Microsoft has announced that DirectStorage API is now available on PC, supporting both Windows 10 and Windows 11. Starting today, developers can release PC games that can take advantage of this API in order to improve overall performance.
As Microsoft noted, this public SDK release will allow developers to more fully utilize the speed of the latest storage devices. Not only that, but all gamers (even those using hard-disk devices) can benefit from it.
“While you may see benefits on any kind of storage device, installing games to an NVMe SSD will maximize your IO performance and help you more fully experience the benefits of DirectStorage. “
This first release of DirectStorage provides developers everything they need to move to a new model of IO for their games. Additionally, Microsoft is working on even more ways to offload work from the CPU. For instance, GPU decompression is next on its roadmap; a feature that will give developers more control over resources and how hardware is leveraged.
Lastly, and from what we know, one of the first games that will support DirectStorage is Forspoken.
Microsoft will present an introduction to DirectStorage at GDC on March 22nd!
John is the founder and Editor in Chief at DSOGaming. He is a PC gaming fan and highly supports the modding and indie communities. Before creating DSOGaming, John worked on numerous gaming websites. While he is a die-hard PC gamer, his gaming roots can be found on consoles. John loved – and still does – the 16-bit consoles, and considers SNES to be one of the best consoles. Still, the PC platform won him over consoles. That was mainly due to 3DFX and its iconic dedicated 3D accelerator graphics card, Voodoo 2. John has also written a higher degree thesis on the “The Evolution of PC graphics cards.”
Contact: Email
It’s party time!!!!
Makes me think of SOAD.
Gonna listen to that shlt NOW!!!
m/
Only improves loading times, no change in FPS (talking about SSDs, ofc).
Loading times in SSDs are short enough anyways without this thing
There was never a talk about that, about improving FPS in games, no matter if SSD or some advanced “SSD tech” behind it. In all my newer games which I tested on multiple HDDs and even some newer “high end” SSDs, the FPS is basically the same…..
Today its for the most parts yes… tomorrow it will no doubt improve fps when for instance open world games will clock in assets bigger than the vram of most cards can handle
Same was said for RAM multiple times, since DDR2 came out, I remember very clearly and nothing happend unless specific “CPU arhitecture” where you can benefit from it (some “AMD eco-systems”), but mostly, you just need the required amount of RAM (speaking MB/GB for desired game and you will be fine). Ofc unless the game is so demanding that even the RTX, 6900 couldnt handle it or what not.. BTW contrary to popular belief, sequential reads can actually improve the loading times even on HDDs…. and its confirmed again with this “DirectStorage API” speak. The game just need to me programmed from the ground up with that in mind…… sure, SSDs would benefit the most but every storage device should benefit from it and if RAM is used properly and wisely, the whole “PC eco-system” would benefit from it.
The thing is vram will be more and more utilized for the actual renders and not keep as much assets loaded, that’s why DS was invented. Allow for good performance even when the gpu handles loads above its vram capacity and storage speeds will affect that down the line once it starts to get fully utilized
It surely can in the future, but this DS thing they’re talking about is pure marketing bullshlt, imo.
There was never a talk about that, about improving FPS in games, no matter if SSD or some advanced “SSD tech” behind it. In all my newer games which I tested on multiple HDDs and even some newer “high end” SSDs, the FPS is basically the same…..
These guys are trying to sell me some new “magic” hardware, innit?
Well, it ain’t gonna happen! I’m only upgrading AFTER ruskis move away from Ukraine!
Not gonna spend thousands to die from radiation later.
It’s literally free software for devs lmfao
Reminds me of the “fastest storage ever, even beats PCs” on the PS5…that endup being absolute bullcrap.
It was true for a little while until the samsung 980 pro and others pcie 4.0 nvme released
Sweet, another reason i can keep being on w10 until they dont support it anymore lol
Im more that M$ optimize this horrendous windows memory management and remove all spyware and bloatware first, but man cant have it all…
It’s kinda funny that this is finally happening. Cause back in 2017 AMD came out with a Pro tier GPU based around Vega 64 that used SSD’s. It was known as the Radeon Pro SSG. Anyways this new API works very much like how AMD’s GPU works only it’s a lot cheaper alternative. Very much cheaper and less time consuming for a developer to optimize for.
