RAID-0 Benchmarks: SSD vs Raptor

Capsaicin · Post by **Capsaicin** » Thu Jun 28, 2007 7:37 am

Source:
http://www.watch.impress.co.jp/akiba/ho ... p_ssd.html

Translation:
http://forum.hardmac.com/index.php?showtopic=1645

No real surprises: Raptor still wins sequential I/O and write tests (random writes of 64K blocks are very close, though). SSD wins random reads.

m^2 · Post by **m^2** » Sat Jun 30, 2007 6:40 am

It's only a partial translation.
WorldLingo translates well from japanise ("You must have 1 posts before you can post URL's/Links.")

freka586 · Post by **freka586** » Wed Jul 11, 2007 11:27 pm

What I find really interesting is the enormous increase in write performance when going RAID0. Does anyone have a good reason for this?

The real bottleneck of current SSDs is the poor random write performance. This is reflected in their standalone tests, 2.75 MB/s at 64k random writes.

But when going RAID0 this bumps up to a whopping 22 MB/s, and 64k reads at 84 MB/s. How is this possible? We are talking a factor of 10 for the write performance here! Could it be that the performance-stealing load-leveling stuff of SSDs gets masked by dual writes or something?

Now if only there were manufacturers selling ready-baked RAID0 SSD sandwiches all would be swell

Aris · Post by **Aris** » Thu Jul 12, 2007 1:10 am

i would have prefered real world testing and not synthetic benchmarks.

easiest way is to time how long it takes to do certain things like:

Boot
Open a program
Load a small file
Load a large file
Compress a file
Uncompress a file
Encode a raw file into .wave/.mpeg2/.mpeg4/.mp3 etc

Because if you think about it, thats all that a faster HD really does for you is make things quicker. Synthetic benchmarks are pretty much useless. I'd take a drive that a synthetic bench said was slower if it "felt" faster when i used it.

aaa · Post by **aaa** » Fri Jul 20, 2007 6:13 pm

freka586 wrote:What I find really interesting is the enormous increase in write performance when going RAID0. Does anyone have a good reason for this?

The real bottleneck of current SSDs is the poor random write performance. This is reflected in their standalone tests, 2.75 MB/s at 64k random writes.

But when going RAID0 this bumps up to a whopping 22 MB/s, and 64k reads at 84 MB/s. How is this possible? We are talking a factor of 10 for the write performance here! Could it be that the performance-stealing load-leveling stuff of SSDs gets masked by dual writes or something?

Now if only there were manufacturers selling ready-baked RAID0 SSD sandwiches all would be swell

Perhaps the drives are better at writing 32k :shrug:

They already make 'ready-baked RAID0 SSD', any SSD that reads greater than 31MB/s does this. There is one SSD that does 100MBread,80MB write according to it's specs.

freka586 · Post by **freka586** » Sun Jul 22, 2007 1:03 pm

aaa wrote:
Perhaps the drives are better at writing 32k :shrug:

They already make 'ready-baked RAID0 SSD', any SSD that reads greater than 31MB/s does this. There is one SSD that does 100MBread,80MB write according to it's specs.

Now this sounds interesting!
Any URL to more info or suggestions on manufacturer / brand / product name?

aaa · Post by **aaa** » Mon Jul 23, 2007 1:23 pm

http://www.dvnation.com/nand-flash-ssd.html

Expensive as usual.

freka586 · Post by **freka586** » Thu Jul 26, 2007 7:55 am

No kidding, $2k-3k is a hefty premium compared to the price of 2 x ordinary SSDs. Still very cool though, and perhaps an indication of things to come?

Moogles · Post by **Moogles** » Thu Jul 26, 2007 3:28 pm

DVnation's prices are ridiculous to begin with, so the $2k pricetag on that drive might only be something like $1200 if a store like newegg carried it. Still, it's a glimpse into the future and hopefully prices start to drop really soon.

vortex222 · Post by **vortex222** » Mon Jul 30, 2007 2:50 pm

I have seen adapters on Ebay that allow one to connect several Compact Flash devices to an IDE interface. Seamed cheap. Not sure where to get them or how much they are or how fast they are otherwise.