• 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.

Next-Gen PS5 & XSX |OT| Console tEch threaD

Status
Not open for further replies.

Greggy

Member
Well previously in pc gpu competitions I heard one of the vendors cut corners on their implementations to get higher benchmark scores. Same might be happening here.
It's not just int the Control RT benchmatk that the XSX is besting the PS5. I do agree however that Ms customesrs probably didn't buy a console to play it in photo mode. They should expect the power advantage long touted by MS to become more regularly perceptible in next gen only multiplat games, once everyone is familiar with the tools.

 

ethomaz

Banned
It's not just int the Control RT benchmatk that the XSX is besting the PS5. I do agree however that Ms customesrs probably didn't buy a console to play it in photo mode. They should expect the power advantage long touted by MS to become more regularly perceptible in next gen only multiplat games, once everyone is familiar with the tools.


In DMC V is barely 5% difference... not exactly 10-20%.

10-20% is the advantage of the 120fps mode for PS5.

I corrected him.
 
Last edited:
PS5 has been smashing XSX on every game
Are You Sure Super Troopers GIF by Searchlight Pictures
 

kyliethicc

Member
It's not just int the Control RT benchmatk that the XSX is besting the PS5. I do agree however that Ms customesrs probably didn't buy a console to play it in photo mode. They should expect the power advantage long touted by MS to become more regularly perceptible in next gen only multiplat games, once everyone is familiar with the tools.


Tell me, how does one PLAY a game in Photo Mode?
 
I would say the data is handled differently during gameplay than photomode. Some things just don't need to be calculated during actual gameplay. It's all that I can think of that would explain the differences between the two modes.
It's simple, the CPU. During cutscene and photomode the CPU is usually barely being used. When the CPU is being used during gameplay others factors (like latency) are also very important. And we know Sony have done plenty of customizations (we don't know yet all the details) to reduce overall latency.

It's not just int the Control RT benchmatk that the XSX is besting the PS5. I do agree however that Ms customesrs probably didn't buy a console to play it in photo mode. They should expect the power advantage long touted by MS to become more regularly perceptible in next gen only multiplat games, once everyone is familiar with the tools.


I agree, in DMC5 the XSX was often beating the PS5 by 2 to 10% in the 60fps modes, but more often in the cutscenes as it was much harder to have like for like scenes during gameplay. In gameplay the performance was often very similar. But in the 120hz mode the PS5 was actually beating XSX because the Xbox game was heavily dropping in the 60s in that mode.
 
Last edited:
It's simple, the CPU. During cutscene and photomode the CPU is usually barely being used. When the CPU is being used during gameplay others factors (like latency) are also very important. And we know Sony have done plenty of customizations (we don't know yet all the details) to reduce overall latency.

1CCX Vs 2CCX is the only thing that I can think of. Without a die shot we really can't confirm if it's true or not.
 

ethomaz

Banned
1CCX Vs 2CCX is the only thing that I can think of. Without a die shot we really can't confirm if it's true or not.
Just remember even the exactly same CPU runs better on PS4 with lower clock than Xbox One.

DirectX API even after all the optimizations in DX12 is still a heavy CPU overhead API.

That is way abstraction layers works... you have a single binary that works in all types of hardware and to archive that there is a lot of conditions decisions for each hardware specific... all these “ifs” are CPU processing.... each GPU call has them and games do several hundred of GPU calls per second... all them doing these CPU specific processing to decide the best code path for that hardware.
 
Last edited:
Just remember even the exactly same CPU runs better on PS4 with lower clock than Xbox One.

DirectX API even after all the optimizations in DX12 is still a heavy CPU overhead API.

Pretty much the backwards compatibility curse. Microsoft API does wonders for BC but at the expense of performance. Also why the GDK works with a wide variety of systems (XSX, XSS and PC).
 

assurdum

Banned
It's not just int the Control RT benchmatk that the XSX is besting the PS5. I do agree however that Ms customesrs probably didn't buy a console to play it in photo mode. They should expect the power advantage long touted by MS to become more regularly perceptible in next gen only multiplat games, once everyone is familiar with the tools.


