• 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.

Forspoken PC Insane Load Speed using DirectStorage 1.1

Guilty_AI

Member
if this a ram thing i guess another solution is to just splurge and get the largest kit you can find... that way the chance for RAM to be exceeded will never happen
maybe something like 32 or 64gb is good enough
I think people are misunderstanding here what virtual memory is, which isn't anyones fault since the material available on the subject is confunsing and even misleading.

What the benchmark video is most likely showing as virtual memory usage is virtual addressing, and that shouldn't worry be of any concern here on any level.
What Bo_Hazem Bo_Hazem is refering to is paging, which is in fact use of storage devices as an extension of RAM. If that degrades SSDs is a bit controversial, some people say it doesn't because of the way modern SSDs works and that this is a non-issue nowadays. Either way, that shouldn't be relevant to gaming because if your game is using so much ram to the point it needs to rely on paging, the game wouldn't even be playable due to how slow that memory is, even on SSDs. As long as the RAM isn't maxing out while playing, this is a non-issue.
 

Hoddi

Member
Wait.... so Directstorage actually can be dangerous to your hardware??? Why would anyone use this, then?
thinking about it, it's more likely that it's on forspoken than it is Directstorage... Denuvo being installed alongside the game doesn't help either
So I got curious and tried running the demo with just 8GB of RAM. It paged out ~3GB at the very beginning to free up memory but that was about it. It still played fine and didn't continue writing out more than a couple of hundred megabytes over 5-10 minutes or so. And it didn't page out anything worth mentioning at 16GB so I don't see it being an issue.

This is with an 11GB GPU though. Might be different with less.
 

Three

Member
what load speeds do not depend on:
#1: How good a game looks
#2: How well a game runs
I'm not so sure about that. Load times absolutely depend on how good a game looks if we're talking about more/higher quality assets loaded.
 
Last edited:

Kilau

Member
Would really like a tech breakdown of how directstorge is working in this game. Seems a bit of a mess overall. The windows store reviews all say the game won't even load lol.

rNhHxVa.jpg
 

01011001

Banned
I'm not so sure about that. Load times absolutely depend on how good a game looks if we're talking about more/higher quality assets loaded.

assets can be huge in size yet look like ass. or maybe there are a shitload of really ugly assets on scene at the same time.

if the game loads in 1000 really shitty looking objects, all with unique textures and normal maps etc, then it would load a big chunk of data, but still look like ass.

proof of that are many games that look like ass but also don't run well even tho you'd think they're multiple generations behind.
why do they run badly? because they stress the GPU or CPU in ways that makes them run slow.
this could be due to copious amounts of overdraw without any proper culling or because they are poorly threaded (see The Witcher 3)

and just like it doesn't automatically mean that if a game runs badly that it's a graphical showpiece, it also doesn't that if a game looks awful that it runs well or uses only a small amount of data.
 
Last edited:

Three

Member
assets can be huge in size yet look like ass. or maybe there are a shitload of really ugly assets on scene at the same time.

if the game loads in 1000 really shitty looking objects, all with unique textures and normal maps etc, then it would load a big chunk of data, but still look like ass.

proof of that are many games that look like ass but also don't run well even tho you'd think they're multiple generations behind.
why do they run badly? because they stress the GPU or CPU in ways that makes them run slow.
this could be due to copious amounts of overdraw without any proper culling or because they are poorly threaded (see The Witcher 3)

and just like it doesn't automatically mean that if a game runs badly that it's a graphical showpiece, it also doesn't that if a game looks awful that it runs well or uses only a small amount of data.
Well, yeah I guess if the artist is bad for a game then nothing can help but an identical artist making a game with lower quality assets (less asset variety, lower poly, lower res) means quicker loading times whereas better graphics (high asset variety, higher poly, higher res assets) means longer load times.

So loading speed does depend on how good a game looks too.
 

DavidGzz

Member
How did the PC performance thread turn into a console war thread?

