• 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.

SquareEnix releases video showcasing Forspoken technologies, including DirectStorage

Bo_Hazem

Banned
Direct Storage is only on the PC version, the PS5 doesn't need it.

Please don't turn everything into fanboy-talk please, lol. I'm objectively talking about Direct Storage here.

It's most likely using a 7.4GB/s SSD resulting in an actually 4.8GB/s with Direct Storage, which is plausible.
 

adamsapple

Or is it just one of Phil's balls in my throat?
Maybe it’s just better to compare the consoles I/O with each other? I know that PC will have a different setup so maybe it wouldn’t be a fair comparison.

Sure but that would be a separate topic, the article used in the OP and post # 2 are only talking about the direct storage API integration for this game, it's the first game on PC to use it.

Got nothing to do with console. Console folks can relax and take a seat for a minute :p
 

Md Ray

Member
The question is whether that requirement of an NVME they were mentioning earlier this year and last was for when Direct Storage transitions to having the GPU handle decompression.
The requirement of an NVMe is because of "NVMe queues" (hardware data access pipes) which, according to MS's first DirectStorage blog post, is particularly well suited to gaming workloads.
To get data off the drive, an OS submits a request to the drive and data is delivered to the app via these queues. An NVMe device can have multiple queues and each queue can contain many requests at a time. This is a perfect match to the parallel and batched nature of modern gaming workloads. The DirectStorage programming model essentially gives developers direct control over that highly optimized hardware.
In addition, existing storage APIs also incur a lot of ‘extra steps’ between an application making an IO request and the request being fulfilled by the storage device, resulting in unnecessary request overhead. These extra steps can be things like data transformations needed during certain parts of normal IO operation. However, these steps aren’t required for every IO request on every NVMe drive on every gaming machine. With a supported NVMe drive and properly configured gaming machine, DirectStorage will be able to detect up front that these extra steps are not required and skip all the necessary checks/operations making every IO request cheaper to fulfill.

For these reasons, NVMe is the storage technology of choice for DirectStorage and high-performance next generation gaming IO.
And also because every Series console comes with NVMe storage. It goes without saying that consoles define the baseline for game development which in turn means NVMe is/will be the baseline.
 
Sure but that would be a separate topic, the article used in the OP and post # 2 are only talking about the direct storage API integration for this game, it's the first game on PC to use it.

Got nothing to do with console. Console folks can relax and take a seat for a minute :p

Nah that's more of a you problem. I know the video was just about Direct Storage on PC. Consoles don't really have a part in this since neither of them use it to my knowledge.

I'm more curious as to how this will work with RTX I/O and if AMD is going to do something similar.
 
RTX IO is something like PS5 IO system/Xbox Velocity Architecture

I believe RTX I/O is just one part of it. You need to combined multiple aspects (like DirectStorage for example) to achieve I/O efficiencies. At least that's how I understand because PCs unlike console are made up of many different parts from different manufacturers. Fixed hardware allows you to take a different approach when it comes to the I/O.
 

M1chl

Currently Gif and Meme Champion
I believe RTX I/O is just one part of it. You need to combined multiple aspects (like DirectStorage for example) to achieve I/O efficiencies. At least that's how I understand because PCs unlike console are made up of many different parts from different manufacturers. Fixed hardware allows you to take a different approach when it comes to the I/O.
It is one part of it, the second part is Direct Storage. On PC there is probably never going to be a fixed HW for decompression, but GPU can do this well (on it's tensor cores, which isn't really utilized without DLSS), so it's good to use them to do the job.
 

Mister Wolf

Member
Sure but that would be a separate topic, the article used in the OP and post # 2 are only talking about the direct storage API integration for this game, it's the first game on PC to use it.

Got nothing to do with console. Console folks can relax and take a seat for a minute :p

I'm certain that once GPU decompression becomes available it will make its way to the to the Xbox. Right now the only decompression hardware accelerated on the Xbox is decompression of textures. The rest is handled the traditional was using its CPU.
 

Utherellus

Member
Strawman as PC is not a fixed box. Nobody said PC would never be able to compete here, but it is a nice proof that consoles raised the bar ;).

Consoles set the medium, a standard for every developer to follow.

PC was already offering SSD-powered ultrasonic planetary travels in Star Citizen, back when players in God of War 2018 were travelling through asset loading corridors to stream a new level of another corridor.



But that was not the point. The point was that the modern, performance-oriented I/O API has arrived on PC, giving similar results as console custom solutions. And it's about to get better with RTX IO
 
Last edited:

Panajev2001a

GAF's Pleasant Genius
Consoles set the medium, a standard for every developer to follow.

