-
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.
Ok, We need to know this: Is the XSX more powerful than the PS5 or not?
- Thread starter //DEVIL//
- Start date
Zathalus
Member
What has the 5500 being low end got to do with anything? It is 12 CU per shader array and performs just as well as the 5700 when comparing relative specifications and performance between the two. Nothing in benchmarks between the two cards indicate that the 12 CU configuration is causing any issues. Also, being excellent at GPU compute does not mean it is not great for gaming either, just look at how the recent Geforce architectures work, they are all amazing for compute as well, and do just fine in general gaming workloads.5500 is a 150 bicks card, its low end.
Point still stands, VEGA and large shader arrays are excellent for GPU compute, the large arrays are not visible anymore on RDNA1 or 2 except the lowest end 22 CU 5500.
XSX silicon was designed with a server grade CPU and has large VEGA like shader arrays. Go figure.
If you want to know why sharing LDS and other other parameter cache / store on a shader array gives bottlenecks, the more CU sharing for gaming work loads I wonder how that goes. and Cerny himself tells you why in his patent to overcome this.
Even for 10 CU, ps5 patent compresses the oarameters for LDS.
If you want benchmarks, check out Ps5 vs XSX.
As for benchmarks between the Ps5 and XSX, that is rather useless in determining relative hardware performance between the two, as there are multiple difference between the consoles from the OS layer, API involved, as well as the differences in development environment. For the moment the only conclusive determination to make is that the PS5 is running games better than the XSX as of now. It's impossible to say if this is due to any specific hardware advantages or other contributing factors. The raw hardware performance between the two is so close, that any of the areas I mentioned would have a bigger impact in how games run between the two.
Cherrypepsi
Member
Anyway, congrats, you made the link for the trap (nobody else noticed)
The VEGA architecture is excellent for servers and CDNA
Server optimised VEGA 20, can you now make the connection ? Does larger shader array now compute ?
I dont think this is a problem at all for the xsx.
you are the only one pushing this shader array argument, I have not read about this elsewhere.
They didnt design a GPU just to not utilize it's full power. come on man. There will be games taking advantage of this, you'll see.
I just hope they will make games specifically taylored for xbox again, or else the box will just run a PC version at fairly good settings.
Gudji
Member
I know that's not what I meant. He and others have been making fun of the cerny talk about the I/O complex on PS5, I'm just pointing out it's not as complex as what's inside PS5.
Anyway there's a really good blog post from RAD tools on this matter for everyone interested:
How Oodle Kraken and Oodle Texture supercharge the IO system of the Sony PS5
The Sony PS5 will have the fastest data loading ever available in a mass market consumer device, and we think it may be even better than yo...
cbloomrants.blogspot.com
Make sure to read the comments because there's an RAD Tools employee giving some tidbits too.
Lol this is literally the exact opposite of the entire point of the dx12u GDK - to unify pc and Xbox development and eliminate tailoring for a specific platform.
Last edited:
geordiemp
Member
So your argument is a VEGA like shader array will be as efficient as an RDNA2 one ?
Ps5 with 36 CU vs XSX with 52 CU, what looks more efficient to you ?
I dont know what to say, Good luck with that.
I noticed 5700 Xt and 6000 series run DX12u perfectly fine if your going down the tools route, again good luck with that.
Thats not to say in some GPU compute heavy games XSX wont do better, or if the engine is optimised for GPU compute, but lets face it common dev environment with PC now so....
Last edited:
geordiemp
Member
I have read this front / back end, CU array and wavefront / packer discussion in lots of places, Check out B3D discussion XSX front and back end,.
Respected AMD leakers such as Kitty is likely correct, the front end is wavefronts and packers / geometry per shader array and not shader engine, Compute unit is likely referring to VEGA like with RT unit added and no fine gated clocks but take from it what you will....
There are lots of tech discussion on XSX server like arrangement if you look.
Does not matter anymore what you or I believe, we can see the performance per CU there are enough XSX and ps5 benchmarks, so whatever.
NOTE - I said this in august after XSX hotchips and predicted ps5 would be same or better than XSX 4 months ago once I saw the XSX VEGA like arrangement, on GAF I got mocked by the usual lot, none of them came back to discuss.....funny that.
Last edited:
three_muffins
Gold Member
That is one of the worst comparisons i have ever seen.
Besides...what is wrong with regular sized pizza?
assurdum
Banned
"Higher specs". Dat juicy 12 TF. Have higher bandwidth and CUs not automatically guarantee better performance especially considered the whole soc setup. The more discouraging thing it's to see so many people convinced about it.
Last edited:
Zathalus
Member
RDNA2 scales extremely well with CU count, the Radeon 6800 has 60 CU units vs the Radeon 6900 XT with 80 CU units, an increase of 33%. In a game such as Doom Eternal which runs on the low level API Vulkan and is extremely well optimised, the 6900XT is roughly 33% faster. A linear increase in performance relative to the CU count. Is a smaller GPU more efficient? Inherently yes, but GPU workloads are embarrassingly parallel so it scales well with increased CU counts.
Once again, if the CU count per shader array was such a massive difference between 14/12 and 10, why do we not observe this behaviour with the 5500 vs the 5700? If going 12 CU per shader array vs 10 CU per shader array was such a drawback, then surely this should be easy to test? Also, if this causes such a difference why did Microsoft and AMD design it that way, they would obviously know the implications themselves?
As for your comment about tools, what exactly are you referring to here? Do you think the APIs, OS layer, and development do not have an impact on how well a GPU performs? Which API between the PS5 and XSX is faster? Which OS has less overhead? Which development kit is easier to master and the developers are more familiar with? What console has been worked on the longest?
Without answers to the above you cannot make a firm judgment on which console is more powerful. You will notice that not once did I claim the XSX was faster in this thread. Just that a number of factors go into how well a game runs on a console. I personally believe it's way too early to say with certainty which console is more powerful than the other, and whichever does eventually end up on top, the differences are going to be minor.
Last edited:
geordiemp
Member
Ps5 has 10 CU per shader array, 4 shader arrays.
6800 has 10 CU per shader array and 6 shader arrays
6800 XT has 10 CU per shader array and 10 shader arrays
Yes GAMING work load performance scales very well per shader array not CU count in RDNA2
If XSX had 6 shader arrays and fine clock gating we would not be having this conversation, XSX would be in 6800 performance bracket. Its not, MS chose to go a different route in architecture.
Yes Ps5 api and drivers, written and optimised around 1 die will always be better, it all adds up.
Which is why in August I predicted there would be no 20 % difference at all and the current performance was no surprise to me at least.
Last edited:
JackMcGunns
Member
I can’t believe we’re still on the tflop train.
We got caught. It’s fine, we shouldn’t be that surprised I mean shit we’ve seen nvidia vs amd with totally different flop counts perform close to each other for YEARS (usually lower AMD tflop equalling higher nVidia flip counts in real world) so we should accept that there is precedent for tflops not telling the whole story anyways.
That being said, wide vs narrow is still a valid area of speculation IMO.
We got caught. It’s fine, we shouldn’t be that surprised I mean shit we’ve seen nvidia vs amd with totally different flop counts perform close to each other for YEARS (usually lower AMD tflop equalling higher nVidia flip counts in real world) so we should accept that there is precedent for tflops not telling the whole story anyways.
That being said, wide vs narrow is still a valid area of speculation IMO.
Iced Arcade
Member
Check the speculation thread
EverydayBeast
thinks Halo Infinite is a new graphical benchmark
Which console is more powerful gets very personal and it’s been less than a hundred days since launch.
Afro Republican
Banned
Poor people here, in this thread. They are so confused.
XSX/PS5 have already nearly 300 games, most o those games run better on XSX, just because most games don't have 500 million dollar ad campaigns and take over most YouTube ad slots don't mean those games don't exist.
People here are acting like XSX/PS5 together only have 50 or 60 games, over a month after launch.
XSX/PS5 have already nearly 300 games, most o those games run better on XSX, just because most games don't have 500 million dollar ad campaigns and take over most YouTube ad slots don't mean those games don't exist.
People here are acting like XSX/PS5 together only have 50 or 60 games, over a month after launch.
New Yorks Finest
Banned
Don't waste your time bro believe me. You have to be jobless, have no life, or breathe GAF 24/7 to go back n forth with these closed minded lunatics..they would go 6 feet under for their preferred plastic if they have to, their mind is set on blue..PERIOD, nothing else exist.
But Especially when you dealing with this type of caliber Individuals:
GTFO
Last edited:
Vtecomega
Banned
Can you name and give an example of some of those said games? How exactly are they better than their PS5 counterpart?
Last edited:
Blue Spring
Read my tears about xbox here --->
PS5 is stronger of course. Did you not see the comparisons?
ethomaz
Banned
More lies.
There are today 79 games released for Xbox Seriex X|S including all old games with the label "Optimized for Xbox Series X/S'.
PS5? There are today 88 games released for PS5.
It is impossible because both system didn't release that many games.
Last edited:
yurinka
Member
Both in paper and in real world results, it is more powerful in some areas and is weaker in some other areas.
The multiplatform comparisions we saw are pretty tied, with sometimes a lead for a console and sometimes a lead for the other one. Even inside the same games sometimes in some things a console or the other one is better, but we saw more frequently a lead for PS5 in native games and for XSX in BC games.
