• 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.
  • The Politics forum has been nuked. Please do not bring political discussion to the rest of the site, or you will be removed. Thanks.

Analysis Hardware [DF] Can the Slowest PS5 SSD Upgrade Run Ratchet and Clank: Rift Apart?

Loxus

Member
Sep 18, 2020
722
2,947
345
The Caribbean
You can't be this ignorant. Sorry for bothering you. Enjoy your bliss
Mods, I don't appreciate someone calling me dumb just because he doesn't agree with what I've said.

It starts with a small insult, then evolves and ruin thread.
 
  • Like
Reactions: chilichote

phil_t98

Member
Oct 10, 2014
5,138
5,511
735
I wasn't around back then, so I don't really know much about that discussion.

For all we know, we both could be right about our views about Unreal Engine 5. Kinda wish PC had some real realistic games. Not what we have now.

you need to go back and read the thread to be honest, a lot of misinformation back then because Tim said only possible because of the advancement of there PS5 SSD
 
  • Like
Reactions: MonarchJT

Snake29

Member
Jul 30, 2013
2,536
4,132
725
The Netherlands
Did you not see the unreal 5 demo thread? Plenty said it wouldn’t run on other hardware because of the PS5 SSD, again we dont know how much data was being transfers at any one time. We do know the Xbox has more grunt than the ps5 but the ps5 has the faster I/o SSD, now if the game is only streaming 3gbs then there is no reason it couldn’t run better on the series x. If its streaming at a full 5.5gbs then the Xbox may not be able to keep up

I mean, it's not the speed alone that makes a difference between the 2 consoles. The XSX doesn't have the same I/O complex as the PS5. You assume, because Rift Apart doesn't use more than 4GB/s that it would naturally run better on the XSX? What are you basing that on? In other words, every launched game for the PS5 right now will always run better on the XSX?

It doesn't make any sense what you're saying here.
 
Last edited:

Loxus

Member
Sep 18, 2020
722
2,947
345
The Caribbean
you need to go back and read the thread to be honest, a lot of misinformation back then because Tim said only possible because of the advancement of there PS5 SSD
Well I don't think Tim Sweeney was wrong for what he said about it only be possible on PS5.

Think about it like this.
PS5 only has about 13.5GB GDDR available for games while PC has (depending on GPU) about 10GB GDDR + about 24/36GB DDR.

PS5 doesn't has the extra 24/36 DDR like PC to keep stuff that maybe used.

So to work around having more RAM for that purpose, Cerny made dedicated hardware to handle stuff rapidly coming of the SSD that would otherwise cripple a Zen 2 CPU.

So what Tim was saying is if you took a game that is design around the PS5s architecture and put it on a PC. It would cripple the CPU trying to get stuff straight into VRAM. It how the game code works that makes it not possible on PC, not how powerful the GPU is.
 

phil_t98

Member
Oct 10, 2014
5,138
5,511
735
I mean, it's not the speed alone that makes a difference between the 2 consoles. The XSX doesn't have the same I/O complex as the PS5. You assume, because Rift Apart doesn't use more than 4GB/s that it would naturally run better on the XSX? What are you basing that on? In other words, every launched game for the PS5 right now will always run better on the XSX?

It doesn't make any sense what you're saying here.

first up I was talking about the unreal 5 demo not rift apart, we don't know how much data was being streamed on the unreal 5 demo. I said IF it was about 3gbs there was no reason why it couldn't be run better on the series X. which is a bit more powerful
 
Last edited:

phil_t98

Member
Oct 10, 2014
5,138
5,511
735
Well I don't think Tim Sweeney was wrong for what he said about it only be possible on PS5.

Think about it like this.
PS5 only has about 13.5GB GDDR available for games while PC has (depending on GPU) about 10GB GDDR + about 24/36GB DDR.

PS5 doesn't has the extra 24/36 DDR like PC to keep stuff that maybe used.

So to work around having more RAM for that purpose, Cerny made dedicated hardware to handle stuff rapidly coming of the SSD that would otherwise cripple a Zen 2 CPU.

So what Tim was saying is if you took a game that is design around the PS5s architecture and put it on a PC. It would cripple the CPU trying to get stuff straight into VRAM. It how the game code works that makes it not possible on PC, not how powerful the GPU is.

but as we seen it is running on pc and its running in the editor mode which people have said is more intensive than just running the demo
 

Loxus

Member
Sep 18, 2020
722
2,947
345
The Caribbean
but as we seen it is running on pc and its running in the editor mode which people have said is more intensive than just running the demo
Running in editor mode doesn't really tell us any.

