[SOLVED] Samsung 970 EVO plus showing different performance readings.

Nov 27, 2019
2
0
10
Hi everyone,
I have recently installed an M.2 drive (via PCIe3) into my motherboard as the boot / system drive.
According to Samsung Magician, the read figure is 3626 MB/s (which is what I would expect. But, using Crystal Disk Mark, it is coming out at 2445 MB/s!

Does anyone know if I am doing something wrong since both of these can't be correct (can they?)

Thanks loads,

Steve.
 
Solution
Yes, they can both be correct, depending on what processes or tasks Windows happens to be running in the background at the time, and those benchmarks don't matter anyhow because you're never going to see those speeds unless you are reading ONE very large file anyhow. Typically Windows is reading MANY very small or medium-ish files, which is going to drastically reduce the performance. Random and sequential performance aren't even measured in the same way.

I would run, if anything at all, the AS SSD benchmark with the default configuration of 1GB test file, and as long as the results are at least marginally within reasonable distance of the advertised specifications, forget about it from that point forward unless there is a compelling...
Yes, they can both be correct, depending on what processes or tasks Windows happens to be running in the background at the time, and those benchmarks don't matter anyhow because you're never going to see those speeds unless you are reading ONE very large file anyhow. Typically Windows is reading MANY very small or medium-ish files, which is going to drastically reduce the performance. Random and sequential performance aren't even measured in the same way.

I would run, if anything at all, the AS SSD benchmark with the default configuration of 1GB test file, and as long as the results are at least marginally within reasonable distance of the advertised specifications, forget about it from that point forward unless there is a compelling reason to be wondering why you lack performance.

For writes, you'll never see anything like that, ever, unless you are writing to or reading from another drive with the same speed capabilities, so those will be significantly lower in real world use as well.
 
  • Like
Reactions: SteveOJames
Solution
Nov 27, 2019
2
0
10
Yes, they can both be correct, depending on what processes or tasks Windows happens to be running in the background at the time, and those benchmarks don't matter anyhow because you're never going to see those speeds unless you are reading ONE very large file anyhow. Typically Windows is reading MANY very small or medium-ish files, which is going to drastically reduce the performance. Random and sequential performance aren't even measured in the same way.

I would run, if anything at all, the AS SSD benchmark with the default configuration of 1GB test file, and as long as the results are at least marginally within reasonable distance of the advertised specifications, forget about it from that point forward unless there is a compelling reason to be wondering why you lack performance.

For writes, you'll never see anything like that, ever, unless you are writing to or reading from another drive with the same speed capabilities, so those will be significantly lower in real world use as well.

Thank you for the reply Darkbreeze.:giggle:

I ran three tests just now and got the following (sequential) results:

Crystal Disk Mark = 2580(R) & 2597(W)
AS SSD = 2779(R) & 2877(W)
Samsung Magician = 3564(R) & 3024(W)


Since it is the system drive, there are always going to be programs/apps/processes using the drive, so a true figure cannot be realistically obtained (I guess unless going into a testing app directly on boot-up, or if the drive in question is a data drive (cloud services disabled etc.)
 
Plus, Magician always shows inflated results. We've all off us who have Samsung NVME and SATA SSDs noticed this. Obviously they don't want to show bad results for their own devices so it's clear they are using test metrics which favor the areas where these drives are strong and forego test metrics that have a weak showing.

What I meant by the other comment is that you can eliminate some of that if you, for example, temporarily (Or permanently, since I like to run these things myself on my own schedule rather than the system deciding when it wants to run them, usually at a time that is not convenient for me) disabling things like Windows update, automatic disk optimization/degragmentation/TRIM, System restore, disk cleanup, any OTHER application automatic update routines, indexing service, Windows defender full system scans, etc.

Of course there will always be background processes running, but these are actual tasks, which do not have to be run according to Windows scheduler or randomly.
 
  • Like
Reactions: SteveOJames