
Well we may already have likely found out the reason for this. If the game heavily relies on AVX jobs, notably AVX256 then PS5 CPU could either throttle or at the very least perform worse than XSX's CPU. But that's really the only game we know off right now that performs worse on PS5 due to CPU bottlneck. For instance Baldur's Gate 3 performs better on PS5 and nobody were really curious about why.
 
Slight, yes, but we are talking 25% here. Shouldn't journalists interview developers to get to the bottom of this?

In compute limited games without HW RT (like those UE5 games) XSX do often performs 15-20% faster than PS5, as expected. Why would PS5 performs 25% faster in CPU limited scenarii with 10% less CPU available (as XBX CPU is clocked higher with more cores available)? And PS5 CPU clocks were supposed to be dynamic and often go lower than 3.5ghz too.


Well we may already have likely found out the reason for this. If the game heavily relies on AVX jobs, notably AVX256 then PS5 CPU could either throttle or at the very least perform worse than XSX's CPU. But that's really the only game we know off right now that performs worse on PS5 due to CPU bottlneck. For instance Baldur's Gate 3 performs better on PS5 and nobody were really curious about why.
but it doesn't explain why it would perform identically to a comparable PC CPU. They basically had the same graph setup as per DF.

Many games are optimized for PS5, which is why you're seeing PS5 well saturated and performing well. This isn't the case so much for XSX, what you're seeing is enough optimization to get it to PS5 level and calling it a day.
 
but it doesn't explain why it would perform identically to a comparable PC CPU. They basically had the same graph setup as per DF.

Many games are optimized for PS5, which is why you're seeing PS5 well saturated and performing well. This isn't the case so much for XSX, what you're seeing is enough optimization to get it to PS5 level and calling it a day.
Why are you throwing PC here. We are comparing PS5 and XSX in either CPU or GPU limited scenes. PC adds too much complexity to the problem.

This isn't the case so much for XSX, what you're seeing is enough optimization to get it to PS5 level and calling it a day.
So the usual explanation, right? PS5 performs better because it's more optimized and XSX performs better because it has the better hardware? BS.
 
Why are you throwing PC here. We are comparing PS5 and XSX in either CPU or GPU limited scenes. PC adds too much complexity to the problem.


So the usual explanation, right? PS5 performs better because it's more optimized and XSX performs better because it has the better hardware? BS.
Throwing a comparable PC in there because they managed to get a PC config run parallel to PS5 on the DF video.

XSX should perform better if it’s been optimized to perform its maximum on XSX. I don’t think that should be too controversial to say. They are the same hardware running very different APIs. How they solve the same problems are very different, and code that is efficient for PS5 may not be efficient for XSX. You can follow along the developer docs on how different the 2 solve the same problems.

Even with Starfield we saw how much better AMD performed at launch until nvidia cards were patched. They use the same API! Software is one of the largest reasons why performance is the way it is. Anything poorly coded can bring any hardware to its knees. Only once you’ve freed up as many bottlenecks as possible is hardware allowed to breathe.
 
Last edited:
I am actually expecting Space Marine 2 to have its framerate noticeably improved on PS5 Pro using boost mode (or at least quite more so than the small CPU boost). If I am right I'll explain my theory later.
 
Different Ryzen gens tested in this video here.


Near as I can tell, the game isn't just CPU limited but also very bandwidth limited. Switching between CCDs on 7950x3D often shows 30-40% difference while Vtune reports bandwidth bottlenecking my older 9900k. This might also factor into Xbox outperforming PS5 in this game.


hFmSZKI.png
 
Different Ryzen gens tested in this video here.


Near as I can tell, the game isn't just CPU limited but also very bandwidth limited. Switching between CCDs on 7950x3D often shows 30-40% difference while Vtune reports bandwidth bottlenecking my older 9900k. This might also factor into Xbox outperforming PS5 in this game.


hFmSZKI.png
I’ll be damned.
That means I’ll have to show the red carpet to @davis.anthony
 
It's finally happened. PS5 Pro is real, announced and detailed by none other than lead system architect, Mark Cerny. John, Rich and Alex sit down to discuss the reveal, how Sony chose to market it, the Big Three tech innovations - larger GPU, ML upscaling and RT enhancements - and of course, the shocking price-point.

0:00:00 Overview
0:01:20 How is Sony marketing the PS5 Pro?
0:09:43 Big three features: larger GPU, machine learning and ray tracing enhancements
0:25:10 The missing disc drive
0:33:12 $700 USD sticker shock: does it make sense?
0:44:23 What do we want to see out of PS5 Pro?
0:49:43 Supporter Q1: Was Microsoft right to skip an enhanced console this generation?
0:54:06 Supporter Q2: Is GT7’s 8K mode as stupid as it sounds?
0:57:26 Supporter Q3: What PC GPU is the best match for PS5 Pro?
1:00:27 Supporter Q4: Will Sony encourage developers to stick to a single PS5 Pro mode, or have multiple?
1:02:58 Supporter Q5: Why doesn’t the PS5 Pro have a larger CPU upgrade?
 
Back
Top