Tests to detect SMR dives?

Recommended Posts

How would one determine by tests whether a HDD uses SMR?

What I gathered is:
 - 4K writes (like in CrystalDiskMark) benchmark gives about 10 times higher values that reads ( 10MB/s vs 1MB/s)
 - write access time (like measured by HDTune) is 10 time less than "normal" (e.g. 0.8 ms)
 - sequential writes are 3 times slower than reads (I could not confirm this yet conclusively on my drive)

Any other ideas?

I did my tests on a Seagate ST5000DM000 5TB drive taken out of an eclosure (Seagate Desktop Expansion 5TB USB3.0 drive - STEB5000200 ).

Edited by xerces8

Share this post

Link to post
Share on other sites

SMR drives should behave slower than normal, not faster, once they run out of buffer space. I don't think the CDM run time is long enough to exhaust it or not? (I have no idea...) The performance improvements you're seeing smell like cached writes, not SMR/PMR differences.


Also note that Seagate does seem to be refining their SMR implementation, as Anandtech found here, so earlier tests that could exhaust the buffer space might no longer work with improved SMR drives:


Share this post

Link to post
Share on other sites
1 hour ago, Brian said:

What you're normally see is very low random write performance. 

Also this scientific paper* claims the opposite:


Cache type and size: The drives use a persistent disk
cache of 20GiB and 25GiB on the 5TB and 8TB drives,
respectively, with high random write speed until the cache
is full
. The effective cache size is a function of write size
and queue depth.


* https://www.usenix.org/conference/fast15/technical-sessions/presentation/aghayev

Share this post

Link to post
Share on other sites

From your link:


PDF page 20, random write latency shows and explicitly says the real SMR drive is actually is high (slow), averaging somewhere between 25 and 30ms (hard to tell, the graphs are crude) with a massive spike out to 325ms (!) for random write latency, while a standard PMR drive is more around 10ms. Page 21 explains the necessary movement of the heads out to the cache map.


The goughlui.com review you linked doesn't seem to actually exhaust any of the PMR cache so the benchmarks on it, while accurate, are of limited use if you actually want to evaluate SMR's full performance impact-- neither ATTO nor CDM are terribly useful here. To be fair I'm not sure what an exceptional CrystalDiskMark 4K write number is going to be?

Link below shows up to about 5 to 6MB/sec for 4K writes for most drives in that link (which are all PMR) and similar for reads; the 10x difference you mention makes sense given that the writes would be in a cached area (PMR) vs. reads could probably be in a SMR area.


Share this post

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now