01011001
Member
runs surprisingly well
sadly I can't find a PS3 video, so no comparison possible at this time.. a shame now we may never know which console is truly more powerful!
Last edited:
-
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.
GhostRunner PS5/XSX/XSS DF Performance Analysis
- Thread starter Mr. PlayStation
- Start date
runs surprisingly well
sadly I can't find a PS3 video, so no comparison possible at this time.. a shame now we may never know which console is truly more powerful!
DonJuanSchlong
Banned
I meant more because there was one area in that game that... Gave me trouble
runs surprisingly well
. I'm having flashbacks lol
01011001
Member
I meant more because there was one area in that game that... Gave me trouble
ONE area gave you trouble? ONE... I think you should consider yourself blessed of only one area really gave you trouble in this game
DonJuanSchlong
Banned
Ok, so it was more than one area...times a couple of occurrences lol.ONE area gave you trouble? ONE... I think you should consider yourself blessed of only one area really gave you trouble in this game
Me playing Sonic 06
Last edited:
But that trainwreck is the only native 4k 60fps trainwreck with Ray tracing on consolesso I see people are still so dumb and argue about console capabilities based on a game that is a trainwreck on every system? damn people are really beyond stupid these days
someone should make a thread about the performance profile of Sonic 06 on PS3 vs 360, and I bet people would argue about which system is more powerful based on the findings
Last edited:
01011001
Member
and it runs like trash... also the RT in this console port is so extremely paired back from the PC version that it is no wonder that it manages it.
on PC the RT is overkill and on console it's basically the lowest end you can have.
just a comparison of the RT, PC vs Console: the PC has RT AO, GI, reflections and shadows... on console only reflections.
on PC the reflections run in full resolution, on console in quarter res (maybe even lower)
and even on PC with the full res reflections they look like ass.
and this brings me to the fact that "4K60fps with raytracing" is not a benchmark that tells you anything about the actual GPU/CPU resources needed to run such a title.
the raytracing in this game is not 4K for example, it's 1080p (if even that), and surfaces that actually use RT reflections in it are extremely limited (I recently played through the game with RT on on PC and the RT implementation was not only bad but also spotty as hell)
if this game ran the full PC raytracing settings at full 4K60, that would be impressive. running sparsely used and low res reflections only is nothing to write home about
edit: with UE4 Unlocker you can try to emulate Console raytracing settings... I might do that later just for fun to see how much it takes to only run RT reflections and at quarter res. I already tried this back in the demo on my old PC using a GTX1070 and I actually got it running at 900p with quarter res reflections at about 40 to 60 fps depending on load, which was not bad for a non-RT accelerated card
Last edited:
Shmunter
Member
As basic as it is, it still hits XsX like a sledgehammer. Finished PS5 with RT, flicked it on and off a few times - couldn’t see any glaring differences to be honest, maybe a bit more depth to the scene in places.
01011001
Member
like I said, consoles only run RT reflections and at a relatively low res. and the game is very weird about it's reflections, even with RT on many surfaces will still either use screenspace reflections instead (sometimes the same object placed in 2 different levels will have RT reflections on or off at random)
so the difference is that reflections will look better, but not all of them and not as good as on PC.
BUT back to my Unreal Engine 4 Unlocker test, I did it, here are the results:
the op is the Console equivalent RT settings (~90-100fps), the bottom is full PC raytracing enabled (~35-49fps)
(the game has the typical UE4 shader compilation stutters on PC, so the 99% percentile number you see here is due to that... there was a significant shader stutter right before I took the screenshot)
edit: because some might ask, the perspective is not perfectly lined up because this game is prone to crash as soon as you alt+tab on it... so I had to reboot the game to do the second screenshot. like I said, not at all stable on any system.
Last edited:
DonJuanSchlong
Banned
The bottom image is soooooo much shaper, detailed, etc.like I said, consoles only run RT reflections and at a relatively low res. and the game is very weird about it's reflections, even with RT on many surfaces will still either use screenspace reflections instead (sometimes the same object placed in 2 different levels will have RT reflections on or off at random)
so the difference is that reflections will look better, but not all of them and not as good as on PC.
