• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

960 EVO 4k slow down

ochadd

[H]ard|Gawd
2FA
Joined
May 9, 2008
Messages
1,464
I've been running this Samsung 960 EVO 500 GB SSD since March of 2017. 4k read performance has slowed 46-48 mbps down to 28-30 MBps. Write performance has slowed from 146-176 MBps down to 128-144 MBps. Have tried quite a few things to figure out the cause but I'm nearly out of ideas. My 850 pro in the same system can do about the same 4k read speeds now. I benchmark with As-SSD 2-3 times a year and the slowdown happened between May and October.

Tried:
Samsung NVME drivers 2.3
Samsung NVME driver 3.0
Microsoft NVME driver
Disabled C states
Power management to high performance
Killed every service and process using the disk.
It was the boot drive then and the boot drive now.
Tried testing right away at boot and waiting 5 minutes for the system to calm down.
Disabled Windows defender realtime protection
Same firmware version.
Tried the original AS-SSD version and the latest available as of today.
M.2 slots are set properly in bios.
Case temps are very low and the system is idling.
- I don't think it's temps as the 4k reads is the only metric that has fallen off so I doubt it's throttling.
Windows 10 Pro 64 bit 1803 w/ all updates installed (Would have been running different patches earlier)

Things that changed that I haven't been willing to try yet is a full format, older BIOS version, older Windows 10 version, and older firmware reversion. Any ideas would be welcome. Anyone else having similar problems?



960compare.JPG
 
Whenever you bench your boot OS drive you are at the mercy of any and every app (Especially the OS itself) that might be doing something in the background. Boot from another volume or USB and run a benchmark from that OS and see if you get the same results.
 
I've heard many complaints about SSD and NVMe I/O perforamnce after micro-code patches for Meltdown and Spectre. I do not know if it really impacted anything, here is mine for reference made fresh this early morning. Windows 10 ver1809, 250GB 960EVO, firmware 3B7QCXE7, 3.0 driver, C-States enabled, ASPM enabled, balanced power configuration, cache flushing disabled, no over-clock, no multicore enhancement, 6700K/2133 Dual Rank RAM/Z100

https://imgur.com/a/zAXCaBy

These are the result back in 2016 from Techpot using AS-SSD: https://www.techspot.com/review/1281-samsung-ssd-960-evo/page5.html

Here is another sample from SSD Review, again in 2016: http://www.thessdreview.com/our-reviews/samsung-960-evo-m-2-nvme-ssd-review-250gb1tb/3/
 
Last edited by a moderator:
I've heard many complaints about SSD and NVMe I/O perforamnce after micro-code patches for Meltdown and Spectre. I do not know if it really impacted anything, here is mine for reference made fresh this early morning. Windows 10 ver1809, 250GB 960EVO, firmware 3B7QCXE7, 3.0 driver, C-States enabled, ASPM enabled, balanced power configuration, cache flushing disabled, no over-clock, no multicore enhancement, 6700K/2133 Dual Rank RAM/Z100

https://imgur.com/a/zAXCaBy

These are the result back in 2016 from Techpot using AS-SSD: https://www.techspot.com/review/1281-samsung-ssd-960-evo/page5.html

Here is another sample from SSD Review, again in 2016: http://www.thessdreview.com/our-reviews/samsung-960-evo-m-2-nvme-ssd-review-250gb1tb/3/

The first 14 months of usage I was getting nearly those same numbers. Now it's 40% slower and I can feel the slow down during certain activities. I don't find it useful to boot from another drive for benchmarking since I have 5 prior benchmarks with it as a boot drive performing much better. There's also no other disk activity when I begin bench.
 
Seems silly but could you disable defender real time protection or equivalent in any AV?
 
... I don't find it useful to boot from another drive for benchmarking since I have 5 prior benchmarks with it as a boot drive performing much better. There's also no other disk activity when I begin bench.
You should still do it which will remove your OS instance from the equation. If the benches go back to the baseline they were at previously, then it is your OS install or something on top of it. If not, then you can look at the hardware or how it is configured.
 
I did disable real time protection before each run.
Ah sorry I missed it in the attempted fixes! A couple of thoughts, could you check out its health, may need to use the probe in HWiNFO as some tools cant read SMART from NVME. The other is memory degradation causing throughput issues. Could you remove a DIMM if in dual channel and rebench?
 
Ah sorry I missed it in the attempted fixes! A couple of thoughts, could you check out its health, may need to use the probe in HWiNFO as some tools cant read SMART from NVME. The other is memory degradation causing throughput issues. Could you remove a DIMM if in dual channel and rebench?

Memory is still working pretty well. I've lowered my overclock slightly due to sleep and noise issues @ 5.1 ghz. Can't imagine 100 mhz and mem timings going from 15-15-15-35 to 16-16-16-35 would drop it that much. My northbridge is set higher now which I would think if anything, better for SSD speeds. Samsung Magician says I'm at 6% used and healthy, 5.5 TB written. The screen shot is before on the left and current on the right.

upload_2018-10-26_14-34-20.png
 
Back
Top