-70% NVMe SSD IOPS performance by 4 Acronis services (Samsung 990 & 980 Pro 2TB)
It took me quite a while to figure out why my Samsung NVMe drives never reached their advertised IOPS performance.
Finally I tried manually switching off Windows services and found out that 4 Acronis services are responsible for slowing down (any faster) NVMe SSDs IOPS.
This is 100% reproducible on different machines. Older NVMe SSDs are affected as well for instance Samsung 960 Pro, WD SN730, etc.
Disabling these services causes some features not working anymore like backup cleanup which then doesn't show previous backups anymore.
I'm really wondering why these services have such negative impact while any usual anti-virus programm doesn't (Bitdefender shows almost no impact on the Samsung 990 Pro).
This performance issue is known since 2017 but unresolved: https://forum.acronis.com/forum/acronis-true-image-2017-forum/acronis-m…
I hope that this helps anyone wondering why the NVMe performance is worse than expected while having Acronis installed.
Of course I now opened a ticket with support but expectations are low.
FS,
Reviewing your posted screenshots suggests that you are talking about a performance drop in random 4K file size tests. Frankly, this is to be expected as high performance IOPS and MB/s of these small files is difficult if not nearly impossible to achieve across all use cases. 4K bench marking is best measured by setting CrystalDiskMark to a Profile of Real World + Mix of 50% read 50% write and then only running it at 4K IOPS by ticking that button rather than the All tests button. Below are results from a 980 Pro 1TB disk.
Real World RND4K (IOPS) 50%-50% test
Default Profile RND4K (IOPS) 50%-50% test.
G. Uphoff wrote:What would improve (which application or program) if you had more IOPS?
Hi G. Uphoff, sry for my late reply. Somehow the forum software didn't email notify me about your post.
I agree that there won't be a big difference between 1.5 Mio IOPS and 600k IOPS in real world usage the user can "feel".
But the question is: why are these 4 Acronis software services reducing the NVMe drive IOPS by that much? Is this necessary ? No other tool, even anti-virus tools like Bitdefender, are reducing the drives IOPS performance.
Enchantech wrote:Reviewing your posted screenshots suggests that you are talking about a performance drop in random 4K file size tests. Frankly, this is to be expected as high performance IOPS and MB/s of these small files is difficult if not nearly impossible to achieve across all use cases. 4K bench marking is best measured by setting CrystalDiskMark to a Profile of Real World + Mix of 50% read 50% write and then only running it at 4K IOPS by ticking that button rather than the All tests button. Below are results from a 980 Pro 1TB disk.
Hi Enchantech,
thanks a lot for posting your performance measurements. Very interesting indeed as I do use the 980 Pro as well as the 990 Pro. My CystalDiskMark measurements are matching yours. But this also shows: Your drives performance might might be limited by these 4 Acronis services being active. Maybe you can try out disabling them via "msconfig" => services => unselect them => restart => measure.
With the new 990 Pro the performance drop is much more pronounced than with the 980 Pro.
Unfortunately I didn't explain the issue well enough in my initial post and I'll do so in the following one below.
I'd like to clarify the motivation of my thread using the initial screenshot again (ACPHO Services vs. Samsung 990 Pro NVMe.png)
- Left column: SSD performance with all Acronis Services activated
- Right column: SSD performance with 4 specific Acronis Services deactivated
- These measurements (negatively) reflect in real time benchmarks like 3DMark Storage Benchmark (launches Games, etc.)
- While I agree that this isn't the end of the world I do not understand why these 4 Acronis services have such a massive impact on the drives IOPS performance.
- What is Acronis doing with these IOPS ? What's the users benefit?
- After deactivating these 4 Acronis services the application still runs just fine.
- That's why I contacted Acronis support suggesting that there might be a code issue unnecessary wasting IOPS for no actual benefit.
- Unfortunately after many friendly mails, log file FTP uploads, etc. the support team still insists that we are talking about "high CPU usage" which we are obviously not. Eventually I had to give up on the support case.
I just want to highlight a potentially unnecessary problem with Acronis Cyber Protect and offer a obviously working solution.