PC was offering SSD-powered ultrasonic planetary travels in Star Citizen, while players in God of War 2018 were travelling through asset loading corridors to stream a new level of another corridor.



But that was not the point. The point was that the modern, performance-oriented I/O API has arrived on PC, giving similar results as console custom solutions. And it's about to get better with RTX IO

1 WIP game, Star Citizen out of all, is your point?

Now, the point that thanks to improvements in the Windows 11 I/O stack and DirectStorage the future is rosier and rosier for PC’s well yes I agree with that.
 

Panajev2001a

GAF's Pleasant Genius
I'm certain that once GPU decompression becomes available it will make its way to the to the Xbox. Right now the only decompression hardware accelerated on the Xbox is decompression of textures. The rest is handled the traditional was using its CPU.
True, on PC’s this will put more emphasis on compute, bandwidth, and VRAM allocated to manage this (the consoles being able to stream data on demand and uncompressed in GPU accessible RAM avoids all those three costs).
 

Utherellus

Member
1 WIP game, Star Citizen out of all, is your point?

Now, the point that thanks to improvements in the Windows 11 I/O stack and DirectStorage the future is rosier and rosier for PC’s well yes I agree with that.

Yes, one game, which shows what was possible with Sata SSDs back in 2017 when they introduced huge scale planetary travels.

Speaking of APIs, Digital Foundry actually explained multiple times how SC developers actually cleverly bypassed out-of-date Win32 API bottlenecks to grasp the "fast SSD seamless open world streaming" philosophy 5 years ago, that today became a reality and we just recently started taking it for granted.

Timestamped.
 
Last edited:

ToTTenTranz

Banned
I keep seeing this "freeing CPU resources" argument as reason to adopt GPU decompression.
How valid is this concern though? Are the 8-core Zen2 CPUs in the consoles a bottleneck for anything but a 120FPS target? Same goes for PC gamers with 8-core Zen2 CPUs or better.

I'm not sure how important it is to offload general processing tasks from the CPU to the GPU nowadays, considering the latter is the bottleneck most of the time.


RTX IO is something like PS5 IO system/Xbox Velocity Architecture
RTX IO proposes to use shader processors. It's a software implementation. The PS5 and Series consoles use dedicated, fixed function hardware for decompression.


I'm certain that once GPU decompression becomes available it will make its way to the to the Xbox.
It would be pretty stupid to waste GPU resources on something that is already 100% optimized in the SoC through the presence of dedicated hardware.


Right now the only decompression hardware accelerated on the Xbox is decompression of textures. The rest is handled the traditional was using its CPU.
The Velocity hardware already handles texture BCPack for textures and Zlib for general data.
 

onesvenus

Member
And in small letters, it says 2.0 is in development.
It also says it's using DirectStorage, should we think that's coming to the PS5? 🤦‍♂️

I mean, it's obvious FSR 2 will work on all platforms, switch might even be included, so it will reach PS5. Using this video as any kind of confirmation is really grasping at straws
 

M1chl

Currently Gif and Meme Champion
I keep seeing this "freeing CPU resources" argument as reason to adopt GPU decompression.
How valid is this concern though? Are the 8-core Zen2 CPUs in the consoles a bottleneck for anything but a 120FPS target? Same goes for PC gamers with 8-core Zen2 CPUs or better.

I'm not sure how important it is to offload general processing tasks from the CPU to the GPU nowadays, considering the latter is the bottleneck most of the time.



RTX IO proposes to use shader processors. It's a software implementation. The PS5 and Series consoles use dedicated, fixed function hardware for decompression.



It would be pretty stupid to waste GPU resources on something that is already 100% optimized in the SoC through the presence of dedicated hardware.



The Velocity hardware already handles texture BCPack for textures and Zlib for general data.
RTX IO requires RTX+ card and it's hw based on Tensor Cores (it's written through CUDA, what does it utilise inside the GPU is meaningless in this sense). There is some solution from AMD, which requires only shader model 6+, which is software solution. However it's using GPU. You can't really speed something in better software without using some HW capabilities, because in that case, we would already have better solutions.

And yeah it's technically not the same, but it works the same so for illustration my words are not incorrect.
 

jroc74

Phone reception is more important to me than human rights
Just like that FSR 2.0 is confirmed for PS5 in an actual game.
I didnt check to see if that was a new talking point. Like how FSR 1.0 was used in a game that ran on PS5.

Its cool to talk about what tech is coming where, where its going to be used ....but there is an appreciation for just letting it show up, happen and we found out like this.

