Support NeoGAF

[lucbr]

To me, this illustrates once again the priority of development. Xbox consoles, for obvious reasons, will always be last on the list. Time and money are limited.

 

[GHG]

To me, this illustrates once again the priority of development. Xbox consoles, for obvious reasons, will always be last on the list. Time and money are limited.

It's more the issue that they have to split resources across 2 SKU's on the Xbox side.

 

[SlimySnake]

He finally fucking paired those CPUs with a 6700. Only took 2 years lol.

This was a great video. I hope to see more of this stuff from DF. It is very curious to see xsx struggle with vrs on and vysnc off. Vsync has a big hit on my GPU so had they disabled vsync, im guessing the PS5 wouldve performed even better. And we see this in almost every game. For whatever reason, MS demands that vsync not be engaged in these games.

The AMD PC setup performed very close to the PS5 so not much in the way of secret sauce for the PS5 doing all the heavy lifting. So maybe not because of Cerny's IO block, but it seems the PS5's high GPU clocks and XSX's wide and slow design are the reason why so many XSX games have struggled to keep up with the PS5 this gen.

I do wonder why other games perform better on the XSX. If the GPU is indeed poorly designed then surely it would affect every game the same way. I understand some engines will always prefer PC over PS libraries, but you never see a 6700 get outperformed by a 6600 on PC.

Regardless, maybe MS shouldve just done a 40 CU GPU at 2.23 Ghz. They wouldve probably topped out at 11 tflops instead of 12 but they wouldnt have had these embarrassing comparisons for sure. Not a good look for their engineering teams.

 

[shamoomoo]

Benchmark is also clearly non-deterministic given that there are cars in the XSX view and not in the PS5 one. So you have to find shots that are roughly equivalent if you want to do a fair comparison.

There are pedestrian and some cars in the PS5 version,the pedestrians are missing on the left-hand side on the Series X.

 

[analog_future]

It does to a degree but I'm not sure how big of a gap 1008p vs 1152p is.

Basic math, but 45fps is 75% of 60fps whereas 1008p is 87.5% of 1152p. So it still doesn't account for all of those lost frames.

 

[Mr Moose]

Basic math, but 45fps is 75% of 60fps whereas 1008p is 87.5% of 1152p. So it still doesn't account for all of those lost frames.

Is 1152p 2048x1152? If it is, the difference between them is 27% (fps and resolution).

 

[adamsapple]

It's more the issue that they have to split resources across 2 SKU's on the Xbox side.

I doubt it's that big of an issue, otherwise it would effect every single released game.

This game (Cyberpunk)'s always ran better on average on the PS5, the iteration of the Red Engine seems to favor the PS5 architecture more.

 

[GHG]

I doubt it's that big of an issue, otherwise it would effect every single released game.

This game (Cyberpunk)'s always ran better on average on the PS5, the iteration of the Red Engine seems to favor the PS5 architecture more.

It does a lot of games, just not to this degree.

 

[Darsxx82]

There are pedestrian and some cars in the PS5 version,the pedestrians are missing on the left-hand side on the Series X.

Then, in the video you can see visual inconsistencies in the PS5 version. I don't know if the game generates details randomly or Is popping but there are more visual details in XSX.

I still believe that the correct comparison would have been to use Phantom Liberty scenes, only due to the fact that it is the most demanding and recently launched and in it the performance of both is supposedly the same.

 

[Vergil1992]

The interesting thing about this comparison is that unlike The Witcher 3 in Novigrad, here the limitation is not the CPU. The Witcher 3 next-gen runs almost the same in performance mode on both platforms but at first PS5 had a 5-10fps advantage in Novigrad, XSX had an advantage in other areas (in addition to a resolution advantage), but after the updates the performance was identical. Except in RT mode, which Xbox Series X performed better.


But here it is a GPU limitation; XSS demonstrates this by having stronger performance than XSX and closer to that of PS5. It's true that XSX runs at a minimum 25% higher resolution, but personally, unless the PS5's DRS is MUCH more aggressive, I think going from 1008p to 1152p isn't going to make it match PS5 performance. Maybe yes, each engine and each game is different, there are games where resolutions have a much bigger impact on performance than others, especially depending on the memory implications.



I think this game just works better on PS5, at least the base game (its expansion works the same). It is possible if they completely match the resolutions that the performance will come close or, hopefully, the same. It's a very strange decision that a version has a higher DRS overall (if we take into account VG Tech and their old analyses, XSX had a resolution advantage on average) but a clearly worse framerate.

 

[Mr Moose]

Then, in the video you can see visual inconsistencies in the PS5 version. I don't know if the game generates details randomly or Is popping but there are more visual details in XSX.

I still believe that the correct comparison would have been to use Phantom Liberty scenes, only due to the fact that it is the most demanding and recently launched and in it the performance of both is supposedly the same.

Looks like a type of dynamic lighting, some parts have lit windows on PS5 and not on Series X and sometimes the other way around. The fog doesn't help.

 

[Del_X]

Doesn't the Xbox have a thick layer of shit between its OS and the game running that the PS5 doesn't have to contend with? Kind of why suspend and resume works or whatever.

 

[SkyHighTrees]

The walls arrived

 

[SlimySnake]

Doesn't the Xbox have a thick layer of shit between its OS and the game running that the PS5 doesn't have to contend with? Kind of why suspend and resume works or whatever.

Shouldnt matter because the XSX OS already reserves 2.5 GB of ram and 1 CPU core. PS5 reserves 3.5 GB and 1.5 CPU cores according to DF.

This is a GPU bottleneck which the OS shouldnt really have an impact on.

 

[shamoomoo]

Basic math, but 45fps is 75% of 60fps whereas 1008p is 87.5% of 1152p. So it still doesn't account for all of those lost frames.

I'm referring to total pixel difference and not an part of the composing resolution.

 

[Flabagast]

Just checking, does this qualify as a "real next-gen" game?

Geometric density is "intermediary gen" I would say (higher polycount that last gen games, but not on par with UE5 or Remedy mesh shaded models), and AI is behind, but otherwise yes it's pretty next gen I would say.

 

[DinoD]

 

[FIREKNIGHT2029]

• Performance can be up to 33% better on the PS5 in matched frames with Series X having lows of 45fps

 

[VolticArchangel]

Rdna 1.5 delivering the goods.

9 TFLOPS go fpsfpsfpsfpsfpsfps

 

[jm89]

Shouldnt matter because the XSX OS already reserves 2.5 GB of ram and 1 CPU core. PS5 reserves 3.5 GB and 1.5 CPU cores according to DF.

Do you know where they mentioned this?

 

[SlimySnake]

Do you know where they mentioned this?

it was in one of their DF directs either last year or maybe the year before that.

 

[Mr Moose]

Do you know where they mentioned this?

DF Direct Weekly on Kratos' comedic Xbox debut - and a memory boost for Series S

Rich Leadbetter, Alex Battaglia and John Linneman react to the unlikely arrival of Kratos on Xbox consoles and the memory upgrade headed to Series S developers.

www.eurogamer.net

And CPU:

 

[Gaiff]

It's roughly the difference between going from say DLSS/FSR Quality (0.66x) to Balanced (0.58x), 0.66/0.58 = 1.137. Just ran it and it's solidly ~20% faster on my 3080. Not surprising, GPU-limited perf tends to scale fairly linear with resolution.

I asked a mate to run a test on his 6700 using 64%-80% DRS with an 1800p output using FSR2 and another test using 56%-80% DRS. Basically, a floor of 1152p and 1008p. SX and PS5 DRS. Basically no difference. He also restarted the game just to make sure there wasn't some screw up with the settings.

He used console settings and got pretty much the same results. This suggests that perhaps the DRS on SX isn't working like it should? Could be not aggressive enough.

He also ran a test using FSR2 with an 1800p output and set the DRS to 56% min and max, and 64% min and max with a 30fps target frame rate. The resolution shouldn't budge since it's always above 30.

Here is what he got:

Performance was 13% higher at 1152p compared to 1008p in the benchmark.

 

[Physiognomonics]

Yep. PS5 is rock solid in that sequence. A couple of frame drops here and there but by and large, sticks very close to 60fps. Series X is all over the place with a lot of screen tearing and tons of drops into the 40s. This is even more puzzling because a PC with DX12 and what should be significantly more overhead actually outperforms the Series X at matched settings and DRS.

A console should at the very least match an identically-specced PC. It loses a lot of its appeal if it just performs worse.