Unfortunately, Acronis support (at initial level at any rate) far too frequently fails to understand the problem being reported. On one occasion I resorted to making a video documenting the problem, once I did this the issue was escalated. Not sure if it was ever solved; too many other things going on and it was not mission critical to me.
Einige Samsung 990 Pro scheinen ein Problem zu haben, wurden die S.M.A.R.T Werte schon mal überprüft?
--------------------------
Some Samsung 990 Pro seem to have a problem, have you checked the S.M.A.R.T values?
https://www.tomshardware.com/news/samsung-990-pro-health-dropping-fast
@IanL-S: Indeed! While I do understand why support staff behaves like this it really hurts, as a former IT product manager, to experience this kind of behavior to well meaning customer feedback.
I like the screencast video idea but then why invest the time if it never arrives at the development team. Most likely the issue would need much more visibility by complaining customers to get any capacity.
Personally I feel disappointed as the new subscription based business model promised much more development activity but didn't really deliver on it imho.
@G. Uphoff True! My Samsung 990 Pro 2TB is severely affected as I invested 30 TBW into finding a solution to this Acronis problem. I'm down -12% in S.M.A.R.T health. On the other hand I can reproduce the Acronis IOPS issue on my other NVMe (Samsung 980 Pro 2 TB) which is still installed.
My 1TB 980 Pro 1 TB is not showing any adverse impact after 18+ months that I use on my Ryzen 5950X system - it is limited to OS and applications. [Runs Windows 11 Pro and 32GB ram.] The T-Force TM8FP8001T drive I use for processing video files is in perfect condition. Cannot explain why my experience differs from yours.
Hi IanL-S
thanks a lot for testing this on your machine and 980 Pro ! Very interesting indeed, good that you're not affected but also surprising for me.
So your 980 Pro reaches about the specified 1.000.000 Random Write IOPS in Samsung Magician like mine (after deactivating the 4 Acronis services) ?
I can also reproduce the issue on different hardware like a Dell XPS 15 laptop which shipped with a Western Digital SN730 NVMe SSD:
Just checked and I had not run Samsung Magician benchmarks on the PC since I built the PC. That would be before I installed ACPHO. I have not tried disabling the Acronis services.
I see what you mean about the random read writes and random writes. It is possible that the benchmarking triggers the malware monitoring when random writes ad reads are taking place. If that is the case, it may not reflect real-world performance. I will white-list Samsung magician and rerun the benchmarks. Well, performance got worse when I whitelisted Samsung magician. Something very weird going on. When I get a chance I will investigate further.
Ian
Interesting further result: disabling protection improves things a bit, but still not where I would expect. Clearly the issue is related to core activity.
Hallo FS,
Wurde schon mal CrystalDiskInfo zum auslesen der S.M.A.R.T Werte verwendet?
Mich interessieren die "Schreibvorgänge gesamt" in GB und die "Betriebsstunden".
---------------------------------------------
Has CrystalDiskInfo ever been used to read the S.M.A.R.T values?
I am interested in the "total writes" in GB and the "Power on hours".
IanL-S wrote:triggers the malware monitoring when random writes ad reads are taking place. If that is the case, it may not reflect real-world performance.
Thanks a lot for testing Ian! If you're referring to the Acronis Cyber Protect "Active Protection" I already uninstalled it long ago as it led to application freezing since upgrading to W11 22H2.
It would be interesting if you'd also test disabling the 4 Acronis services I mentioned in my start (via msconfig) post and benchmark again. I'd bet you will see much improved write IOPS as well.
G. Uphoff wrote:Hallo FS,
Wurde schon mal CrystalDiskInfo zum auslesen der S.M.A.R.T Werte verwendet?
Mich interessieren die "Schreibvorgänge gesamt" in GB und die "Betriebsstunden".
---------------------------------------------
Has CrystalDiskInfo ever been used to read the S.M.A.R.T values?
I am interested in the "total writes" in GB and the "Power on hours".
Sure, all my drives are automatically monitored by Hard Disk Sentinel, here are todays values:
The new 990 Pro is suffering by the S.M.A.R.T health bug, already lost 12% health after using it for 1 month. How did I generate 28 TB writes on this new drive? ;) Mostly by testing my way out of this ugly Acronis IOPS bug... Days of trial and error.