But it's also true that in both consoles the games we're comparing are launch games, which traditionally have been developed without having access to the final hardware or even specs during a big portion of its development time, which means these game developers and engine developers still didn't have the knowledge and expertise to take full advantage of the hardware. Traditionally the 2nd gen of the games for a console see a big jump because the devs know better everything since the start of the development and have both the engines and games more optimized for the consoles.
TLDR: Maybe it's too soon, but as we saw until now and on paper they are pretty tied, each console has a lead in certain areas. In the available comparisions see more frequently a small lead on PS5 for native games but in some cases a small lead on XSX too.
Last edited:
Superayate
Member
That's not the case for cyberpunk for example. IT will be the cas when the next gen update will be out, that's not considered as optimized for Xbox Series X/S now (this is durango gdk version, not scarlett one).
tichamac
Banned
Don't you mean the opposite? And it's a valid comparison because both consoles are using the same cpu/gpu manufacturer.
Yeah.
And no it’s not necessarily a valid comparison because the gpu architectures are not the same... we know that each chip differs significantly in their approach of wide vs narrow as well as supplemental capabilities.
alabtrosMyster
Banned
What you wrote would make sense if, and only if, Microsoft did not use the EXACT same excuses the last time around (with esram, cloud and dx12).
Sure developers can and will get more acquainted with the hardware, how the new ide works and integrate all of Microsoft's platforms (mostly the same as before, but slightly different), but the same can be said for Sony's console, which seems to have more issues when it comes to sharing cross generation code. Sure Microsoft's OSes are bloated, but don't expect that to change in the next 7 years. Sure, you can say it's too early, but in all honesty with the relatively familiar setup these consoles have, there should not be too much expectations for higher rendering performance (beyond the extra data the SSDs can bring to the table... See cyberpunk 2077 in slow hdd mode vs high speed drives on PC).
Thirty7ven
Banned
Only thing that matters is performance and Series X is showing no signs of being more performant than PS5.
For the dozen of you that bought a XSX over a PS5 due to the expectation that it would be more performant, well it sucks to be you.
For the dozen of you that bought a XSX over a PS5 due to the expectation that it would be more performant, well it sucks to be you.
ethomaz
Banned
Yeap I checked it now.
The point is where is he finding these 300 games?
If you include BC it is over 2k... so 300 is basically a lie.
Last edited:
there is no point to discuss with 15+ years sony fanboys will you ever come to a conclusion even when there is evidence (just read geordie thelastword ethomaz james sawyer...etc etc and all the others olds in Xbox One x vs ps4 pro comparisons when what they "defended" was technically losing on all fronts) just give it the the time ... time will itself silences them.
Last edited:
JackMcGunns
Member
Developers had PS5 kits way before Series X, so it stands to reason that the lead platform for these games was the PS5. If you optimize your game to run well on higher clocks and lower CU count, why would anyone expect the same engine to run well on Series X? Just look at AC: Valhalla, there's a section that runs perfectly smooth on PS5, but the screen hitches when panning on Series X, must be that the PS5 is more powerful, right? but wait, the same thing happened on a GTX 3090, now either you've lost your mind and believe the PS5 is more powerful or more performant than a high end PC with a 3090 GPU, or Ubisoft optimized for the PS5 and everything else later. Historically the lead platform has always performed the best, even when you had huge discrepancies like PS2 vs OG Xbox with games like Metal Gear Solid 2.
Hezekiah
Banned
What a crock of nonsense, just amalgamated gibberish copied from elsewhere. At least others have tried to give actual explanations
.
Last edited:
thicc_girls_are_teh_best
Member
To MS's credit, one area Series X should see some notable advantage is texel fillrate. They have more TMUs on their GPU, which helps with a texel advantage even if their clocks are lower.
The texel advantage should allow them to do more with leveraging texels for texture and non-texture tasks, outside of simply some ray-tracing performance benefits. But it's also as you mention; whether something like that advantage shows up readily comes down to if the hardware is specifically targeted for.
Microsoft going with a focus on (hopefully) optimizing the high-end language APIs is a step in the right direction because truth be told, you don't want to be messing with assembly or hand-written assembly unless you HAVE to, either because the high-level tools aren't that good/compatible with the hardware, or you want every single last bit of performance out of the hardware. But the latter is generally not something devs do until in the latter years of a console's lifecycle.
That said MS's issue could potentially be the fact that their high-level tools are standardizing themselves for a suite of platforms. We don't know how generalized the GDK tools are but I'd like to think they are taking the approach of high-level API tools that are still tailored for the specific platform hardware, even if they share root commonalities. If it's just flush standard with little tailoring for specific platforms in the ecosystem the GDK covers, that could be a potential problem, especially if low-level assembly/hand-written assembly access is not provided easily due to obfuscation through abstraction (or at all).
Le Big Mac
Member
Xb1S has the raw power for sure and a year from now a better picture will emerge on platform performance.
This does present the interesting throwback for PS3/360 days where PS3 on paper with it's exotic architecture was hyped up a lot. Obviously we know that the 360 was far easier to develop for and late into lifecycle because of its scalability (and the copious usage of UE3) that it continued to exceed PS3 simply because it was easier to develop for.
This does present the interesting throwback for PS3/360 days where PS3 on paper with it's exotic architecture was hyped up a lot. Obviously we know that the 360 was far easier to develop for and late into lifecycle because of its scalability (and the copious usage of UE3) that it continued to exceed PS3 simply because it was easier to develop for.
Now it's true people.
You should have continued to disappear from the forum ... but a vagabond gives a break, he doesn't stop.
Last edited:
XXL
Member
rnlval
Member
Anyway, congrats, you made the link for the trap (nobody else noticed)
The VEGA architecture is excellent for servers and CDNA
Server optimised VEGA 20, can you now make the connection ? Does larger shader array now compute ?
NAVI 10 XT RDNA v1 has 40 active CU within the dual shader engine layout with 64 ROPS and 4 MB linked L2 cache. NAVI 21 is just near 2X scale from NAVI 10!
Hawaii Pro GCN has 40 active CU within the quad shader engine layout with 64 ROPS.
X1X GCN has 40 active CU within the quad shader engine layout with 32 ROPS and 2 MB render cache.
Tonga GCN has 32 active CU within the quad shader engine layout with 32 ROP
From
Figure 3 (bottom of page 5) shows 4 lines of shader instructions being executed in GCN, vs RDNA in Wave32 or “backwards compatible” Wave64.
Vega takes 12 cycles to complete the instruction on a GCN SIMD. Navi in Wave32 (optimized code) completes it in 7 cycles.
In backwards compatible (optimized for GCN Wave64) mode, Navi completes it in 8 cycles.
So even on code optimized for GCN, Navi is faster., but more performance can be extracted by optimizing for Navi.
Lower latency, and no wasted clock cycles.
--------
With Vega GCN's wave64 instructions, that's about 33% reduction in render time's shader processing.
NAVI 10 XT with 9.75 TFLOPS is effectively about 12.96 TFLOPS Vega GCN. Major efficiency gain via shader pipeline latency reduction.
NAVI 10 XT's (nearly) "everywhere" (AMD PR) delta color compression reduces the need for Radeon VII's HBMv2, hence why NAVI 10 XT is close to Radeon VII's gaming results.
Last edited:
rnlval
Member
Fire Pro W5000 (Pitcairn) has 12 CU with 156 GB/s memory bandwidth GDDR5 2GB VRAM is still inferior to R7-265 (Pitcairn Pro) or HD 7850 (Pitcairn Pro).
MS's ESRAM and DX12 arguments don't address XBO's less TFLOPS problem when compared to PS4.
XBO's 32 MB ESRAM dependant on CU processing ROP"s read and write I/O.
XBO's 32 MB ESRAM bandwidth is effectively close to 256 bit GDDR5-5500 which is not like NAVI 21's super-fast 128 MB L3 cache Zen design.
Last edited:
Shmunter
Member
The hitches aren’t framerate drops however, just some oversight in camera movement unrelated to performance. Check the analysis.
Here’s the thing. Wider, narrower, faster, etc, Gpu’s don’t need special coding, their parallel processing nature is transparent and basic bread and butter stuff. Sticking a better GPU into a pc has immediate benefits.
Specific feature sets do however need coding, and the api’s need to support those e.g. raytracing, dlss, etc. It is unclear what special features are being ignored on XsX to make it perform lower than PS5. Indeed, logically the simple answer is the system is just less performant. Likely due to the split speed memory setup or lacking cache, which indeed may require special coding to not get trapped by the XsX setup.
Coulomb_Barrier
Member
rnlval
Member
XSX GPU has additional CUs for raytracing and texture operations. TMUs can be used as an ROPS workaround.
ethomaz
Banned
Last edited:
The hell is this?
ethomaz
Banned
MisterXMedia level of delusion.
Last edited:
LordBlodgett
Member
This is my thought. I still don't expect much difference though. This is the closest these two consoles (Sony & Microsoft) have ever been spec-wise. Even on paper we are talking about only a 17% difference in the GPU and a negligible CPU power difference. One X had a 50% more powerful GPU than PS4 Pro, and the PS4 had 50% more powerful GPU than base Xbox One. PS3 was way more powerful than Xbox 360 on paper but was extremely hard to develop for, and then original Xbox was much more powerful than PS2 but came out much later.
I guess the 25TF when ray-tracing line got him lol
rnlval
Member
You both missed GDC 2014 lecture. Try again.
Last edited:
rnlval
Member
That's BS.
Try again.
I'm talking about this "XSX GPU has additional CUs for raytracing and texture operations"