RAID-5 On The Desktop - Hardware vs. Onboard Compared

[notoriousformula]

GamePC has completed a comparison of onboard "software" RAID controllers versus add-in hardware-based controllers. Is the performance difference worth the added cost?

While RAID-5 controllers are dropping in price, becoming more readily available and arrays are getting easier to setup, RAID-5 still is a bit out of range for the typical home user. None of the three RAID-5 controllers provided us with an utterly seamless RAID-5 experience that we were looking for. Some have setup issues, some have software issues, and some have performance issues. Nobody is a clear winner here, although each controller does have advantages.

Link (http://www.gamepc.com/labs/view_content.asp?id=r52005&page=1)

[Player0]

That is a great review, and on a topic I would love to see more of.

But I find it to be completely and utterly innaccurate.

HDTach is an awful benchmark. Its like using Sandra or something. Dont know much about DiskSpeed32. However, both of these seem to be more accurate than ATTO for some reason.

I've tested several RAID5 controllers and run the 3Wares on this webserver, my home file server, and several other servers. I have not personally used the RAIDCore no, because I find it to be a cheaper software based card and I cant agree that it performs better than a Promise TX or a 3Ware.

Anyway. Write performance on any RAID5 array is pretty poor. He only measures this with 2 tools, and one of them is quite obviously giving misinformation. Instead of dumping those results as flawed, or trying something else, he leaves them up there and favors the RAIDCore.

Hardware RAID relies heavily on drivers, drivers which arent always compatible with disk performance tests, which often use their own disk accessing functions (HDTach for one). This might even have the affect of disabling all the hardware accelleration.

I dont know. I dont think the result is accurate, I can only speculate why. I think the 3Ware is highly underrated in these tests. Although I dont think the 3ware will do much ebtter than the other two in performance. I think it will be more even.

But put it in a server environment on a PCI-X slot and THEN see whos flying.