So is it a hardware issue on my side? No as the 980 Pro drive, which is still installed, shows the same IOPS drop with the 4 Acronis services activated.
Es wird mit 47 GB pro Betriebsstunde auf der SSD geschrieben, was viel ist.
--------------------------
It writes to the SSD at 47GB per hour of use, which is a lot.
Correct, it's a lot and for a reason (thanks to Acronis)
"How did I generate 28 TB writes on this new drive? ;) Mostly by testing my way out of this ugly Acronis IOPS bug... Days of trial and error."
Ich denke, das die SSD auch ohne Acronis laufend auf die SSD schreibt.
Die Samsung 980 Pro hatte auch ein Problem.
-----------------------------------
I think that the SSD is constantly writing to the SSD even without Acronis.
The Samsung 980 Pro also had a problem.
https://www.tomshardware.com/news/samsung-980-pro-ssd-failures-firmware…
Indeed, that's why I'm running the proper 980 Pro firmware since release: 5B2QGXA7 .
No constant writes detected with my 990 Pro and 980 Pro when idling Windows 11.
Also this Acronis bug is not related to a specific NVMe vendor. I can reproduce this on Western Digital NVMe drives as well.
The fact that disabling 4 Acronis services solves the issue instantly is pretty convincing to me that Acronis software code is the culprit.
Update:
- Acronis Cyber Protect Home Office Build 40338
- SSD NVMe slowdown persists with no improvements yet
- My proposed fix via switching off 4 Acronis services still works
Update:
- Acronis Cyber Protect Home Office Build 40901
- SSD NVMe slowdown persists with no improvements yet
- My proposed fix via switching off 4 Acronis services still works
FS wrote:Update:
- Acronis Cyber Protect Home Office Build 40901
- SSD NVMe slowdown persists with no improvements yet
- My proposed fix via switching off 4 Acronis services still works
Hello!
I suggest you to raise a new ticket informing about that situation with those updates.
Happens the last one was from the last year and it's important to raise tickets so our support can see the volume for known issues and accelerate a solution. If nothing is reported the issues won't be identified. https://kb.acronis.com/content/8153
Best regards.
Hi Jose Pedro,
I agree but last time the support experience was very disappointing as the contact person never understood what I meant even though I was trying very hard and patiently to explain the issue. I agree that most users won't "feel" the performance impact and so this isn't the most important bug.
At least I presented a working solution for anyone affected. If I find time and patience I'll try the support again.
Best regards
FS wrote:Hi Jose Pedro,
I agree but last time the support experience was very disappointing as the contact person never understood what I meant even though I was trying very hard and patiently to explain the issue. I agree that most users won't "feel" the performance impact and so this isn't the most important bug.
At least I presented a working solution for anyone affected. If I find time and patience I'll try the support again.
Best regards
Hello!
Thanks for updating.
If you decide to raise a ticket afterward, please let me know. Also, thanks for sharing the workaround that helped you; this may assist other users.
Best regards.
A hearty cheer for FS!
I've been chasing this problem ever since I got my new computer. I couldn't find a Linux benchmark program, so I could never get Windows out of my way. And since the problem is random, and I need my computer, I couldn't just strip it to the bare bones and sit there running Samsung Magician all day long.
Thank you for doing the work and documenting it so thoroughly.
Hi Jerry,
very happy to hear that I'm not the only one bothered by this unnecessary performance impact.
I'm still wondering why Acronis doesn't really want to learn more about it but maybe the responsible code of these services needs a complete rewrite (while there is little ROI regarding the usual customer).
Nevertheless paying for a subscription software now my expectations are higher compared to the former licensing model.
Hello,
I have emailed the team regarding that issue. Hopefully, this will be addressed and improved as soon as possible. I will update the thread with any news as soon as I receive it.
Best regards.
Hi Jose Pedro,
awesome news, thanks for helping out on this issue. Hoping for the best. For sure fixing this will take time but I believe it's very worth it (in times of PCIe5 NVMe drives etc.)
Thanks again