As usual PS5 performs as it should be. The real anomaly, again, is the XSX that performs quite worse than a similar PC / GPU. Thing is, there are no RDNA2 GPUs on PC with so many CUs per Shader Array. Maybe there is a good reason?

To me, this illustrates once again the priority of development. Xbox consoles, for obvious reasons, will always be last on the list. Time and money are limited.

The Xboxs were actually the first versions to have the next-gen update with more NPCs, better geometry and assets. PS5 got it later.

Shouldnt matter because the XSX OS already reserves 2.5 GB of ram and 1 CPU core. PS5 reserves 3.5 GB and 1.5 CPU cores according to DF.

This is a GPU bottleneck which the OS shouldnt really have an impact on.

Sure but beware as things could have changed from there. Sony almost always increased memory and even CPU available to games (seen on PS3, PSVita and PS4), even years after the release.

 

[Darsxx82]

Looks like a type of dynamic lighting, some parts have lit windows on PS5 and not on Series X and sometimes the other way around. The fog doesn't help.

I don't know for sure, but there it's not just a question of some light from some window. These are buildings and stage elements in the distance completely absent in the PS5 version.

Whatever the reason, it is certainly shocking because of how obvious it is at first glance.

 

[Mr Moose]

I don't know for sure, but there it's not just a question of some light from some window. These are buildings and stage elements in the distance completely absent in the PS5 version.

Whatever the reason, it is certainly shocking because of how obvious it is at first glance.

Do you know where that is on the map? I could boot up the game and check.

 

[poppabk]

PlayStation is unbelievably poweful. I might get one instead of a pc next gen.

Did you miss the part where it performs similar to a PC spec'd to an Xbox - which would be considered a pretty low end PC.

 

[Darsxx82]

Do you know where that is on the map? I could boot up the game and check.

I haven't played CP2077 in a long time. I do not remember. Can it be part of a predetermined mission?

It would be great if you can verify it, only to explain in first hand such a shocking visual discrepancy in that video.

 

[Vergil1992]

I asked a mate to run a test on his 6700 using 64%-80% DRS with an 1800p output using FSR2 and another test using 56%-80% DRS. Basically, a floor of 1152p and 1008p. SX and PS5 DRS. Basically no difference. He also restarted the game just to make sure there wasn't some screw up with the settings.

He used console settings and got pretty much the same results. This suggests that perhaps the DRS on SX isn't working like it should? Could be not aggressive enough.

He also ran a test using FSR2 with an 1800p output and set the DRS to 56% min and max, and 64% min and max with a 30fps target frame rate. The resolution shouldn't budge since it's always above 30.

Here is what he got:

Performance was 13% higher at 1152p compared to 1008p in the benchmark.

It probably doesn't make a difference because it stays close to the target framerate (60fps) more often. Even so, it is difficult for running 1008p vs 1120p to have such a difference in performance. It is true that running at the same resolution as PS5 should help, but I personally think it would still not work like it does on PS5.


The Red Engine thing is curious, because The Witcher 3 (next-gen patch) had superior performance on PS5, at least in performance mode, but then they were even (there was a difference of 5-10fps), but later XSX had the lead in performance, especially in RT mode.

When there are GPU limitations, in TW3 it seems that XSX has improved performance over PS5.

I'm not sure the Red Engine works better on PS5 overall. I think that simply the performance of CD Projekt games varies a lot over time: with The Witcher 3 we saw how in Novigrad both worked "badly", with frequent drops to 45-50fps; In one of the updates, performance got worse on both PS5 and XSX (worse on XSX), and then had a massive improvement that made XSX work the same or better than PS5.


Cyberpunk is simply better optimized for PS5. In fact, I think the fact that the XSX is losing the battle against a 6700 should make that clear. I know there are people who could argue that it is due to the frequencies, but the truth is that if you use a 6800XT and with Adrenalin you lower its frequency to 1800mhz, it will still perform better than a 6700XT.


In this game PS5 is working as it should, or even better. Xbox Series X is performing below what it should. However, in other games like Alan Wake 2, a 6700XT vastly surpasses the PS5, but the performance of the Xbox Series with PS5-XSX graphics settings in performance mode, at 1440p and with FSR in balance, which is 1:1 with the console version, my 6700xt did not maintain 60fps as solid as XSX).