BUT back to my Unreal Engine 4 Unlocker test, I did it, here are the results:
the op is the Console equivalent settings, the bottom is full PC raytracing enabled
01011001
Member
it's really bad in motion with the Sword skin I have equipped here as the sword is full with RT reflections (since it's basically fully golden), so on console equivalent settings you really see some nasty shimmering on it. I wonder if this looks as bad on console. the screenshot doesn't really bring it across because it's an issue with their de-noising and in a still image the shimmer is therefore not really visible
Last edited:
DonJuanSchlong
Banned
Not saying it looks the best, but it definitely suffers from that. PC release is a mess, still, and console release isn't any better. But the bottom image looks much better. Nothing like cyberpunk, but it still looks much better.
Hopefully it gets fixed on ALL platforms, and it's an end to these stupid fucking console wars. I'm sick and tired of that shit.
01011001
Member
I just looked up a gameplay video of the PS5 version and tried to get a shot from roughly the same spot.
due to youtube compression it's not that reliable to judge a game based on a yt video, but the reflections here look even lower res than what I set them to in my image above.
this is 4K on PS5 with raytracing, and I captured 1440p... so if rt resolution set to a value of 50 is sharper than this, then I suspect the PS5 version actually running way lower than what i set it to.
but like I said, hard to really be sure based on a YT video, there is almost no fine detail visible in the reflections tho...
due to youtube compression it's not that reliable to judge a game based on a yt video, but the reflections here look even lower res than what I set them to in my image above.
this is 4K on PS5 with raytracing, and I captured 1440p... so if rt resolution set to a value of 50 is sharper than this, then I suspect the PS5 version actually running way lower than what i set it to.
but like I said, hard to really be sure based on a YT video, there is almost no fine detail visible in the reflections tho...
Last edited:
Loxus
Member
I believe people like Mark Cerny, Tim Sweeney and actual game developers.
You guys didn't believe the PS5 couldn't do ray tracing even though Mark Cerny said it can multiple times. The actual lead Architect of the console or Tim Sweeney when talking about the PS5 SSD, who knew more about the PS5 and UE5 than any of us.
You guys don't even bring reputable sources to the thread. You guys just say it's a bug or the developers fault.
We see the PS5 performing equally, outperforming or performing slightly worst, there is no 15-25 or 30% performance advantage for XBSX if the PS5 matches or outperform it.
If you guys don't see that, you are just console warriors. That 18% is only computational performance.
Teraflop performance values don’t always accurately reflect performance, especially when comparing two different custom GPUs. Better texture filtering, compression and tessellation, etc. or extra features such as image sharpening can result in some GPU architectures doing more than others with the same TFLOP numbers.
There are only based on RDNA 2, from there, these are two different and custom GPUs with customizations not done on the other console.
It's just the same fanboy nonsense.
I can only imagine how hard it is for the Mods to read through these threads at times seeing the same thing over and over in every Game Performance Thread.
Zathalus
Member
What the hell are you on about? Who is this nebulous "you guys"? I have never once claimed the PS5 does not have ray tracing, nor have I claimed the performance of the XSX is this particular game was a bug. I just made the statement that the XSX is outperforming the PS5 is most games released this year so far, as you incorrectly stated that more games are performing the same on both with the PS5 having the advantage in details.I believe people like Mark Cerny, Tim Sweeney and actual game developers.
You guys didn't believe the PS5 couldn't do ray tracing even though Mark Cerny said it can multiple times. The actual lead Architect of the console or Tim Sweeney when talking about the PS5 SSD, who knew more about the PS5 and UE5 than any of us.
You guys don't even bring reputable sources to the thread. You guys just say it's a bug or the developers fault.
We see the PS5 performing equally, outperforming or performing slightly worst, there is no 15-25 or 30% performance advantage for XBSX if the PS5 matches or outperform it.
If you guys don't see that, you are just console warriors. That 18% is only computational performance.
Teraflop performance values don’t always accurately reflect performance, especially when comparing two different custom GPUs. Better texture filtering, compression and tessellation, etc. or extra features such as image sharpening can result in some GPU architectures doing more than others with the same TFLOP numbers.
There are only based on RDNA 2, from there, these are two different and custom GPUs with customizations not done on the other console.
