• 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 Weekly: If Xbox Series X is more powerful, why do some PS5 games run better? We finally have some answers.

Thick Thighs Save Lives

NeoGAF's Physical Games Advocate Extraordinaire
AdobeStock_558020453_Editorial_Use_Only.jpeg

On paper, the Xbox Series X is a considerably more powerful piece of hardware than PlayStation 5, so why are we looking at a console generation where the Sony and Microsoft machines often deliver like-for-like results? It's a puzzle that has confounded many over the last few years and after discussions with multiple sources, including some of the highest profile triple-A developers in the business, we finally have some answers - and it's all laid out in this week's edition of DF Direct Weekly.

In terms of technical specifications, Xbox Series X is undoubtedly the most powerful console on the market - at least until the mooted PlayStation 5 Pro arrives. Its GPU hits 12 teraflops, augmented by a handsome amount of memory bandwidth: a max of 560GB/s. PlayStation 5? 10.23TF of GPU compute is available, fed by 448GB/s of bandwidth. It's a clear divide then. If we were to see that spec comparison in the PC space, you'd expect to see the former handily outperform the latter.
Except that hasn't happened this generation. By and large, it's the closest console generation we've ever seen and in the bulk of multi-platform releases, the differences are fairly minimal. One machine may outperform the other, or vice versa. Sometimes where there are differences, it comes down to variances in dynamic resolution scaling - which often looks invisible in A to B comparisons. It's a far cry from the Xbox One X vs PS4 Pro face-off, where the Microsoft machine commanded an obvious advantage - or the PS4 vs Xbox One comparison, where Sony typically held a similarly noticeable lead.

So, after a few years of conversations with developers, what's the explanation? How can a less capable machine outperform the more powerful one? As seen recently in our Elden Ring: Shadow of the Erdtree coverage, the frame-rate difference in favour of PS5 is surprising, bearing in mind it's apparently the less capable machine.
We've heard some interesting theories over the years - such as how Xbox Series X's split-memory set-up (560GB/s of bandwidth on 'fast' memory, 336GB/s on 'slow' memory) may be impacting performance, but the number one reason we've heard from developers concerns the nature of Sony's development environment. More than one key triple-A developer tells us that the PlayStation GPU compiler is significantly more efficient than the Microsoft alternative, meaning that there's better utilisation of the graphics hardware. In general, we understand that the lower level API access afforded to PlayStation development means game makers get more from the hardware.

The second most common explanation we've received from developers concerns the nature of the GPU itself. Mark Cerny himself discussed this way back in March 2020 when he revealed the technical specifications of the PlayStation 5. While the console may have fewer compute units than Xbox Series X - 36 vs 52 - the entire GPU runs faster, meaning that certain tasks will complete faster, better suiting certain game engine designs. The extent of this advantage was always an unknown as PS5 operates with a boost clock on both CPU and GPU - maximum clock speeds will be limited by a power limit on the processor.
However, we've yet to hear any complaints from developers about the boost clock unduly impacting GPU performance. And in terms of raw numbers, there's an interesting titbit from the recently leaked PlayStation 5 Pro developer documentation: engaging the CPU boost option, which increases clock speeds by ten percent, only sees GPU performance drop by one percent - effectively negligible then.

So, based on our conversations, the combination of a more efficient GPU compiler, lower-level APIs and higher clock speeds allows PlayStation 5 to match or even exceed the outputs of Xbox Series X in some scenarios. That said, of course there is value in the approach Microsoft has taken: by standardising on DirectX 12 and the DXR ray tracing API, there is a commonality with PC development that obviously helps game makers. And of course, Xbox does still have more compute throughput - so game engines that tap into that will see advantages. Additionally, there are situations where the Xbox ecosystem and feature set yields dividends. For example, while Elden Ring may run faster on PlayStation 5, Sony's limited implementation of variable refresh rate support means that we'd much rather play the game on Xbox Series X - it's just a smoother and more consistent experience.
At this point, having heard the same set of points from a range of entirely unrelated developers means it's effectively case closed on this particular topic - and we expect to see a similar situation play out across the rest of the generation. PlayStation 5 Pro? That's going to be interesting. We'd expect it to lean into the same strengths that made PS5 competitive and go beyond - PSSR machine learning-based upscaling should act as an effective 'multiplier' of sorts on its increased GPU performance.
 
Last edited:

Tsaki

Member
Every square millimeter of silicon is expensive in 2024, and when console makers take a bath selling this hardware, they better make sure the software can tap as much of it and as efficiently as possible.
These kind of optimizations in the software and hardware level can go a long way; making one console perform on par with another that costs up to $200 more to create is quite the feat.
I'm eagerly awaiting for the man's Pro presentation.

jMa9JpZ.jpeg
 
So they're actually admitting they didn't have a clue what they were talking about during the early part of this generation?
Or admitting whoever wrote the script out of Xbox HQ didn't have a clue

Again I go back to watching the Road to PS5 and Ybarra and Jason Ronald literally laughing at the PS5 SSD solution saying Sony spent more time and money customizing that and were going to lose every head to head game matchup because of the power advantage the Series X had (I don't recall their exact wording but it was close to this)
 

K' Dash

Member
So they're actually admitting they didn't have a clue what they were talking about during the early part of this generation?

well, it is complex, the hardware is definitely better, but if you don't have a way to leverage that power, then it is meaningless. I'm guessing some studios are more invested than others, since Gears 5 looks amazing while running at 60fps, you can create a framework to give precise instructions and take advantage of that extra power but it doesnt make sense for multiplat studios to do it, it is a waste of time, just like the PS3 years ago, thats why you got amazing looking games like Uncharted and TLOU and the rest looked and ran like crap.
 

GHG

Member
It makes me laugh the amount of times during the early gen comparisons they would say something along the lines of "we are reaching out to the developers to try and understand why the Series X version isn't outperforming the PS5, we hope this can be addressed in a future patch".

Meanwhile the Series X is sat there like:


destitute i have nothing GIF by Breaking Bad
 
Last edited:

Topher

Identifies as young
"So, based on our conversations, the combination of a more efficient GPU compiler, lower-level APIs and higher clock speeds allows PlayStation 5 to match or even exceed the outputs of Xbox Series X in some scenarios. "

To me, that just sounds like an overall better engineered system.

pretty-cool-cerny.gif
 
Last edited:
“On paper, the Xbox Series X is a considerably more powerful piece of hardware than PlayStation 5”


Uh……No, it is not

Why do DF continue to think an 18% higher TF number is in any way considerable?
In compute, yes it is. In next-gen compute workloads Series X will beat PS5 every time... and that's despite their compiler being worse.. aka "teh toolz"

Surprised DF also didn't communicate the obvious... which is that PS consoles clearly (and understandably) get priority.. as the console with the bigger player-base by far.
 
Last edited:

Magic Carpet

Gold Member
I have one question: How has DirectX come to this?

• Are there ancient, legacy sins to consider? Does it need a fundamental rewrite?
• Why isn't it treated like with top priority when being developed?

Would be super interesting to have a peek behind the curtains of how DirectX development team operates and what exact problems cause the inefficiency.
No Nixes needed to port to PC. I can't think of a single other upside for DirectX.

Which begs the question. If all those Asian games are released on PC it should be trivial to port to Series X. But time and again It's just too expensive or too much trouble.
I guess Direct X is a big bag of trouble.
 

Topher

Identifies as young
Not surprising honestly. Visual Studio (if that’s the software the tools are targeted at) is many things, some of them great, but an efficient compiler or great at optimization it is not.

It used to be much better than it is. It has become incredibly bloated these days. I think efficiency has taken a back seat to maximizing features.
 

Dorfdad

Gold Member
Do you ever think we will see a modular console design? Not a PC but a console that has removable parts. Let's say Microsoft and AMD make a deal for 10 years. That they offer upgrade paths every 3 years. 2 different GPU configurations / 2 CPU configurations / and Ram / SSD Upgrade paths?

This would ensure that customers can upgrade as they see fit or need. This also allows developers to have a closed "known" set of variables unlike PC where you can many different specs that could cause issue?

This would allow more PC centric flow and compatibility with games, while offering a slightly more advanced console experience?
 
Oh, you mean like ray-tracing in games such as Ghostwire Tokyo and Callisto Protocol where PS5 bests Series X?
There's nothing next gen about the ray-tracing in those UE4 games :messenger_tears_of_joy:

And again, that was already explained by DF.. The Series X/S RT API shares a lot in common with the PC DXR API.. They don't easily take advantage of specific features the hardware COULD afford them.. because it then becomes harder to co-develop between Xbox and PC.

Basically PC APIs are holding them back from reaching their true potential on Xbox consoles.
 
