RS results for misc old hard drives (1 SSD, 1 hybrid, 4 spinning)

  • Be sure to checkout “Tips & Tricks”
    Dear Guest Visitor → Once you register and log-in:

    This forum does not automatically send notices of new content. So if, for example, you would like to be notified by mail when Steve posts an update to his blog (or of any other specific activity anywhere else), you need to tell the system what to “Watch” for you. Please checkout the “Tips & Tricks” page for details about that... and other tips!

    /Steve.

mfalkvidd

IoT consultant
Sep 24, 2020
8
3
Code:
Driv Size  Drive Identity     Location:    0      25%     50%     75%     100
---- ----- ---------------------------- ------- ------- ------- ------- -------
 81  4.0TB ST4000DM005-2DP166            177.6   181.4   161.8   133.7    89.3
 82  256GB Samsung SSD 840 PRO Serie     537.8   538.2   538.3   538.7   538.1
 83  2.0TB WDC WD20EARS-00S8B1           110.6   103.6    89.6    72.3    49.3
 84  500GB Hitachi HDP725050GLA360        89.7    87.3    81.2    69.3    49.3
 85  4.0TB ST4000DX001-1CE168            150.4   163.0   145.5   116.4    77.5
 86  500GB Hitachi HDP725050GLA360        90.8    89.3    81.1    70.3    50.6

                  Benchmarked: Monday, 2021-01-18 at 23:17

81 has 21,378 power-on hours according to the SMART data. It is a 7200rpm drive rated for 190MB/s max and 161MB/s average. I use it as a target for automatic backups of work files. SATA/600. Purchased 2017-07-13.
82 has 41,487 power-on hours and serves as C: on my main work station. It has written a total of 52,129 GB (203x its size). SATA/600. Purchased 2013-04-07.
83 has 59,419 power-on hours. SATA/300.
84 has 38,968 power-on hours. SATA/300.
85 has 29,548 power-on hours. It is a hybrid disk, with a 8GB MLC cache. This is where I store all my work files. SATA/600. Purchased 2015-06-28.
86 has 45,025 power-on hours. It serves as a target for automatic Windows backups. SATA/300.

Conclusions/thoughts
* The Samsung 840 PRO really delivers. No slowdown, even after many years of use. The disk is usually 85-90% full, so there is not much slack space.
* I think the SSD cache in the hybrid disk doesn't contain the regions tested by RS, so what we're seeing is the performance of the spinning disk. If I would run RS multiple times, the disk would probably cache the regions that RS reads and we would see the performance of the SSD cache instead.
 
Last edited:

DarkwinX

Well-known member
Sep 17, 2020
57
9
Melbourne, Australia
The Samsung 840 PRO really delivers. No slowdown, even after many years of use. The disk is usually 85-90% full, so there is not much slack space.
@mfalkvidd what kind of usage do you have with this disk? Is it doing alot of writing to all parts of the disk? What we have been finding is that a write triggers the SSD to "fix itself" by doing something behind the scenes. Anecdotally on the SSDs I've seen as O/S disks; the front part gets slower over time and because the back never gets written to, the firmware knows it's blank and therefore just returns the fastet possible result.

I think the SSD cache in the hybrid disk doesn't contain the regions tested by RS, so what we're seeing is the performance of the spinning disk. If I would run RS multiple times, the disk would probably cache the regions that RS reads and we would see the performance of the SSD cache instead.

I know much less about hybrid drives so can't make too much comment here. On a more traditional HDD, ReadSpeed does a large enough transfer in one go to force the drive not to cache the read. So perhaps in this case since it's a 1GB transfer per location, it's just smoothing the inner sectors where the reads are slower..
 

mfalkvidd

IoT consultant
Sep 24, 2020
8
3
@mfalkvidd what kind of usage do you have with this disk? Is it doing alot of writing to all parts of the disk? What we have been finding is that a write triggers the SSD to "fix itself" by doing something behind the scenes. Anecdotally on the SSDs I've seen as O/S disks; the front part gets slower over time and because the back never gets written to, the firmware knows it's blank and therefore just returns the fastet possible result.
As I said, this is the C: drive of my workstation so it has the typical writes for such usage. Windows was installed on 2013-04-20 and has applied all updates over the almost 8 years it has been used. I use the machine almost daily. It has booted 1,239 times. There are 67,944 hours since Windows was installed, so the machine has been on for about 61% of the time.

The disk has been close to 100% full at some times, so I am pretty sure all parts of the disk have been written to. Unless the disk is doing some things to actively avoid writing to some sectors, I think it would be very surprising if there are sectors that have not been written after writing more than 200x the disk size.

There is a paging file on the disk as well, but it doesn't get used much because I rarely use up all of the 32GB ram my system has.

I know much less about hybrid drives so can't make too much comment here. On a more traditional HDD, ReadSpeed does a large enough transfer in one go to force the drive not to cache the read. So perhaps in this case since it's a 1GB transfer per location, it's just smoothing the inner sectors where the reads are slower..
The SSD cache is 8GB large, and RS reads 4x1GB out of a total of 4TB (0.1%). Statistically, the impact of the cache on such reads should be very small since the cache is optimized for my usage, not RS's usage.

But if I run RS a lot of times, the disk might evict everything else and prioritize the 4GB that RS reads and put all of those 4GB in the cache. I don't want to try that though, since that would evict the data I use the most.
 

DarkwinX

Well-known member
Sep 17, 2020
57
9
Melbourne, Australia
The disk has been close to 100% full at some times, so I am pretty sure all parts of the disk have been written to. Unless the disk is doing some things to actively avoid writing to some sectors, I think it would be very surprising if there are sectors that have not been written after writing more than 200x the disk size.
Right, well fair enough. And from installed from 2013 too, hopefully as this work continues we'll learn a bit more about what's happening under the hood and why we see slowdowns vs why we don't.

The SSD cache is 8GB large, and RS reads 4x1GB out of a total of 4TB (0.1%). Statistically, the impact of the cache on such reads should be very small since the cache is optimized for my usage, not RS's usage.
From the maximum read performance on these drives it looks like your numbers are pretty spot on, especially at the front of the disk. I would say the cache is pretty close to doing all it can do.

Is that also how the SSD cache on these disks work? Do they permanently cache information based on usage? I guess it's a good time to read up on this for me :).
 

miquelfire

I like red!
Sep 26, 2020
51
5
www.miquelfire.red
Is that also how the SSD cache on these disks work? Do they permanently cache information based on usage? I guess it's a good time to read up on this for me :).
I think it depends on the drive. There was a period where the hybrid drives basically had a section that was solid state, and the rest as spinning. Those drives would show SSD speeds in region 0, and the rest was spinning. (Those drive might have had enough SSD to almost reach the 25% region with the smaller versions, though this was a period where 1TB SSDs were around $200-$300 for a cheap model)
 

mfalkvidd

IoT consultant
Sep 24, 2020
8
3
I wouldn't call the caching permanent. The disk will evict cached sectors that are not used and replace them with sectors that are used more frequently. But maybe it's just a case of semantics. I would call the caching persistent (the cache will still be there after the disk has been powered off) but not permanent.

The specs contains the following numbers:
  • Average read speed of the NAND media: 190MB/s
  • Max sustained data rate, OD read: 180MB/s
  • Average data rate, read, all zones: 146MB/s.

RS's numbers say 145.5MB/s for the middle of the drive, which closely matches the average data read for all zones. RS's top performance for this disk is 163MB/s which indicates that the cache isn't helping much; if we had a lot of cache hits we should get close to the 190MB/s the NAND can deliver.

I am not sure what Seagate mean by OD read though.
 
Last edited: