Conroe shows dodgy RAID performance anomalies

PRELIMINARY CONROE results have been really impressive, but as time goes by and the platform gets tested in more and more environments, interesting things are popping up.

Right now, it seems that Intel has some serious problems when using RAID5 configuration. If you are using an Intel's own baby, D975XBX motherboard and put four drives in RAID5 array, an interesting overhead and a slowdown will occur on all upcoming Conroes.

We're talking about all the combinations, including Core 2 Duo E6300 with 2MB of L2 cache, E6700 with 4MB of L2 cache and the X6800 - Extreme Edition as well. An odd thing indeed.

When comparing the identical disk configuration several different combinations on AMD's side, nothing happens. We tried using with older Athlon 64 on an nForce4-SLI platform or the new ones, Athlon 64 5000+ and FX-62 on new Socket AM2 with both nForce 590SLI and CrossFire Xpress 3200 chipsets. Result? Works flawless, without any suspicious CPU overhead and so on. So, either we're crazy, or someone really screwed up in the SouthBridge department.

The performance in benchmarks is there, but the performance in real world isn't. While synthetic benchmarks will do the thing and show RAID5-worthy results, CPU utilisation will go through the roof no matter what CPU is used, and the hiccoughs will occur every now and then. It remains to be seen whether this can be fixed via BIOS or micro-code update. Or does Intel have a bigger problem at hand?

It seems to us that Intel's Core 2 is suffering from a specific overhead when on-board RAID controllers are used. If you're unfamiliar with current on-board RAID5 controllers, let's just say that they really look like soft-modems and soft-sound cards, utilising power of the CPU for everyday work.

I only wonder how will a two-meg cache Conroe work on a fully integrated motherboard, with Gigabit Ethernet, software-based sound-card - ever popular AC'97 codec and so on. Probably not as advertised. µ