The Control RT "benchmark" is tested in a static scene. It's surely a curious stuff but Christ hearing series X could handle 60 FPS with raytracing and the developers have to deliver in this state just because they need to release the game as faster as possible, it's really embarrassing. Fine, let's take as valid the series X cans run at steady 60 FPS Control in an empty corridor when nothing happens (which is not even true). It is already stated dips to 33 FPS are commons when the action is chaotic. What's the point to lock to 60 FPS with such premises? Did they even forgot already FPS it's more problematic on Xbox console? Leadbetter should use a bit better his brain before claim such absurdity.
 
Last edited:

Bo_Hazem

Banned
Control is an interesting case when it tries to do everything, but do we really need all raytracing now? If this is only software-based GI, only RT reflections are needed for full realism:

unreal-engine-5-ps5-demo-xbox-series-x-gaming-tech-performance-game-developer.jpg


Control isn't using Geometry Engines and many other performance-efficient tools, nor expecting much from a small studio. Insomniac managed 1600-1080p with RT reflections at 60fps, many particles going around, and still with great city density. This is a muddy phase into this great generation, Death Stranding would've been shockingly realistic if made around PS5 from the ground up.
 
Last edited:
1CCX Vs 2CCX is the only thing that I can think of. Without a die shot we really can't confirm if it's true or not.
Yes of course that should be enough to explain most of PS5 advantages here during gameplay. We know this from PC benchmarks. There are actually others ways to reduce system latency but unfortunately without more details about their hardware we can only guess.
 
Last edited:

roops67

Member
Do you have a link to the bolded?

Without having seen the APU die shot, I can't imagine how it wouldn't have the same issue. It's still fundamentally a UMA system design with single common pool of memory. Perhaps a clever memory bus arrangement? But even then, both CPU and GPU will still access the individual memory chips across the the same common lanes.
Bear with me I'll look for it. It was either Matt Hargett or the other Matt and backed up by MLiD heard from devs about 6 months ago, it was in a talk in regards to how 'balanced' the PS5 is opposed to the XSX (this always gets debated every gen lol). If anyone else here can find a link that would be appreciated

As far as I can remember the gist of it was the bandwidth to the APU was sufficient to feed both CPU an GPU without hitting contentions, due to the efficient design of the caches ie CPU unified L3 and GPU cache scrubbers etc they minimise going to main memory cos it's very expensive on clock cycles

Remember that CPU (and DMA's) uses a fraction of the bandwidth of the GPU needs. Rough example the GPU can run efficiently with 400GB/s and the rest of the system has just over 40GB/s, keep that balance without hiccups then won't tripup over each other. The Tempest engine can be the wild card though
 
Last edited:

Bo_Hazem

Banned
The most annoying thing about this photo mode benchmark stupidity is that this nonsense is going to run and run for weeks until we get the next proper comparison. There is nothing big coming out any time soon, as far as I know.

Suddenly they forgot that PS5 beats XSX in Graphics and Performance modes, actual gameplay, but then photo mode is still the benchmark! They need to please their "sponsors".
 

DeepEnigma

Gold Member
Suddenly they forgot that PS5 beats XSX in Graphics and Performance modes, actual gameplay, but then photo mode is still the benchmark! They need to please their "sponsors".
If you can’t win in a straight up gameplay environment where it actually counts, formulate a controlled test to support your narrative.

It’s what dictators do, after all.
 
Last edited:
In matter of RT, yes XSX can theoretically cast 18% more rays at any given scene/time but PS5 can generate 22% more (again theoretically) bounces per ray cast or processe any given bounce 22% faster since this is tied to CU frequency. Some implementations of RT can also heavily depend on cache performance and PS5 has an advantage there. So i can't see a clear-cut advantage in this area for XSX, specific implementations can slightly favor either machine.

As to memory bandwidth XSX's split bandwidth (with only one part being 25% faster and the other around 30% slower compared to PS5's) solution sharing the same adress space is far from ideal. Which is important is the real world whole system bandwidth, which system has more data readily available to be processed in the caches by the CUs, with as less interruptions/cache misses in the data flow as possible. In this, cache speed/bandwidth, architectural differences in cache subsystem (eg. cache scrubbers) also play a very important role. In game examples like the one you provided PS5 often fares slightly better in alpha heavy situations and i don't think this is due to higher pixel fill rate of PS5 only since often fill rate and bandwidth have to work together in alpha heavy scenarios.