It's just the same fanboy nonsense.
I can only imagine how hard it is for the Mods to read through these threads at times seeing the same thing over and over in every Game Performance Thread.
A list of games that have higher resolution or performance on the XSX over the PS5 released this year:
Watch Dogs: Legion 60FPS mode
Outriders
Control
RE: Village
Far Cry 6
Metro Exodus
Hitman 3
Tales of Arise
Subnautica + Subnautica Below Zero
Doom Eternal
Avengers
Games that the PS5 has an advantage in:
Scarlet Nexus
Little Nightmares 2
Ghostrunner
The Touryst
Tony Hawk's Pro Skater 1+2
Die Namek Ability
Member
The list is a bit misleading, for instance the Medium has higher res on the PS5 but they made concessions to make it happen. Some of the games on your XsX list have other advantages on PS5. You did point out it was strictly when talking about resolution OR framerate but you using this as evidence of XsX outperforming PS5 more often. In that case you need to consider more than just res and fps. My .02
I’d just add that the Touryst was remade for PS5, the XSX was a port of the Switch version. Those are essentially two different games. DF even pointed out how the end image is different because they remade the game to be a native PS5 game not a switch port version.
Nikana
Go Go Neo Rangers!
someone should make a thread about the performance profile of Sonic 06 on PS3 vs 360, and I bet people would argue about which system is more powerful based on the findings
Sosokrates
Report me if I continue to console war
You know something is up with the XSX version when a 5700xt is outperforming it.
The 5700xt is getting 140-160fps @1080p max settings.
It could probably get 120fps with DRS being 1296p the majority of the time.
The 5700xt is getting 140-160fps @1080p max settings.
It could probably get 120fps with DRS being 1296p the majority of the time.
it's logic etho. The sx have most powerful hardware so a right brain would say that yes there are problem in the code. Why are you so surprised?
and by the way there's no trend stop it here
Last edited:
Darius87
Member
your logic is that all GPU silicon is made of CU's? right brain would say no.
it run better in rt mode as it shouldn't and the advantage it could have (for the clock difference) it might be in 120fps mode and it doesn't. It is little optimization. period.
Loxus
Member
Dude, you literally just proved my point by listing not one but five games the performs better on PS5.
But here is a proper list of games the perform better on PS5.
Updated October 31, 2021
Every PS5 Game That Outperforms The Xbox Series X Version
- Assassin’s Creed Valhalla (PS5 boasts locked 1440p resolution in Performance Mode whereas Xbox Series X drops as low as 1188p)
- Call of Duty: Black Ops Cold War (In 120 FPS mode, the PS5 version sticks far closer to that target, with Xbox Series X dropping as many as ten frames more)
- Call of the Sea (Improved resolution and framerate, better shadow quality and loading times)
- Control: Ultimate Edition (More stable frame rate in both graphics and performance modes)
- DiRT 5 (Better filtering and texture detail in Image Quality Mode. Much higher level of detail in 120 FPS Mode)
- Ghostrunner (120 FPS and RT modes both have much more consistent performance on PS5 compared to Xbox Series X)
- Greedfall (Sticks much closer to 60 FPS on PS5 than on Xbox Series X)
- Immortals Fenyx Rising (PS5 has a locked 4K native resolution compared to the dynamic 4K seen on Xbox Series X in Resolution Mode)
- MLB The Show 21 (PS5 has a clear framerate advantage at the same resolution over Xbox Series X)
- NBA 2K21 (Xbox Series X version suffers minor performance dips when compared to PS5)
- Need for Speed: Hot Pursuit Remastered (Xbox Series X framerate is nowhere near as consistent as PS5)
- Scarlet Nexus (Sticks to 60 FPS more closely on PS5)
- Shadow of the Tomb Raider (Though resolution is slightly lower on PS5, the framerate sticks much closer to 60 FPS whereas Xbox Series X drops far more often)
- The Elder Scrolls V: Skyrim Special Edition (Maintains 60 FPS much more consistently than Xbox Series X with 60 FPS mod applied)
- The Touryst (Supports 60 FPS at 8K – Xbox Series X runs 60 FPS at 6K and has poorer depth of field)
- Tony Hawk’s Pro Skater 1 + 2 (PS5 version has faster loading times than Xbox Series X version)
With that many games, how can one say XBSX has a performance advantage. Some games will perform better on either console depending on many other scenarios, not just 18% better teraflops.