They could of shown us gameplay with frame rate unlocked instead, but didn't.

No one is questioning why they didn't do the flying sequence on PC to get a like for like comparison.

No seeing the flying sequence on PC is the reason everyone is saying running the demo in gameplay is impossible on PC.

Seeing is believing for many here.
 

Hoddi

Member
Nov 3, 2020
283
403
305
Running in editor mode doesn't really tell us any.

They could of shown us gameplay with frame rate unlocked instead, but didn't.

No one is questioning why they didn't do the flying sequence on PC to get a like for like comparison.

No seeing the flying sequence on PC is the reason everyone is saying running the demo in gameplay is impossible on PC.

Seeing is believing for many here.
FWIW, it's easy enough to show the amount of reads in the PC demo. Here's a video below that shows how it only reads ~9GB in total and includes all the important metrics such as MB/s and the total amount of reads as they come off the disk. You'll find both of these following the 660p heading:


Considering that the full playthrough is slightly below 3 minutes then that only gives it an average read rate of about 50MB/s. Even the peak is only about 500MB/s and that's during a 'dimensional shift'.
 
Last edited:
Oct 26, 2018
22,739
31,950
885
Well I don't think Tim Sweeney was wrong for what he said about it only be possible on PS5.

Think about it like this.
PS5 only has about 13.5GB GDDR available for games while PC has (depending on GPU) about 10GB GDDR + about 24/36GB DDR.

PS5 doesn't has the extra 24/36 DDR like PC to keep stuff that maybe used.

So to work around having more RAM for that purpose, Cerny made dedicated hardware to handle stuff rapidly coming of the SSD that would otherwise cripple a Zen 2 CPU.

So what Tim was saying is if you took a game that is design around the PS5s architecture and put it on a PC. It would cripple the CPU trying to get stuff straight into VRAM. It how the game code works that makes it not possible on PC, not how powerful the GPU is.
PC gamers have 24-36gb system ram?
 

staticshock

Member
Jun 24, 2020
100
339
340
FWIW, it's easy enough to show the amount of reads in the PC demo. Here's a video below that shows how it only reads ~9GB in total and includes all the important metrics such as MB/s and the total amount of reads as they come off the disk. You'll find both of these following the 660p heading:


Considering that the full playthrough is slightly below 3 minutes then that only gives it an average read rate of about 50MB/s. Even the peak is only about 500MB/s and that's during a 'dimensional shift'.
Please stop using logic. Loxus Loxus and Snake29 Snake29 are allergic to that.The valley demo using 3 GB system ram and 7 GB VRAM?
The valley demo having slightly less Nanite data and more texture data according to the developers? No way. We can't listen to the engineers.
Nanite creator literally posts the memory stats on his Siggraph presentation? No way we can't listen to the guy who created Nanite.



No no no no no the PS5 demo couldn't run on a pc using just 3GB system ram and 7GB VRAM.
Impossible, every one is LYING!!!! You need a hyper speed SSD and god tier IO or you need 64 GB Ram.
 
Last edited:

phil_t98

Member
Oct 10, 2014
5,138
5,511
735
Please stop using logic. Loxus Loxus and Snake29 Snake29 are allergic to that.The valley demo using 3 GB system ram and 7 GB VRAM?
The valley demo having slightly less Nanite data and more texture data according to the developers? No way. We can't listen to the engineers.
Nanite creator literally posts the memory stats on his Siggraph presentation? No way we can't listen to the guy who created Nanite.



No no no no no the PS5 demo couldn't run on a pc using just 3GB system ram and 7GB VRAM.
Impossible, every one is LYING!!!! You need a hyper speed SSD and god tier IO or you need 64 GB Ram.

exactly what I been saying, while the pc may not have the I/O the same as PS5 which is super impressive don't get me wrong they can handle it slightly differently with more system ram and Video ram
 

Loxus

Member
Sep 18, 2020
722
2,947
345
The Caribbean
FWIW, it's easy enough to show the amount of reads in the PC demo. Here's a video below that shows how it only reads ~9GB in total and includes all the important metrics such as MB/s and the total amount of reads as they come off the disk. You'll find both of these following the 660p heading:


Considering that the full playthrough is slightly below 3 minutes then that only gives it an average read rate of about 50MB/s. Even the peak is only about 500MB/s and that's during a 'dimensional shift'.
I don't understand what your trying to show me, but my point was about RAM usage.