People so triggered that someone prefers Xbox achievements over the lame Steam ones that no one cares about. To the point that they call people pathetic for it, lmao! Over this POS game no less. Can't make it up.
 

DenchDeckard

Moderated wildly
I speak german.

Edit: This one?



Brand new tech, first game that's also a poorly looking developed piece of shit I can see it having issues. The fact its in a retail product and smokes the ps5 is pretty damn amazing though and bodes well for the future. Imagine when it's tweaked and we hit 2.0 etc it's gonna be nuts and put PC in a massive distant lead in every front, like it always used to be.
 

RoadHazard

Gold Member
Brand new tech, first game that's also a poorly looking developed piece of shit I can see it having issues. The fact its in a retail product and smokes the ps5 is pretty damn amazing though and bodes well for the future. Imagine when it's tweaked and we hit 2.0 etc it's gonna be nuts and put PC in a massive distant lead in every front, like it always used to be.

It doesn't "smoke" the PS5 though? From what I've seen they're about on par.
 
Last edited:

Md Ray

Member
To get in the ballpark of the theoretical max of the PS5 SSD you need a 4090 paired with a gen 4 nvme. You can see the results from a variety of PC's in the link I posted above to the direct storage thread.

Make of that what you will.


Also, apparently has ~10% perf loss with DirectStorage. I suspected this would be the case when they announced they'll be using GPUs for decompression task on PC. Consoles have dedicated decompression HW likely for this reason.
 

winjer

Gold Member


Also, apparently has ~10% perf loss with DirectStorage. I suspected this would be the case when they announced they'll be using GPUs for decompression task on PC. Consoles have dedicated decompression HW likely for this reason.


The issue is not losing 10% performance for DS1.1 The real problem is that even with this tech, the game looks like a PS4 game.
Remember that PC GPU's are much more powerful than what consoles have.
But when a game is using so much power, and it's textures look so bland, it's just a waste.
 

winjer

Gold Member
So I tested the demo on my Sata SSD and got very similar frame rate.
Loadings were noticeably slower. But the frame rate seems identical to having it installed on the nvme drive.
Once again, I wonder if this has to do with some overhead with Denuvo.

Can someone that has the full game, test it against the demo?
 

RoboFu

One of the green rats
I wouldnt trust any off the cuff “ reports” with out thorough testing. These people thrive off “ being first “ and then you get stuff running wild like nvidia power cables and 6800 xt burning out that turned out to not be what the first reports claimed.
 
Last edited:

M1chl

Currently Gif and Meme Champion
Because it uses GPU power:

geforce-rtx-30-series-rtx-io-announcing-rtx-io.jpg


Unlike PS5's dedicated decompressor and I/O complex.
But this isn't even ready yet. Commits · microsoft/DirectStorage (github.com) The latest one, has reference support for GDeflate, which is building stone of RTX IO, reference means that is it "preparing codebase for this feature". And that's from month ago, I doubt that they are using it, because for it they would have to change their whole authoring, which is the most annoying thing with game release.
 


Also, apparently has ~10% perf loss with DirectStorage. I suspected this would be the case when they announced they'll be using GPUs for decompression task on PC. Consoles have dedicated decompression HW likely for this reason.

Its funny to see Sony fans jumping to conclusions to fuel their "god tier SSD/IO" none-sense that has already been debunked.
The 10% less performance has NOTHING to do with DirectStorage as DS isn't used in Gameplay in this game. They are clearly not streaming things.
 

Bo_Hazem

Banned
yamaci17 yamaci17 64bitmodels 64bitmodels

Had to come back to you guys as to show the behavior of Apple Macs using "swapped" memory or virtual RAM via internal SSD even with 16GB of RAM. Timestamped:




That improves performance over not doing so, but you put your data in that SSD at risk.
 
Because it uses GPU power:
Unlike PS5's dedicated decompressor and I/O complex.
Looks like Forspoken doesn't use the GPU Decompression of DirectStorage 1.1. You people are jumping to conclusions to fuel your bias rather than waiting on accurate analysis and testing.

"I still have yet to find evidence of GPU decompression in use here - when checking GPU compute usage when doing loads, it is a rather constant percentage and does not spike at all. IMO, I think it is safe to assume it is perhaps not in use." DF Dictator
 

Bo_Hazem

Banned
Looks like Forspoken doesn't use the GPU Decompression of DirectStorage 1.1. You people are jumping to conclusions to fuel your bias rather than waiting on accurate analysis and testing.

"I still have yet to find evidence of GPU decompression in use here - when checking GPU compute usage when doing loads, it is a rather constant percentage and does not spike at all. IMO, I think it is safe to assume it is perhaps not in use." DF Dictator

nVidia jumped into conclusions. Here, contact them and tell them how you feel about it:

 

Tarin02543

Member
I've moved my install from the m2 to a sata ssd, it's the same load speeds. i think the direct storage support is a hoax, it's just a matter of smart asset loading.
 

yamaci17

Member
I've moved my install from the m2 to a sata ssd, it's the same load speeds. i think the direct storage support is a hoax, it's just a matter of smart asset loading.
Not necessarily a hoax but game loads a wide range of map to the RAM to begin with. I've fast travelled between two farthest points in the demo and I see no "ssd" utilization at all yet instant loading. It simply unnecessarily puts a lot of data into RAM and VRAM and calls it fast loading. At least for the demo.
 

Thebonehead

Banned
So I tested the demo on my Sata SSD and got very similar frame rate.
Loadings were noticeably slower. But the frame rate seems identical to having it installed on the nvme drive.
Once again, I wonder if this has to do with some overhead with Denuvo.

Can someone that has the full game, test it against the demo?

What sort of bonehead could you get to do that?

I guess i could always purchase on Steam and test under the 2 hour timeframe for refund
 


Also, apparently has ~10% perf loss with DirectStorage. I suspected this would be the case when they announced they'll be using GPUs for decompression task on PC. Consoles have dedicated decompression HW likely for this reason.



"So the average FPS isn't actually reduced with an NVMe, it's just that the loading screens run at a higher frame rate which means SSD and HDD get skewed results if you measure it continuously across all test scenes of the benchmarks. Basically the longer the load, the higher the reported avg fps in the end. If you cut those bits out, performance is essentially the same."
 

GHG

Member
Not necessarily a hoax but game loads a wide range of map to the RAM to begin with. I've fast travelled between two farthest points in the demo and I see no "ssd" utilization at all yet instant loading. It simply unnecessarily puts a lot of data into RAM and VRAM and calls it fast loading. At least for the demo.

And that's why the specs are over the top with the RAM recommendations.

This game has been cobbled together by a bunch of hacks.
 

Spukc

always chasing the next thrill
yamaci17 yamaci17 64bitmodels 64bitmodels

Had to come back to you guys as to show the behavior of Apple Macs using "swapped" memory or virtual RAM via internal SSD even with 16GB of RAM. Timestamped:




That improves performance over not doing so, but you put your data in that SSD at risk.

that is one clickbaity video haha
and i will NOT click on it :pie_smfaceyes:
 
Last edited:

Gaiff

Member


"So the average FPS isn't actually reduced with an NVMe, it's just that the loading screens run at a higher frame rate which means SSD and HDD get skewed results if you measure it continuously across all test scenes of the benchmarks. Basically the longer the load, the higher the reported avg fps in the end. If you cut those bits out, performance is essentially the same."

Jesus Christ, if this is true the other guys (Tom's Hardware) are retarded. Didn't they look at the frame rate in real time or look at the graph to see that the beginning had much higher fps? They can't be this incompetent.
 

Xdrive05

Member
What's the load time differences between NVMe PCIe 4.0 and NVMe PCIe 3.0? Difference versus SATA quite a bit as expected.
 
Top Bottom