- May 19, 2011
- 17,706
- 9,567
- 136
I'm puzzling over the very hit-and-miss performance of a drive I've just bought for a customer. Here's a benchmark I did just now:
The drive on the left is my own Seagate portable drive that I bought a few years ago that I have no problems with. The drive on the right is the one I've just bought and TBH that ATTO benchmark is the new drive on a particularly good day (and the drive in very general terms seems to be improving).
I was first alerted to a problem when I was transferring a backed-up user profile (so lots of small files in AppData followed by large image files in say the Pictures folder). I was seeing delays in robocopy that I'd normally associate with copying to a flash drive: tiny files taking a second or two each, then a second or two delay between each file from time to time. But then even more alarmingly the images were taking a good second or two each as well rather than the 'multiple files transferred per second' scenario I'd normally expect.
As I've recently migrated to Linux I was wondering if this might be related to that, but I had similar results in Windows 10 (I'm dual-booting and I have Win10 for gaming mostly). I'd see odd behaviours like transfer performance improving if I stopped it for a few minutes and let the drive settle down for a while. Sometimes I'd see a transfer rate of 60MB/sec for images, sometimes 6MB/sec for similar-size images.
That ATTO result is among the best I've seen for the new drive. One of the first ATTO benchmarks I did came out with 118 BYTES/SEC write and 3 BYTES/SEC read at the start of the test. It literally took a couple of minutes to post the first bar.
As that graph shows, it's like the drive is busy doing something else, but in Windows I'm watching the Task Manager graph for that particular drive, wait for it to completely flatline for 10 seconds then run the benchmark.
SMART data for the drive is fine, neither Windows or Linux reported anything in the logs to suggest a drive failure.
I had to transfer about 130GB of data to the drive for this customer. I stopped the transfer a fair few times and attempted the same transfer with the same cable and USB port to my own drive which was pushing things through at at least twice the speed for small files. Eventually I just let the newer drive get on with it, and it did at a fairly reasonable rate for the images (80-90MB/sec), but the random performance is kind of like how a friend put it, as if the drive was somehow having to format sectors on the fly before starting the copy. Just in case something bizarre was done to the file system, I have wiped that drive more than once (quick format NTFS) before doing things under Windows in earnest (like the final transfer). The wonky performance of that drive also showed up with a brand-new laptop with a clean install of Win10.
If I had one solid bit of evidence that strongly suggested a faulty drive I would have started the returns procedure on it without question, but unfortunately I don't. I think the next time I need to buy such a drive I'll be scrutinising its performance though.
The drive on the left is my own Seagate portable drive that I bought a few years ago that I have no problems with. The drive on the right is the one I've just bought and TBH that ATTO benchmark is the new drive on a particularly good day (and the drive in very general terms seems to be improving).
I was first alerted to a problem when I was transferring a backed-up user profile (so lots of small files in AppData followed by large image files in say the Pictures folder). I was seeing delays in robocopy that I'd normally associate with copying to a flash drive: tiny files taking a second or two each, then a second or two delay between each file from time to time. But then even more alarmingly the images were taking a good second or two each as well rather than the 'multiple files transferred per second' scenario I'd normally expect.
As I've recently migrated to Linux I was wondering if this might be related to that, but I had similar results in Windows 10 (I'm dual-booting and I have Win10 for gaming mostly). I'd see odd behaviours like transfer performance improving if I stopped it for a few minutes and let the drive settle down for a while. Sometimes I'd see a transfer rate of 60MB/sec for images, sometimes 6MB/sec for similar-size images.
That ATTO result is among the best I've seen for the new drive. One of the first ATTO benchmarks I did came out with 118 BYTES/SEC write and 3 BYTES/SEC read at the start of the test. It literally took a couple of minutes to post the first bar.
As that graph shows, it's like the drive is busy doing something else, but in Windows I'm watching the Task Manager graph for that particular drive, wait for it to completely flatline for 10 seconds then run the benchmark.
SMART data for the drive is fine, neither Windows or Linux reported anything in the logs to suggest a drive failure.
I had to transfer about 130GB of data to the drive for this customer. I stopped the transfer a fair few times and attempted the same transfer with the same cable and USB port to my own drive which was pushing things through at at least twice the speed for small files. Eventually I just let the newer drive get on with it, and it did at a fairly reasonable rate for the images (80-90MB/sec), but the random performance is kind of like how a friend put it, as if the drive was somehow having to format sectors on the fly before starting the copy. Just in case something bizarre was done to the file system, I have wiped that drive more than once (quick format NTFS) before doing things under Windows in earnest (like the final transfer). The wonky performance of that drive also showed up with a brand-new laptop with a clean install of Win10.
If I had one solid bit of evidence that strongly suggested a faulty drive I would have started the returns procedure on it without question, but unfortunately I don't. I think the next time I need to buy such a drive I'll be scrutinising its performance though.
Facebook
Twitter