I think there is not much mystery: it depends exclusively on the game, the API and the optimization.

 

[twilo99]

To me, this illustrates once again the priority of development. Xbox consoles, for obvious reasons, will always be last on the list. Time and money are limited.

This is a game from a big studio, so they should've had enough time to optimize for both systems, but overall, yes, if you had to pick one to optimize for it makes perfect sense to do so for the one with 4x the install base or whatever it is.

All of these comps between PS5 and xsx are pointless because the difference in performance is 99% down to software optimization, not the hardware.

 

[SKYF@ll]

Cyberpunk 2077 benchmark one year later
PS5(10.28TF)≧RX6700(limit 2.23GHz)>Xbox Series X(12.15TF)

The reason why there is a big drop in XSX may be because it is during night battles.
For XSX, it may be better to disable VRS and improve IQ.
PS5 and Xbox Series X have different benchmark scores depending on the game and situation.
Each has its pros and cons.

 

[Mr Moose]

I haven't played CP2077 in a long time. I do not remember. Can it be part of a predetermined mission?

It would be great if you can verify it, only to explain in first hand such a shocking visual discrepancy in that video.

I found the area, it does show lights and buildings but they look a bit different than in the video, I am not on the latest version though so I am downloading that to check. I have to say... The fog around the street lights looks absolute shite.
1.610.000

Updated it to 2.100.000 or whatever and the buildings and lights across the water look the same as the Xbox.

 

[Darsxx82]

I found the area, it does show lights and buildings but they look a bit different than in the video, I am not on the latest version though so I am downloading that to check. I have to say... The fog around the street lights looks absolute shite.
1.610.000

Updated it to 2.100.000 or whatever and the buildings and lights across the water look the same as the Xbox.

Buildings and constructions are created randomly at night perhaps? Or different builds on each console? It's strange because in XSX it is the same building as in the DF video ?

Anyway, I don't think it's anything relevant either.

Thanks for the effort

 

[Mr Moose]

Buildings and constructions are created randomly at night perhaps? Or different builds on each console? It's strange because in XSX it is the same building as in the DF video ?

Anyway, I don't think it's anything relevant either.

Thanks for the effort

Yeah seemed to change with the update lol weird. I wonder what else they changed/added from 1.610 to 2.100. The fog around the lights looks slightly less shit in the latest version but that might just be the angle.

 

[playsaves3]

He finally fucking paired those CPUs with a 6700. Only took 2 years lol.

This was a great video. I hope to see more of this stuff from DF. It is very curious to see xsx struggle with vrs on and vysnc off. Vsync has a big hit on my GPU so had they disabled vsync, im guessing the PS5 wouldve performed even better. And we see this in almost every game. For whatever reason, MS demands that vsync not be engaged in these games.

The AMD PC setup performed very close to the PS5 so not much in the way of secret sauce for the PS5 doing all the heavy lifting. So maybe not because of Cerny's IO block, but it seems the PS5's high GPU clocks and XSX's wide and slow design are the reason why so many XSX games have struggled to keep up with the PS5 this gen.

I do wonder why other games perform better on the XSX. If the GPU is indeed poorly designed then surely it would affect every game the same way. I understand some engines will always prefer PC over PS libraries, but you never see a 6700 get outperformed by a 6600 on PC.

Regardless, maybe MS shouldve just done a 40 CU GPU at 2.23 Ghz. They wouldve probably topped out at 11 tflops instead of 12 but they wouldnt have had these embarrassing comparisons for sure. Not a good look for their engineering teams.

I was always tried of them pairing their pc gpus with like a 12900k then doing console comparisons so it’s nice it changed and good on you for noticing

 

[playsaves3]

He finally fucking paired those CPUs with a 6700. Only took 2 years lol.

This was a great video. I hope to see more of this stuff from DF. It is very curious to see xsx struggle with vrs on and vysnc off. Vsync has a big hit on my GPU so had they disabled vsync, im guessing the PS5 wouldve performed even better. And we see this in almost every game. For whatever reason, MS demands that vsync not be engaged in these games.