I think I’m wrong but I remember that we need something related to motherboard to get this feature like PCIE version 4 or something
There’s always some trickery, innit?
The thing is they’re trying to sell more hardware, and that’s all.
Dude I played Ghosts of tsushima on ps4pro and ps5. The loading times being around 1 Sec. Is so god damn awesome I would definitely upgrade for this feature alone
Yeah, loading times are awesome on SSDs, and they’re doing it for quite a while now in PCs. I can’t be hyped over this, cause I know this industry and its promises. This looks like some fairy tale console players use to believe.
PS5 has GPU decompresson, and a ton of hoopla was made about its SSD.
But PS5 games don’t load any faster than their PC counterpart.
So I wouldn’t get too excited about this.
Some games load faster on PS5. However, e.g Hhosts of Tsushima was loading within a second. This wouldn’t work on pc without proper Directstorage.
The thing is that its basically a chicken and hen situation, dev’s (beside a few exclusives) won’t really utilize this tech fully until the mainstream has it… and the current DS api don’t have what i personally see as the big piece…. the dma like transfer/decompression directly from storage to vram… IE it don’t utilize rtx i/o etc yet… but rather bog down the cpu etc still
And on another side-note… by the time the devs starts to utilize that we likely will have pcie5 storage on PC that have raw what the pcie4 + decompression etc has on the conslows… then add the next gen gpu i/o handling and i would not be surprised if we see 2x the i/o when streaming assets on an high end PC vs conslows.
The tech will allow insane quality/huge open world games for instance so im excited but that’s for the future… not today.
Cool, but without games that support the feature is useless. Which will be the first? forespoken? LOL
Final Fantasy Origins: Stranger of CHAOS will be the first.
Well look at that, it’s almost like there’s virtually nothing stopping Windows 10 from utilizing the same newer features of 11 and Direct X12 Ultimate, almost as if…it’s practically a new coat of UWP paint with some changes for Intel’s 12th Gen chips. Hmm…
The Win10 version is slower than W11 version. According to MS its because of the changes under the hood
SSDs are not supposed to improve Fps, they’re supposed to improve loading times – nothing more.
It will improve FPS if your game is waiting for assets… or at the very least make thing’s pop in much less – Later when DS will start to get utilized it will no doubt start to affect fps/pop-ins far more in for instance open world games that have more data than the vram can handle
How absolutely braindead do you have to be to think SSD’s are supposed to improve FPS?
Do you seriously think that’s what DirectStorage is for? Improving storage = faster load times, however load times are already similar even on HDD’s because devs haven’t fully taken advantage of modern (NVMe) SSD’s yet. This will help in achieving that.
“DLSS is trash”
No doubt you’re a moron but didn’t know you were blind too.
You have zero technical knowledge of the things you’re talking about and it’s obvious with the utter nonsense you spew.
Nearly every single claim you just made is factually wrong.
I cannot understand how anyone takes what you say seriously.
For the sake of anyone else who reads the above nonsense, here’s the technical background on DLSS, AA, (Anti-Aliasing), and DirectStorage:
DLSS (Deep Learning Super Sampling) is a machine learning based upscaling algorithm that is trained on the Saturn V supercomputer at high-resolution (4K). Once the algorithm learns what a frame should look like at 4K, it can intelligently take low resolution data and upscale it to a target resolution. It is vastly superior to standard upscaling because of this. MANY games look better than Native with DLSS enabled, even at lower resolutions like 1440p or 1080p (Control, Cyberpunk 2077, and more). The main benefit of this however, is that it dramatically boosts performance while doing so, because the game is running internally at a lower resolution.
DSR (Dynamic Super Resolution) is NVIDIA’s way of doing super-sampling. They recently released DLDSR which is a deep learning method of super sampling that is faster than standard supersampling. It uses similar tech to DLSS and achieves higher quality.
MSAA is an ancient anti-aliasing algorithm that only works on geometry (non transparent items like grass and foliage cannot be antialiased without other techniques). This is why all modern game engines moved past it via Fullscreen AA techniques like SMAA or FXAA. TAA is just like a fullscreen AA technique but over time (many frames). It brings very good specular stability (and overall image stability) at the cost of a slight blurriness, which can be alleviated with sharpness filters. Fast moving objects can ghost, but this is heavily game and implementation dependent. However, this is not a significant issue with DLSS, as the output data is already sharp enough. God of War (2018) looks better with DLSS On and sharpness set to zero because of this.
