Poor performance - PERC 6/i - Server R710

I would also use iOmeter to test instead of Atto.

Just curious, I know you said you think this is raid controller related, but what BRAND hard drives are in this server? Seagates or Hitachi or Tosihiba drives??

Does seem low. Though I do not use ATTO as a benchmark, I ran it on a client's production server, 2900III, 8 meg ram. On a raid 1,same size disk/speed, produced an average just under 150 B for 16k and up. For raid 5 (4 disk) array an average of just under 300 MB for 16k and up. Have not really benchmarked in a couple years, and do not trust Atto...as I see it, it is only useful for seeing relative differences between servers, good for judging cache speeds. Could be Atto does not interact with your newer motherboard properly, (considering you at least have write back enabled on the raid controller). In the bios you should have all unnecessary devices disabled, In OS disable all unnecessary services, in device manager disable all unnecessary devices. Possibility one of the devices/ drivers or services interfering.

I/Ometer is good, bit rough to get good results util you get the correct parameters, you might also try Cosbi, the version just for disk benchmarking, simple DOS based and adds no code to the OS....

"File Copy HERE Ver. 0.52 HDD test."

http://4peeps.com/ivb/index.php?showtopic=4108

The PERC 6 controller may not the best set up to make disk benchmarking, the disk are not shown to the OS, you need to build a virtual disks for OS to run any tests.  A RAID 0 on one drive may be an option but, is this close enough?

"the disk are not shown to the OS, you need to build a virtual disks for OS to run any tests."

The disks individually do not show, but their combined IO does, is that not the point ?.  Using raid 0...the designers of raid hardware/firmware place very little effort/thought/tweaking into running an array with 1 disk,.... for that matter up until a few years back they ignored even raid 1 performance, For real world tests, it  would make no sense to test an individual disk or array without an OS(and or data) on it, unless your just going for the highest disk benchmark points.... pointless as it does not resemble the real world benchmarks/performance results once the OS/data (and it's cache/formatting) are in place.

Anyone have a fix to this? I know this is a really old post but I too am having almost this issue except I have a dell perc 6/i in my personal machine and only getting around 200 mb/s read and write.