Zathalus
Member
I said this year, that list includes games from last year and is factually incorrect on several of the games.
Control now performs better on the XSX ever since Microsoft fixed the VM bug that was causing the stutters. Immortals Fenyx Rising is identical on both consoles due to a series of patches for it on Xbox.
Need for Speed, Skyrim, and Tomb Raider are all BC games and should not be used in any real comparison. Especially Tomb Raider, considering it is running at twice the resolution on the Xbox.
That equals more native games that released this year that perform better on the XSX.
Die Namek Ability
Member
Thats the problem with list wars.
XsX has more games that launched this year, that are native apps that run perform better or have a better resolution that are not BC that are named Forza Horizon 5 than PS5....
Zathalus
Member
Yes, Xbox has more native games that launched this year that perform better on it. That's it, no further statement needed.
Do all games run better on it? No, but that was not the point I was making.
Loxus
Member
Having more still don't prove me wrong.
It would be nice if you would link sources backing up your claims.
Zathalus
Member
Watch Dogs Legion:
0.01% better average FPS on the PS5, 10-20% resolution advantage for the XSX. AF has been patched to be the same.
Outriders FPS drops have been fixed since the demo, and the launch version of the game had no issues:
Same performance, 10% resolution advantage.
Control, no more performance drops on XSX:
Benchmark shows average 16% performance advantage:
Metro Exodus:
0.03% advantage in average performance for PS5, 20% advantage in resolution for XSX.
The performance advantage that the PS5 had due to the slight FPS stutter issue the XSX version had was also patched:
Metro Exodus Update 1.002 Patch Notes 1.002.000
Metro Exodus update 1.002 Patch Notes released for PS5 and Xbox Series X players. Read Metro Exodus version 1.002.000 patch notes.
updatecrazy.com
Hitman 3, 44% advantage in resolution, a single area of the game that has 40% less performance:
Tales of Arise:
RE: Village:
Far Cry 6:
Doom Eternal:
Zathalus
Member
Avengers:
Subnautica Below Zero:
Immortals Fenyx Rising being the same after some patches:
DenchDeckard
Gold Member
I guess receipts were provided?
Loxus
Member
Clearly you don't get it do you.
When I side claims, I meant provide technical information why the XBSX isn't outperforming PS5 in all games.
The point you don't understand, is the PS5 is outperforming XBSX in too many games to claim a performance advantage.
Me and others in this very thread were saying some games perform better on PS5, while others on XBSX. But you can't except that, can you?
Zathalus
Member
I can totally accept that, or did you miss me mention some games performing better on PS5? I was just pointing out that it is simply not true that most games are performing the same or better on PS5. Which your original claim was.
i think most of this games arent native code, they are running lastgen code just enhanced on nextgen consoles its why they have trouble performing well on ps5 most of the times and even on series x sometime, i would personally wait for nextgen only games to do a fair comparison cause weve seen what native games can perform on ps5 like the touryst achieving 8k 60 with raytracing because they rewrote the code to fully utilise ps5's api.. there is no reason for a lastgen looking game like hitman to run 1800p on ps5 or outriders, far cry 6 even resident evil village when games like ratchet and clank have more detail well above those games from raytracing to geometry to particles or demon souls per scene and achieve full 4k.Watch Dogs Legion:
0.01% better average FPS on the PS5, 10-20% resolution advantage for the XSX. AF has been patched to be the same.
Outriders FPS drops have been fixed since the demo, and the launch version of the game had no issues:
Same performance, 10% resolution advantage.
Control, no more performance drops on XSX:
Benchmark shows average 16% performance advantage:
Metro Exodus:
0.03% advantage in average performance for PS5, 20% advantage in resolution for XSX.
