I have a trio of 1TB Sata SSDs I'm using with a USB adapter and all of the drives suffer very poor write speeds under FFS of less then 50MB/s when benchmarks of the drives while on Sata with CDM 6 shows a minimum of 150MB/s.
Anyone want to provide thoughts on what is happening as Windows does seem to handle the drives faster if I'm doing a normal copy.
Slow Writes to USB connected SSD
- Posts: 9
- Joined: 11 Oct 2020
- Posts: 1037
- Joined: 8 May 2006
What is CDM 6? [CrystalDiskMark]USB adapter ... benchmarks of the drives while on Sata with CDM 6
Are you saying you're comparing writes a USB connected drive vs. the same drive directly connected via SATA cable in your computer?
What version USB?
What type of USB adapter?
(I ran [WD] diagnostics on a USB connected external HDD. My computer only has USB 2.0. Other computers had USB 3.0. Reading online, people were saying that USB 2.0 would take ~24 hours for a full diagnostic on the drive. Others were reporting ~4 hours for the same.
So my thought was to take my external HDD & plug it into a system that had USB "3" on it & run the diags from there. Tried one computer, & it was estimating 18 hours. Let it run, figured I had to leave (with my drive) before it would ever finished, so I just aborted the run. (Absconded) a second computer with USB 3 on it & took it home to run on that, figuring "it's got to be faster". Well that computer reported ~18-24 hour range & sure enough it did take that long... [If I knew that was what it was going to turn out to be, it would have simply used my computer & let it run & run & run.]
Actual writes to the drive, via USB 3, were in the 50 MB/s range, I was getting variable numbers, though in any case that was ~2x what I can get with my USB 2.0.
So, those people who were reporting 4 hour diag times, must have USB 3.2's or something like that.)
- Posts: 9
- Joined: 11 Oct 2020
Yes I'm comparing the benchmarks using Crystal Disk Mark 6 on drives that were benchmarked on Sata. Same set of tests so I have a known baseline to work from.
The issue is that FFS isn't as fast as Windows in copying/transfering files to the same drive (about half as fast) and that has me scratching my head as Windows is generally slower; so I'm scratching my head trying to figure out what is causing FFS to be so much slower then Windows instead of being faster as normal.
As to whether it's on a USB 2.0 or 3.0 port, it's a B450 board so no USB 2.0 at all other then backwards compatibility and yes I have the needed drivers.
Not sure what's going on as I just checked the CDM 6 benchmark against what I'd recorded and the drive is performing faster on the USB then it did on Sata. Am I chasing a Phantom or just got my knickers in a twist for no reason?
The issue is that FFS isn't as fast as Windows in copying/transfering files to the same drive (about half as fast) and that has me scratching my head as Windows is generally slower; so I'm scratching my head trying to figure out what is causing FFS to be so much slower then Windows instead of being faster as normal.
As to whether it's on a USB 2.0 or 3.0 port, it's a B450 board so no USB 2.0 at all other then backwards compatibility and yes I have the needed drivers.
Not sure what's going on as I just checked the CDM 6 benchmark against what I'd recorded and the drive is performing faster on the USB then it did on Sata. Am I chasing a Phantom or just got my knickers in a twist for no reason?
-
- Posts: 4056
- Joined: 11 Jun 2019
CDM benchmarks aren't realistic speeds, good tool, but not real. Considering USB vs SATA, there's a huge difference. SATA has better shielding/less interference, better commands, and doesn't rely on adapter specs. You can get a big performance increase by using a USB-SATA adapter that supports UASP. USB 3.x is prone to 2.4GHz interference as well. Native SATA commands are also going to be faster than USB commands through a driver and adapter.
As far as FFS being slower than Windows transfer, that's odd. That can be caused by so many things though. AV software may handle USB different than internal drives, overhead on file operations may be compounded by the process of FFS behavior, who knows...
As far as FFS being slower than Windows transfer, that's odd. That can be caused by so many things though. AV software may handle USB different than internal drives, overhead on file operations may be compounded by the process of FFS behavior, who knows...
- Posts: 9
- Joined: 11 Oct 2020
What's AV Software? Haven't used any of that snake oil in years since it's proven to be worse then simply nuking and reinstalling. I've gotten things down to the point a Clean INstall with all of my software only takes about an hour thanks to backup software and the extra time is simply to ensure I didn't forget any files.
As to CDM being useless, most of the Benchmark tools are useless unless you use them correctly. I find that running CDM with the same settings actually gives me an indication of the systems overall performance and by simply using the defaults, I'm not introducing any variations by myself in how it runs so that helps drastically.
On the FFS being slower then a Windows Transfer, it surprised me when it happened as Windows can handle multiple reads from Backup to multiple drives (C,D,E,F) at the same time as I do have plenty of memory. Diving into it in more depth, what I determined is that Windows has no issues reading from the backup SSD and writing those files to multiple drives due to write caching. As stated, I have C, D, E, F drives and the files I'm restoring from backup are going to D, E, F so there are three threads copying files to the drives while FFS only does a single copy at a time.
It's not an issue when I'm doing a system backup as the reverse simply isn't true for Windows due to the speed limit of the backup drive and as FFS is only copying changed files/folders after the initial backup, it runs pretty quick as on a weekly basis, it's less then 100GB of data being backed up.
As to CDM being useless, most of the Benchmark tools are useless unless you use them correctly. I find that running CDM with the same settings actually gives me an indication of the systems overall performance and by simply using the defaults, I'm not introducing any variations by myself in how it runs so that helps drastically.
On the FFS being slower then a Windows Transfer, it surprised me when it happened as Windows can handle multiple reads from Backup to multiple drives (C,D,E,F) at the same time as I do have plenty of memory. Diving into it in more depth, what I determined is that Windows has no issues reading from the backup SSD and writing those files to multiple drives due to write caching. As stated, I have C, D, E, F drives and the files I'm restoring from backup are going to D, E, F so there are three threads copying files to the drives while FFS only does a single copy at a time.
It's not an issue when I'm doing a system backup as the reverse simply isn't true for Windows due to the speed limit of the backup drive and as FFS is only copying changed files/folders after the initial backup, it runs pretty quick as on a weekly basis, it's less then 100GB of data being backed up.
-
- Posts: 4056
- Joined: 11 Jun 2019
AV = AntiVirus. I do agree as I don't use antivirus, and even disabled defender. It is definitely not harder than a reinstall and restore though. Unless you have images/snapshots I guess.
Glad you were able to figure more info out and share it!
Glad you were able to figure more info out and share it!