Take a look at the video and watch the RAM usage.
Dedicated VRAM: 5.3 GB
Allocated VRAM: 6.8 GB
System RAM: 20.2 GB

VRAM Allocation vs. VRAM Usage – What is the Difference?
Dedicated VRAM is the amount of VRAM that is actively involved in rendering a scene that is being displayed on the screen or any subsequent scenes for which the game needs immediate access to assets and textures.

Allocated VRAM is basically, when a game claims the entirety of the card's memory buffer and stores assets and textures on it that it might be needed later.

On System RAM, games read and write data to RAM because it's orders of magnitudes faster than accessing data on a storage device.

On PS5 it doesn't have the luxury of having all that RAM to store assets and textures that it might need later.

So to solve this issue, Mark Cerny rearchitected the way SSDs work.




Transcribe - The Road to PS5 - Mark Cerny's a deep dive into the PlayStation 5 -
On PlayStation 5 though the SSD is very close to being like more RAM.

f:id:keepitreal:20200329140011j:plain


Typically it's fast enough that when you realize you need a piece of data you can just load it from the SSD and use it there's no need to have lots of data parked in system memory waiting to potentially be used.

A different way of saying that is that most of Ram is working on the game's behalf.



Tim Sweeney Explains Exactly Why the PS5’s SSD and I/O Architecture Is Way More Efficient Than PC’s
On PS5
Systems integration and whole-system performance. Bringing in data from high-bandwidth storage into video memory in its native format with hardware decompression is very efficient. The software and hardware stack go to great lengths to minimize latency and maximize the bandwidth that's actually accessible by games.

While on PC


Those PC numbers are theoretical and are from drive into kernel memory. From there, it's a slow and circuitous journey through software decompression to GPU driver swizzling into video memory where you can eventually use it. The PS5 path for this is several times more efficient. And then there's latency.

On PC, there's a lot of layering and overhead. Then you have the issue of getting compressed textures into video memory requires reading into RAM, software decompressing, then calling into a GPU driver to transfer and swizzle them, with numerous kernel transitions throughout.


So in short, PC currently can't get assets and textures directly from storage into VRAM as fast as the PS5.

Also PC SSDs aren't low latency/high bandwidth like the PS5's SSD with the help of the decompression I/O. The low latency/high bandwidth allows the games to have access to data already stored on the SSD similar to how games access stuff stored in RAM, not that data is copied somewhere esle on the SSD to be used.

So a game design around the PS5 storage architecture can not run on PC. Not because of GPU performance or SSD speed but SSD low latency/high bandwidth performance.
The game would have to have two versions, with the other version designed the old way.
 
Last edited:

Snake29

Member
Jul 30, 2013
2,536
4,132
725
The Netherlands
Please stop using logic. Loxus Loxus and Snake29 Snake29 are allergic to that.The valley demo using 3 GB system ram and 7 GB VRAM?
The valley demo having slightly less Nanite data and more texture data according to the developers? No way. We can't listen to the engineers.
Nanite creator literally posts the memory stats on his Siggraph presentation? No way we can't listen to the guy who created Nanite.



No no no no no the PS5 demo couldn't run on a pc using just 3GB system ram and 7GB VRAM.
Impossible, every one is LYING!!!! You need a hyper speed SSD and god tier IO or you need 64 GB Ram.

Jonah Hill Snl GIF by Saturday Night Live


Go do something useful with your life instead of worrying so much. Go worry in the UE5 thread if it's hurting you so much.

exactly what I been saying, while the pc may not have the I/O the same as PS5 which is super impressive don't get me wrong they can handle it slightly differently with more system ram and Video ram

You never have tried to say anything. You kept going from one subject to another. With only more ram and Vram, it doesn't reduce the latency at all, and the CPU still needs do all that work and thats the bottleneck right here. Nobody is says that it CAN'T run, but HOW compared to the PS5 dedicated chip...we don't now yet and 1 test system doesn't claim others can run it at with the same performance because there is no 1 pc fixed system.
 
Last edited:

Loxus

Member
Sep 18, 2020
722
2,947
345
The Caribbean
exactly what I been saying, while the pc may not have the I/O the same as PS5 which is super impressive don't get me wrong they can handle it slightly differently with more system ram and Video ram
Imagine buying a GPU with 64 GB of VRAM.
System RAM is still going to be a bottleneck because it isn't as fast as GDDR RAM, so you still have to get the data into VRAM. Which will hurt CPU performance when trying to decompress at those streaming speeds. That's why the PS5 has dedicated hardware for that purpose.



