The Test - Real World First

Before we get to our usual benchmarks I put together a few hand timed tests to help convey the experience the Momentus XT offers. Seagate badly wants the world to think that the Momentus XT is SSD-like in its performance. In my opinion it performs more like a fast hard drive rather than an SSD, but regardless of what you want to call it the tests below are designed to help convey the user experience.

To aid in our comparison I've thrown in two SSDs: the Intel X25-V and the Patriot Inferno. The X25-V is a small, but affordable SSD while the Inferno is based on the SandForce SF-1200 controller making it one of the fastest and most expensive consumer SSDs we've tested. We've also got the new 600GB VelociRaptor and Seagate's own Momentus 5400.6, a fairly standard 5400RPM 2.5" drive for comparison. While a better comparison would have been a modern 7200RPM 2.5" drive, we didn't have one handy in time for this review. The 5400.6 should give you a general idea of how all 2.5" drives will perform though; 7200RPM drives will be faster but not by a huge margin.

These are mature results, measured on the third of three runs so the Momentus XT has more than enough time to learn (between each run I rebooted the machine).

CPU Intel Core i7 965 running at 3.2GHz (Turbo & EIST Disabled)
Motherboard: Intel DX58SO (Intel X58)
Chipset: Intel X58 + Marvell SATA 6Gbps PCIe
Chipset Drivers: Intel 9.1.1.1015 + Intel IMSM 8.9
Memory: Qimonda DDR3-1333 4 x 1GB (7-7-7-20)
Video Card: eVGA GeForce GTX 285
Video Drivers: NVIDIA ForceWare 190.38 64-bit
Desktop Resolution: 1920 x 1200
OS: Windows 7 x64

Our first test is a timed boot. You are looking at the time from initial OS load (not POST) to getting a cursor at the Windows 7 desktop.

The Windows 7 boot process is a very read intensive task that's easy for the Momentus XT to learn. Seagate may be on to something here, the Momentus XT is much more like a SSD than a HDD in this test. Even the VelociRaptor is no match for it. The SSDs are still faster, but they don't offer nearly the same capacity as the hybrid drive. Compared to a standard 2.5" notebook drive the Momentus XT is a no brainer, it's in a league of its own.

Adobe's Photoshop CS4 is a great application launch test as it can take a fairly long time to load. Here we're just timing the period from clicking on the CS4 icon to ending up at the Photoshop workspace.

Application Launch - Adobe Photoshop CS4

The new VelociRaptor is slightly, but consistently faster than both of the SSDs in this chart. It does still have a few tricks left up its sleeve. The Momentus XT is about half the speed of the SSDs and 10K RPM HDD here, but much faster than the 2.5" 5400RPM notebook hard drive. It's noticeably faster than your standard notebook HDD but it does feel slower than a SSD. That is an important distinction as throughout my experience with the drive it felt like a fast hard drive, rather than an SSD. Part of the reason there is because the NAND isn't used for writes, so you still get the high latency response time whenever there are any random writes happening in the background.

Our standard Photoshop CS4 performance test in our reviews is optimized for comparing CPUs and as such it's set to only record the previous action. With only one level of history, the benchmark is not nearly as disk intensive as Photoshop can get. For this test I ran our CS4 benchmark but left the application at its default setting of remembering the previous 20 actions. With more to keep track of, performance goes down while the reliance on a fast storage subsystem goes up.

Photoshop Performance - Retouch Artists Speed Test

The X25-V is penalized by having a very low sequential write speed. The combination of heavy reads and writes means that the Momentus XT behaves more like a normal hard drive and less like an SSD. The 4GB SLC read cache does help though, while the hybrid drive isn't faster than the VelociRaptor it's pretty close for being a 2.5" notebook drive. The 100GB SandForce drive is much faster than anything else here, but it does offer 1/5 the space at 2.5x the cost of the Momentus XT.

For the next test I wanted to see how well the Momentus XT would cache a single large file. I timed how long it took to load a 70MB Excel sheet:

File Open - 70MB Excel Spreadsheet

The results are a lot closer than you'd expect. Despite having a 4GB SLC NAND read cache, the Excel load time isn't any better than the VelociRaptor (although it is faster than the 2.5" 5400RPM drive). The upper bound in performance is set by the SSDs so it looks like the Momentus XT's cache does as much as physically possible to improve performance, we're simply bound elsewhere.

My favorite test for showing the usefulness of an SSD is to load a bunch of applications immediately after booting to the desktop. On a system with a hard drive there are usually a lot of disk accesses right after you hit the desktop that will slow down any application launches. A well made SSD acts like nothing is going on in the background. But what about the Momentus XT?

To test it I threw Internet Explorer, Outlook 2007, Access 2007, Excel 2007, PowerPoint 2007, Word 2007 and Photoshop CS4 in the startup folder of my testbed. The times below measure how long it took to load all of those applications immediately after boot:

Load 6 Applications Immediately After Boot

I'm impressed. While it's not quite as fast as an SSD, Seagate's Momentus XT is faster than any hard drive here. It's fast enough in this test where you might even consider using one in a desktop instead of a VelociRaptor. There is still the issue of random write performance (it is a 7200RPM 2.5" drive after all) but this is very good. It's because of this that Seagate believes the Momentus XT is delivering a SSD-like experience and in this test I'd have to agree.

More 918 Spyder than Prius Sequential & Random Read/Write Speed
Comments Locked

120 Comments

View All Comments

  • enderwiggin21 - Tuesday, May 25, 2010 - link

    Almost 40% more at idle and ~33% more at load is most definitely "considerably more."

    Under 10% and *then* you're into the statistically insignificant category.
  • svend - Monday, May 24, 2010 - link

    Will the NAND still be effective when using the XT with an encrypted partition/filesystem? What affect would encryption have on the performance of the XT?
  • GullLars - Monday, May 24, 2010 - link

    Thumbs up to seagate for the sollution, the one used in previous hybrids that relied on the OS was not very effective.
    I actually made a thread on XS forum a couple of months back discussing a sollution like this, a HDD with a NAND read-only cache using LBA read pattern mapping as caching algorithm.
    In my discussion, i suggested using 8-16GB (1-2 dual plane dies) MLC ONFI 2.x NAND on a single channel for lower complexity. This would give up to about 7500 4KB random read IOPS (30MB/s), and 90-180MB/s sequential read (for cache hits). At a cost of roughly 2-3$/GB + a NAND controller, the added price tag would be around 25-50$.
    I think this could be interresting for higher capacity 3,5" "black" drives as well as high-performance high capacity notebook HDDs.
  • auhgnist - Monday, May 24, 2010 - link

    Hybrid HD seems similar to Intel Turbo Boost technology, in terms of both SSD capacity and NAND type, except that Turbo Boost allows certain level of control such as caching only data accessed some pre-given programs. Any idea how these two compare in terms of performance?
  • joshv - Monday, May 24, 2010 - link

    I mean seriously. Add another 4GB of RAM and let the OS do the caching. Sure, it won't help with boot, but for everything else it should be a lot faster than some SLC NAND on the HD.
  • enderwiggin21 - Tuesday, May 25, 2010 - link

    Or how about both?
  • mpx999 - Monday, May 24, 2010 - link

    It would be interesting to compare performance of hybrid SSD with software-based flash read cache schemes. There are basically two of them for Windows: built-in Readyboost and third-party eBoostr. Other operating systems have even better mechanisms, eg. Solaris has L2ARC caching scheme for reads and ZIL for writes.

    http://blogs.sun.com/brendan/entry/test

    http://dotnet.agilekiwi.com/blog/2008/10/how-eboos...

    Both Windows caching mechanisms can work with either USB sticks, or memory cards in built-in readers. Second option is very interesting as it means SLC flash cards can be used.

    I suspect hardware controlled read-cache as implemented in this hybrid drive will totally beat Windows software -based mechanism in one important application: running virtual machines. It's because hardware cache operates on disk blocks so it doesn't matter it these blocks are from base machine or virtual machine, while software Windows mechanisms rely on file-based statistics, so they only speed up base machine. One would have to create separate cache for each VM to get speedup. Btw. software-based Solaris caching doesn't have this limitation.
  • vision33r - Monday, May 24, 2010 - link

    It's nowhere need a true SSD in performance and just offers the cheap space of a spindle drive.

    The the same has having a SSD internally and carrying a spindle drive externally.

    Once SSDs gets cheaper, solution like this will be less interesting.

    I expect SSD prices to drop from $99 for a 40GB to like $60, then we'll see 200GB SSD for like $150 or less.
  • jmv2009 - Monday, May 24, 2010 - link

    I don't think writing through the ssd is that important, as it can be effectively cached using the main memory of the system. Only for sustained writes >1 Gb this would become slow if the caching works well.
  • nsravan - Monday, May 24, 2010 - link

    The site had not much activity since you went on vacation. So I took one too :)

Log in

Don't have an account? Sign up now