I’ve been real busy recently, juggling work with studies and its a really fine line balancing them. I’ve not played with hardware in a long time, so I decided to get an OCZ Apex off eBay, since local stores are selling them at jacked up prices.
Although the shipping costs could be lower, it is still a lot cheaper than buying it locally (note to self: get the retailer to ship them in a smaller box). Anyway, here’s another shot.
After popping the OCZ Apex into my system, the first thing I did was to format it in NTFS (which, in retrospect, is unnecessary since HD Tune runs benchmarks using raw disk access) and run HD Tune to get information on this drive.
Let’s run the read benchmarks, since the write benchmark is going to take quite a while:
Doesn’t look too different from other sites, and so does the random read speeds:
Well, that’s all for the read tests. Now here comes the torture test – sequential and random write speeds:
That’s quite bad. Let’s take a look at the IOMeter benchmark, which is the main problem with all SSDs using the JMicron 602 controller, as reported by Anandtech. However, IOMeter is able to do benchmarks either by using raw disk access, or using the filesystem access. I did both tests for 30 minutes each using the latest stable version with queue depth of 3 and 100% 4K random writes, using various filesystems:
At this point, let me explain the ext2 benchmarks – I installed the ext2 IFS driver for Windows originally for my (sucky) Transcend 32GB SSD last year, and now that the SSD has been put to better use elsewhere (which I can’t say or I’ll be arrested), I figured, why not just run it just for the heck of it. As for the initial and second run results, I decided to put them up because I came across this problem whereby after benchmarking in IOMeter, I accidentally clicked on the “start test” button in IOMeter, and found out that the results were drastically different. It seems that on the initial run, a file is created by IOMeter on the disk, and once that file is created, the 30-minute benchmark begins. On subsequent runs however, the file is not deleted and re-created, and it is re-used for the benchmark, hence the reason I had to include the two runs. Raw disk benchmark is pointless as its not a filesystem, and running the NTFS benchmark twice gave me results that were close to each other.
Again, more strange results – ext2 is actually alot faster than NTFS and raw? This is weird!
Disappointing response times on the initial and second runs, but the write speed is pretty weird:
Finally, CrystalDiskMark using the two different file systems:
With this mixed bag of results, is ext2 or NTFS the ultimate winner? It’s pretty hard to say right now, though I can fairly say that ext2 “feels” much more faster than NTFS right now. Stay tuned for more parts of my review of the 64GB Super Talent MasterDrive OX, RAID0 results and mixed-media results.