It also says it's using DirectStorage, should we think that's coming to the PS5? 🤦‍♂️

I mean, it's obvious FSR 2 will work on all platforms, switch might even be included, so it will reach PS5. Using this video as any kind of confirmation is really grasping at straws
Was it grasping at straws when a game using FSR 1.0 was also found in a game running on a PS5?

Kinda getting de ja vue here...
 
Last edited:
I think Luminous Team needs to hire some strong technical artists.

The particle VFX for FF games are usually the strongest of most games historically, but I'm not sure whether its the tech or the artistry that holding this game back but the particle effects are just nasty grainy awful.
 
RTX IO requires RTX+ card and it's hw based on Tensor Cores (it's written through CUDA, what does it utilise inside the GPU is meaningless in this sense). There is some solution from AMD, which requires only shader model 6+, which is software solution. However it's using GPU. You can't really speed something in better software without using some HW capabilities, because in that case, we would already have better solutions.

And yeah it's technically not the same, but it works the same so for illustration my words are not incorrect.

Aren't tensor cores too low precision for this type of workload?
 

Md Ray

Member
I'm certain that once GPU decompression becomes available it will make its way to the to the Xbox. Right now the only decompression hardware accelerated on the Xbox is decompression of textures. The rest is handled the traditional was using its CPU.
No, GPU decompression isn't coming to Xbox, it's PC only since there's no dedicated ASIC for asset decompression on the PC side. The only other way is to use the CPU.

Xbox Series consoles on the other hand, already have custom decompression HW for asset decompression work, it doesn't require its GPU to do that job.
 

winjer

Gold Member
RTX IO requires RTX+ card and it's hw based on Tensor Cores (it's written through CUDA, what does it utilise inside the GPU is meaningless in this sense). There is some solution from AMD, which requires only shader model 6+, which is software solution. However it's using GPU. You can't really speed something in better software without using some HW capabilities, because in that case, we would already have better solutions.

And yeah it's technically not the same, but it works the same so for illustration my words are not incorrect.

Can't find anything implying that RTX IO uses the tensor cores.
Where did you find that?
 

Tripolygon

Banned
Could very well just be for the PC version for all we know.
You are well within your right to think what you want. It’s obvious to me though.
It also says it's using DirectStorage, should we think that's coming to the PS5? 🤦‍♂️

I mean, it's obvious FSR 2 will work on all platforms, switch might even be included, so it will reach PS5. Using this video as any kind of confirmation is really grasping at straws
Are you incapable of reasoning and simple comprehension?

Fidelity Fx are AMD platform agnostic technologies which FSR, FCAS, Fx Variable Shading, Fx AO, Fx SSR, Fx denoiser, etc falls under. Those are being used on both PC and PS5 in the game.
t3Qi3gx.png



Direct Storage is Microsoft technology being used on PC. Direct Storage is not part of AMD Fidelity Fx. Does that make it clearer?
 
Last edited:

M1chl

Currently Gif and Meme Champion
Can't find anything implying that RTX IO uses the tensor cores.
Where did you find that?
In my testing with pre-alpha. There is some CUDA implementation (well their api), which suggest that you can use only data types which are compatible with Tensor interface. I cannot verify that, but it's more than likely. The documentation states only that it's using their powerful CUDA programming language, but when you work in that, you can spot the pattern.

Granted you can use the inaccuracy of Tensor cores could be used to represent much more complex data, if you throw multiple units at it at once.
 
Last edited:

winjer

Gold Member
In my testing with pre-alpha. There is some CUDA implementation (well their api), which suggest that you can use only data types which are compatible with Tensor interface. I cannot verify that, but it's more than likely. The documentation states only that it's using their powerful CUDA programming language, but when you work in that, you can spot the pattern.

Granted you can use the inaccuracy of Tensor cores could be used to represent much more complex data, if you throw multiple units at it at once.

Have you tried using Nsight or Nvprof, to check if there is any activity on the Tensor Units?
 

SSfox

Member
Game changer. It's so early in this gen and I'm already spoiled by it.

When Cerny first spoke of the PS5 and that the console was built around the drive itself I thought he was talking crazy or just trying to throw MS off.

It's turning out to be one of the biggest features of the new gen.

How could you doubt Mark Cerny though, this guy is to console Hardwares what Bruce Lee was for Martial Arts.
 
Last edited:

vpance

Member
If regular ass Win32 is actually capable of being that fast then I can live with my DirectStoragless Win10 for another few years I guess.
 
Last edited:

Mister Wolf

Member
No, GPU decompression isn't coming to Xbox, it's PC only since there's no dedicated ASIC for asset decompression on the PC side. The only other way is to use the CPU.

Xbox Series consoles on the other hand, already have custom decompression HW for asset decompression work, it doesn't require its GPU to do that job.

You're right. I wonder how much better those decompressors are at doing the job vs the GPU once they have the method available to developers.
 
Last edited:

onesvenus

Member
Was it grasping at straws when a game using FSR 1.0 was also found in a game running on a PS5?
Can you read? I said that it's obvious that it will get to the PS5. What's grasping at straws is reaching the conclusion that it will, based on a video that shows a multiplatform game using FSR 1.0
 

BennyBlanco

aka IMurRIVAL69
Square Enix has been dropping dog shit PC ports for a while now. They no longer have Nixxes to do their dirty work. I wonder why of all games they chose this one to showcase all this new tech.
 

Tygeezy

Member
1.9 seconds is some wild shit!

We are going to see a great deal of open world games this gen and I want to see how all of them use that storage.

Ubisoft out here foaming at the mouth over this tech lol

I also wonder how the next Final Fantasy will use this tech. The way I see this, its not even JUST about saving time, its about deep immersion in being able to go thru a game with near seamless transition from a scene to in game, to a new town, to interior all without any loading keeping you immersed in the game.
This should also remove asset streaming stutter that plagues a lot of games on pc.
 
When Cerny first spoke of the PS5 and that the console was built around the drive itself I thought he was talking crazy or just trying to throw MS off.

It's turning out to be one of the biggest features of the new gen.
And it will still be different experience for a lot of people. No fixed hardware, not everyone has M2 in their PC. Most people still have SATA (3.7 sec is still quick enough) or even harddrives.

This just shows that all those people who made fun of Mark Cerny and called him a liar are now suddenly waking up.
Well, it's going to be interesting to see how this directly compares with the PS5 version. This will likely be the first "demanding" game available on both which utilizes DirectStorage and the PS5's storage architecture as efficiently as possible... so I'm looking forward to that.

The PC isn't going to be as efficient as PS5 is with regards to the architecture and I/O bottlenecks.. we're still a long way away from being as efficient... but now, PC's with drives faster than 5GB/s, could see brute force push past that efficiency.

You people are still trying to push the "ps5 god tier sdd" immaculate "i/o complex", "6 lane" none-sense which has already been debunk a hundred times.

Its funny because the only people that were calling anyone a liar was people saying MS were lying about Direct Storage and Velocity Architecture.

People were saying that the Xbox and PC version of games would need elevators and that Xbox and PC wouldn't be able to run certain games like the 2020 PS5 demo.

All of which have been debunked. Not only can PC run the 2020 PS5 demo, it runs it at a higher resolution and FPS WITHOUT DirectStorage and using old windows 10 storage system.

Then we have the matrix demo which proves yet again that there no god tier SDD or myth-tical i/o complex as xbox series and PS5 runs identically.

Not just that but the loading speed is also the same. Actually it loads faster on the xbox series. Guess what the matrix demo will be released on PC in less than 3 weeks and we will see it perform with an even better resolution and FPS and load the same or faster on PC.

Then you have the loading speeds of dozens of games with completely identical loading times or difference that is around a second or less. This completely debunks the myth-tical ssd and i/o complex or that other platforms need elevators.

This DirectStorage and Win 11 Storage is further proving that. You couldn't take that so you have to comment about the PS5.

You people then result to glorifying Rachet & Clank, which was also debunked by DF by taping off the SSD and running it on 1.7 GB/s SSD.

But people still couldn't give up on the house of cards so they make up things like. "Oh the i/o complex is doing all the work and the ssd doesn't really matter."

So which is it, the magically i/o complex or the ssd? Cause XS with slower SSD is performing better or virtually the same in ALL games.

So if its "mythical i/o complex"...why can't it be a "mythical direct storage and velocity architecture"?

Whats next? claiming that the current games are not taking full advantage of the PS5 ssd/io? that's a joke. Developers have and are already overhauling their engine to take advantage of the ssd/up for loading.

Remember the viral quote "Epic game had to completely redesign UE5 to leverage PS5 SSD/IO". Yet Matrix loads faster on the Xbox Series...Now you are claiming there hasn't been a game that fully leverages the SSD/IO? So which is it?

In fact Control for example uses PS5’s kraken compression and XS still loaded faster.

The article states it's GPU decompression using DirectStorage.
"we saw cutscene and gameplay examples where Forspoken took 1.9 seconds, 1.7 seconds, and 2.2 seconds to get us there, thanks to the way DirectStorage offloads the streaming of graphical assets from the CPU to the GPU."