It's either RTX I/O or a GPU with a large amount of VRAM to allocate data to help with high speed streaming of in-game data.

More system RAM isn't going to cut it at those speeds and a 64 GPU would be insanely expensive.

Very high GDDR6 costs blamed for such high GPU prices
 
Oct 26, 2018
22,739
31,950
885
It's just an example.
8, 16, 32, 64GB.
If your running with a 3090, you might as well have 32-64 GB of RAM.
Most PC gamers on here seem to have high end gaming PCs.

What GPU and RAM setup do you have?
I'm on a 3 year old laptop. I'm not a core PC gamer. Was just asking since youre comparing PS5 to a PC and going to the extreme end of PC gaming at 3090 at 32/64 gb system ram.
 
  • Like
Reactions: Loxus

Kenpachii

Member
Mar 23, 2018
9,468
11,770
815
I don't understand what your trying to show me, but my point was about RAM usage.


Take a look at the video and watch the RAM usage.
Dedicated VRAM: 5.3 GB
Allocated VRAM: 6.8 GB
System RAM: 20.2 GB

VRAM Allocation vs. VRAM Usage – What is the Difference?
Dedicated VRAM is the amount of VRAM that is actively involved in rendering a scene that is being displayed on the screen or any subsequent scenes for which the game needs immediate access to assets and textures.

Allocated VRAM is basically, when a game claims the entirety of the card's memory buffer and stores assets and textures on it that it might be needed later.

On System RAM, games read and write data to RAM because it's orders of magnitudes faster than accessing data on a storage device.

On PS5 it doesn't have the luxury of having all that RAM to store assets and textures that it might need later.

So to solve this issue, Mark Cerny rearchitected the way SSDs work.




Transcribe - The Road to PS5 - Mark Cerny's a deep dive into the PlayStation 5 -
On PlayStation 5 though the SSD is very close to being like more RAM.

f:id:keepitreal:20200329140011j:plain


Typically it's fast enough that when you realize you need a piece of data you can just load it from the SSD and use it there's no need to have lots of data parked in system memory waiting to potentially be used.

A different way of saying that is that most of Ram is working on the game's behalf.



Tim Sweeney Explains Exactly Why the PS5’s SSD and I/O Architecture Is Way More Efficient Than PC’s
On PS5
Systems integration and whole-system performance. Bringing in data from high-bandwidth storage into video memory in its native format with hardware decompression is very efficient. The software and hardware stack go to great lengths to minimize latency and maximize the bandwidth that's actually accessible by games.

While on PC


Those PC numbers are theoretical and are from drive into kernel memory. From there, it's a slow and circuitous journey through software decompression to GPU driver swizzling into video memory where you can eventually use it. The PS5 path for this is several times more efficient. And then there's latency.

On PC, there's a lot of layering and overhead. Then you have the issue of getting compressed textures into video memory requires reading into RAM, software decompressing, then calling into a GPU driver to transfer and swizzle them, with numerous kernel transitions throughout.


So in short, PC currently can't get assets and textures directly from storage into VRAM as fast as the PS5.

Also PC SSDs aren't low latency/high bandwidth like the PS5's SSD with the help of the decompression I/O. The low latency/high bandwidth allows the games to have access to data already stored on the SSD similar to how games access stuff stored in RAM, not that data is copied somewhere esle on the SSD to be used.

So a game design around the PS5 storage architecture can not run on PC. Not because of GPU performance or SSD speed but SSD low latency/high bandwidth performance.
The game would have to have two versions, with the other version designed the old way.

It can because PC has far larger memory pools so it doesn't have to address the storage device constantly to feed game data.

This is why u see EU5 run a demo that's less optimized and more demanding while running on potato ssd's perfectly fine.
 
Last edited:

Hoddi

Member
Nov 3, 2020
283
403
305
I don't understand what your trying to show me, but my point was about RAM usage.


Take a look at the video and watch the RAM usage.
Dedicated VRAM: 5.3 GB
Allocated VRAM: 6.8 GB
System RAM: 20.2 GB

VRAM Allocation vs. VRAM Usage – What is the Difference?
Dedicated VRAM is the amount of VRAM that is actively involved in rendering a scene that is being displayed on the screen or any subsequent scenes for which the game needs immediate access to assets and textures.

Allocated VRAM is basically, when a game claims the entirety of the card's memory buffer and stores assets and textures on it that it might be needed later.

On System RAM, games read and write data to RAM because it's orders of magnitudes faster than accessing data on a storage device.

On PS5 it doesn't have the luxury of having all that RAM to store assets and textures that it might need later.

So to solve this issue, Mark Cerny rearchitected the way SSDs work.




Transcribe - The Road to PS5 - Mark Cerny's a deep dive into the PlayStation 5 -
On PlayStation 5 though the SSD is very close to being like more RAM.

f:id:keepitreal:20200329140011j:plain


Typically it's fast enough that when you realize you need a piece of data you can just load it from the SSD and use it there's no need to have lots of data parked in system memory waiting to potentially be used.

A different way of saying that is that most of Ram is working on the game's behalf.



Tim Sweeney Explains Exactly Why the PS5’s SSD and I/O Architecture Is Way More Efficient Than PC’s
On PS5
Systems integration and whole-system performance. Bringing in data from high-bandwidth storage into video memory in its native format with hardware decompression is very efficient. The software and hardware stack go to great lengths to minimize latency and maximize the bandwidth that's actually accessible by games.

While on PC


Those PC numbers are theoretical and are from drive into kernel memory. From there, it's a slow and circuitous journey through software decompression to GPU driver swizzling into video memory where you can eventually use it. The PS5 path for this is several times more efficient. And then there's latency.

On PC, there's a lot of layering and overhead. Then you have the issue of getting compressed textures into video memory requires reading into RAM, software decompressing, then calling into a GPU driver to transfer and swizzle them, with numerous kernel transitions throughout.


So in short, PC currently can't get assets and textures directly from storage into VRAM as fast as the PS5.

Also PC SSDs aren't low latency/high bandwidth like the PS5's SSD with the help of the decompression I/O. The low latency/high bandwidth allows the games to have access to data already stored on the SSD similar to how games access stuff stored in RAM, not that data is copied somewhere esle on the SSD to be used.

So a game design around the PS5 storage architecture can not run on PC. Not because of GPU performance or SSD speed but SSD low latency/high bandwidth performance.
The game would have to have two versions, with the other version designed the old way.
Here I've added amount of system RAM in use by the demo. The first number ('MEM') is the amount of VRAM consumed and the latter ('RAM') is the amount of system RAM consumed. This removes any editors or browser windows from the RAM count and shows how much memory the demo itself is using.

1

2

This demo only uses up to about 6.5GB of VRAM (at 1080p) and then another 3-4GB of system RAM.
 
Last edited:
Jun 1, 2016
2,865
3,817
795
Here I've added amount of system RAM in use by the demo. The first number ('MEM') is the amount of VRAM consumed and the latter ('RAM') is the amount of system RAM consumed. This removes any editors or browser windows from the RAM count and shows how much memory the demo itself is using.

1

2

This demo only uses up to about 6.5GB of VRAM (at 1080p) and then another 3-4GB of system RAM.
I ran that compiled fine with a 3060ti and 16gb Ram at 1440p. In editor though I would run out of Ram and it was crash.
 
Last edited:
  • Thoughtful
Reactions: DonJuanSchlong

Loxus

Member
Sep 18, 2020
722
2,947
345
The Caribbean
It can because PC has far larger memory pools so it doesn't have to address the storage device constantly to feed game data.

This is why u see EU5 run a demo that's less optimized and more demanding while running on potato ssd's perfectly fine.
Clearly you don't get it do you.
In-game movement speed is another advantage PS5 will have.

PC solution for movement speed is to have more data in VRAM not system RAM. System RAM would be to slow and the more VRAM your GPU has, the more expensive it's going to be. 3080 10GB of VRAM is not going to be enough and not everyone can afford a 3090 let alone a GPU with 32-64 GB VRAM.

Take a look at this video to get a better understanding of what I mean.

On PC you will have more assets and textures in VRAM than usual base on distance. Those assets and textures have to be insanely ready when traveling at those speeds through a city or you would get manger hitches.

On PS5 your streaming everything directly from low latency/high bandwidth storage without the need to have large amounts of RAM.
 

Hoddi

Member
Nov 3, 2020
283
403
305
Clearly you don't get it do you.
In-game movement speed is another advantage PS5 will have.