DirectStorage is designed to handle large numbers of I/O requests easier, since data is streamed in many small pieces rather than big chunks in modern games. Newer drive interfaces like NVMe offer significantly increased bandwidth that has not been taken advantage of, as all data still needs to be passed through an API that incurs an overhead for I/O requests.
Yes, SSD’s are already much faster at loading than HDD’s. But NVMe SSD’s offer an insane bandwidth (different from speed) increase that isn’t being used currently. With the extra bandwidth, more I/O requests can be executed in parallel, allowing data to be streamed to the GPU MUCH faster. This can not just reduce loading times even further, but eliminate it all together or allow for new visual features enabled by streaming in huge amounts of data from the SSD.
“I expect FPS boost because of reduction of CPU work being done due to the RAM to GPU path removed in the process if it doesn’t achieve it then this garbage tech is useless.”
^ This is sincerely one of the dumbest statements you have ever made. The work that is offloaded from the CPU will now be done by the GPU. It will be done faster, but that’s just a nice side-effect, not the main purpose. You are measuring the improvements DirectStorage offers by looking at framerate. You are an idiot.
The comment wasn’t for you, you’d know that if you could read. Its for any poor fellow who sees your nonsense and believes it.
Also for anyone else, to understand why something can look better than native, think about how machine learning algorithms are trained: they use data that is native at a much higher resolution than normal games.
If you train an algorithm with 8K data for example, it will know what a frame should look like at 8K, so when it reconstructs an image from a lower res like 720p, it will try to replicate how it looks at 8K, which can quite possibly look better than 4K native as an example.
As for you latex, you’re literally too stupid to convince otherwise.
It’s a human thing to make mistakes and be wrong from time to time, because no one knows everything. Normal people can admit if they were wrong, but not Latex. Latex will rather attack you than admit he was wrong.
For example my crysis 2 disscusion with him. He tried to tell me textures in the original game (around 512×512 with sharpening filter) looked better than in the remaster (4K up to 8K). Of course someone can still prefer original textures because of different art direction, but from technical point of view 8K textures are massive upgrade compared to 512×512, not downgrade.
When it comes to your DLSS discussion with him, he also wrote totally innacurate claims (like saying DLSS is just upscaling with sharpening filter).
DLSS is reconstruction method powered by AI, that can reconstruct missing details. DLSS quality preset at 4K use 1440p internally, but it doesnt mean the final output is anything like 1440p. Not only AI is used to reconstruct missing details, but also data from previous frames, so the final picture can indeed approach native quality. The are some downsides, like ghosting, and small glitches on certain surfaces, but overall this is a good trade off considering massive fps boost (especially if RT is enabled). Native + TAA has even more ghosting than DLSS, therefore I think it’s fair to say DLSS offers supperior picture quality native + TAA.
DLSS cant however match native + MSAA picture quality for sure. MSAA picture is a little bit sharper, and has no artifacts and ghosthing problems on top of that. There are however limitations like shading aliasing, and unfiltered bitmaps, although AA transparency settings in Nv panel usually filter 2D bitmaps with good results and then pretty much entire picture remains clean and sharp.
But the thing is MSAA has MASSIVE performance penalty, and it’s also not easy to implement it on modern engines without destroying VRAM budged. The reality is, developers dont want to use it, therefore this disscusion about MSAA vs DLSS is pointless, because most games will continue to use TAA anyway, and in such games DLSS can indeed offer supperior picture quality, because TAA alone blurs picture to the extreme.
Personally I prefer to use just FXAA as my favorite AA option. Yes FXAA blurs textures in the distance slightly, and there’s also some shimmering during motion (especially on vegetation), however with small downsampling (especially with AI downsampling filter Nv recently added) textures sharpness is restored, and shimmering is almost nonexistent. The final picture looks as good as MSAA to my eyes, but without MSAA shortcomings and it also works in pretty much every game
Braindead? I would think the opposite… what do you think will happen once a games assets are 2x the size of the vram? Yes the game engine will starts to stall or you will face insane delayed pop-ins at the very least. Today its no biggie but once the tech starts to be utilized you could actually start to really gain fps by storage speed (if that’s the bottleneck to begin with)
This is an entirely different issue. Assets are already bigger than many GPU’s vram. This is why data is streamed in and out in small pieces.
