SSHD - Performance comparison with HDD and SSD

I have removed the original posting from here because it was polluted with Adware. But you can still access it from my One Drive - without Adware. I don't want my good name associated with advertisement for questionable products.

Link: SSDH - Perfornance comparison.

Interesting results but not very meaningful for typical real world conditions. The test was basically measuring only transfer speed which is the ideal case for a conventional HD. Under typical workloads there is a great deal of reading of small sections of files spread over the disk surface. That requires a large number of head seeks, a weakness of conventional drives and a strength of SSds. SSHD drives will have also have an advantage here but to a lesser extent then an SSD. How much will depend a great deal on the nature of the workload.

That's what I said at thje end of my evaluation. One day I will image the OS on the SSHD and compare it to my SSD. But as I said, that is a bigger job because I will have to physically change disks in my laptop (that's what I would use because I do not fancy to crawl under my desk where my desktop is).

Any other useful comments?

My experience with an SSHD was somewhat different. When I first installed it in my notebook, boot times didn't improve much but program loading was much faster and access of large blocks of data by a couple of programs (MediaMonkey and caliber) improved quite a bit. However, mostly because I kept deleting and adding large blocks of data in a short period of time, the NAND in the SSHD apparently became overwhelmed and everything slowed back down to HDD levels.

This is not to say SSHDs are not any good. They just aren't for everyone. If one uses a limited number of programs and pretty much does the same thing every time they use the computer, then they will see a dramatic improvement. I my case, I was better off getting an SSD.

LF, this is valuable input. I was sure that the SSHD is going to behave differently in different situations.

After the holidays I am going to image the OS from my laptop's OCZ Vertex III to the SSHD. Then I am going to run some benchmarks on the Vertex first. Then I switch the disks and rerun the same benchmarks on the SSHD.

Problem is that the OS does not register a lot of performance data. Apart from the boot and shutdown times in Events 100 and 200 in the Event Viewer, I know no other performance data that is recorded. Maybe I can find some benchmark program for the job.

The problem with benchmarks is they do not always reflect real use performance. Reports of seat of the pants performance are far more meaningful. Benchmarks are useful, however, as an indicator of how performance later compares to performance earlier; a drop in the benchmark provides a tangible indicator of when something is cause a drop in performance; basically, a diagnostic tool..

I think anything you measure is going to be a bit artifical because the next guy may have a completely different setup and different requirements. I just like to provide those numbers for whatever they are worth - it's better than armwaving.

Benchmarks can be useful but creating a good benchmark and properly interpreting the results is difficult. Unless very carefully done the results may be unduly influenced by unrelated factors. And unless the results are properly interpreted they may have little validity in real world situations. I would expect the proper evaluation of an SSHD to be particularly difficult.

whs said:

That's what I said at thje end of my evaluation. One day I will image the OS on the SSHD and compare it to my SSD. But as I said, that is a bigger job because I will have to physically change disks in my laptop (that's what I would use because I do not fancy to crawl under my desk where my desktop is).

Click to expand...

The Seagate sshd uses the nand for a cache, not for permanent file storage. The 1tb has 16gb nand and the 500gb has 8gb nand. I have 2 of the 500 gb sshd's and they have learned to load windows almost as fast as my 840 pro, but seldom used programs are the same speed as a straight spinner.

I bought the first one for $50 as a factory refurb for my lap top, after using for a few weeks I bought a second one for the ball and chains desk top.

Your experience confirms the caching nature of the SSD part. It essentially works like the caching of Superfetch in RAM. Once you used a function/program and it has found a place in the cache, the SSD speed applies. But stuff that is used for the first time has to be first found and cached from the HDD.

I paid 90 odd Dollars for the 1TB model.

I used a 750GB Seagate MomentusXT SSHD with an 8Gb SSD cache as the data drive where I have the OS on a 128GB Crucial M4 SSD. I didn't keep any detailed stats on the set up. I ran it for a few days but I could tell no difference in app loading, accessing data, saving data.

After about a week I run AS SSD on the SSHD and it scored considerably lower than the HDD I had been using, Seagate Momentus. (Both are 2.5" drives)

My conclusion is that a SSHD would likely do much better as an OS drive than for a data drive. There's not much of the same data accessed on the data drive so the SSHD controller I suspect would be swapping data in and out all the time. An OS on the other hand re-uses data often, e.g. boot files, browsers, email and so on.

Britton30 said:

I used a 750GB Seagate MomentusXT SSHD with an 8Gb SSD cache as the data drive where I have the OS on a 128GB Crucial M4 SSD. I didn't keep any detailed stats on the set up. I ran it for a few days but I could tell no difference in app loading, accessing data, saving data.

After about a week I run AS SSD on the SSHD and it scored considerably lower than the HDD I had been using, Seagate Momentus. (Both are 2.5" drives)

My conclusion is that a SSHD would likely do much better as an OS drive than for a data drive. There's not much of the same data accessed on the data drive so the SSHD controller I suspect would be swapping data in and out all the time. An OS on the other hand re-uses data often, e.g. boot files, browsers, email and so on.

Click to expand...

That was close to the experience I had and came to the same conclusion, only I had the SSHD partitioned off for data and OS and, initially, app and OS loading was faster. It would have worked out better if my data had been far more static.

Loading an application is a very demanding task for a conventional hard drive. Modern operating systems do not initially load all of the application into memory but only those portions when they are needed. Loading an application means loading a few blocks of the exe, a few blocks from a dll, back to the exe for a few more blocks, and so it continues until the application is running. Modern operating systems have many optimizations to make the process more efficient but the basic principles remain the same. Once data is in memory the OS will try to keep it there for as long as possible, consistent with other demands on memory. This saves a great deal of memory. Modern large applications contain a great deal of code that is rarely used, and for some individual users not at all. Keeping all that unused code in memory all the time is wasteful and would impair performance. This method is not new, having been used in virtually all versions of Windows for the last 20 years.

This means that a conventional drive will be doing a large number of seeking to the data needed. Modern drives have high transfer rates but that doesn't matter much when you read only a few blocks at a time, then it is off to read something else. An SSD is in it's element for this kind of thing as seek times are almost non-existent. That means the drive can spend much more of it's time transferring data, not moving a mechanical head to where it is found.

Reading data tends to be more serial in nature so there is less head movement. An SSD will still be a benefit but not so much. And considering the current cost of SSDs putting all of your data on on such a device is rather expensive.

In addition to the caching on the SSD part of the SSHD there is also caching in RAM by Superfetch. If you have a lot of RAM on the system, that can be considerable. It is probably impossible to divide the SSD effect from the RAM effect.

The disc speed of the 1TB SSHD is 7200 rpm not 5400 rpm, the 4tb version is 5900 rpm.
http://www.seagate.com/staticfiles/support/docs/manual/desktop%20sshd/100726566b.pdf checkout page 11.