PC solution for movement speed is to have more data in VRAM not system RAM. System RAM would be to slow and the more VRAM your GPU has, the more expensive it's going to be. 3080 10GB of VRAM is not going to be enough and not everyone can afford a 3090 let alone a GPU with 32-64 GB VRAM.

Take a look at this video to get a better understanding of what I mean.

On PC you will have more assets and textures in VRAM than usual base on distance. Those assets and textures have to be insanely ready when traveling at those speeds through a city or you would get manger hitches.

On PS5 your streaming everything directly from low latency/high bandwidth storage without the need to have large amounts of RAM.
Sorry, I thought you were talking about the UE5 demo.
 

Loxus

Member
Sep 18, 2020
722
2,947
345
The Caribbean
Here I've added amount of system RAM in use by the demo. The first number ('MEM') is the amount of VRAM consumed and the latter ('RAM') is the amount of system RAM consumed. This removes any editors or browser windows from the RAM count and shows how much memory the demo itself is using.

1

2

This demo only uses up to about 6.5GB of VRAM (at 1080p) and then another 3-4GB of system RAM.
Did you not watch the video and read what I posted?

You guys only ever talk about the amount of VRAM that is actively involved in rendering a scene.

What about the assets and textures stored just in case it's needed?

Or how is the PS5 displaying the same amount of fidelity with only 13.5 GB of available RAM.
 

staticshock

Member
Jun 24, 2020
100
339
340
Did you not watch the video and read what I posted?

You guys only ever talk about the amount of VRAM that is actively involved in rendering a scene.

What about the assets and textures stored just in case it's needed?

Or how is the PS5 displaying the same amount of fidelity with only 13.5 GB of available RAM.

What are you talking about?
Allocated VRAM is the amount of VRAM you have reserved for a particular application and in this case the game.
Used VRAM is the amount of VRAM you are actively using out of the reserved VRAM.

So if you allocate 7 GB and only use 5 GB. That doesn't come out to be 12.
It means you only used 5 GB out of the 7 you reserved.
IF you had 7 cookies and only ate 5. That doesn't mean you had 12 cookies.

Like i said and others have CONFIRMED. The compiled packaged version of the Valley Demo. You know the .exe.
Uses only ~7GB ram and 3GB System ram, way less than PS5's 13.5 GB of available RAM.

This memory usage is also confirmed by DF. But keep spreading misinformation.
 
Last edited:

staticshock

Member
Jun 24, 2020
100
339
340
Jonah Hill Snl GIF by Saturday Night Live


Go do something useful with your life instead of worrying so much. Go worry in the UE5 thread if it's hurting you so much.



You never have tried to say anything. You kept going from one subject to another. With only more ram and Vram, it doesn't reduce the latency at all, and the CPU still needs do all that work and thats the bottleneck right here. Nobody is says that it CAN'T run, but HOW compared to the PS5 dedicated chip...we don't now yet and 1 test system doesn't claim others can run it at with the same performance because there is no 1 pc fixed system.
Since UE engineers say Valley's nanite performance cost is 2x more than the PS5 demo. that means a pc will run the ps5 demo at double the resolution and fps than the valley demo.
 
Last edited by a moderator:

Snake29

Member
Jul 30, 2013
2,536
4,132
725
The Netherlands
Since UE engineers say Valley's nanite performance cost is 2x more than the PS5 demo. that means a pc will run the ps5 demo at double the resolution and fps than the valley demo.

Nanite is just one part of the whole UE5 engine, how does that make you feel? And the whole demo was totally different from the PS5 demo, still doesn't say anything in the end. As long as we have not seen the gameplay demo from the PS5 running on pc, we don't know the difference of the performance of the last part of the demo.

It's not like a SATA SSD will run it any better then the PS5, so the majority of the PC users won't experience the same......
 

Shmunter

Member
Aug 25, 2018
10,784
24,453
835
Imagine buying a GPU with 64 GB of VRAM.
System RAM is still going to be a bottleneck because it isn't as fast as GDDR RAM, so you still have to get the data into VRAM. Which will hurt CPU performance when trying to decompress at those streaming speeds. That's why the PS5 has dedicated hardware for that purpose.



It's either RTX I/O or a GPU with a large amount of VRAM to allocate data to help with high speed streaming of in-game data.

More system RAM isn't going to cut it at those speeds and a 64 GPU would be insanely expensive.