The AMD PC setup performed very close to the PS5 so not much in the way of secret sauce for the PS5 doing all the heavy lifting. So maybe not because of Cerny's IO block, but it seems the PS5's high GPU clocks and XSX's wide and slow design are the reason why so many XSX games have struggled to keep up with the PS5 this gen.

I do wonder why other games perform better on the XSX. If the GPU is indeed poorly designed then surely it would affect every game the same way. I understand some engines will always prefer PC over PS libraries, but you never see a 6700 get outperformed by a 6600 on PC.

Regardless, maybe MS shouldve just done a 40 CU GPU at 2.23 Ghz. They wouldve probably topped out at 11 tflops instead of 12 but they wouldnt have had these embarrassing comparisons for sure. Not a good look for their engineering teams.

Tbf this game favors nvidia cards so it may explain why the ps5 isn’t punching above its spec like normal

 

[mckmas8808]

But I thought Series X was Microsoft's mid-gen console?

How is it possibile that it performs worse than the base PS5???

This will be Microsoft's worst nightmare! They should have never told people this.

 

[Mr.Phoenix]

I'm confused with the SX results though. The PC version using the PS5 settings easily beats it, and using the Series X settings, it still wins comfortably, albeit by a slimmer margin. I thought maybe it was some DX quirk but this doesn't appear to be the case.

Whats really interesting is that even the XSS performs better than the XSX.

 

[Vergil1992]

I don't think they are using different settings. I think simply on PS5 at that angle the fog covered the buildings.
I think it's fair to say that Cyberpunk is a cross-gen game, it's a really early-gen game that has received updates, but it's probably relatively old code. I don't think it's a better example than games like Avatar, Alan Wake 2 and other more recent games; Even its expansion seems to have very similar performance between PS5 and XSX.

In fact, when the game did not have a patch for PS5/XSX, the game had superior performance on PS5, and I think that difference has remained.


But something happens here that doesn't make sense: in general, a DRS system is more aggressive on the platform that has the worst performance. Another strange thing I see is that the performance of XSX seems to have degraded with the patches, at least in the VG Tech video the metrics didn't seem that different:



The driving parts on XSX are almost as locked to 60fps as on PS5, and in most areas where there are framerate drops, they perform very similarly:

On the other hand, the resolutions seemed closer according to VG Tech (the minimum resolution DF said was now lower):

PS5 in Performance Mode uses a dynamic resolution with the highest resolution found being 3840x2160 and the lowest resolution found being approximately 2062x1160. Pixel counts at 3840x2160 seem to be very rare on PS5 in Performance Mode.Xbox Series X in Performance Mode uses a dynamic resolution with the highest resolution found being 3840x2160 and the lowest resolution found being approximately 2062x1160. Pixel counts at 3840x2160 seem to be very rare on Xbox Series X in Performance Mode.


Below are some example pixel counts for certain scenes on PS5 and Xbox Series X in Performance Mode. Note that these figures are approximate and not necessarily representative of how the entirety of a given area will render.Kabuki Entrance - PS5 2176x1224, Series X 2304x1296Near Police Station - PS5: 2435x1370, Series X: 2560x1440Outside Tom's Diner - PS5: 2506x1410, Series X: 2631x1480Corpo Start Building - PS5: 2656x1494, Series X: 2744x1544Streetkid Start - PS5: 2062x1160, Series X: 2062x1160


I believe that some update has worsened the performance of XSX, something that would not be strange for CD Projekt, it has already happened with The Witcher 3:

In one of the updates, performance was severely degraded in performance mode, both on PS5 and XSX (although it affected XSX more), fortunately, with another update they improved the performance of both platforms, running better than at launch.
It's also possible that simply the Xbox's higher DRS is to blame, or it works better on PS5, or maybe both.


In any case, with CD Projekt's track record, I wouldn't use Cyberpunk as a cross-platform performance metric today. It's not normal behavior at all to have higher DRS, and at the same time, have worse GPU-limited performance.

 

[Mr Moose]

Whats really interesting is that even the XSS performs better than the XSX.

With a low of 648p though.

 

[Physiognomonics]

He finally fucking paired those CPUs with a 6700. Only took 2 years lol.

This was a great video. I hope to see more of this stuff from DF. It is very curious to see xsx struggle with vrs on and vysnc off. Vsync has a big hit on my GPU so had they disabled vsync, im guessing the PS5 wouldve performed even better. And we see this in almost every game. For whatever reason, MS demands that vsync not be engaged in these games.

The AMD PC setup performed very close to the PS5 so not much in the way of secret sauce for the PS5 doing all the heavy lifting. So maybe not because of Cerny's IO block, but it seems the PS5's high GPU clocks and XSX's wide and slow design are the reason why so many XSX games have struggled to keep up with the PS5 this gen.

I do wonder why other games perform better on the XSX. If the GPU is indeed poorly designed then surely it would affect every game the same way. I understand some engines will always prefer PC over PS libraries, but you never see a 6700 get outperformed by a 6600 on PC.

Regardless, maybe MS shouldve just done a 40 CU GPU at 2.23 Ghz. They wouldve probably topped out at 11 tflops instead of 12 but they wouldnt have had these embarrassing comparisons for sure. Not a good look for their engineering teams.

This is not a game (or engine) that showcases PS5 custom I/O. We can see the effect in Sony first party games (Spider-man and TLOU).

 

[OverHeat]

I love hobo’s who fight for scrapes

 

[Denton]

I wonder what framerate with unlocked vsync would PS5 and XSX get in the PC integrated official benchmark. I just ran it on those PS5 settings and got 164fps average.

 

[LiquidMetal14]

That piece of shite Plantation 5

 

[Grildon Tundy]

So, some cool stuff in there. Rich has save files on PC meant to be benchmark runs. They're all scripted in-game and real time sequences to stress the PC. He used the cross-save feature to export those saves to Series consoles and PS5. The scripting remained intact and they serve as perfect 1-to-1 benchmarks.

The interesting bits.

• The benchmarks for Quality/RT and Performance mode are locked to 30 and 60fps respectively
• Both consoles output at 1800p using FSR2 with a maximum internal resolution of 1440p
• Series X in performance mode has a lower bound of 1152p vs 1008 for the PS5
• In the RT mode, the performance difference is irrelevant. They both lock to 30
• First benchmark and least demanding one: Series X drops 36/3600 frames (1%), PS5 25/3600 (0.7%)
• Second benchmark: Series X drops 74/3100 frames (2.4%), PS5 34/3100 (1.1%)
• Third benchmark: Series X drops 1264/15500 frames (8.2%), PS5 32/15500 (0.2%)
• Performance can be up to 33% better on the PS5 in matched frames with Series X having lows of 45fps
• Rich also used an AMD 4800S desktop kit which has a CPU identical to the Series X paired with a 6700 downclocked to 2.23GHz to match the PS5's clocks
• They tweaked the PC version to be a match for the PS5's settings and DRS range
• They ran the third benchmark on all three: Series X 1264/15500 (8.2%) dropped frames with a 55.1fps average, PC 218/15500 (1.4%) for a 59.2fps average, PS5 leading with 32/15500 (0.2%) with a 59.9fps average, and finally series S 432/15500 (2.8%) with a 58.3fps average
• Pairing the 4800S with a 7900 XTX resulted in no dropped frames, confirming that the results on consoles are GPU-limited
• Also tweaked the 4800S/6700 to match the slightly higher DRS of the Series X. PC was still ahead
• Series X/S seem to be using VRS

Edit: DRS windows

Thanks for the summary!

 

[MasterCornholio]

They seem about the same to me.

No idea why people are trying to damage control this so badly. It really isn't a big deal.

 

[ChiefDada]

Kudos to DF - this is a methodically sound benchmark analysis compared to many of their other platform comparisons. I would like to see Richard spearheading more of these analysis.

That piece of shite Plantation 5

 

[Rubim]

This is not a game (or engine) that showcases PS5 custom I/O. We can see the effect in Sony first party games (Spider-man and TLOU).

And would that make a difference on PC?

Does Ratchet and Clank shows that? Cause we got faster loading without direct storage.

 

[Zathalus]

I think there is not much mystery: it depends exclusively on the game, the API and the optimization.

Bingo.

 

[DenchDeckard]

Awesome performance by ps5. When there is this kind of delta I wish digital foundry would reach out to the developers to see if they can comment on what the reasons are, or to give some kind of insight.

I can totally see the ps5s increased frequency helping when traversing the city in a car etc, streaming in assets etc with the sad and fast gpu.

Big gap for the ps5 here.