[081213]
So I’m up to 508 now. Hope to get maybe 10-15 more done over the break.
I was going to write one last entry on buying and selling stuff (this is, ultimately, somewhat related, though), but on Wednesday afternoon, I chanced upon a really good deal on the NBR forums – and I mean like a once or twice in a few years really good deal. It was a SAMSUNG MCCOE64G5MPP-0VA00 2.5″ 64GB SATA II SSD. As you can see this is a very expensive piece of hardware, almost $100 more than my HP and just about half the price of my ’64. Why I nearly shit in my pants was because I bought one for $110, shipped to boot.
Initially I feared this guy might have ripped me since this really does fall into the realm of “too good to be true”, but my fears were quelled when this thing arrived in my mailbox today in perfect working condition. I mean, this isn’t the biggest or fastest SSD out on the market anymore, but it’s still 64GB of SLC with a good controller (as opposed to the jmicron crap you might find in cheaper SSDs) and solid IO speeds; I honestly couldn’t believe my luck – it’s usually quite bad.
And of course I took some time to put the Samsung through its paces; I pulled out the 80GB, 5400RPM Toshiba in my ’64 and loaded up XP (in 15 minutes!) on the Samsung.
As you all know, I’ve got this thing for quiet computing and this solid state business doesn’t disappoint. I’ve played with a Lenovo X300 and while it comes with a (slower – AH HA HA HA) Samsung SSD, the small form factor gives it a somewhat weaker thermal solution coupled to a whiny-ass fan – which makes it less than quiet. Despite the LV processor, that whiny-ass fan comes on even if the machine is idling.
I’m pleased to say that my ’64 with an SSD (and a smattering of AS5) is completely silent at idle and a light load. It’s not as good as my HP, but the fuckers at Intel won’t let me push the P8400 below 0.9250 volts – and the fuckers at the nVidia don’t let me touch the Quadro’s VIDs at all (or at least no one has figured out how to?). Nonetheless, it’s all very good; a definite improvement over the spinner.
Sadly, bootup times don’t really improve that much (though they did significantly on my HP when I went to CF), but the machine was definitely faster overall in loading drivers, installing/loading programs, etc. While the raw performance numbers aren’t as amazing as the $600, 80GB Intel X25-M, they are, as I’ve said, nothing to sneeze at.
http://colette.trianglesoft.net/2008/img/0812/HDTune_Benchmark_TOSHIBA_MK8052GSX.png
http://colette.trianglesoft.net/2008/img/0811/HDTune_Benchmark_TRANSCEND.png
http://colette.trianglesoft.net/2008/img/0812/HDTune_Benchmark_SAMSUNG_SSD_RBX_Seri.png
So the first graph is for the stock Toshiba MK8052GSX in my ’64; this is generally representative of the speed “curve” you get for most mechanical HDDs (the spikes are cuz I’m reading from the disk with WMP whilst doing the test). I mean this ain’t too bad; the access times are a bit slower, but the average throughput the is actually on par with the 2003-2004 era 7200RPM desktop drives I have in Colette and Motoko.
The Transcend 300x CF card I have in the HP has a slower average throughput, but due to the nature of SSD, the throughput is consistent across the entire drive. And of course, due to the nature of SSD, the access times are through the floor. 0.3ms is a full 62 times faster than the MK8052 (and it might actually be faster – I’m not sure what’s the smallest access time HDTune can accurate detect).
Of course, the Samsung has a similar throughput curve – effectively a straight line – save that’s twice as fast, averaging 85MB/s. Access times are similar to the other solid state solution, but the increased burst rate indicates that there’s probably a small cache on the Samsung drive – though I’m not quite sure of the specifics here.
But HDTune isn’t that good an indicator of overall drive performance as it only measures sequential read speeds and access times, and this is really less than half the story. The Jmicron MLCs I pointed out earlier are very good at these sequential and reading things – they just happen to fail at writing. So I’m going to introduce another set of numbers provided by a utility I recently discovered: crystaldiskmark
Sequential Read : 51.705 MB/s
Sequential Write : 51.001 MB/s
Random Read 512KB : 22.201 MB/s
Random Write 512KB : 25.963 MB/s
Random Read 4KB : 0.297 MB/s
Random Write 4KB : 0.948 MB/s
Test Size : 100 MB
Sequential Read : 94.794 MB/s
Sequential Write : 86.774 MB/s
Random Read 512KB : 92.378 MB/s
Random Write 512KB : 68.579 MB/s
Random Read 4KB : 15.583 MB/s
Random Write 4KB : 5.387 MB/s
Test Size : 100 MB
The first set of numbers for the MK8052, the second set for the Samsung. Sequential reads and writes for both drives are pretty much right on the average throughput as depicted by HDTune, 51MB/s & 51MB/s for the Toshiba and a whopping 94MB/s & 86MB/s for the Samsung. The gap really starts to widen as the SSDs negligible access time comes into play on the random reads and writes. From reads sequential to 512KB to 4KB, the Samsung goes from being about twice as fast to 4 times as fast to fifty times as fast. From writes of sequential to 512KB to 4KB, the gains are less impressive, but still in the range of 2-5 times as fast.
There you go, SSD performance in a nutshell. Til next time then.
Nonsense Wars 
