RIP DirectStorage. It was nice knowing ya

Well, unfortunate but according to Tom's Hardware site, Microsoft DirectStorage tech tested on Windows 11/RTX 4090 shows 10% hit to GPU performance in Forspoken. Tech is dead on arrival I tell you faster load times at the cost of performance? No thanks.

Taxing on even the most powerful GPU. Imagine putting this tech on a console lmao.

*Take note they are testing with RTX4090. Pending lower spec GPU reviews.

DirectStorage Causes 10% Performance Hit On RTX 4090 In Forspoken

By Zhiye Liu

published about 10 hours ago

Microsoft's DirectStorage technology has brought lightning-fast loading times to games. However, it would seem that not everything is sunshine and rainbows. DirectStorage may induce a slight performance loss, according to the latest test by the German publication PC Games Hardware(opens in new tab).

DirectStorage 1.1 moves game asset decompression from the processor to the graphics card. Microsoft's demo showed that scenes load almost 3X faster on a graphics card than the processor. Of course, your mileage will vary and depend hugely on your hardware. Nevertheless, it's an exciting technology that will make loading screens a thing of the past. While many news outlets are focusing on benchmarking the impact of DirectStorage on loading times, PC Games Hardware took it a step further and measured the effect that Microsoft's technology has on frame rates.

Forspoken, from Square Enix, is the first game to leverage DirectStorage, the foundation for PC Games Hardware's tests. The publication's system consisted of a Core i9-12900K (Alder Lake) processor and a GeForce RTX 4090, one of the best graphics cards, at 4K. The reviewer tested with a SATA SSD and two M.2 SSD but didn't mention the exact models.

DirectStorage Frame Rate Benchmarks

Swipe to scroll horizontally

Configuration	Average FPS	1% Percentile	0.2% Percentile
Core i9-12900K + RTX 4090 (SATA SSD)	83.2	54.3	36
Core i9-12900K + RTX 4090 (M.2 SSD)	75.3	53.8	37.3
Core i9-12900K + RTX 4090 (M.2 SSD)	74.4	53.5	35.1

According to the results, the configuration with the SATA SSD offered the best performance in terms of frame rates. However, switching to an M.2 SSD resulted in up to 10% lower average frame rates. While the number looks significant, it's less than a 10 FPS difference between the configurations. Furthermore, the percentiles were very close to each other, meaning the gameplay was smooth despite the lower average frame rates.

It's too early to tell if DirectStorage only contributes to a 10% game performance degradation. The number could be significantly more considerable. PC Games Hardware tested at 4K, so it's plausible that the impact is more visible on lower resolutions, such as 1080p or 1440p. The GeForce RTX 4090 also possesses a lot of firepower; therefore, it'll be interesting to see how DirectStorage affects slower graphics cards. What would be even more intriguing is to quantify the performance loss and compare it between AMD, Nvidia, and Intel graphics cards.

It was to be expected that DirectStorage would take some toll on the graphics card. There is still a lot of testing to be done, though. PC Games Hardware's results are just a stepping stone. In any event, DirectStorage is a game-changer for game developers and gamers, and we'll likely see it in many upcoming titles.

"It was to be expected that DirectStorage would take some toll on the graphics card. There is still a lot of testing to be done, though. PC Games Hardware's results are just a stepping stone. In any event, DirectStorage is a game-changer for game developers and gamers, and we'll likely see it in many upcoming titles."

Huh, if that's RIP it must be the Walking Dead. lol :P

PC gaming been dying for like 20+ years man, so they say 😋🤣

"DirectStorage is dead"
"DirectStorage is a game changer"

Not sure which one is the most dumb.

Another thread backfired. 😂

It's not dead, and it's not alive. It's

It's DOA for mid range cards, 100%. New theory: Jensen and MS did a deal so the only logical GPU to buy would be a 4090.

I don't think they got the right conclusion. DirectStorage still works on SATA SSDs. Forspoken also has the 1.1 DLL but GPU decompression isn't enabled for this game, it's strictly CPU. More than likely, a CPU bottleneck occurs with the nvme drive because more CPU threads are being used to decompress the assets.

Doubt it's GPU-related.

R.I.P. this thread. It was nice knowing ya. 😅

Yeah, that article is frankly awful. We need a reputable source like Hardware Unboxed or Gamers Nexus. Tom's Hardware just sucks.

A better question is what does this mean for the much much much weaker Xbox Series hardware. Fans were claiming this MS software solution would somehow nullify Sony's hardware architecture advantage in this area. What about now?

Hardware advantages are always >>> Software after thoughts

The problem with Xbox/MS fans who were trying to use this software implementation as some type of "Ah ha gotcha" response to the architectural hardware tech that Sony created is that, Sony can, at any time, introduce new software solutions ON TOP OF their own hardware tech to create superior results.

Sony's proprietary SSD & I/O architecture is a 0 performance cost technology

It's the first game with the technology, it'll be optimized over time.

Also, Forspoken clearly isn't the most optimized game.

As said, Forspoken is the first game that takes advantage of DirectStorage. And while the loading times are really fast, we encountered some awful frame pacing/stuttering issues. For some reason, the game suffered from a lot of stutters when we installed it on both our Samsung 970 Pro Plus and Samsung 980 Pro NvME SSDs.

and later on:

By moving the game folder to our SATA SSD, we were able to get better frametimes

We are not sure what is going on here, however, we currently recommend installing the game on a SATA SSD. Yes, the loading times may be a bit longer but you’ll get a smoother gaming experience.

Something something terrible Japanese game port.

Seriously this is borderline on the same page as Dark Souls being locked 720p resolution.

@above_average: You do know this about PC and not Xbox right?🤭

Wait. I thought Forspoken only used DS 1.0 (no GPU decompression).

Anywho, whenever a game does use GPU decompression I do expect it to take a toll on the framerate. I dont think E-sports or highly competitive FPSs will use it for that reason (not the GPU decompression anyway). Or they might only use it in a brute force way to speed up loading during a loading screen. but no GPu decompression during play.

Does that mean the feature is DOA? I don't think so. The pay off could still be worth it. It's good to fully leverage the capabilities of a system. With the consoles: I remember an Xbox engineer saying that it would take 12 Xbox Series X cores (so zen 2 at 3.2GHZ ish) to match the decompression rate of the decompression hardware in the XSX. Sounds like a job for a GPU on the PC front to be honest.

No wonder the tools have never arrived. The X was a lie you just like the DX12 and the Cloud. This is sad and pathetic.

Sounds like another technical fail by the developer on this game. I look forward to seeing other games using DS and seeing how it performs.

Actually just thinking about it some more (Mein Deutch is nicht seh gut so i haven't looked at the orignal source): I wonder did the framerate drop precisely because they are not using GPU decompression.

It gets worse with a faster SSD according to the article. This could mean that the faster SSD is placing a higher burden on the CPU to decompress data at a faster rate since it's coming in faster compared to the Sata SSD.

Is Forspoken even using GPU decompression? Sh*tEnix themselves said they are using CPU, the last time they talked about this. Most probably the confusion is because they are using the DS 1.1 SDK/libraries on the installed system but not actually utilizing the 1.1 feature set.

I'll wait to see how MS games work with the tech this year.

Not sure about using a 5/10 game to judge anything.

Tell that to the xbox fans who were hyping this software add on for xbox consoles since launch

It's not a fail, it's working as designed. But this is SOFTWARE implementation, it's not free to be used without pulling from other system resources as if it was it's own hardware based solution.

People acting surprised over this is funny to see.

Every single thing you said is wrong. Both Xbox and PS5 has a dedicated fixed function hardware decompression block and this is precisely why neither MS nor Sony can introduce new software solutions ON TOP OF their own hardware tech to create superior results. And there's nothing proprietary about Sony's SSD or IO architecture.

Alright, so here is what is likely the real story.

The game has a frame rate increase during the fade to black loading screen that lasts longer on the sata ssd resulting in higher average fps. Otherwise in-game, they perform the same. Nothing to see here but terrible journalism. This is simply a case of dumbshits wanting a story out as fast as possible without thorough research or testing.

The game doesn't even use GPU decompression which Tom's Hardware erroneously reported that it did.

It is not a software addon for Xbox and this thread is not about Xbox. You don't even understand the topic you are talking about.

This is why the PS5 has a dedicated chip for compression/decompression. No performance hit to the rest of the system and the game is free to use all resources.

Outside of missing Juub's comment, all these resources and no performance hits you are citing didn't prevent the game from running at 720p on the PS5 😎

you ever see on the sotrage drive boxes.

up to speed. its not legal guarantee speed.

Well, I pointed out the obvious fact that DirectStorge had not yet been proven in a video game when some PC fanboy was desperately hyping it, and I also pointed out another obvious fact that there is a difference between the theoretical values that MS and their media henchmen were bragging about "for clicks and views" and the actual implementation in video games.

lol but it's not "running at 720p" you just quoted the lowest bound possibility from a dynamic range resolution. You are referring to the performance mode only here with a 1440p target. The Quality and RT modes both have lower bounds that are around 1080p at the lowest.

Not sure why you feel the need to lie, just shows how much BS artist you actually are.

@st_monica: Read my post. The OP is completely wrong.

