I've finally jumped aboard the Momentus XT wagon. Unlike Seagate's desktop drives, these still have a 3 year warranty. I'm installing a 500GB SATA2 model (running SD28 firmware) into someone's laptop in a few days. In the meantime I'm running benchmarks and other tests on the drive, to make sure it holds up.
It's currently installed inside my AMD X6 1055T AM3 build (Sabertooth 990FX - SB950) running Catalyst 12.8 drivers, in AHCI mode.
First, standard HDTune read tests - First Run:
Caching kicks into full for the second run, delivering astounding access times thanks to HDTune's random read test always selecting the same sectors. (That makes it more comparable between drives, but the Momentus XT just loves it) I think access times will be somewhere in between for real usage.
HDTune's RandomAccess test can reveal a lot about how well a drive performs - First Run:
The Momentus XT seems to focus its efforts on caching smaller reads, which normally drag the head away for very tiny amounts of data, slowing the whole drive down. Caching them is smart, and is probably the best use of the limited 4GB cache. It doesn't cache any 1MB+ reads, which perform fine thanks to it's reasonable sequential speeds.
When I run Short-stroke tests in HDTune, it selects different sectors to read from, which requires new data to be cached.
But again, after a single run it seems to be performing at full efficiency. Granted, HDTune is an easily detectable synthetic benchmark - for application benchmarks I've heard it takes 1-2 runs to reach full speed. (I have not yet run any such tests) Likewise, rebooting 3 times is the way to trigger it to cache your boot data as optimally as possible.
I ran CrystalDiskMark on the XT, but for some reason it decided not to cache anything at all. There was no difference between runs beyond a few percentage points, which is within the margin of error.
I know from testing my WD Green drives (with 64MB cache) that if the test set fits within a drive's RAM cache, performance is vastly improved. The 500MB Momentus XT only has 32MB of RAM cache, but these numbers do indicate improved performance when working with single files. (Such as when editing a single word document, and not doing much else...)
I can think of very few scenarios where you ever see a drive's cache speed. I suppose accidentally closing a word document, then instantly re-opening it would be the main one. The program will technically still be in RAM, and the document is still in HDD cache. It should re-open almost instantly.
My WD20EARX 64MB 2TB Green, for interest:
If you believe those numbers, my WD Green stomps all over many older SSDs!
Actually, I suppose technically it does... when working with a single file that fits in its cache.
I ran AS SSD for good measure. It didn't seem to be able to directly access the drive for the read access time benchmark, but the other numbers look good for a 2.5" 7200RPM laptop drive:
I've heard that the SLC cache on the Momentus XT is only used for read caching. After concluding my read tests, I wiped out the drive's partition and ran some HDTune write tests. The results varied little from run to run, with access times never dropping below a few milliseconds. That supports the SLC being used only for read caching.
Write speeds look good for a laptop drive. I ran some short-stroke tests, to get an idea how the drive would perform with a smaller OS partition, under write heavy workloads:
Overall I think my customer will be quite happy with this drive.
Performance is reasonably good for a 2.5" drive, even without the caching. (It's only slightly behind a 500GB Scorpio Black.) As soon as it figures out his most frequently used programs and tasks, it'll be even faster.
An SSD would stomp all over this drive in performance, but wouldn't actually suit him. For one, it's going into an older Mac that doesn't support TRIM. Also, he has more than 200GB of Data (and increasing), making SSDs an expensive proposition. 256GB is too small, but 512GB SSDs still cost $350+; I picked up this Momentus XT online for $69.99 with free shipping.
Unprecedented value by comparison.