Ok, I know this is something that should have been posted on my Tumblr blog, but I think my comments will be too much there.
24 solid state drives (SSD) are truly a geek’s dream. Let’s check out the setup:
- 2 x Intel QX9775 CPUs
- 2 x AMD 4870×2 in crossfire
- 4GB
- 800MHz FBDIMMs (Assume from the speed, it’s DDR2. And I have no idea what’s so “custom” about that.)
- 2 x 1kW PSUs
- 6 x SFF-8087
I have no idea what’s an SFF-8087. A google search showed me this and that, so I’m guessing the setup is using some RAID card that accepts SAS. Or maybe since the motherboard is probably a server board, it would have onboard SAS controllers.
But from the video, it is obviously a marketing gimmick. That “IT guy” in the video is either someone who is not too bright, pressured by marketing or bribed. Or simply a paid actor. Because the choice of tests done is not really that tailored for showing off the performance from 24 SSDs. It is also not mentioned what type of RAID is it on, but for the sake of the video, let’s just assume it’s RAID-0.
The throughput is definitely awesome. 2 gigabytes per second! At that speed, performance is no longer bound by the drive, but the type of controller you have.
Opening programs at the same time super fast is impressive, but not that extreme. Running apps is more bound by the amount of RAM. It will be infinitely more impressive if the PC just has 512MB of RAM and the applications open that fast. And then try to switch using one application to another. Only then will we see the kickass SSDs in action, trying to do loads of swapping to and fro the RAM and SSDs.
The defrag test is really awful, because solid state drives do not require defragmenting. Defragmentation is a process of placing the contents of a file together so that the number of seeks is reduced. Seek time for a normal spinning platter HDD is expected to be about 14ms for 7200RPM drives. Faster 10000RPM HDDs can reduce that to 7-9ms. From benchmarks I’ve seen on the web, an SSD has a seek time of less than 1ms. This extremely low seek time makes it completely unnecessary to perform a defrag. Another awful thing about it is that defragging a relatively empty disk and one that is nearly full with a large number of fragmented files are two very different ball games.
I regularly empty gigabytes of data from my recycle bin as well. And on my lousy 7200RPM HDDs, I also am able to perform it that quickly. It’s just removing the pointers to the file. Not a good test. In fact, it’s a really stupid one. 
The copying test is really cool. 700MB copied in 0.8 seconds! (Lucky they didn’t choose to do a “move-a-large-file-from-one-location-of-the-drive-to-another” test. 
)
Playing Crysis in high resolution has got absolutely zero effect from drive performance. Loading Crysis maybe, but playing it? Zilch! Playing it at that high resolution smoothly is due to the 2 Intel QX9775s, 2 4870×2 graphics cards in crossfire and the 4GB RAM. The drive makes no impact at all. Don’t believe me? Reduce the specs of the machine to 1 Intel E7300, a single ATI 4380 graphics card and 2 GB of RAM. Try playing Crysis at 1920×1200 with textures set to “Very High” and see the frame rate.
I’m not really sure what the point of the image loading test is. Isn’t that just a replication of the first test?
The bouncing on the trampoline test is not that impressive. At that bounce rate, even normal spinning HDDs can still be working. If something goes wrong there, I would expect it’s due to a loose connector rather than a failed drive.
The thing that got my excited was the throughput and only the copying data test seemed to be a valid one. The music and video presentation was nice, but when you’re trying to impress geeks, those are irrelevant. Sorry Samsung, you gotta do a bit better than that.
Oh, and if I went wrong somewhere, do point it out in the comments! 
Tags: fail, samsung, SSD, viral video