And again here.
"the current implementation of DirectStorage in Forspoken is only removing one of the big I/O bottlenecks — others exist on the CPU."

I assume the one big I/O bottleneck is decompression based on this.

The win32 results kind of makes Direct Storage and GPU decompression questionable.
Maybe that 5900X is really good at decompression compared to the 6800XT.
Its not using GPU decom. This has already been confirmed. Not surprising, you jumped straight to the mythical i/o complex which has already been debunked.

I remember the days when PS5 was supposed to be 22GB/s and destroy everything to the point games on other platform would either not work, be lower quality or need elevators and that the solution like DirectStorage, Velocity Archetecture were PR marketing myths.
 
Last edited:

DaGwaphics

Member
Good to see them unlocking the potential of the NVMe drives with this. Nearly a 2,000MB throughput improvement on the file IO, that's quite nice.
 

M1chl

Currently Gif and Meme Champion
The decompression is done using CUDA cores not Tensor cores.
Have you tried using Nsight or Nvprof, to check if there is any activity on the Tensor Units?
Alright guys I loaded up the Nsight, had to update it to some nightly build and it appears as tho, it's using actually both units. However based on profiler I am getting pretty significant load (around 20% on Tensor) while CUDA cores is around 50% utilized. This is paired by RAID0 of SN850 (1TB units, both load at 95%), made some benchmark with data from KC: D. I cannot really post it, or show it to you, both of these things are private property (and second one is in NDA, I can only say percent), but I am preparing visual benchmark when RTX IO went live in UE 5. Obviously I am trash at graphics, so it's going to be ugly, but functional. I am aiming to show the user if the SSD or GPU is the weakest link.

Also since the percentage, my PC is 3900x R9, 64GB of 3600 DDR4 RAM and RTX3900.

Also fun fact, CPU was at 50%, just due to load on PCIe bus.

So thanks for engagement, I've learnt something : )
 
Last edited:

Snake29

RSI Employee of the Year
Typin You people are still trying to push the "ps5 god tier sdd" immaculate "i/o complex", "6 lane" none-sense which has already been debunk a hundred times.

Its funny because the only people that were calling anyone a liar was people saying MS were lying about Direct Storage and Velocity Architecture.

People were saying that the Xbox and PC version of games would need elevators and that Xbox and PC wouldn't be able to run certain games like the 2020 PS5 demo.

All of which have been debunked. Not only can PC run the 2020 PS5 demo, it runs it at a higher resolution and FPS WITHOUT DirectStorage and using old windows 10 storage system.

Then we have the matrix demo which proves yet again that there no god tier SDD or myth-tical i/o complex as xbox series and PS5 runs identically.

Not just that but the loading speed is also the same. Actually it loads faster on the xbox series. Guess what the matrix demo will be released on PC in less than 3 weeks and we will see it perform with an even better resolution and FPS and load the same or faster on PC.

Then you have the loading speeds of dozens of games with completely identical loading times or difference that is around a second or less. This completely debunks the myth-tical ssd and i/o complex or that other platforms need elevators.

This DirectStorage and Win 11 Storage is further proving that. You couldn't take that so you have to comment about the PS5.

You people then result to glorifying Rachet & Clank, which was also debunked by DF by taping off the SSD and running it on 1.7 GB/s SSD.

But people still couldn't give up on the house of cards so they make up things like. "Oh the i/o complex is doing all the work and the ssd doesn't really matter."

So which is it, the magically i/o complex or the ssd? Cause XS with slower SSD is performing better or virtually the same in ALL games.

So if its "mythical i/o complex"...why can't it be a "mythical direct storage and velocity architecture"?

Whats next? claiming that the current games are not taking full advantage of the PS5 ssd/io? that's a joke. Developers have and are already overhauling their engine to take advantage of the ssd/up for loading.

Remember the viral quote "Epic game had to completely redesign UE5 to leverage PS5 SSD/IO". Yet Matrix loads faster on the Xbox Series...Now you are claiming there hasn't been a game that fully leverages the SSD/IO? So which is it?

In fact Control for example uses PS5’s kraken compression and XS still loaded faster.


Its not using GPU decom. This has already been confirmed. Not surprising, you jumped straight to the mythical i/o complex which has already been debunked.

I remember the days when PS5 was supposed to be 22GB/s and destroy everything to the point games on other platform would either not work, be lower quality or need elevators and that the solution like DirectStorage, Velocity Archetecture were PR marketing myths.

Billy Crystal Reaction GIF by MOODMAN


angry typing GIF by David Firth
 
Last edited:
Top Bottom