Using more bandwidth to increase the streaming rate would allow loading to complete faster but it will have zero effect on the time it takes to render a frame.
Game engines, especially modern ones don’t block execution while waiting for data to stream in. This is why pop-in even exists. Of games waited for all assets to stream in before rendering, there would be no pop-in but that defeats the purpose of streaming, as the frame rate is now tied to the streaming rate, which is a horrible solution.
Some of the data within the frame may not be up to date but that is by no means tied to the frame time.
The idea behind DS is to allow way more complex/bigger assets than we have to day, utilize the ssd as viritual vram is one very simplistic way to see part of what it will be utilized for and it will no doubt affect frame-rates down the line when way more assets will get streamed. Kind of like an gpu that’s vram starved on a title… then we see an fps drop beside insane popins.
Benchmarks will no doubt show that once the tech will reached that point
I’ve stated this already, yes faster storage can have the side affect of giving a higher frame rate but frame time is not a function of storage speed.
A frame is drawn by the GPU however, there are things that can take time over several frames, mainly CPU work that is scheduled. Nearly every game engine in use does not block the GPU from working while waiting for data to be streamed in, it will just get there late, hence pop-in. Because the GPU can continue working, the time it takes to draw a frame isn’t impacted.
good to see you back. Your were missed.
BTW, some more recent news and development related to this tech.
Actually, Nvidia and its partners IBM and Cornell University, have found a way to make GPUs seamlessly work with SSDs ”without’‘ a proprietary API.
The method, called Big Accelerator Memory (BaM), promises to be useful for various compute tasks, but it will be particularly useful for emerging workloads that use large datasets. Essentially, as GPUs get closer to CPUs in terms of programmability, they also need direct access to large storage devices.
The concept seems interetsing but a bit complicated. Check the link given below, once it is approved. It was posted today on “TheRegister” website.
BaM mitigates the I/O traffic amplification by enabling the GPU threads to read or write small amounts of data on-demand, as determined by the compute.
As you all may already know, that any modern GPU isn’t just for graphics; it is also used for various heavy-duty workloads like analytics, artificial intelligence, machine learning, and high-performance computing (HPC). And obviosuly, in order to process large datasets efficiently, GPUs either need vast amounts of expensive special-purpose memory (e.g., HBM2, GDDR6, etc.) locally, or efficient access to solid-state storage.
BaM essentially enables Nvidia GPU to fetch data directly from system memory and storage without using the CPU, which makes GPUs more self-sufficient than they are today.
At least that is the concept given by NVIDIA.
BaM is a step by Nvidia to move conventional CPU-centric tasks to GPU cores. Rather than relying on things like virtual address translation, page-fault-based on-demand loading of data, and other traditional CPU-centric mechanisms for handling large amounts of information, BaM instead provides software and a hardware architecture that allows Nvidia GPUs to fetch data direct from memory and storage and process it without needing a CPU core to orchestrate it.
Compute GPUs continue to use local memory as software-managed cache, but will move data using a PCIe interface, RDMA, and a custom Linux kernel driver that enables SSDs to read and write GPU memory directly when needed.
Commands for the SSDs are queued up by the GPU threads if the required data is not available locally.
As mentioned earlier, BaM does not use virtual memory address translation and therefore does not experience serialization events like TLB misses, translation look aside buffer. Nvidia and its partners plan to open-source the driver to allow others to use their BaM concept.
Modern compute GPUs already carry 80GB–128GB of HBM2E memory, and next-generation compute GPUs will expand local memory capacity. But dataset sizes are also increasing rapidly, so optimizing interoperability between GPUs and storage is important.
To a large degree, Nvidia’s BaM is a way for GPUs to obtain a large pool of storage and use it independently from the CPU, which makes compute accelerators much more independent than they are today.
https://www.theregister.com/2022/03/14/nvidia_gpu_data/
Welcome back
Kiitos friend !
Nice, you remember ?
Hey, I’m doing good. Thanks for asking. Nah, I did not buy any new hardware till now. Still using my OLD gaming rig, RX 480 GPU and core i7 4790.
It’s a Potato PC, but does the job for 1080p gaming, hehe. Dropped the plan for any upgrades after this whole COVID pandemic started. It has already disrupted the whole PC market, and we still can’t buy any new GPU closer the MSRP as well.
Also, there was little point in upgrading to Alder lake as well, since this was a new hybrid architecture, and prone to issues as well. Also, getting a DDR5 Memory these days is kind of very expensive.
I don’t want to go for DDR4 as well for Alder Lake or any other CPU arch, but maybe the upcoming ZEN 4 AM5 socket CPUs might bring huge improvement as well. I might upgrade to Zen 4 assuming this new CPU arch brings a lot of benefits to the table.
We need to wait and see. BTW, I did hear about
wasn’t aware ofAlder lake’s ILM bending design flaw. Looks like some trouble coming to consumers in the long run.But I guess there is one MOD for this as well.
EDIT:
Yeah, IgorsLab tech news outlet posted a method via which we can mod this ILM thing. Sounds interesting, but I doubt an average gamer/user can mod this issue easily.
Yeah, I just edited my previous reply. It appears that IgorsLab tech news outlet posted a method via which we can mod this ILM thing. Sounds interesting, but I doubt an average gamer/user can mod this issue easily.
I will check the AnandTech review as well. Looks like a serious problem at least in the long run for any Alder lake CPU.
Hi! glad to see you again!!! How are you?
Hey, I’m fine. Just being kind of busy these days. How are you doing ?
I’m fine man…please come back to be present in this community, because there is a need for serious people … unfortunately it’s frequented by many stupid people now!
I’m sure your i7 can still deliver 60fps gaming, maybe not with max settings in every game, but close to it. If you arnt planning to play at 120fps+ I would keep 4790 for another year.
GPU prices went downhill lately, but are still far from good :P. If your 480 RX still can run games at 1080p at med/high settings with around 60fps, then I would also use this GPU a little bit longer.
I’m however forced to upgrade, because my old GPU has died lately (1080ti). I’m planning to buy something like 3060ti, because I feel like xx80 series from Nvidia are no longer a good investment. I used to buy only high end series cards from Nvidia for the last 20 years, but with current prices even 3060 ti is not cheap (MSRP for 3060ti shuld be 2200z? in my country, but I have to pay for it 3500 z?, and that’s 777 US dolars, similar amont that I have paid for 1080ti in 2017 :P). Overall 3060 ti is a little bit faster than my previous 1080ti (well, in RT the difference can be really big, up to 6x times even), but I worry if 8GB will suffice in the near future. Standard 3060 has 12 GB, but this GPU is slower than my previous 1080ti, therefore I dont want it.
Hi Paul, good to see your comment. How are you doing these days ? Yes, I fully agree with you. For 1080p gaming, my current CPU and GPU both are very good. Sorry to know that your old GPU died.
I know it SUCKS pretty much these days, since it’s impossible to get a new GPU at least on the MSRP value.
But things are slowly improving though, but I’m still not very optimistic. 780 USD is still a very steep price for the 3060 Ti. But as gamers, we all are helpless.
And since your GPU died, then obviously you are forced to upgrade if you plan to game in future as well. 8GB VRAM should be okay for now, but it also depends on the game being played, and the in-game video/graphic settings and resolution applied.
These days most of the AAA games eat up a lot of VRAM, if not all. But for future-proofing we obviously need for VRAM.
BTW, how are things in Poland these days ? Might be a very tense atmosphere.
I heard almost more than a million Ukrainian war refugees have arrived in Poland safely, after Russia attacked their country. I really appreciate the Polish government and people for their help and support.
*respect*
Long time no see, also saw that announcement. Hope it gets adopted from the data center to gaming gpu’s. Can see tons of potential in that
Hey, good to see you as well ! How are you guys doing ?
Hey @thartist, just read your comment now.
If it solves stutter im in
Until devs bother to include it in all their games going forward, it’s about as meaningless as a wet fart (like most of MS’s PCC tech that’s boasted).