Weird performance issue

Associate
Joined
12 Jul 2008
Posts
1,313
My desktop PC has a 500GB Samsung 870 SSD as a boot drive, which I backup by imaging to two SSDs in a USB dock twice a week.

The first image, to another 500GB Samsung 870 SSD takes 7mins.
The second image, to a 500GB Fanxiang S101 SSD, takes 30mins!

I assumed this was just down to the Fanxiang being a much cheaper drive than the Samsung but, for backup images, I could live with it.

I bought a 2TB Crucial BX500 for my laptop at the weekend and decided to check it was ok in my Desktop USB dock and took the opportunity to benchmark all 3 drives.

I was very surprised to see that all 3 benchmarked with very similar numbers (all in the USB3 dock)...

500GB Samsung 870
02-500-GB-Samsung-870.jpg


500GB Fanxiang S101
03-500-GB-Fanxiang-S101.jpg


2TB Crucial BX500
01-2-TB-Crucial-BX500.jpg


So why does that Fanxiang take 30mins to make an image (using Macrium) whilst the Samsung and Crucial both take 7mins?!?!
 
So why does that Fanxiang take 30mins to make an image (using Macrium) whilst the Samsung and Crucial both take 7mins?!?!
Is the position in the dock something that could change the transfer speeds?

If I was to guess why one SATA SSD can be lower than another, it might have a smaller SLC cache that gets exhausted quicker with large transfers, or a different chip config using large die(s) which are harder to spread the writes over.

Digging into the specs: according to Fanxiang's website, it uses SM2259XT2 and YMTC TLC flash.

In theory, it shouldn't have the issue that QLC drives do (with slow writes after the cache is used up), but I believe the SM2259XT2 is a downgrade on the SM2259XT used in the BX500, since it is only 2 channel (versus 4 channel) and only supports up to 800 MT/s per channel, which is much slower than modern flash is capable of.

Many drives downgrade the write specs according to size too, 2TB drives are capable of writing faster than 250GB drives, for example.
 
Last edited:
All 3 benchmarks and the images are always done in the same slot of the USB dock.
I've had this setup for 18months and it has performed like this from day 1 but, as it's just my backup images, I've not looked into it any deeper.

The image files are currently around 150GB each, so each drive has my last 3 images on.

In my mind, it indicates something lacking in the Crystal benchmarking, as given the 3 results above, you would expect very similar real world performance.
It has also persuaded me to stick to the big, well-known brands for my storage requirements in the future :)

I did verify all of the images and they are all ok, according to Macrium, so I'm happy that my backup regime is secure, just wondered about the vastly different performances.
 
In my mind, it indicates something lacking in the Crystal benchmarking, as given the 3 results above, you would expect very similar real world performance.
It has also persuaded me to stick to the big, well-known brands for my storage requirements in the future :)
For large file transfers, you have to look at reviews that show you how big the cache is and how it handles it when exhausted, but most drives have pretty small write caches (nowhere near 150GB) and after that you're relying on how fast you can write directly to the flash.

In that circumstance, I think the quality of the controller, the flash and the die configuration is very important to the speeds you get, which is why small drives are unlikely to be optimal for large files. The 870 is a higher-end drive with a good controller, DRAM, fast flash and firmware optimisations that help to push it up in the benchmarks.

Your usage can expose the limitations on any drive really (even good ones), but cheap drives in a small capacity are likely to be exposed the worst. At least the Fanxiang doesn't use QLC.
 
Thanks for such comprehensive responses... took me out of my comfort zone on how SSDs work!

I guess you wouldn't really see such a performance dip if the drive was being used 'normally' and not just for writing out a big backup image.
 
I guess you wouldn't really see such a performance dip if the drive was being used 'normally' and not just for writing out a big backup image.
That's my assumption yeah, though I don't know how 'cheap' the Fanxiang really is, it looks lower-end than the BX500 (which is itself lower-end than the 870).

Thanks for such comprehensive responses... took me out of my comfort zone on how SSDs work!
This particular problem (large file transfers) is often an unpleasant surprise, so you're not alone :D

This is a good example of what happens with a QLC drive, you can see in the graph that after 42GB (the write cache) it goes from 500MB/s to just 100MB/s:

The 870 Evo (TLC, 24GB write cache) or the MX500 (TLC, 64GB write cache) both sustain 400MB/s once the cache is used up:

In regards to the size, the 870 Evo says it has a 1.75 higher quoted write speed for the 1TB and higher models versus the lower ones. I think they can compensate for this problem, by using more flash chips of a lower die density.
 
I received a 500GB Crucial BX500 today, as I plan to move to Win11 on my desktop and prefer to keep the old SSD as a working backup and start with a new Boot SSD.

The 500GB Crucial BX500 benchmarked much like its 2TB brother...
500-GB-BX500.jpg


but when used for writing a backup image to, it was woefully slow again, taking 27mins!

The packaging was identical for the 2 drives and the 500GB even quotes faster speeds of "up to 550MB/s" on the box (2TB says "up to 540MB/s").

This experience has certainly opened my eyes on the SSD front!
 
Back
Top Bottom