You were also testing and empty array before and thus testing the extreme outside 32mb of the array and thats not representative of what you get further into the disk also expect a 5k to 10k loss there alone with smaller drives raided.
When you tested your software raid also... was it on a disk with a good bit of data or empty drives?
Hi Tex, thanks for stepping in.
I sure can use some advise not because of the scores (I can care less) but to understand the differences.
All tests were run with an empty array and the configuration is the same for all (NTFS 16/16)
I've tried converting to dynamic when using HPT374 and also used your latency guide but the scores don't change.
They only get better when I move the WD array to SATA.
Next, I'm going to try my Maxtors on the HPT374 to see how they react to the change.
ps: I haven't use software raid at all
0
Mt_GoatHead Cheezy KnobPflugerville (north of Austin)Icrontian
edited October 2003
Equito,
First, I have been trying to tell people that the SATA controllers are just faster. Point, I had thw WD SE's on a HPT RocketRaid 133 (HPT-372 controller core) and the combo did well. Then I got a RocketRaid 1520 SATA card (HPT-372 controller core) and it did about 12 - 14% better with the same drives. Since tweaking on my S.I. onboard controller it pretty well keeps up with my 1520 now.
Was the SATA you used an onboard or another PCI card? What other variables could be possible between the two? Were these raw or tweaked?
First of all, let me apologize to Gargoyle for hijacking his thread... :o
Larry, to answer your Q's:
I'm using the onboard SATA controller, no other variables that I can think of and no tweaks at all.
Since tweaking on my S.I. onboard controller it pretty well keeps up with my 1520 now.
How do you tweak the Si controller?
Now, check this out: the Maxtors (2MB cache) on the HPT374, again NTFS 16/16 empty array. Notice that they scored better than the JB's (8MB cache), specially during 'reads' but the pattern is the same when it comes to 'writes' under HPT...
Mt_GoatHead Cheezy KnobPflugerville (north of Austin)Icrontian
edited October 2003
All I meant on the S.I. tweaking was how to config everything. Nailing down the latency and the stripe/cluster.
Have you tried adjusting the latency on the HPT just to see what it does? I don't think the controller is bad since you get good reads but you may need a new bios and or drivers. I would work with the latency first though and remember the HPT's really like the high side. My S.I. likes 96, my 1530 likes 200 and my RR133 likes 248 just like the onboard HPT's did.
I must be doing something wrong then cause I've tried a bunch of different latency values but none of them make any significant difference. I'm using latest bios and drivers btw.
Thanks for the tips. Cheers!
0
Mt_GoatHead Cheezy KnobPflugerville (north of Austin)Icrontian
edited October 2003
As one lase hoorah, have you tried the HPT in different slots or even another machine? I am just as puzzled as you are. After you said that adjusting the latency had nil affect I am even more puzzled and am now wondering if something in the card has broken down.
I'm sorry, I should've mentioned that is not a PCI card, is the HPT374 onboard controller of the Abit AT7.
No sweat dude, it ain't mine and the customer is happy with the current performance.
I was just trying to understand why the same array behaves so differently under these two controllers and why these cheap Maxtors are faster than JB's.
EQuito said Hi Tex, thanks for stepping in.
I sure can use some advise not because of the scores (I can care less) but to understand the differences.
All tests were run with an empty array and the configuration is the same for all (NTFS 16/16)
I've tried converting to dynamic when using HPT374 and also used your latency guide but the scores don't change.
They only get better when I move the WD array to SATA.
Next, I'm going to try my Maxtors on the HPT374 to see how they react to the change.
ps: I haven't use software raid at all
And I hadnt seen your posts when I replied. It was meant for the original guy.
When you say you get no differance adjusting the latency? What bios and drivers are you running? From the looks of the atto its sharing a IRQ or needs the latency increased big time for the writes to come up. This is just guessing from the ATTO though.
Ok, midterms are over and I finally got around to working with my drives again. For the last 2 weeks I've been running Windows practically crippled, and the only program installed was MS Word!
Anyway, I've got my Windows running on the 80gb WD 8mg cache. I've just formatted the 2x40's and ran an atto on the array. Doesn't the sudden jump from about 30mb/s read to 70/mb seem a little strange?
Tried it at 192 and 248 latency and got the same results. Here's the 248 screenie. Gained a few mb/s in writes from default
I'm still a little suspicious of the sudden jump of read speed, and a little irritated that my 32k read is so low... but after I get some things installed we'll see how it feels
Mt_GoatHead Cheezy KnobPflugerville (north of Austin)Icrontian
edited October 2003
Try moveing it up one slot to 3rd from the top. I'm pretty sure that PCI slot won't have anything else. You are trying to make the HPT the only thing on an IRQ.
Comments
When you tested your software raid also... was it on a disk with a good bit of data or empty drives?
Tex
I sure can use some advise not because of the scores (I can care less) but to understand the differences.
All tests were run with an empty array and the configuration is the same for all (NTFS 16/16)
I've tried converting to dynamic when using HPT374 and also used your latency guide but the scores don't change.
They only get better when I move the WD array to SATA.
Next, I'm going to try my Maxtors on the HPT374 to see how they react to the change.
ps: I haven't use software raid at all
First, I have been trying to tell people that the SATA controllers are just faster. Point, I had thw WD SE's on a HPT RocketRaid 133 (HPT-372 controller core) and the combo did well. Then I got a RocketRaid 1520 SATA card (HPT-372 controller core) and it did about 12 - 14% better with the same drives. Since tweaking on my S.I. onboard controller it pretty well keeps up with my 1520 now.
Was the SATA you used an onboard or another PCI card? What other variables could be possible between the two? Were these raw or tweaked?
Larry, to answer your Q's:
I'm using the onboard SATA controller, no other variables that I can think of and no tweaks at all.
How do you tweak the Si controller?
Now, check this out: the Maxtors (2MB cache) on the HPT374, again NTFS 16/16 empty array. Notice that they scored better than the JB's (8MB cache), specially during 'reads' but the pattern is the same when it comes to 'writes' under HPT...
Have you tried adjusting the latency on the HPT just to see what it does? I don't think the controller is bad since you get good reads but you may need a new bios and or drivers. I would work with the latency first though and remember the HPT's really like the high side. My S.I. likes 96, my 1530 likes 200 and my RR133 likes 248 just like the onboard HPT's did.
Thanks for the tips. Cheers!
No sweat dude, it ain't mine and the customer is happy with the current performance.
I was just trying to understand why the same array behaves so differently under these two controllers and why these cheap Maxtors are faster than JB's.
When you say you get no differance adjusting the latency? What bios and drivers are you running? From the looks of the atto its sharing a IRQ or needs the latency increased big time for the writes to come up. This is just guessing from the ATTO though.
Tex
Anyway, I've got my Windows running on the 80gb WD 8mg cache. I've just formatted the 2x40's and ran an atto on the array. Doesn't the sudden jump from about 30mb/s read to 70/mb seem a little strange?
I'm still a little suspicious of the sudden jump of read speed, and a little irritated that my 32k read is so low... but after I get some things installed we'll see how it feels
Start > Run > "msinfo32.exe", then expand "Hardware Resources" and click on 'IRQs"
Then look to see for your controller and see if anything else has the same IRQ assigned.
/edit: well maybe the USB doesn't count. Under Device Manager it says "No Conflicts."