Brian

Seagate Momentus Hybrid Hard Drive Review

Recommended Posts

Hard drive manufacturers have seen the writing on the wall; embrace flash storage or suffer the consequences. Seagate is one of the few remaining hard drive manufacturers who haven’t released a line of consumer SSDs. Today though, they have come out with a cross between flash and magnetic storage known as a solid state hybrid drive (SSHD). The hybrid hard drive concept takes a standard spindle hard drive and adds flash and pattern recognizing software to boost drive performance with repetitive tasks. But will 4GB of flash memory sitting on top of a 7200RPM drive offer a big performance gain? Read the full Seagate Momentus XT review to find out.

Full Review

Share this post


Link to post
Share on other sites

Hard drive manufacturers have seen the writing on the wall; embrace flash storage or suffer the consequences. Seagate is one of the few remaining hard drive manufacturers who haven’t released a line of consumer SSDs. Today though, they have come out with a cross between flash and magnetic storage known as a solid state hybrid drive (SSHD). The hybrid hard drive concept takes a standard spindle hard drive and adds flash and pattern recognizing software to boost drive performance with repetitive tasks. But will 4GB of flash memory sitting on top of a 7200RPM drive offer a big performance gain? Read the full Seagate Momentus XT review to find out.

Full Review

Our second real-life test covers disk activity in a productivity scenario. For all intents and purposes this test shows drive performance under normal daily activity for most users. This test includes: a three hour period operating in an office productivity environment with 32-bit Vista running Outlook 2007 connected to an Exchange server, web browsing using Chrome and IE8, editing files within Office 2007, viewing PDFs in Adobe Reader, and an hour of local music playback with two hours of additional online music via Pandora.

'real world tests' lol. &seems like you throw together a bunch of commonly used Windows progs, then say that is the way people use their computers in the 'real world'. From a stand point of consistence, for testing purposes, you get that, just like the more 'static' tests.

What I would like to see is a set recording tests that casual users can do, that can be put into a SR database. People the buy a certain drive, note their system hardware/software setup, then install these progs that record their usage over time periods, like say a week or month. Then you get 'real world' usage considerations with usable "averages" <probably end up with a wide range of statistical averages, but that would be good to know, not 'made up' scenarios...who really runs *all* of those above, at the same time, everyday???

http://news.cnet.com/8301-17938_105-20005669-1.html

With these specs, the new Momentus XT offered great performance in our tests, even faster than that of some SSDs. The best part about the drive is the price: at just around $135 for the 500GB version, the Momentus XT costs just a fraction of the price of an SSD that has just one fourth of its storage space.

^I'll assume the prices quoted in the SR article are direct from Seagate marketing?

Since the 750GB 7.2k Momentus should be shipping soon, I'd hold off on a purchase of the hybrid 500GB, to see if the gap in performance closes much vis a vis the 500GB XT vs 500GB 7200.4...many of those performance test, 'real world' included as shown in the SR review; don't seem as dramatic an increase as you would see with an SSD (keeping mind some people perceive the SSD's being only 'somewhat' faster that HDD's in their own 'real world' usage...YMMV).

Share this post


Link to post
Share on other sites

WOW this is like Déjà vu. Didn't we have this discussion before. I would have gone with a 4GB DIM + a capacitor then the NAND memory then at least you could have the precived performance increase on the writes, at least up to a point.

Share this post


Link to post
Share on other sites
^I'll assume the prices quoted in the SR article are direct from Seagate marketing?

$156 is the MSRP for the 500GB, but NE has them now for $129+$6 shipping.

Since the 750GB 7.2k Momentus should be shipping soon, I'd hold off on a purchase of the hybrid 500GB

We have one en route for review as well.

Share this post


Link to post
Share on other sites

Does the 320 GB unit use 1 or 2 platters? The 500 GB one uses 2 I guess, so the 320 GB one would be faster. Much worse price / GB, though.

And generally: kudos to Seagate for bringing us such a drive. The notebook market seems to be the perfect place to start such a product.

And as Anand says: write caching would be sweet, but also requires a much more complex controller. Other companies already have them, but Seagate preferred the simple route here.

MrS

Share this post


Link to post
Share on other sites

So what happens when you take the circuit board off the XT and stuff it on the 7200.4? :)

We have been toying with that idea to prove the point, but what if the new drive had a slightly newer revision of some part that ended up getting fried? We were hoping to get an additional drive for further testing, if we had a backup we might consider it. We just didn't want to nuke the drive before the review was finished and additional questions answered.

Share this post


Link to post
Share on other sites
MrSpadge' date='25 May 2010 - 06:29 PM' timestamp='1274826585' post='261989']

Does the 320 GB unit use 1 or 2 platters? The 500 GB one uses 2 I guess, so the 320 GB one would be faster. Much worse price / GB, though.