The performance advantage that the PS5 had due to the slight FPS stutter issue the XSX version had was also patched:
Metro Exodus Update 1.002 Patch Notes 1.002.000
Metro Exodus update 1.002 Patch Notes released for PS5 and Xbox Series X players. Read Metro Exodus version 1.002.000 patch notes.updatecrazy.com
Hitman 3, 44% advantage in resolution, a single area of the game that has 40% less performance:
Tales of Arise:
RE: Village:
Far Cry 6:
Doom Eternal:
basically its simply developers not being bothered to rewrite the code simply because it takes time and they are comfortable with the results and they have to release the games on the budget and time limit. the "control" devs have said this so have crysis devs and metro exodus devs theyve all said there games arent running native code or fully using the low level api's console. only devs to clearly state that they are using native code are doom eternal devs, demon souls, ratchet and clank, the touryst. so very few games are native that renderes most of this comparisons meaningless.
for reference heres control devs statement: "Whenever you're in this cross-generational point, to be blunt, it sucks," Puha said. "You have to support the previous gen, make sure that sings, and then whatever you bring to next-gen is still limited by the choices you made years ago for the previous generation. It's not a very realistic thing, that this old game, we're just going to remake everything and then bring it to next-gen. It's just not like that. It's not a reality for us, because you're literally taking away resources that are building the future games and improving the engine for the future." Control Developer: Making Games For Two Generations of Consoles at Once 'Sucks' - IGN
heres crysis dev statement on df interview :In a recent Digital Foundry interview with Crytek, it was revealed that there will be no native next-gen console versions of Crysis 2 and Crysis 3 – they will just be playable via backwards compatibility, with the PS5 offering a resolution of 1440p and a target of 60 fps. The Xbox Series X should offer a higher resolution. You can check out some footage of Crysis 2 on PS5 in Digital Foundry’s full video below. Crysis Remastered Trilogy Won’t Run Natively on Next-Gen Consoles, PS5 is Only 1440p (wccftech.com)
Did you bring pizza?Hey guys I'm back .. what the? ...
Last edited:
MasterCornholio
Member
Did you bring pizza?
rnlval
Member
FYI, XSS/XSX and PC RDNA 2 has variable-rate shading capable pixel shading-ROPS path. PC RDNA 2 further evolved the ROPS path with a large L3 cache.I believe people like Mark Cerny, Tim Sweeney and actual game developers.
You guys didn't believe the PS5 couldn't do ray tracing even though Mark Cerny said it can multiple times. The actual lead Architect of the console or Tim Sweeney when talking about the PS5 SSD, who knew more about the PS5 and UE5 than any of us.
You guys don't even bring reputable sources to the thread. You guys just say it's a bug or the developers fault.
We see the PS5 performing equally, outperforming or performing slightly worst, there is no 15-25 or 30% performance advantage for XBSX if the PS5 matches or outperform it.
If you guys don't see that, you are just console warriors. That 18% is only computational performance.
Teraflop performance values don’t always accurately reflect performance, especially when comparing two different custom GPUs. Better texture filtering, compression and tessellation, etc. or extra features such as image sharpening can result in some GPU architectures doing more than others with the same TFLOP numbers.
There are only based on RDNA 2, from there, these are two different and custom GPUs with customizations not done on the other console.
It's just the same fanboy nonsense.
I can only imagine how hard it is for the Mods to read through these threads at times seeing the same thing over and over in every Game Performance Thread.
Mesh shader/Amplification shader, variable-rate shading and deep learning dot math features have been verified for XSX.
Layout for PS5 is closer to RX 5700 XT but with RT cores attached TMUs.
The following picture is for RX 5700 XT.
The following picture is for NAVI 22 (PC RDNA 2, RX 6700 XT)
Notice NAVI 22's non-CU hardware is mostly on one side like XSX's RDNA 2 GPU design.
NAVI 22's memory controllers location with sides and top is similar to XSX GPU's memory controllers location with sides and top.
XSX's GPU design is effectively scaled up NAVI 22.
RDNA v1 and RDNA v2 have similar IPC. PC/XSX/XSS RDNA v2 has extra resource conservation and NVIDIA influenced DX12U features.
Last edited:
rnlval
Member
FYI, Doom Eternal RT has the advantage for XSX.
Sosokrates
Report me if I continue to console war
Its not that people are disturbed by it, people just dont agree on what the reasons may be.
rnlval
Member
XSX is showing its RT advantages over PS5. Most cross-gen games are not using DirectX12U's features.
XSX Doom Ethernal used DirectX12U's Variable-Rate Shading and RT.
If a game run better on Xbox Series X people accept that it's because of the higher compute performance but when it run better on PS5 no one wants to entertain the possibility that it could be the higher pixel fill rate , higher geometry output , higher internal bandwidth or better data management?
It has to be the devs showing more love to PS5 SMH
Sosokrates
Report me if I continue to console war
Probably because, compute and memory bandwidth have been the determining factors for performance for a long time.
"higher geometry output , higher internal bandwidth or better data management?"
Where are you getting this from? Im pretty sure the PS5 does not have higher geometry output and higher internal bandwidth, apart from the ssd+io setup. (But series consoles have there own unique solutions (hardware based SFS) which may alleviate that advantage
The seriesX Has considerably higher differences in the areas you mentioned (as well as gpu clockspeed) compared to the SeriesS, but the seriesS performs what the tflops suggest, the seriesX does not have some significant performance gain because of its higher clocks, and higher pixel fillrate.
Also I dont think its a case of devs showing more love to the PS5, I think its a case of the xbox versions being less Optimised, which could happen for a number of reasons.
In the case of ghostrunner its not only about it being compared to the PS5 version, a 5700xt gets 140-180fps @1080p ultra , achieving 120fps at an with DRS OF 1080-1296p does seem likely, whats your explanation for that? In terms of hardware theres no reason why the 5700xt should outperform the XSX, the XSX's GPU is suprior in every aspect.
Loxus
Member
So you just gonna forget about Spider-Man and Ratchet RT implementation?
5700XT actually have a small advantage in Pixel fill rate , Geometry output & internal bandwidth over Xbox Series X for the same reason that PS5 has those advantages because of the higher clock.
PS5 has a even bigger advantage because of the older RBE setup that they used because of more depth ROPS.
Sosokrates
Report me if I continue to console war
Yeah the 5700xt would, like the xbox one had over the PS4....
But a few percent higher clock on the 5700xt is not going to cause like a 20-50% performance increase over a 12tflop RDNA2 based GPU with 560gb/s
Last edited:
DonJuanSchlong
Banned
Forgot about that completely lol. Good point.
Actually no the Xbox One had 16 ROPS while PS4 had 32 ROPS
Sosokrates
Report me if I continue to console war
Still had the fillrate advantage or does that not count now?
Loxus
Member
FYI, XSS/XSX and PC RDNA 2 has variable-rate shading capable pixel shading-ROPS path. PC RDNA 2 further evolved the ROPS path with a large L3 cache.
Mesh shader/Amplification shader, variable-rate shading and deep learning dot math features have been verified for XSX.
Layout for PS5 is closer to RX 5700 XT but with RT cores attached TMUs.
The following picture is for RX 5700 XT.
The following picture is for NAVI 22 (PC RDNA 2, RX 6700 XT)
Notice NAVI 22's non-CU hardware is mostly on one side like XSX's RDNA 2 GPU design.
NAVI 22's memory controllers location with sides and top is similar to XSX GPU's memory controllers location with sides and top.
XSX's GPU design is effectively scaled up NAVI 22.
RDNA v1 and RDNA v2 have similar IPC. PC/XSX/XSS RDNA v2 has extra resource conservation and NVIDIA influenced DX12U features.
Take a very good look at this.
How in the world can the PS5 be closer to a 5700XT?
CU are not the same,
ALU are not the same, and
ROPs are not the same.
PS5 components are laid out the same way as Navi 10, but when looking closely, XBOX CUs and ALU looks the same as Navi 14.
If their was a 6700 (not XT), it would be a PS5.
RDNA2 ROPs are there for compatibility with DirectX VRS. PS5 does not use DirectX API, so why implement RB+ ROPs?
PS5 has their own VRS solution.
"Using primitive shaders on PlayStation 5 will allow for a broad variety of techniques including smoothly varying level of detail addition of procedural detail to close up objects and improvements to particle effects and other visual special effects."
Yes according to their patent Sony seem to have another strategy with VRS (particularly for VR purpose). Instead of reducing the resolution of (the ideally) less visible parts of the image they want to increase the resolution of the most visible parts of the image and render only the visible polygons (thanks to their geometry engine).
You end up with varying levels of quality either way.
As for Machine Learning, this link says it all.
PlayStation 5 and Machine Learning: An Analysis
Being based on RDNA 2 means your using RDNA2 instruction set.
https://developer.amd.com › ...PDF Web results "RDNA 2" Instruction Set Architecture: Reference Guide
Ray Tracing support includes the following instructions:
-IMAGE_BVH_INTERSECT_RAY
-IMAGE_BVH64_INTERSECT_RAY
Dot product ALU operations added accelerate inferencing and deep-learning:
V_DOT2_F32_F16 / V_DOT2C_F32_F16
V_DOT2_I32_I16 / V_DOT2_U32_U16
V_DOT4_I32_I8 / V_DOT4C_I32_I8
V_DOT4_U32_U8
V_DOT8_I32_I4
V_DOT8_U32_U4
Edit:
As for Infinity Cache.
You should read this.
Infinity Cache for APU? AMD is probably planning something better
Infinity Cache might not have been ready for APUs.
Last edited:
Sosokrates
Report me if I continue to console war
Take a very good look at this.
How in the world can the PS5 be closer to a 5700XT?
CU are not the same,
ALU are not the same, and
ROPs are not the same.
PS5 components are laid out the same way as Navi 10, but when looking closely, XBOX CUs and ALU looks the same as Navi 14.
If they was a 6700 (not XT), it would be a PS5.
RDNA2 ROPs are there for compatibility with DirectX VRS. PS5 does not use DirectX API, so why implement RB+ ROPs?
PS5 has their own VRS solution.
"Using primitive shaders on PlayStation 5 will allow for a broad variety of techniques including smoothly varying level of detail addition of procedural detail to close up objects and improvements to particle effects and other visual special effects."
Yes according to their patent Sony seem to have another strategy with VRS (particularly for VR purpose). Instead of reducing the resolution of (the ideally) less visible parts of the image they want to increase the resolution of the most visible parts of the image and render only the visible polygons (thanks to their geometry engine).
You end up with varying levels of quality either way.
As for Machine Learning, this link says it all.
PlayStation 5 and Machine Learning: An Analysis
Being based on RDNA 2 means your using RDNA2 instruction set.
https://developer.amd.com › ...PDF Web results "RDNA 2" Instruction Set Architecture: Reference Guide
Ray Tracing support includes the following instructions:
-IMAGE_BVH_INTERSECT_RAY
-IMAGE_BVH64_INTERSECT_RAY
Dot product ALU operations added accelerate inferencing and deep-learning:
V_DOT2_F32_F16 / V_DOT2C_F32_F16
V_DOT2_I32_I16 / V_DOT2_U32_U16
V_DOT4_I32_I8 / V_DOT4C_I32_I8
V_DOT4_U32_U8
V_DOT8_I32_I4
V_DOT8_U32_U4
O onQ123
PS5 having more ROPS does not equal a performance advantage, the reason it has double the ROPS is because it uses the older RDNA1 RB's.
XsX uses RDNA2 RB+ which doubles the output per unit, so only half is needed.
Comparing the PS5 and Xbox Series X SoCs in Detail: Area, Similarity and Differences
The PlayStation 5 and the Series X consoles launched last year, with largely similar underlying CPUs and GPUs. While the former features fewer Compute Units (36) in its GPU, it had higher adaptive clocks (2.33 GHz peak). On the other hand, the Series X has more ALUs (52), its operating clocks...
The PS5 has four RBs per Shader Array (like RDNA 1), with a total of 72 on the entire SoC (only 64 are functional). Each RB has four color ROPs and sixteen depth ROPs, once again much like RDNA 1.
The Series X, on the other hand, uses the RB+ design seen on RDNA 2 which doubles the throughput from four to eight 32-bit pixels per cycle, along with sixteen depth samples, in addition to advanced H/W VRS support. The PS5 relies on a software/API variant of VRS. The use of the RB+ design has allowed MS to reduce the area consumption of the RBs in half compared to the PS5, although the real-world benefits will vary from application to application.