Samsung SSD 840 Low write

Stankenestor

Honorable
Aug 19, 2015
80
2
10,635
Hello!

I recently ran peformance benchmark in samsung magician 5.1 using my 840 EVO SSD (bought october 2013). I have a good 80k read score but my write score is 30k. ACHI mode is activated, im using a 6b/s SATA III cable and TRIM is enabled. Sometimes when i run performance optimization under the "advanced" tab in samsung magician (it takes 50 minutes) i get about 60k write, but then it's low again like a week after. I have no clue why my SSD is behaving like this.

Any suggestions?

Thanks!
 

Stankenestor

Honorable
Aug 19, 2015
80
2
10,635


Before i had 80k Read and about 80k write, now i had 80k read and 20k write. But i tested another SATA III port, and now i had 75k read and 75k write. This seems reasonable right?
 

Stankenestor

Honorable
Aug 19, 2015
80
2
10,635


I dont use RAID, i have 85 GB free storage (the drive is 250 GB) and here is a userbenchmark http://www.userbenchmark.com/UserRun/5603192

Everything seems fine about my SSD except the "Deep queue 4k" read and write, whatever that is. Here is a random user with also an 840 EVO SSD: http://www.userbenchmark.com/UserRun/5602862

 

TJ Hooker

Titan
Ambassador
Alright, free space looks alright.
RAPID mode has nothing to do with RAID, it's a Samsung proprietary RAM-caching utility that increases write speeds (unless you're writing enough data that the cache fills up before it can flush data to the SSD).

High queue depth performance is irrelevant to typical desktop performance. The deep queue 4K test in that benchmark measures read/write performance when 64 read/writes are queued up at once, something which will pretty much never happen outside of a server.
 

Stankenestor

Honorable
Aug 19, 2015
80
2
10,635


Sorry, yes i meant i dont use RAPID mode. Im currently in ACHI mode.
 

TJ Hooker

Titan
Ambassador
The fact that you brought up AHCI makes me think you may still be confused as to what RAPID is...
Drives can typically be configured as RAID, AHCI, or IDE, usually in the BIOS. The RAPID mode I'm talking about has nothing to do with any of that. It's a setting that you can enable/disable from within the Samsung Magician software.
 

Stankenestor

Honorable
Aug 19, 2015
80
2
10,635


No i know what they are, i just wanted to say that for the record, im using ACHI mode (like you should with an SSD). Rapid mode is not enabled in samsung magician. Any more suggestions, or does things seem like they are in order? Maybe my motherboard's SATA port was slower than the others, seems like i have consistent 80-75k write and read speeds now, which are similiar to what other people have.
 

Stankenestor

Honorable
Aug 19, 2015
80
2
10,635


Using performance benchmark from samsung magician, im now up in 90k read and 60k write in one test and 90k read 40k write in another, with no big programs running in task manager. It seems to fluctuate quite a bit.

Edit: I installed the latest windows version 1709 and now my test results look like this:

Before windows update:
koXsqQm.jpg


After windows update:
9IxQQzT.jpg


Seq write is lower?
 

TJ Hooker

Titan
Ambassador
Again, I'd recommend downloading some standardized benchmark that you can compare to review benchmark results. The problem with the Userbenchmark results you were looking at earlier is that you don't necessarily know what the rest of their setup looks like, whether they had RAPID enabled and inflating results, etc.
 

Stankenestor

Honorable
Aug 19, 2015
80
2
10,635


edited my post

a 3rd performance test shows:
oykBO3o.jpg


Why does it fluctuate so much? am i fine or is something wrong?
 

TJ Hooker

Titan
Ambassador
That is odd...
Maybe there's other stuff going on the background? You could try looking in task manager for anything with high disk usage.
To guarantee the system is in the same state every time, you could restart the PC between successive tests. For each test, wait the same amount of time after reaching desktop for stuff to finish loading, and then run the benchmark. Maybe you'll get more consistent results that way. Finally, you could maybe try manually initiating TRIM a little before running the benchmark each time.

I'll admit, I'm grasping at straws here.
 

Stankenestor

Honorable
Aug 19, 2015
80
2
10,635


I did exactly this and i still have very inconsistent write results. sometimes its 60k write, sometimes 30k etc. Seems strange