And generally: kudos to Seagate for bringing us such a drive. The notebook market seems to be the perfect place to start such a product.

And as Anand says: write caching would be sweet, but also requires a much more complex controller. Other companies already have them, but Seagate preferred the simple route here.

MrS

I had to double-check the platters with Seagate, here's the breakdown by capacity:

500GB - 2 platters (250GB each)

320GB - 2 platters (one 250GB platter, and just one side of the second platter)

250GB - 1 platter

Share this post


Link to post
Share on other sites

I'm unclear-- did SR test the drive both in a clean, "un-learned" state as well as say after running the tests 3 times to test the "learned" state?

Share this post


Link to post
Share on other sites

I'm unclear-- did SR test the drive both in a clean, "un-learned" state as well as say after running the tests 3 times to test the "learned" state?

Correct. The slowest recorded speed mentioned in the text is the unlearned state and the score shown in the benchmark charts are those which were seen after it had reached its peak "learned" state. Our traces are expansive enough that the drive started to show signs of caching during the first test.

The standard IOMeter and CrystalDiskMark benchmarks had no learning phase. They could be repeated over and over without any improvement since they have no defined pattern to work from.

Share this post


Link to post
Share on other sites

Thanks for the clarification, Brian. If the new 750 GB 7200.4 uses 2 platters (it's not said to be any taller than standard drives) I wonder why they're not using these higher density platters for the Momentus XT. Could be as trivial as "development time", though ;)

MrS

Share this post


Link to post
Share on other sites

Correct. The slowest recorded speed mentioned in the text is the unlearned state and the score shown in the benchmark charts are those which were seen after it had reached its peak "learned" state. Our traces are expansive enough that the drive started to show signs of caching during the first test.

The standard IOMeter and CrystalDiskMark benchmarks had no learning phase. They could be repeated over and over without any improvement since they have no defined pattern to work from.

is there any indication how much the drive can learn? for example, say a guy booted his laptop and play wow the same amount of times... the usage patter is this; boot up, load wow, close wow, turn off. boot up, load wow, close wow, turn off. over and over again.

would it benefit from neither or show slight benefits to both? for instance, did you run StorageMark 2010 (SM10) HTPC until it learned it, then you ran SM10 Productivity until it learned that... then you ran SR10 HTPC again and found it had forgot all it learned from its first encounter?

Share this post


Link to post
Share on other sites
MrSpadge' date='27 May 2010 - 12:29 PM' timestamp='1274988541' post='262042']

Thanks for the clarification, Brian. If the new 750 GB 7200.4 uses 2 platters (it's not said to be any taller than standard drives) I wonder why they're not using these higher density platters for the Momentus XT. Could be as trivial as "development time", though ;)

MrS

i searched around for the 750GB 7200.4 and only found one reference to it. i found it mentioned on Seagate Product Family Overview PDF. however, i couldnt find it anywhere else. there was no mention of it in Seagates 7200.4 Product Manual.

the Product Family Overview did say the 750GB was an advanced capacity drive using 4k sectors.

that on top of the fact i cant find it in any retail channels (where the XT is readily available) leads me to believe the 750GB is still under development.

more then likely the wanted to push out the XT on the current 7200.4 platform instead of waiting for the RD completion of the 750 7200.4.

in short, waiting for the 750 7200.4's completion would have delayed the release of the XT.

Share this post


Link to post
Share on other sites

The 750GB Is indeed a 9.5mm drive and is Advanced Format (4K) and will start shipping in June. There are issues still with SF, so that's why I'm sure they didn't want to push more new tech into the XT and risk compatibility problems.

@pico - we don't know their model for data, but it depends on the frequency of the files accessed, and to some degree their size. What you might see is a caching of some of both types of files, system and WOW. Now...I asked Seagate about ways to manually flush the cache, and there's not one. However, they did indicate that if we ran the same benchmark three times for instance, that it should be fully cached. So in that way we could make sure the cache was full and then we could run something else for the first time, knowing it was a clean first run.

Share this post


Link to post
Share on other sites

this sounds like the drive i have been looking for, for my HTPC. it sounds perfect for that. the system stays on most all the time with only seldom reboots and only runs a few programs routinely. i am really looking forward to getting one.

Share this post


Link to post
Share on other sites

And now they're $130 for 500GB. I'd say for most notebook users, this is really a nice improvement over standard HDDs.

Share this post


Link to post
Share on other sites

"Phillip’s head screws" (I know these as "Phillips", sans apostrophe), "Windbond DDR RAM cache module" (last time I checked the company was still called Winbond, though IIRC the parent company is another these days), "The area the Seagate Momentus XT really shined " (cringe) - methinks a pinch of proofreading wouldn't have hurt for this article.

