Inconsistent benches from my 2x Performance Pro 256 in RAID 0

[RandomUser]

I'm setting up a new computer with 2 Performance Pro 256 GB SSDs in a RAID 0 configuration and ran into some weird numbers when I was doing some benchmarks.

 

My first question is: is it normal for random 4k read/write performance to drop when using RAID 0 vs individual drives? (My guess would be that it's because I was using msahci drivers before I set up the RAID, then went to the IRST drivers after...)

 

Individual Drives:

http://img818.imageshack.us/img818/6449/individial.png

 

RAID 0 (before/after enabling write caching):

http://img507.imageshack.us/img507/8388/asssdraid.png

 

More importantly, I'm wondering if the anomalies I'm seeing in ATTO are anything to be concerned about. Sometimes 16K, 32K reads and 2048K writes seem to be a LOT slower than they should be. If I limit my test range to between 16K and 2048K the numbers seem fine though, so I don't know if it's a 'real' problem or just something having to do with ATTO.

http://img228.imageshack.us/img228/7438/attoafterirstchipset1.png

http://img20.imageshack.us/img20/7264/attoafterirstchipsetsub.png

http://img84.imageshack.us/img84/7256/attoafterirstchipsetlat.png

 

I used both drives individually (one Windows 7, one Windows 8 beta) in my older system for about a month before putting together my new system, and they seem to be working fine. And yes, I did a secure erase on both before RAIDing and formatting in the new system.

Any thoughts? (Thanks in advance...)

[Toasted]

The ATTO results look normal.

 

http://forum.corsair.com/forums/showthread.php?t=106206

[RandomUser]

So are you saying the dips in 16K/32K reads and 2048K writes are normal? Personally I don't remember seeing an ATTO screenshot with dips that large. 16K reads are nearly 1/3 the speed of the 8K reads in one of those pictures...

 

I'm running a clean Windows 7 install with only the motherboard chipset drivers and the IRST software installed. I don't have an antivirus or anthing else installed. I know that what you have installed and what services/etc are running can make a difference, but I didn't see it have as much an effect as people usually suggest. In my first picture I showed AS-SSD results for both drives individually -- The one labeled as "F:" is faster than the other, even when it's the OS drive and the other was just secure erased. (Here's a picture of that drive as the OS drive, write cache disabled obviously.)

http://img195.imageshack.us/img195/4516/asssdbenchintelsystemdr.png

 

Unfortunately I wasn't testing the individual drives with ATTO so I don't know if the 16K/32K read speed is lower on one, both, only when used as a system drive, or only in RAID 0.

[Toasted]

I would leave it for a while as benchmarking the SSD may temporarily cripple the drives.

[RandomUser]

I appreciate the advice, but the ATTO screenshot that I wrote "another full test, much later" on was done after at least 20 hours idle on-time. In fact, the only times I've gotten 'normal' results (no dips in transfer rate on ATTO) were when I retested immediately after getting the lower scores -- twice on just the 16K-2048K range, once over the entire range.

 

Honestly, if I can get 'only' 370 MB/s for 16K chunks and 500 MB/s for 32K reads, and writes of 725 MB/s for 2M that's still really fast and I'll probably never notice it in real world usage. I just want to make sure it's not an early warning sign that something is wrong with one of the drives.

[Toasted]

If you want, Secure erase the drives again, create a new RAID array then run one single ATTO benchmark. Don't run the AS SSD benchmark.