• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

RAID performance tweaks

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

DaedalusX

Member
Joined
Jan 24, 2002
Location
McConnell Air Force Base, Kansas
Okay so I just got myself a Fasttrak100 TX card, and I have two 20gb 7200rpm ATA133 maxtors hooked up. Acutally I've use RAID striping many a time.

My question is, how do I get it to perform well? I've always just hooked up to RAID-0 and gone along with the slight peformance increase from a single 7200rpm drive.

Up until now I just used onboard RAID on the motherboards I've purchased (MSI K7T266Pro2-RU, ASUS A7V333 RAID), but in an effort to help with performance I bought a motherboard that didn't have onboard RAID (Abit KD7) and then picked up a Fasttrak100. I've always known things to perform better when they're not onboard, using the system's clock cycles.

Now I've noticed a very slight increase in performance over previous RAID endeavors, but I would really like to see the monster performance a lot of people seem to have on overclockers, futuremark, etc.

With SiSoft Sandra when I benchmark my RAID, I don't even get close to the 7200rpm 2 hard drive RAID-0 reference that it puts me up against. Whereas I've seen pics where people have just SMOKED it.

Is there anything I am doing wrong? what are some good tips, or good sites that talk about tweaking to enhance RAID performance? Any help is appreciated, thanks.
 
for one thing, run the drives on seperate channels (if you can). that shouild improve speed a bit. other than that. I dunno sorry
 
Use ATTO, HDTach or IOMETER for disk benchmarks. Sandra is not a reliable disk performance source (wish I had a penny every time I've said that here). :]
 
But just because a benchmark doesn't accurately tell how well something perfroms doesn't mean that you are actually getting better performance compared to other people. Anybody that smoked SiSoft SANDRA probably has a better RAID setup than people that don't. Now, it may mislead you as to the real performance of your array, but that doesn't mean that you are running faster or closer to the person that smoked it. I'd bet that somebody that smoked SANDRA would outperform DX in any other RAID benchmark as well. I can't think of any tips, but for comparison purposes, I think any benchmark is fine. If you want to know how fast it's really going, go elsewhere (benchmark-wise, I mean . . .don't leave the forums:D ).

Z
 
unlike the retarded mistake i made...make sure you have anough power to run both of the drives ;)
 
Jon said:
Use ATTO, HDTach or IOMETER for disk benchmarks. Sandra is not a reliable disk performance source (wish I had a penny every time I've said that here). :]

While Sandra may not be the best benchmark, it certainly isn't bad by any means. The worst thing I understand about it is it doesn't score SCSI drives as well as it should. (Peeps running SCSI really has a disliking for Sandra.) Otherwise I find it scoring HDs pretty fairly. If you will notice the aggragate scores you get in Sandra, they will be pretty close to the overall average read score you get in HD Tach. Although certain situations could make this vary.

I find the worst performer in benching HDs to be ATTO.
While I like the way ATTO functions, I find no consistency over a given time with it. Plus ATTO is very sensitive to HD buffers and sizes of the various variables concerning RAID.

But I digress, I use them all and I try to balance out, to average out the information from all 3 to an average idea of how my RAID and HDs are working.

As for any tips for the fellow who asked. All I can say is your Hard Drives has MUCH, no MOST everything to do with how well you score. Sure you can set stripe sizes and such to tweak your RAID but it really shouldn't be necessary. Also you should consider that a lot of peeps who takes this seriously, benches their setups on bare drives with nothing running in Windows that would or could possibly hamper performance. OTOH when I benchmark something I try to have everything I normally use or run, running in Windows so I can get a good idea of what I'm really experiencing in real life usage. (Although I do disable AV because of the far reeaching effects it has.)

Cheers,
Mike
 
1) Make sure your RAID controller is not sharing IRQ's with other devices (took half my write scores until I fixed it).

2) Adjust your PCI latency upwards from the mediocre default 32. Many people have benefited from longer bursts using RAID controllers. Test in conjunction with #3 below.

3) Adjust your stripe size for the tasks you want your system to shine at. Default is often 64Kb, but for better random access a smaller size might be better. Larger size may be better for continuous streams, but personally I think 64Kb is large enough already. I use 16K myself.
 
Back