Share this post


Link to post
Share on other sites

Made the copy edits...the job's yours BTW if you want it. :)

We had 4 days with the drive and did better than most of our competition to get a competent review out - with real world testing that some couldn't figure out with the SSD cache and all. But yeah, thanks for the grammar edits.

Share this post


Link to post
Share on other sites

In the productivity and HTPC benchmarks, can you go into a bit more details about the testing methodology? Did you reboot the computer in between runs? Can you post the improvements between runs?

Did you use a similar methodology (ie. multiple runs) when testing the Momentus 7200.4 and the VelociRaptor? They might not have the 4GB cache of the XT but the Windows SuperFetch service also learns your access pattern and pre-fetches data into memory to speed up disk access.

Sorry about all these questions but I'm trying to figure out whether the 4GB NAND cache on the XT is any better than having lots of RAM on the laptop and just letting SuperFetch do its magic.

Share this post


Link to post
Share on other sites

Happy to answer some of those questions. For all drives we use an average of three fastest-in-a-row benchmark results without any reboot. Charts are created, review finalized, and then benchmarks are again retested the night/day before the review is published to make sure results are still at the same level as tested prior. To date we have not had any problems come up but its still one of those odd rituals ;)

The review system is loaded with 6GB of DDR3 memory and we have not seen pre-fetch enter the equation on other drives being tested. The only benchmarking difference between this drive and the Raptor for example was that this drive took more benchmark runs to reach its peak speeds, where others will be at top performance for runs 1, 2 and 3.

Share this post


Link to post
Share on other sites

Thanks for the details TSullivan.

Did you find that the Momentus XT got significantly better after several runs or did the other drives (like the 7200.4 or VelociRaptor) improve by just as much?

After doing a bit more research, it looks like SuperFetch is designed to pre-load boot and/or program files so this is probably why synthetic benchmarks like the Storagemark traces aren't affected.

Share this post


Link to post
Share on other sites

The other issue so I read was that if you do any defragmentation it resets SuperFetch because of the way the drives logic works. In the end the performance is at worst the same as an existing drive of its class, at best some reads show near SSD performance. I will go back what I said originally by the time is makes sense you in effect have an SSD.

Share this post


Link to post
Share on other sites

SuperFetch refers to Windows 7s own prefetching logic, which loads stuff into memory in advance (if there's enough available) because it knows you're usually using it. So it's something which might very well interfere with HDD benchmarks, especially if you run real world tests repeatedly and have a testing machine with lots of RAM.

I will go back what I said originally by the time is makes sense you in effect have an SSD.

So the Momentus XT doesn't make sense for you? If I wanted to get a new high quality laptop now I'd still be looking at 1000+€. And upgrading the HDD to a Momentus XT wouldn't increase the overall cost all that much, but would greatly increase performance. On the other hand going for a pure SSD would either be too small or painfully more expensive. One could argue that a 128 GB SSD is not that much more expensive and that it should really be big enough. But having 500 GB available would certainly make me feel much more comfortable, since I wouldn't have the option to swap things out to another HDD, like on a desktop.

MrS

Share this post


Link to post
Share on other sites
MrSpadge' date='02 June 2010 - 09:20 PM' timestamp='1275510028' post='262115']

He already said the other drives perform virtually identical during runs 1 to 3, so SuperFetch is apparently not interfering.

MrS

Assuming the traces were captures in Vista or Win7, then SuperFetch may have been active when the traces were captured, so some of the recorded accesses will be for loading data into the SuperFetch cache. And that data shouldn't have needed to be read form the hard disks again during the rest of the trace. But that's just part of what makes it a "real world" test, since a real world system would also be using SuperFetch. It doesn't make the trace any less useful or representative.

The encouraging thing about the benchmark runs being the same is that it means that there's NO evidence that SuperFetch is, for example, caching the trace being played back so that next time the benchmark runs, it's mostly reading from the SuperFetch cache instead of from the drive being tested.

As long as the machine on which the trace was captured had a sensible amount of memory, then SuperFetch being enabled is actually better, since it will more accurately represent performance in a real system with a similar amount of memory with SuperFetch enabled. I wouldn't disable SuperFetch on a real system just to stress the hard drive more, and doing so on the test system would just give less realistic results.

However, in the drive being tested, the Momentus XT, it's perfectly reasonable that the drive should be able to, and allowed to, cache whatever it sees fit given the real world pattern of accesses. The result would be that it caches exactly what it would in the real world, and so the benefit of that caching on subsequent runs would also be present if the exact same behaviour was repeated in the real world as well. The only reason you don't see this effect in the other drives is because the cache is small and non-persistent, so while it may help with repeated accesses of the same data within the trace, it's less likely to still have data in the cache from the beginning of the trace when the next benchmark run starts, so it's effectively starting from scratch each time. The Momentus XT, with its big, persistent cache, clearly does still have data cached from the beginning of the trace, and so we see the benefit of that in the subsequent benchmark runs.

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