Very high GDDR6 costs blamed for such high GPU prices
Interestingly that diagram seems to indicate decompression is done during asset loading, no need for a vram scratchpad. That would be ideal, unless the diagram is simplified.
 

Loxus

Member
Sep 18, 2020
722
2,947
345
The Caribbean
Interestingly that diagram seems to indicate decompression is done during asset loading, no need for a vram scratchpad. That would be ideal, unless the diagram is simplified.
I was thinking the same thing. Games may not need to used System Ram either if the diagram is to be believed.

Tim Sweeney had a good explanation about the old way.

Those PC numbers are theoretical and are from drive into kernel memory. From there, it's a slow and circuitous journey through software decompression to GPU driver swizzling into video memory where you can eventually use it.

On PC, there's a lot of layering and overhead. Then you have the issue of getting compressed textures into video memory requires reading into RAM, software decompressing, then calling into a GPU driver to transfer and swizzle them, with numerous kernel transitions throughout.


So my guess it's normally,
Storage --> RAM --> Software Decompression --> GPU Driver --> VRAM

RTX I/O most likely,
Storage --> GPU Decompression --> GPU Driver --> VRAM

While on PS5 it's most likely,
Storage --> Dedicated Hardware Decompression --> RAM

Tim Sweeney also said this about the PS5.

Bringing in data from high-bandwidth storage into video memory in its native format with hardware decompression is very efficient.

So my guess it will be the same for RTX I/O.
 
Last edited:

Shmunter

Member
Aug 25, 2018
10,784
24,453
835
I was thinking the same thing. Games may not need to used System Ram either if the diagram is to be believed.

Tim Sweeney had a good explanation about the old way.

Those PC numbers are theoretical and are from drive into kernel memory. From there, it's a slow and circuitous journey through software decompression to GPU driver swizzling into video memory where you can eventually use it.

On PC, there's a lot of layering and overhead. Then you have the issue of getting compressed textures into video memory requires reading into RAM, software decompressing, then calling into a GPU driver to transfer and swizzle them, with numerous kernel transitions throughout.


So my guess it's normally,
Storage --> RAM --> Software Decompression --> GPU Driver --> VRAM

RTX I/O most likely,
Storage --> GPU Decompression --> GPU Driver --> VRAM

While on PS5 it's most likely,
Storage --> Dedicated Hardware Decompression --> RAM

Tim Sweeney also said this about the PS5.

Bringing in data from high-bandwidth storage into video memory in its native format with hardware decompression is very efficient.

So my guess it will be the same for RTX I/O.
Agreed. The left part of the diagram clearly shows cpu loading asset into ram, then working on decompressing the loaded asset back into ram uncompressed, then sending it to the GPU to load into vram. Lots of steps, lots of latency and cpu load.

The right side shows GPU loading asset directly into vram and we can assume it is decompressing it at the same time when contrast to the left part of the diagram.

Such a single operation does mirror the ps5’s overall movement of assets, bar the PS5 shared ram pool advantage where non GPU assets can also be decompressed for cpu usage, and no extra load on GPU needed for decompression.

Eagerly awaiting on how it plays out in practice and how GPU load / latency is effected. Can’t help but wonder why Xbox and PS5 didn’t just use the GPU if it were an ideal solution, certainly much cheaper bom than having custom I/o.
 

Loxus

Member
Sep 18, 2020
722
2,947
345
The Caribbean
Agreed. The left part of the diagram clearly shows cpu loading asset into ram, then working on decompressing the loaded asset back into ram uncompressed, then sending it to the GPU to load into vram. Lots of steps, lots of latency and cpu load.

The right side shows GPU loading asset directly into vram and we can assume it is decompressing it at the same time when contrast to the left part of the diagram.

Such a single operation does mirror the ps5’s overall movement of assets, bar the PS5 shared ram pool advantage where non GPU assets can also be decompressed for cpu usage, and no extra load on GPU needed for decompression.

Eagerly awaiting on how it plays out in practice and how GPU load / latency is effected. Can’t help but wonder why Xbox and PS5 didn’t just use the GPU if it were an ideal solution, certainly much cheaper bom than having custom I/o.
Holy smoke, I think you just shed some major light on this.

This video explains it all.

This next videos goes into detail about how the CPU bottlenecks the SSD.

This explains how Ratchet can run on a slower SSD. No CPU bottlenecks and the SSD currently is enough. In the future games may require a faster SSD.