@Juub1990: Well, either way my simple fact based statement remains the same. If, by any chance, the DS proves to perform as MS advertises in actual video games, then wake me up.

@above_average: Take up your complaints with your boys at Digital Foundry 😂

It takes a special kind of incompetent dev to screw up so badly where a faster hardware component leads to less perceived performance. Outside of Capcom and Kojima, the rest of the Japanese devs are so sh*t that it baffles my mind. I can understand other devs with limited budgets but Sh*tEnix is just takes the cake.

Do we even need it in the first place? I don’t remember any recent open world game having loading screens besides loading the game for the 1st time

@ivangrozny: It will be interesting to see how much it's used over time if it is. Measuring it's effect may be more difficult too. You wont be able to point at something directly at the screen and say "wow...amazing SSD and direct storage tech"....not unless it's the PS5 anyway :p j/k.

The thing is it's highly probable that devs at least asked for this ability on consoles (much faster tertiary storage). Sony and MS didn't put fast SSDs and decompression silicon into their consoles just for the Lols. They also both did it. It's not like Sony went out on a limb and MS used a SATA drive. MS especially have a long track record of taking on board dev feedback and Sony have learned that lesson too after the PS3.

The approach on the PC is different (using the GPU and/or brute forcing it through the CPU since PCs don't have dedicated silicon for decompression) but it's there to address the same requirement. Again MS wouldn't have bothered to go to all the trouble of developing this stuff for Windows and Nvidia wouldn't bother with developing a decompression algorithm optimised for GPUs (amongst other efforts) if developers weren't asking for it.

I suspect, at the moment, Production capacity is hitting a financial and time barrier before it's hitting a technical barrier. Very few of the new toys on either console are really being used at the moment. A lot of games are just X1/PS4 tech running on steroids.

Regarding open world games: Massive open worlds can certainly be done while storage is very slow. It's been done for years. Hell Xenoblade X on the Wiiu has a colossal open world and that loads straight from a Blu-ray drive AND it has no loading screens unless you trigger a cutscene. No HDD to even cache too yet you can fly from one end of the map to the other. So we don't "Technically" need it. But there is a price being paid to hide the loading going on in the background. E.g. in Spiderman on the PS4, about half of the 4GB available to the game is being used for caching and buffers. In Xenoblade X there is extensive pop in for characters around NLA because the Blu-Ray drive can't keep up.

It wasn't MS, it was Nvidia who developed it and not for game development but for machine learning/enterprise workloads. They just also brought its subset to the consumer space on Windows with MS.

Forspoken DirectStorage on test: Contrary to reports, there's no GPU performance hit on an RTX 4090

https://www.pcgamer.com/forspoken-directstorage-on-test-contrary-to-reports-theres-no-gpu-performance-hit-on-an-rtx-4090/

Or they can use more RAM memory on PC. No SSD will ever beat the speed of DDR memory. It's been a while we have been sitting at 8-16GB of RAM for game requirements. Most people already have 32-64GB in their gaming PCs. I would say developers should just go for using more RAM memory on PC instead of direct storage thingy or what ever.

To be fair, RTX4090 has 24 GB of memory. it can literally load 1/4 of the game into its own GPU memory never needing to access hard drive in the first place.

MS developed direct storage along with Nvidia and AMD. It's also a component of the velocity architecture of the Xbox series consoles.

https://news.xbox.com/en-us/2020/07/14/a-closer-look-at-xbox-velocity-architecture/

RTX IO is Nvidias implementation of direct storage: More specifically the GPU accelerated decompression algorithm (which is thankfully vendor agnostic.....good on ya Nvidia). Maybe it's used for machine learning too...i dunno. Though that could still be interesting to game devs too. But (as per the below link) RTX IO plugs into direct storage.

https://www.nvidia.com/en-us/geforce/news/rtx-io-gpu-accelerated-storage-technology/

@ivangrozny: Why though? What is direct storage affecting badly (note: it looks like the TCs story was a nothing burger. poor methodology in measuring the performance)? It removes a whole bunch of inefficiencies from the old Win32 API which was primarily made for old Hard Drives.

The precursor to all of this is GPUDirect by Nvidia. Hell the RTX IO diagram that Nvidia showed back in 2020 is literally the exact same one with different colors and the name DirectStorage it self is taken from GPUDirect Storage. So no, RTX IO isn't Nvidia's implementation of Direct Storage but DirectStorage it self is a specification build to support RTX IO/GPUDirect in consumer space. Now it remains to be seen if GPUDirect RDMA comes to DirectStorage/RTX IO or not.

@pc_rocks: Ah fair enough. I had not heard of GPU Direct. It does look awfully familiar alright.