Post your 3DMark DirectStorage results

[INSTG8R]

Ran the test with a Gen3 SSD in the same system and I got weird results. DS is of little benefit here unless GPU decompression is used.

Maybe it's because the drive is attached to the chipset? I'll test the same drive attached to the CPU in a different system.

View attachment 376471

My Gen 4 game drive is off the chipset. I thought the numbers looked as expected? But you may be right it may be faster on the CPU. I look forward to your test results.

 

[stahlhart]

On the Strix board, the CPU lanes are ties to a worthless m.2 slot that not only can't keep an installed drive's temperatures in control, it also steals eight lanes from the GPU if you use it. Gen4 through the chipset for all of the m.2 drives is an acceptable compromise.

On the MSI board, the boot drive is using CPU lanes -- that probably explains the score difference?

 

[INSTG8R]

On the Strix board, the CPU lanes are ties to a worthless m.2 slot that not only can't keep an installed drive's temperatures in control, it also steals eight lanes from the GPU if you use it. Gen4 through the chipset for all of the m.2 drives is an acceptable compromise.

On the MSI board, the boot drive is using CPU lanes -- that probably explains the score difference?

Yeah and it’s only a Gen 3 on the CPU. It’s just my OS so it doesn’t need to be lightning fast either. My games are on my Gen4 so it would be the one to actually use DS when available.I decided to check if there were any new firmware for them and my Corsair had one (game drive) so I’m actually rerunning the test to see if there was any changes.
While my OS drive doesn’t steal any lanes like you it can get pretty toasty even with a heat sink

Edit well I didn't even actually look for any notes on the latest firmware but it's pretty much just within the margin of error. I mean I don't know what I was expecting but faster would have been nice...

 

[thestryker6]

I'd really love to see how they designed the DirectStorage benchmark as the numbers seem somewhat odd when looking at the data rate breakdown. I also had the same DS on being slower on my P44 Pro so perhaps something with the way the drive itself operates. I did a run on my 905p and one of my P44 Pro drives:

edit: 3DMark storage bench results (P44 Pro on top 905p on bottom):

 

[INSTG8R]

Well after getting into this benchmark and seeing everyones varying results including mine. I got to thinking I hadn't checked if my MP600 Core had had any FW updates, well it did! so it was painless tho finding any proper notes proved fruitless. So then I thought you know it's been a long time since I ran Crystal Disk on it either. Well lets just say I am a little shocked at the result. It's rated 4900/3700. I'm at a bit of loss seeing how well it performed in this bench to see numbers like this....I mean it's just my game drive so it's gonna be mostly reads but that's not the point....

you've got a driver interfering with Direct Storage; you can check for this by running "fsutil bypassIo state c:\" from an elevated command prompt. I had this same issue, and in my case it turned out to be disk imaging software (O&O). Uninstalling the application corrected the problem. Apparently, third party virus scanners can cause this as well.

Ah I'm glad I went back and read through this thread again because I have the same thing and noticed nobody else did(well except Kapone)Thanks for tip

Well this is just adding insult to injury on this drive at this point...I'm not sure what that exactly means @stahlhart I would appreciate your expertise or suggestions here. I got one drive with gimped writes and the other being hobbled by I don't know what. I am running the Samsung NVME driver but I doubt that would be it, but really at this point this whole drive thing has just snowballed on me...

Well I didn't really change anything but ran it again to see the if the BypassIO was still on, it is but my score went from 131% to this...

alright final Edit it was the Samsung NVME driver blocking BypassIO

So now I gotta run the test again....

 

[rusTORK]

I scored 120.4% faster in DirectStorage feature test

AMD Ryzen 7 5800X3D, NVIDIA GeForce RTX 3080 Ti x 1, 32768 MB, 64-bit Windows 11}

www.3dmark.com

Here is final result (i think), since i can't get higher scores. What was done:

1. Used "Boost lock" in EVGA Precision X1 to lock VRAM frequency on higher clock;
2. Disabled ASPM for PCI Express in Control Panel (Power Options);
3. Run Optimize and Defragment for 990 Pro drive;
4. I also have got enabled "Data Link Feature Exchange" in BIOS (it help boost PCI Express bandwidth for PCIe 4.0 devices).

GDeflate result basicly stuck at 59.99GB\s. BUT i didn't OC VRAM yet.

Also, CPU\MB still auto OC itself (OR it's just a bug).

 

[INSTG8R]

I scored 120.4% faster in DirectStorage feature test

AMD Ryzen 7 5800X3D, NVIDIA GeForce RTX 3080 Ti x 1, 32768 MB, 64-bit Windows 11}

www.3dmark.com

Here is final result (i think), since i can't get higher scores. What was done:

1. Used "Boost lock" in EVGA Precision X1 to lock VRAM frequency on higher clock;
2. Disabled ASPM for PCI Express in Control Panel (Power Options);
3. Run Optimize and Defragment for 990 Pro drive;
4. I also have got enabled "Data Link Feature Exchange" in BIOS (it help boost PCI Express bandwidth for PCIe 4.0 devices).

GDeflate result basicly stuck at 59.99GB\s. BUT i didn't OC VRAM yet.
View attachment 376585

I find it interesting my Gen3 is scoring better than your Gen 4 our CPU and and chipset are the same. You have faster RAM tho and that was before I fixed the BypassIO

 

[kapone32]

Updated:

 

[rusTORK]

I find it interesting my Gen3 is scoring better than your Gen 4 our CPU and and chipset are the same.

You score is higher (191,3%) because high diffirence between DirectStorage On and Off. High result of GDeflate probably because of 7900 XTX VRAM frequency (it's 20Gbps vs 19Gbps on RTX 3080 Ti).

Results in my first score (129%) are lower then current one (120%). It's look like worse, but in reality it's better since most numbers are higher.

and that was before I fixed the BypassIO

I didn't touch this thing. CMD show that drive is BypassIO compatible.

 

[INSTG8R]

You score is higher (191,3%) because high diffirence between DirectStorage On and Off. High result of GDeflate probably because of 7900 XTX VRAM frequency (it's 20Gbps vs 19Gbps on RTX 3080 Ti).

Results in my first score (129%) are lower then current one (120%). It's look like worse, but in reality it's better since most numbers are higher.

I didn't touch this thing. CMD show that drive is BypassIO compatible.

Yeah I guess my VRAM helps and the BypassIO was a problem on my end caused by using the Samsung NVME driver. You shouldn’t have it anyway it was a “self inflicted error“ on my part. Now that I fixed it I don’t know if it will make any tangible difference other than it working properly.

Edit: So this a run with BypassIO working properly and it had no effect on my score.

 

[rusTORK]

my MP600 Core had had any FW updates

It's rated 4900/3700.

According to Corsair page for this drive it's rated 4700\1950 MB/s. CrystalDiskMark results show strange results. Read is higher, but write is lower. When i got lower results on my drive i talk with ASUS and they told me that some devices on motherboard might lower bandwidth of drive. I have to disable (and remove) Wi-Fi since it lowered drive scores. Only after that i got stock speed for drive (even a bit higher).

 

[INSTG8R]

According to Corsair page for this drive it's rated 4700\1950 MB/s. CrystalDiskMark results show strange results. Read is higher, but write is lower. When i got lower results on my drive i talk with ASUS and they told me that some devices on motherboard might lower bandwidth of drive. I have to disable (and remove) Wi-Fi since it lowered drive scores. Only after that i got stock speed for drive (even a bit higher).
View attachment 376595

Running off the chipset so other than a couple other drives there is much to drag it down and I do still have a couple spare lanes
This is right from the specs from Corsair

MP600 CORE 2TB M.2 NVMe PCIe Gen. 4 x4 SSD

The CORSAIR MP600 CORE Gen4 PCIe x4 NVMe M.2 SSD provides extreme storage performance, using Gen4 PCIe technology to achieve blazing fast sequential read speeds of up to 4,950MB/sec.

www.corsair.com

 

[stahlhart]

Yeah I guess my VRAM helps and the BypassIO was a problem on my end caused by using the Samsung NVME driver. You shouldn’t have it anyway it was a “self inflicted error“ on my part. Now that I fixed it I don’t know if it will make any tangible difference other than it working properly.

Edit: So this a run with BypassIO working properly and it had no effect on my score.
View attachment 376596

I noticed the same result with the 13700K build (first score I posted) -- the score remained in the +140% vicinity. I was thinking of that more in terms of having DS set up correctly for eventual actual use as opposed to benchmarking. Maybe the UL code doesn't take that feature into account when testing/scoring, perhaps because it's only supported on Windows 11? Dunno.

 

[INSTG8R]

I noticed the same result with the 13700K build (first score I posted) -- the score remained in the +140% vicinity. I was thinking of that more in terms of having DS set up correctly for eventual actual use as opposed to benchmarking. Maybe the UL code doesn't take that feature into account when testing/scoring, perhaps because it's only supported on Windows 11? Dunno.

