- May 19, 2011
- 19,973
- 14,296
- 136
This is a pet project I'm considering if/when I have the time and inclination. There's also the cost of the equipment to consider.
I don't think much has been done publicly to point out just how poorly a lot of the currently available solutions perform, and also it's one of those times that I like to think I have a fair bit of experience on this topic but experiences can often turn into knowledge stored in bullet points and losing some of the substance over time and I'd like to review my beliefs. The bigger-name reviewers often like to focus on more expensive equipment as well such as reviewing a 4TB SSD when most people are unlikely to buy that size.
This thread is a bit of a fork of a previous thread I started:
forums.anandtech.com
My focus is likely to be on write performance as that's of most interest to me (e.g. backup systems). My beliefs on this topic are as follows:
A typical USB 2.0-connected flash drive can sustain about 6MB/sec.
A half-decent USB 3.0-connected flash drive can do about 17MB/sec (but how well can it sustain it?).
A typical USB 3.0-connected CMR portable hard drive can sustain about 100MB/sec.
A typical USB 3.0-connected SMR portable hard drive's write statistics are all over the freaking shop.
USB-connected SSDs can also be quite variable (e.g. DRAM-less, QLC).
Off the top of my head, I'm thinking of benchmarking the following equipment:
The cheapest 64GB flash drive I can find
The 64GB/128GB flash drive I usually sell to customers (Sandisk Ultra USB 3.0)
Some old CMR 2.5" HDDs I have
A new portable USB HDD (they're all SMR these days)
A new 2.5" CMR HDD in a USB external HDD enclosure
The cheapest SSD I can find and stick it in an enclosure (maybe SATA to keep things consistent, I have both M.2 SATA and M.2 NVMe enclosures handy though)
I'll likely throw in a new and old 3.5" CMR 7200RPM drive in for good measure.
I'll likely have a Samsung 8x0 PRO SATA SSD spare
A USB SSD if I can afford one
Off the top of my head, I'd run a couple of tests:
ATTO benchmark
Script a robocopy write of about 50GB of data consisting of very large files (I'll likely use a portion of my backed-up film collection), screenshot the modern Windows transfer throughput graph, note the robocopy transfer speed stats as well. Maybe the typical disk time usage in Task Manager too?
My own PC is an i5 Haswell build, though I'll likely use an old AM3 SATA 6G USB3 spare PC I have once I've confirmed that the write statistics aren't terribly dissimilar to my own PC.
Thoughts?
I don't think much has been done publicly to point out just how poorly a lot of the currently available solutions perform, and also it's one of those times that I like to think I have a fair bit of experience on this topic but experiences can often turn into knowledge stored in bullet points and losing some of the substance over time and I'd like to review my beliefs. The bigger-name reviewers often like to focus on more expensive equipment as well such as reviewing a 4TB SSD when most people are unlikely to buy that size.
This thread is a bit of a fork of a previous thread I started:
Discussion - External (portable / 2.5") hard drives and how to avoid SMR
I posted this thread a few years ago: https://forums.anandtech.com/threads/does-anyone-have-a-new-seagate-portable-usb-3-0-drive-to-test-with.2553793/ I thought it would be better to start a new one as the old thread title isn't that helpful for this topic. I started using Toshiba Canvio...

My focus is likely to be on write performance as that's of most interest to me (e.g. backup systems). My beliefs on this topic are as follows:
A typical USB 2.0-connected flash drive can sustain about 6MB/sec.
A half-decent USB 3.0-connected flash drive can do about 17MB/sec (but how well can it sustain it?).
A typical USB 3.0-connected CMR portable hard drive can sustain about 100MB/sec.
A typical USB 3.0-connected SMR portable hard drive's write statistics are all over the freaking shop.
USB-connected SSDs can also be quite variable (e.g. DRAM-less, QLC).
Off the top of my head, I'm thinking of benchmarking the following equipment:
The cheapest 64GB flash drive I can find
The 64GB/128GB flash drive I usually sell to customers (Sandisk Ultra USB 3.0)
Some old CMR 2.5" HDDs I have
A new portable USB HDD (they're all SMR these days)
A new 2.5" CMR HDD in a USB external HDD enclosure
The cheapest SSD I can find and stick it in an enclosure (maybe SATA to keep things consistent, I have both M.2 SATA and M.2 NVMe enclosures handy though)
I'll likely throw in a new and old 3.5" CMR 7200RPM drive in for good measure.
I'll likely have a Samsung 8x0 PRO SATA SSD spare
A USB SSD if I can afford one
Off the top of my head, I'd run a couple of tests:
ATTO benchmark
Script a robocopy write of about 50GB of data consisting of very large files (I'll likely use a portion of my backed-up film collection), screenshot the modern Windows transfer throughput graph, note the robocopy transfer speed stats as well. Maybe the typical disk time usage in Task Manager too?
My own PC is an i5 Haswell build, though I'll likely use an old AM3 SATA 6G USB3 spare PC I have once I've confirmed that the write statistics aren't terribly dissimilar to my own PC.
Thoughts?