Last edited:

Darsxx82

Member
It makes me laugh the amount of times during the early gen comparisons they would say something along the lines of "we are reaching out to the developers to try and understand why the Series X version isn't outperforming the PS5, we hope this can be addressed in a future patch".

Meanwhile the Series X is sat there like:


destitute i have nothing GIF by Breaking Bad
It's funny that you find it funny when the vast majority of the time the patches post launch generally improve the XSX version compared to the release. On some occasions the improvement has even been dramatic ( aka The Calisto Protocol)

Then, let's put the context:

1- XSX "enjoys" worse tools (Api) than PS5 that can squeeze the maximum and in less time the capabilities of its hardware (although I don't know if calling them "worse" is correct since the purpose of DX12 is more to be a development tool efficient for a multitude of different hardware and that has its counterpart for specific hardware)

2-As assumed and recognized, PS5 is usually the base development platform for most multiplatform games.

3-XSX is usually the platform that enjoys the lowest optimization time vs PS5 and even XSS.

And despite that, you have to have XSX performing the vast majority of times the same as PS5 and on many many occasions being the console version with the best performance.....

I don't think we have to be very enlightened to know how to draw a conclusion about about the capabilities and power of XSX hardware.
 

StereoVsn

Gold Member
It used to be much better than it is. It has become incredibly bloated these days. I think efficiency has taken a back seat to maximizing features.
Yeah, I would say starting with around 2007-2008 or so (been a while), somewhere around Vista release, features took over optimization.

Thought process became that computers have a lot of hardware power and more can always be thrown at a problem. Features they can sell but it’s hard to quantify optimization of the compiler.
 

Darsxx82

Member
Oh, you mean like ray-tracing in games such as Ghostwire Tokyo and Callisto Protocol where PS5 bests Series X?
You talk as if XSX doesn't outperform PS5 in many other games in its RT modes....

Ps. You should take a look at what Calisto Protocol's situation is currently before using it as an example of something. Nothing to do with the launch situation where we saw the most flagrant case of a Studio prioritizing one version over another and the XSX version released in a regrettable situation.
 
Last edited:

bender

What time is it?
Do you ever think we will see a modular console design? Not a PC but a console that has removable parts. Let's say Microsoft and AMD make a deal for 10 years. That they offer upgrade paths every 3 years. 2 different GPU configurations / 2 CPU configurations / and Ram / SSD Upgrade paths?

This would ensure that customers can upgrade as they see fit or need. This also allows developers to have a closed "known" set of variables unlike PC where you can many different specs that could cause issue?

This would allow more PC centric flow and compatibility with games, while offering a slightly more advanced console experience?

A modular design would be too cost prohibitive in the console space.
 

Topher

Identifies as young
Yeah, I would say starting with around 2007-2008 or so (been a while), somewhere around Vista release, features took over optimization.

Thought process became that computers have a lot of hardware power and more can always be thrown at a problem. Features they can sell but it’s hard to quantify optimization of the compiler.

They have also spent a hell of a lot of time integrating new technology into Visual Studio over the years. The number of templates has increased dramatically. This "one size fits all" approach doesn't just incorporate console and PC gaming, but Azure, Windows, web, mobile.....etc. It is a Swiss army knife that is made up of more Swiss army knives. Here's a screenshot....

2mK5ILe.png


Then compare PlayStation's compiler and APIs.......it's 100% dedicated to PlayStation.
 
Do you ever think we will see a modular console design? Not a PC but a console that has removable parts. Let's say Microsoft and AMD make a deal for 10 years. That they offer upgrade paths every 3 years. 2 different GPU configurations / 2 CPU configurations / and Ram / SSD Upgrade paths?

This would ensure that customers can upgrade as they see fit or need. This also allows developers to have a closed "known" set of variables unlike PC where you can many different specs that could cause issue?

This would allow more PC centric flow and compatibility with games, while offering a slightly more advanced console experience?
The closest thing to your proposal is to me the Steam Deck and other handhelds. No modularity(yet), but a set design that devs can optimise to or not. As you can see the results vary depending on the strength of the support the machine can get. For me there is too much incentive to just brute force with better GPU and CPU than work under constraints unless once again there is a huge market waiting for it. Can Microsoft do it? Maybe. But why would they do it? A console is a cost effective way to play. Modularity beyong a certain point woud be against that purpose, and destroy the greatest advantage consoles have against PC, their ability to be objectively the best may to play recent games for 500$ or less. Maybe it will work if/when cloud gaming is good enough?
 
Top Bottom