Question Help with drastic decrease in new NVME performance.

Jun 26, 2019
10
1
15
0
Hi all, I'm wondering if you can help with an NVME issue. I built a new rig a couple months ago (see my signature), and ran a benchmark a month ago. Everything checked out. Then I ran the same benchmark a week ago, due to noticeably slower start up times, and I had a drastic decrease in NVME performance:from 82nd percentile to 8th percentile. I have an Adata XPG SX8200 NVMe PCIe M.2 480GB . I've used maybe 10 gigs in the past month, and still have close to 300 gigs free, so it's not like I've filled it up.

The other thing is that my SSD (Crucial M500 240 GB) performance is also way down (73rd percentile to 10th percentile). And it's just sitting there, hosting my old Windows installation :) I haven't done anything to it since putting Windows on my NVME and using it as my main (boot) drive. I've attached screenshots below. You can see that the the benchmarks are down across the board.

The only thing I might have installed since the quicker bench is Bitdefender, but I don't see how that would slow down my NVME.

Original bench picture is first, later, slower, bench picture is second.

I'd appreciate any advice you might have!


ORIGINAL BENCH.




NEW, SLOWER BENCH

 

Lutfij

Titan
Moderator
Please stop using Userbench to tell you if you're right or wrong with your hardware. Have you used other app's like Crystal Disk? When posting a thread of troubleshooting nature, please list your specs like so:
CPU:
Motherboard:
Ram:
SSD/HDD:
GPU:
PSU:
Chassis:
OS:
 
Reactions: USAFRet
Jun 26, 2019
10
1
15
0
Sure, thanks. The reason I used Userbench data is that I don't have any Crystaldisk info from a month ago to indicate the dropoff.

Here's the Crystaldisk info. I don't expect to get the rated 5000MB/s, but this is less than a third of that!

Sequential Read (Q= 32,T= 1) : 1422.443 MB/s
Sequential Write (Q= 32,T= 1) : 1346.908 MB/s
Random Read 4KiB (Q= 8,T= 8) : 137.771 MB/s [ 33635.5 IOPS]
Random Write 4KiB (Q= 8,T= 8) : 74.947 MB/s [ 18297.6 IOPS]
Random Read 4KiB (Q= 32,T= 1) : 39.813 MB/s [ 9720.0 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 45.238 MB/s [ 11044.4 IOPS]
Random Read 4KiB (Q= 1,T= 1) : 24.231 MB/s [ 5915.8 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 28.994 MB/s [ 7078.6 IOPS]

Test : 1024 MiB [C: 33.2% (148.4/446.5 GiB)] (x5) [Interval=5 sec]
Date : 2019/06/26 17:14:36
OS : Windows 10 Professional [10.0 Build 17763] (x64


CPU: 9600k
Motherboard: ROG Maximus XI
Ram: 16 GB Adata Gammix 3000
SSD/HDD: Adata XPG SX8200 NVMe PCIe M.2 480GB
GPU: Radeon 7950
PSU: EVGA 850GQ
Chassis:
OS: Win 10 Pro
 
Can you post a screenshot of your crystaldiskmark results?

The mitigations/patches for all of Intel's security vulnerabilities affected storage performance, could explain some or all of the difference in scores. https://www.tomshardware.com/reviews/intel-amd-storage-test-mds-zombieload-ridl-fallout,6146.html

Also, I don't know where userbenchmark got a max speed of 5000 MB/s from, but that is not right. Adata only rates it for 3200/1700 MB/s sequential R/W, 310K/280K IOPS R/W. And those values will be the max values in the absolute best case scenario.
 
Last edited:
Jun 26, 2019
10
1
15
0
Can you post a screenshot of your crystaldiskmark results?

The mitigations/patches for all of Intel's security vulnerabilities affected storage performance, could explain some or all of the difference in scores. https://www.tomshardware.com/reviews/intel-amd-storage-test-mds-zombieload-ridl-fallout,6146.html

Also, I don't know where userbenchmark go a max speed of 5000 MB/s from, but that is not right. Adata only rates it for 3200/1700 MB/s sequential R/W, 310K/280K IOPS R/W. And those values will be the max values in the absolute best case scenario.
My bad. I was quoting that from memory. I see that 3200mb/s is the max. But that still doesn't explain the dropoff. Also, I'm not even at half of the max speed; see screenshot below. I guess maybe there was an OS patch in between May and mid-June. That would explain why my old Crucial M500 SSD is performing worse as well. Is there any way to disable that? From what I understand, the threat is theoretical, and no one is going to be targeting me anyway...

Thanks!

 
Last edited:
The only optimization that an SSD needs is TRIM, which Windows 10 should do automatically. You can manually initiate it my going to the drive Properties>Tools>Optimize.

Are you having any actual performance issue? I wouldn't disable security patches just to get better benchmark scores. That being said, I don't know if those patches fully explain the difference in benchmarks that you're seeing.

Also, could you provide links to your userbenchmark results (old and new), rather than just posting screenshots?
 
Last edited:
Jun 26, 2019
10
1
15
0
The only optimization that an SSD needs is TRIM, which Windows 10 should do automatically. You can manually initiate it my going to the drive Properties>Tools>Optimize.

Are you having any actual performance issue? I wouldn't disable security patches just to get better benchmark scores. That being said, I don't know if those patches fully explain the difference in benchmarks that you're seeing.

Also, could you provide links to your userbenchmark results (old and new), rather than just posting screenshots?
Yes, I am having performance issues. The computer is noticeably slower to boot. I originally timed it at about 4 seconds from bios splash screen to password/lock screen, now it's about 10-12. Other than that, I don't have anything other that subjective impressions, which may be informed by the dropoff in userbench results.

Here's the link to userbench results. I've done many benches in the last week or so, and they are all the same. I don't get anywhere near the first one with either NVME or SSD. CPU and GPU are about the same.

https://www.userbenchmark.com/UserRun/16993179 (first, much faster)

https://www.userbenchmark.com/UserRun/17937674 (one of many slower benches - only a month after first bench)
 
Jun 26, 2019
10
1
15
0
Try putting the m2 in the second m2 slot and run your crystal disk mark see if anything changes. What bios are you running and what model Rog Maximus XI is the motherboard?
Bios 0602 on a ASUS Maximus Hero XI Wifi.

The second m2 slot shares PCI lanes with the GPU, so the NVME is supposed to go in the first slot for best performance.
 
Jun 26, 2019
10
1
15
0
Neither M.2 slot on that board shares bandwidth with the GPU, performance will be the same regardless which one you use.

I'd try updating your BIOS.
I updated to the latest BIOS. Crystal Disk results are a bit better, but still not near where they were. On the plus side, the fan controller is finally managing my 3-pin case fans and my rig is much quieter :)


 
Reactions: TJ Hooker

Fix_that_Glitch

Prominent
Mar 31, 2019
848
65
490
18
I don't think this has been asked, are you running your ssd in ahci mode in bios? And can you just humor me and try the m2 in the second slot and run crystal disk? Also in the ssd tool box use the system optimization and see if it changes anything.
 
Jun 26, 2019
10
1
15
0
I don't think this has been asked, are you running your ssd in ahci mode in bios? And can you just humor me and try the m2 in the second slot and run crystal disk? Also in the ssd tool box use the system optimization and see if it changes anything.
Will do, but I'll have to wait till tomorrow. I'm out of the house all day today.
 

ASK THE COMMUNITY

TRENDING THREADS