Well I play Euro and American Truck Sims and I'm always tweaking and testing because I play in VR and while in the games main bin file I recently noticed there are DS .dlls in there now. Now this is the first time I have seen any kind of implentation of DS or what effect it has on the games if any. I mean it's welcome because eeking out any extra performance is great. I play pretty much daily so any updates that change something would stick out to me like a sore thumb, this slipped under my radar but like you said it's better to have it set up properly for JUST this situation. Unfortunately I can't think of a way to quantify it or measure any tangible effects outside of maybe using a stop watch and measure load times with and without the .dlls
But yeah bottom line is making it sure it's actually working properly now that I have something that is trying to use it.

Lets just say I'm glad this benchmark was brought to my attention because I have reason to be invested in it doing what it supposed to do

 

[freeagent]

My 1TB Black, just a boot drive with a few programs installed:

My 2TB Black that is just a bit over 2/3 full:

 

[QuietBob]

in the games main bin file I recently noticed there are DS .dlls

You can swap them for the newest version 1.2.3 found here. Click download package on the right, extract the contents with 7-zip and you will find dstorage.dll and dstoragecore.dll in native\bin\x64.

Here's another result with an ADATA SX8200 Pro. I'm surprised it does so poorly on transfers in this benchmark. It's one of the fastest Gen3 drives, trading blows with the Samsung 970 EVO Plus. But the SSD->VRAM copy is still 2.2x faster using DS with a budget CPU/GPU combo:

@freeagent You left out the most important part of this benchmark

 

[rusTORK]

result with an ADATA SX8200 Pro. I'm surprised it does so poorly on transfers in this benchmark

ADATA messed with parts in this drive. I think it was in news. Review samples and retail product wasn't same. In fact there was multiple versions.

If i rememper correctly fastest was Samsung SSD 970 Pro. I sold my to buy 980 Pro and then get better version - 990 Pro to stay on PCIe Gen4.

 

[freeagent]

@freeagent You left out the most important part of this benchmark

Dammit.. ok I will run it again, takes a bit of time eh?

Maybe I will change up my lower fan config so I can slot in an SN770..

 

[rusTORK]

I will run it again, takes a bit of time eh?

You can just load this result in benchmark. If you didn't deleted files. =)

 

[INSTG8R]

You can swap them for the newest version 1.2.3 found here. Click download package on the right, extract the contents with 7-zip and you will find dstorage.dll and dstoragecore.dll in native\bin\x64.

Sweet! thanks. Now as I said these have been added recently(games got a big update this month) so I’m really not sure if it’s active/ but I’ll definitely grab those for sur.e I just wish I could tell if it’s actually working. I’ll inquire over on the forum and hopefully a Dev will have some answers. You’d think they’d make a big deal about it…

@QuietBob thanks for the link but turns out they are the same version. Now I am hoping someone over at SCS will give me an answer about their appearance. Both ETS/ATS are perfect candidates for this being that you are driving so the ability to stream assets faster would be a huge benefit. I strictly play them in VR so anything to keep a solid smooth framerate is welcome. I was sure to post my Gen 4 Bench results with the 150% improvement to hopefully get some others curious and drooling over it

 

[QuietBob]

ADATA messed with parts in this drive. I think it was in news. Review samples and retail product wasn't same. In fact there was multiple versions.

True, there were at least ten variants in total! Though mine are the early versions with the same controller as reviewed. You'd think it'd do better with these specs, but its transfers are a full 1 GB/s lower than the 970 Evo Plus in the DS benchmark:

I’ll inquire over on the forum and hopefully a Dev will have some answers ...
I was sure to post my Gen 4 Bench results with the 150% improvement to hopefully get some others curious and drooling over it

Great, I'm wondering what the devs will say. From what I learned the two Truck Simulators use dx11 for rendering, though they're planning to upgrade the engine to dx12 at some point. Dx12 is required for DS, so I'm not sure how they would implement it in the current state.

I'm also surprised that they included the most recent version of DS. Even this benchmark uses the previous version 1.2.2. I already copied the 1.2.3 files over, but saw no difference in the results.

 

[INSTG8R]

Dx12 is required for DS, so I'm not sure how they would implement it in the current state.

Ah so cart in front of the horse, Yes DX12 is planned but that's all we here is it's planned. They put up a tenentive roadmap for both of them yesterday DX12 was not on it...

 

[freeagent]

Ok..

SN850:



SN850X:



SN770:

 

[Launcestonian]

I ran this yesterday on the rig in my system specs, but without accessing that machine atm, I'm pretty sure the benchmark defaults to whatever drive the OS is on & not a preferred drive that actually has a bunch of games on it. I could be wrong but until I can access that PC I don't know for sure at this point however this was the result of a run I did last night with the OS drive ( that has no games on it! )

 

[stahlhart]

It does default to the boot drive, but there is a drop down where you can select the drive to run it with: