Samsung 830 2x 128GB RAID0

I picked up a second 830 128GB and decided to RAID them for giggles. Turned out really nice, although the reduction in 4k random at QD1 has given me a "meh" reaction. I thought folks might like a look at how it turned out for their own consideration.

Prior to RAID0 - single 128GB 830 on a system well into "in use" status in terms of crap installed, background applications etc. System had 16GB at this point.

Sequential Read : 494.923 MB/s
Sequential Write : 316.313 MB/s
Random Read 512KB : 336.395 MB/s
Random Write 512KB : 208.680 MB/s
Random Read 4KB (QD=1) : 22.071 MB/s [ 5388.5 IOPS]
Random Write 4KB (QD=1) : 84.110 MB/s [ 20534.6 IOPS]
Random Read 4KB (QD=32) : 255.911 MB/s [ 62478.3 IOPS]
Random Write 4KB (QD=32) : 111.942 MB/s [ 27329.5 IOPS]

Test : 1000 MB [C: 33.7% (40.1/119.1 GB)] (x5)

Immediately following Windows installation. Basically nothing on the machine. Still at 16GB of RAM when I did this. Breaking the 1GB/s barrier for the first time on a PC that I own was pretty damned exciting.

Sequential Read : 1011.976 MB/s
Sequential Write : 631.958 MB/s
Random Read 512KB : 584.863 MB/s
Random Write 512KB : 312.137 MB/s
Random Read 4KB (QD=1) : 20.739 MB/s [ 5063.3 IOPS]
Random Write 4KB (QD=1) : 81.961 MB/s [ 20010.0 IOPS]
Random Read 4KB (QD=32) : 173.626 MB/s [ 42389.1 IOPS]
Random Write 4KB (QD=32) : 168.625 MB/s [ 41168.3 IOPS]

Test : 1000 MB [C: 4.5% (10.7/238.4 GB)] (x5)

A day or so later now that I have all of my stuff reinstalled. Sequential is still impressive, random has kinda dropped more. System has 32GB at this point.

Sequential Read : 971.503 MB/s
Sequential Write : 626.577 MB/s
Random Read 512KB : 586.604 MB/s
Random Write 512KB : 512.176 MB/s
Random Read 4KB (QD=1) : 20.759 MB/s [ 5068.0 IOPS]
Random Write 4KB (QD=1) : 75.122 MB/s [ 18340.2 IOPS]
Random Read 4KB (QD=32) : 182.802 MB/s [ 44629.4 IOPS]
Random Write 4KB (QD=32) : 167.933 MB/s [ 40999.3 IOPS]

Test : 1000 MB [C: 19.1% (45.5/238.4 GB)] (x5)

Prior to the RAID config I had all environment variables altered to push temp folders to a conventional drive for write reduction (haven't done this on the RAID setup yet, but I probably will soon), excepting the pagefile. I see no reason move it from the SSD since I have it manually config'ed for 512-4096 and I've never seen it get larger than 525MB.

From the usage standpoint, I couldn't tell you I "feel" any difference in this configuration from how I did prior to RAID0. The only serious improvement that is blatantly obvious to me is installation times for applications. In the RAID config, MSO2013 installed in 1m 47s, and VS2012 Ultimate installed in ~6m. Pretty damned fast!

So yea, it's pretty cool n' all, maybe not worth the expense versus having just bought a newer, larger, single SSD. I figured since the additional 830 could be acquired fairly cheaply and it would double my system drive capacity, I might as well do it up. I guess the only negative is that the existing 830 had ~1TB written to it already, so the wear leveling on the two drives is mismatched but, I'm not too concerned about it.

My understanding from Googling on the subject is that 4K reads make up about 60% of the total workload on a drive for standard Windows and applications, while game loading is all about sequential reads.

Game loading happens so fast on even a slow SSD compared to a HDD that 500 MB/s vs 1000 MB/s is probably the difference between loading a level in 2 seconds instead of 4 seconds, while just about everything else you do on a PC feels faster with better 4K reads. Which is why the people who are actually using SSD RAIDs say "don't bother".

evilsofa said:

the people who are actually using SSD RAIDs say "don't bother".

Click to expand...

I think that is probably my primary reaction to the situation. Having tried this, I would in general recommend to someone that they spend the $$$ toward a larger single SSD. It just happened in this instance for me that the cost difference was significant to get a larger single SSD (I'd really like a 512GB Samsung 840 pro), and the real goal wasn't increasing speed as much as getting a bit more space. I also figured the experiment was worth it just because I like to play with things and I figured I'd share the results.