I'm clearly not a GPU expert so I can just agree for the RT part. I know that's in reality, XsX can't reach most the time his raw advantage and we will obtain similar results on PS5. I think that's the same thing we have for the compute part. For GPU cache part, I totally agree with you, I think that the PS5 has a good advantage at this level.

For the memory aspect, I disagree about the XsX split memory. I'll try to fastly explain why (I said that when I will have more time, I will share past experiences I had with many memory configuration, splitted one between them). In reality, you need to not watch the bandwidth has separated, and consider that you have 560 GB/s "max" bandwidth for the whole memory subsystem. The thing is simply that you have in your pool (for the six 2GB chip) half adresses only dedicated to fast pool (+ the four 1 GB chip, to be used by the most demanding component, the GPU for example) and the other half for the slow pool (for the other components that request less bandwidth and "smaller" datas). It's totally conceptual.
We can take the example where the fast pool is dedicated to the GPU and slow pool for the other components. When the CPU access to the memory banks, it will allocate/use the channels exactly in the same way that is the case on PS5. The difference is that on PS5, 8 channels can burst the data whereas on Series X, it will be limited to 6. On Series X, when the GPU needs to access a "big" data, it can use all the channels to burst data from the fast pool in one transaction (at 560GB/s), and then if the CPU for example need to read a data, it can use max 6 channels to burst data from the slow pool immediately afterwards (at 336GB/s). On PS5 these two steps are done with the same 448GB/s bandwidth. In the two cases, the more the CPU access to the ram, the more it reduces the bandwidth allocated for the GPU, but it will also reduce the OVERALL bandwidth on Series X. That's the only difference between the PS5 and the XsX (in addition to the dedicated pools 10GB + 6BG). But we know that the GPU is clearly the only part that can really use/saturate the memory bandwidth. So the overall bandwidth on Series X won't be too much reduced (I have really simplified my explanation).
I can add that two flaws are directly possible with this memory configuration (non uniform). A software/game can be bottlenecked by the 6 GB slow pool, in this case that's simply that the ressources are not distributed as it need to be between the two pools. That's the main problem we can encounter with this configuration, where the devs will have extra-work to mitigate that and where we need good tools (yes I know, a forbidden word :D). The second problem could simply appear when the datas quantity that require high bandwidth (>> 336GB/s) will exceed 10GB. In this case, I think that "technology" as sampler feedback streaming could help to reduce the quantity of datas stored in the memory (if I'm not wrong).
For me, as I said the splitted memory is clearly not a really big problem. But I can be wrong and we will see.
 
Last edited:

assurdum

Banned
I'm clearly not a GPU expert so I can just agree for the RT part. I know that's in reality, XsX can't reach most the time his raw advantage and we will obtain similar results on PS5. I think that's the same thing we have for the compute part. For GPU cache part, I totally agree with you, I think that the PS5 has a good advantage at this level.

For the memory aspect, I disagree about the XsX split memory. I'll try to fastly explain why (I said that when I will have more time, I will share past experiences I had with many memory configuration, splitted one between them). In reality, you need to not watch the bandwidth has separated, and consider that you have 560 GB/s "max" bandwidth for the whole memory subsystem. The thing is simply that you have in your pool (for the six 2GB chip) half adresses only dedicated to fast pool (+ the four 1 GB chip, to be used by the most demanding component, the GPU for example) and the other half for the slow pool (for the other components that request less bandwidth and "smaller" datas). It's totally conceptual.
We can take the example where the fast pool is dedicated to the GPU and slow pool for the other components. When the CPU access to the memory banks, it will allocate/use the channels exactly in the same way that is the case on PS5. The difference is that on PS5, 8 channels can burst the data whereas on Series X, it will be limited to 6. On Series X, when the GPU needs to access a "big" data, it can use all the channels to burst data from the fast pool in one transaction (at 560GB/s), and then if the CPU for example need to read a data, it can use max 6 channels to burst data from the slow pool immediately afterwards (at 336GB/s). On PS5 these two steps are done with the same 448GB/s bandwidth. In the two cases, the more the CPU access to the ram, the more it reduces the bandwidth allocated for the GPU, but it will also reduce the OVERALL bandwidth on Series X. That's the only difference between the PS5 and the XsX (in addition to the dedicated pools 10GB + 6BG). But we know that the GPU is clearly the only part that can really use/saturate the memory bandwidth. So the overall bandwidth on Series X won't be too much reduced.
I can add that two flaws are directly possible with this memory configuration (non uniform). A software/game can be bottlenecked by the 6 GB slow pool, in this case that's simply that the ressources are not distributed as it need to be between the two pools. That's the main problem we can encounter with this configuration, where the devs will have extra-work to mitigate that and where we need good tools (yes I know, a forbidden word :D). The second problem could simply appear when the datas quantity that require high bandwidth (>> 336GB/s) will exceed 10GB. In this case, I think that "technology" as sampler feedback streaming could help to reduce the quantity of datas stored in the memory (if I'm not wrong).
For me, as I said the splitted memory is not a really big problem. But I can be wrong and we will see.
I suspect you look to the bandwidth usage just to the contention side but you don't consider how could affect the processing speed in the whole hardware. I'm not an expert but I use the simple logic. Bandwidth is unified physically, so if you splitted virtually an unified bus, the GPU has to wait the CPU uses and viceversa if they have to work separately. An unified environment can't use two separate speeds at the same time and can't have a minimal impact such approach. There will be surely a conflict when CPU and GPU are involved in the bandwidth in the same instant. I'm quite convinced AC Valhalla, Control and Hitman 3 FPS issues are caused by it.
 
Last edited:

M1chl

Currently Gif and Meme Champion
Control is an interesting case when it tries to do everything, but do we really need all raytracing now? If this is only software-based GI, only RT reflections are needed for full realism:

unreal-engine-5-ps5-demo-xbox-series-x-gaming-tech-performance-game-developer.jpg


Control isn't using Geometry Engines and many other performance-efficient tools, nor expecting much from a small studio. Insomniac managed 1600-1080p with RT reflections, many particles going around, and still with great city density. This is a muddy phase into this great generation, Death Stranding would've been shockingly realistic if made around PS5 from the ground up.
Control is 2019 game, tho. It's best for time being, at least in my opinion. But it's not really next-gen in a sense of what APIs are available with release of the new GPUs/DX Ultimate, etc. For that we need to wait a little : )
 

Bo_Hazem

Banned
Control is 2019 game, tho. It's best for time being, at least in my opinion. But it's not really next-gen in a sense of what APIs are available with release of the new GPUs/DX Ultimate, etc. For that we need to wait a little : )

Not best, I still think Death Stranding is the best looking game. Overall Control is like 8-9 out of 10 in terms of graphics. GOW Ragnarok gonna blow everything in terms of graphics and gameplay, can't for that GOTY game.
 
But again you are talking about the bandwidth as there is only the contention to "solve". Bandwidth is unified so if you splitted virtually an unified bus, the GPU has to wait the CPU load and viceversa if they have to work separately. An unified environment can't use two separate speeds at the same time and something we have to pay . There will be surely an issue when CPU and GPU are involved in the bandwidth in the same instant.

You will encounter exactly the same problem on PS5 than you have on XsX. When you say "the GPU has to wait the CPU load and viceversa if they have to work separately", that's the case with unified memory. The fact that the memory pool is splitted does not change that. Also, there is not separate "speed", the access to each memory bank is done using the same 2*16 bits channel running at the same datarate. The issue you are speaking about will exactly be the same on PS5 and XsX, and the memory controller will manage that exactly in the same way. Simply that on XsX, due to slower bandwidth for the CPU, it will reduce the overall bandwidth.
 
Last edited:

assurdum

Banned
You will encounter exactly the same problem on PS5 than you have on XsX. When you say "the GPU has to wait the CPU load and viceversa if they have to work separately", that's the case with unified memory. The fact that the memory pool is splitted does not change that. Also, there is not separate "speed", the access to each memory bank is done using the same 2*16 bits channel running at the same datarate. The issue you are speaking about will exactly be the same on PS5 and XsX, and the memory controller will manage that exactly in the same way. Simply that on XsX, due to slower bandwidth for the CPU, it will reduce the overall bandwidth.
Absolutely no, on ps5 you can't have such problems. That's data contention what you are talking about, not processing (coherent/uniformed) data speed. Why you don't see the difference between the 2 arguments?
 
Last edited:

M1chl

Currently Gif and Meme Champion
Not best, I still think Death Stranding is the best looking game. Overall Control is like 8-9 out of 10 in terms of graphics. GOW Ragnarok gonna blow everything in terms of graphics and gameplay, can't for that GOTY game.
I really like that art style and stuff, so that might be big contributor, I like brutalism I guess. Probably commie PTSD
 

JonkyDonk

Member
digital foundry is like trying hard to prove xbox series x is stronger than ps5

why they feel the urge to prove that ??

are they suspecting the opposite ?? and trying to convince people otherwise by using these wired tests??

we already know xbox series x is the stronger machine by paper specs

but the real question is ... is it the better machine??
It is strange that they are so determined to find the superiority of XSX in any way they can, but they haven't spent anytime investigating how PS5 is performing equal or even better in many cases, that to me is the real 'story' of this generation so far. Alex has this need to prove that GPU size heirarchy is all that matters despite all the other evidence we've seen. He has no answer as to why those other games performed better on PS5 and hasn't bothered to find one, other than some vague thing about 'tools'. Instead he keeps trying to find increasingly more 'pure' benchmarks to prove that one GPU is bigger than the other, all the while ignoring the actual useful performance and any other nuances related to that.
 
Last edited:

FrankWza

Member
Ok, some of his comments cracked me up.

“Fanboys don’t even work this hard for the dub.”

Facts. 😂
It’s deeper than that. It’s just really hard for them not to have the most powerful console by a long shot. It’s not enough for it to be on paper and theoretical. So this “benchmark “ , even though it’s technically still on paper and still theoretical, is “proof”
I don’t think anyone really doubted the advantage. The systems being this close is an absolutely crushing outcome.
 

onesvenus

Member
Just remember even the exactly same CPU runs better on PS4 with lower clock than Xbox One.

DirectX API even after all the optimizations in DX12 is still a heavy CPU overhead API.

That is way abstraction layers works... you have a single binary that works in all types of hardware and to archive that there is a lot of conditions decisions for each hardware specific... all these “ifs” are CPU processing.... each GPU call has them and games do several hundred of GPU calls per second... all them doing these CPU specific processing to decide the best code path for that hardware.
The Xbox guy at hot chips said that's the biggest missconception regarding DX. When compiling to Xbox, all that abstraction layer is removed so all those "ifs" as you say, shouldn't be there.
 
Absolutely no, on ps5 you can't have such problems. That's data contention what you are talking about, not processing (coherent/uniformed) data speed. Why you don't see the difference between the 2 arguments?

To be honest, I don't see your point.

in you first message, you wrote: "the GPU has to wait the CPU uses and viceversa if they have to work separately". I simply said that's exactly the same on PS5 and XsX, because of interleaved datas. CPU, GPU or other parts of the system are managed in a queue with priority orders to access to the memory.
"An unified environment can't use two separate speeds at the same time and can't have a minimal impact such approach. There will be surely a conflict when CPU and GPU are involved in the bandwidth in the same instant". Same, I don't see your point here, why separate speed ?? why would there be a conflict because of data read with different bandwidth?
 

DeepEnigma

Gold Member
It’s deeper than that. It’s just really hard for them not to have the most powerful console by a long shot. It’s not enough for it to be on paper and theoretical. So this “benchmark “ , even though it’s technically still on paper and still theoretical, is “proof”
I don’t think anyone really doubted the advantage. The systems being this close is an absolutely crushing outcome.
Yep, and your last sentence is what it ultimately comes down to. Embarrassing the lengths they are going with this. All while playing victims in the great console war they help fuel.
 
Last edited:
Status
Not open for further replies.
Top Bottom