If the UE5 PS5 demo is design this way, I can't see it running on PC without RTX I/O. No matter how much RAM you have, CPU will struggle when that massive stream of data starts coming off the SSD and the frame rate will cripple as soon as the player moves.
 
Last edited:

Topher

Gold Member
Sep 25, 2020
4,240
13,330
645
Atlanta, Georgia, USA
Agreed. The left part of the diagram clearly shows cpu loading asset into ram, then working on decompressing the loaded asset back into ram uncompressed, then sending it to the GPU to load into vram. Lots of steps, lots of latency and cpu load.

The right side shows GPU loading asset directly into vram and we can assume it is decompressing it at the same time when contrast to the left part of the diagram.

Such a single operation does mirror the ps5’s overall movement of assets, bar the PS5 shared ram pool advantage where non GPU assets can also be decompressed for cpu usage, and no extra load on GPU needed for decompression.

Eagerly awaiting on how it plays out in practice and how GPU load / latency is effected. Can’t help but wonder why Xbox and PS5 didn’t just use the GPU if it were an ideal solution, certainly much cheaper bom than having custom I/o.

Does AMD have anything like RTX IO in the works?
 

staticshock

Member
Jun 24, 2020
100
339
340
Nanite is just one part of the whole UE5 engine,
Nanite and Lumen are the *only new feature in UE5.
Nanite literally is the only feature that deals with IO/SSD.
Everything else would run better on a PC even if the PC had a hard drive from the 2000s.

So if Nanite is 2x more costly in the new demo. That means you would be able to run that demo at 2x the resolution/fps.
We already know how nanite works. You fans refuse to learn how nanite works. Its like a kid refusing to go to school to learn basic math.
All the info on nanite is available. You refuse to learn how it works. You think its pulling Gb of data per frame from the SSD. Its not. The SSD is BARELY TOUCHED.
The CREATOR OF NANITE (Brian Karis) HAS TOLD YOU THAT.

You are literally arguing with THE CREATOR OF NANITE. Like WTF?
how does that make you feel? And the whole demo was totally different from the PS5 demo
No its the same.
Same chaos physics, same Niagara particles, same character. same character animation system, same animations.
The beetles and birds are literally available for download.
Literally the same animation system and animations. Notice how everything i say is backed up by statements from UE5 Engineers, Nanite creator or developers of both demos.
You on the other hand rebut with absolute bs.




Again if you think this beatle is monumental then drag it into the valley demo, if your theory is right, the FPS should tank and the game should crash immediately afterwards.



, still doesn't say anything in the end.
The CREATOR OF NANITE already commented about the end "fly up to show this vista because i don't want anyone to get the impression that i skipped something on purpose because it doesn't work. This vista still works!"

Literally the entire nanite data required for that vista IS ALREADY IN MEMORY.
Go learn about instance meshes.
As long as we have not seen the gameplay demo from the PS5 running on pc, we don't know the difference of the performance of the last part of the demo.
There's literally no difference in nanite between gameplay in editor versus the fly through he is doing.
Brian Karis the CREATOR OF NANITE already debunked this. He had nanite stats up the entire time to the right, showing it was crunching all that data into ~20 million triangles.
He played through the gameplay ON HIS PC and said it in the livestream. He made statements that would make him a liar if the gameplay doesn't work on PC.
But you ofcourse can't allow anything to tarnish the reputation of your god tier ssd & i/o
It's not like a SATA SSD will run it any better then the PS5, so the majority of the PC users won't experience the same......
The new demo begs to differ. Runs 4k 30/60 fps on PC but 1080p 30fps on PS5.
Why? Because nanite in the new demo is 2x costlier and nanite scales with resolution.
This is basic math.
 
Last edited:

Md Ray

Member
Nov 12, 2016
4,138
13,345
785
Does AMD have anything like RTX IO in the works?
We haven't heard anything from AMD about their version of RTX IO.

But whatever it is it will likely use DirectCompute API to handle the decompression tasks on CUs.
 
  • Like
Reactions: Topher

Rikkori

Member
May 9, 2020
2,704
5,122
470
Does AMD have anything like RTX IO in the works?
On PC it's all just DirectStorage because it has to be done at an OS/API level, so AMD/Nvidia can't really do it just for themselves, it's a collaboration between all vendors (AMD, Nvidia, Intel; soon ARM ones also) and Microsoft (because Windows). RTX IO is just the branding from Nvidia because they do be smart like that. 🤪
 
Last edited: