PCIE SATA RAID controller worth it?

[z3r0-]

My NVraid array with 2x 74gb raptors feels pretty slow compared to when I ran it on ICH8R. Would getting a highpoint rocketraid 2310 PCIE controller or any PCIE controller for that matter give me a substantial speed boost over the NVraid?

[Madwand]

Assuming RAID 0.. for straight-ahead large file transfer or STR benchmarks such as HDTach, no, NV RAID should be faster than a 23xx or 22xx because NV RAID allows changing the stripe size and hence taking greater advantage of concurrency between the drives than the RocketRaid, which only support 64K stripe size. Stripe size <= 64K / # of drives would generally be needed to improve concurrency (under Windows at least).

Caveat: I haven't actually measured RocketRaid's performance with 2-drive RAID 0.

For RAID 1, a good implementation should out-perform NV RAID because (well, at least the one I measured) NV RAID doesn't take advantage of data duplication for read performance improvement, but others, including Intel, sometimes do. I'm not sure about RocketRaid.

For RAID 0 with more complex access patterns, a caching controller could in theory give better performance, but I haven't seen anything even approaching exhaustive testing on this; they typically don't even include more than one stripe size.

[z3r0-]

Yeah RAID 0. Running 64k stripe now. I guess I'll stick with nvraid and get another raptor.

[gjvrieze]

personally i love the nvraid, my new conroe board, a p5b-d, and its intel are not as great IMO, i have tested 4 x320 seagates in raid 5 with a stripe of 128k and only get 370 burst and 160 sustanded in hd tach, but i am adding 2 more 320s, so i will what kind of speeds i get with that..... I have 2x74gb raptors in raid 0 on an nvidia board and love it...

[z3r0-]

I was getting 140mb sustained on intel RAID and 100 on nvraid : :mad: