Fastpath..

[Anvil]

2R0 X25-E 9211 swraid. (1W and 4W)

[SteveRo]

I wouldn't know, i haven't worked with xml files.

I must be opening you attachments incorrectly - most are carrying a .xml identifier

edit - 0ffice 2000 doesn't work - i need to move to a different machine.

[GullLars]

Ah, it's an excell 2007 file, xlsx. It uses xml formatting. You can open it in office 2007/2010, and Open office 3 (with slightly wierd formatting).
I'll have a look at the 2R0 x25-E's. Don't know if 2 of them are enough to get a real benefit of swRAID though.
EDIT: 65K IOPS @ QD 32, and about 76K at QD 76-80. Nice to see you used the detailed config
EDIT2: what's up with renaming the result files to txt btw?

[Anvil]

GullLars,
It's easier to import to Excel when the result files are .txt
(it's not renamed, just select .txt from the drop down box in the file save dialog)

[Computurd]

@steve-o
i believe the difference is with the writing. the cache allows for much better write scores than the 9211 does without cache. in turn, the ssd's perform better when not bogged down with writing. pcmark vantage is a heavily write weighted benchmark. i think it writes 5 gb per run. might be 8 but if i remeber correctly it was 5gb per run, and if you look over tilts post that had a breakdown of the benchmark a while back, almost every single benchmark suite writes like crazy. i dont think stripe size plays into it as much. i have a 128k stripe size (on 9260) and it is good for hall of fame at number 11.

here it is...look at write distribution on windows media.. lol 48 percent!

http://www.xtremesystems.org/forums/...hlight=vantage


what is crazy though is now that we have fastpath the 9260 has cache and is better!

[mbreslin]

Detailed..

[mbreslin]

Doing scaling per Ndrives now, will post here in a bit. Was curious how fast I would hit controller max, 1 worker qd64 3 drives 160k iops.

[mbreslin]

Ndrive scaling 1,6,12 worker

[SteveRo]

@steve-o
i believe the difference is with the writing. the cache allows for much better write scores than the 9211 does without cache. in turn, the ssd's perform better when not bogged down with writing. pcmark vantage is a heavily write weighted benchmark. i think it writes 5 gb per run. might be 8 but if i remeber correctly it was 5gb per run, and if you look over tilts post that had a breakdown of the benchmark a while back, almost every single benchmark suite writes like crazy. i dont think stripe size plays into it as much. i have a 128k stripe size (on 9260) and it is good for hall of fame at number 11.

here it is...look at write distribution on windows media.. lol 48 percent!

http://www.xtremesystems.org/forums/...hlight=vantage


what is crazy though is now that we have fastpath the 9260 has cache and is better!

Good morning Paul, yes i had been through the whitepaper a couple times, there are a couple tests with very large % of writes for sure and yes I agree that cache should help a lot in writes for SSDs that need help with writes, i would have thought that DDR2 based acards wouldn't need that help.

[SteveRo]

Doing scaling per Ndrives now, will post here in a bit. Was curious how fast I would hit controller max, 1 worker qd64 3 drives 160k iops.

Wow, 160k - very nice - this is R0 not dynamic disk? - so it is bootable?

[Anvil]

Steve,

Have you compared the results by subscore, where does the Areca make a difference?

BTW, 4KB iometer random read is not much to go by to determine the best storage system
8KB (workstation pattern) and 16/64KB is just as important, maybe more so than 4KB, although, if 4KB is good then the rest should be good as well.
Write Through is probably the culprit to the 9211s overall score.

[mbreslin]

Same thing we've been doing to test fp, with 1-8 disks. When Gull gets up and makes graphs we should be able to see the sweet spot for number of c300s on 9260+fp.

[SteveRo]

Steve,

Have you compared the results by subscore, where does the Areca make a difference?

BTW, 4KB iometer random read is not much to go by to determine the best storage system
8KB (workstation pattern) and 16/64KB is just as important, maybe more so than 4KB, although, if 4KB is good then the rest should be good as well.
Write Through is probably the culprit to the 9211s overall score.

Yes, i remember all of of the HDD tests on the 9211 were lower (vs the 1231) - did this yesterday - some not as much as others, I will take a look at that and post back what i find.

[Computurd]

@steve-o...i did fail to take into consideration the write speeds of acards are going to be uber...hmm that is a good question, the 9211 is kind of a enigma when it comes to the vantage results, and i guess that was my conclusion in the end ( writes being the issue) but taking into consideration the acards performance with writes i dunno

i am thinking of testing a new device btw...might have it inbound this afternoon if it "checks out" after a little more research.
it does TRIM in raid 0 btw, first device to do so not sure how much that would matter but with a little downtime waiting for the (new) intel drives i am getting bored...

[SteveRo]

ok so 9211 vs 1231 - compare these two pcmv suite scores -

9211 - 29878 - http://service.futuremark.com/compare?pcmv=326725

1231 - 34744 - http://service.futuremark.com/compare?pcmv=308078

overclock's are close, pcie is 105 vs 119, same mobo, 980x and video card also oc'd. 9211 run is 8xR0 drives, 1231 is 5xR0 drives

9211 does MUCH worse in 4 out of the 6 HDD tests and also gets beat in the other two - HDD gaming and HDD application loading - just not as wide a margin.

I would of thought the 9211 would of done better.

[SteveRo]

Paul - anxious to hear about your new acquisition.

[Ao1]

RevoDrive?

[GullLars]

i am thinking of testing a new device btw...might have it inbound this afternoon if it "checks out" after a little more research.
it does TRIM in raid 0 btw, first device to do so not sure how much that would matter but with a little downtime waiting for the (new) intel drives i am getting bored...

It's OCZ's new Z-drive. Only device doing TRIM in RAID-0. If i'm not mistaken, they made custom drivers and firmware for a LSI 92xx card, and i think it uses barefoot controllers to handle the NAND DIMMs. If the rated IOPS numbers are actual performance, i suspect you will be disappointed. (by synthetic benchmarks, maybe not so much in actual usage)

I'm looking into mbreslins results now.

[GullLars]

Here are graphs for your 1-8R0 numbers mbreslin.
IOPS:




IOPS/accesstime:




There is little to no difference between 6 and 12 workers.
The difference between 1 and 6 workers i would speculate would mainly be because of the different QD meassuring points.

I would love it if you could my 4W detailed config for 1-8R0 too, since that would capture the relevant QD area for RAID scaling. QD 4-128. You can see from the graphs that up to QD 4 there is little difference between 1 and 8 drives.

[Computurd]

nope not a z-drive! i have never been impressed by any revision of the z-drive. they suck.

hint: it is sata 6gb/s device! no sas.

[GullLars]

Some sort of raid controller then, LSI or Areca?
Or could it possibly be one of the new ones from highpoint?

[Computurd]

alright, ya got me...
this controller piques my interest for some reason...only four ports, BUT with trim being passed in raid 0 and with sata only interface i am curious as to how it handles, also has dual chips, so...one would think it would be quite capable. as much as i hate highpoint with their 4320 series....they really burned me on that one. they had compatibility issues with barefoot drives, and i arranged for OCZ representatives (tony) to send them ten vertex for testing to iron the bugs out...me and ten other guys bombarded highpoint with emails for a month LOL...and they never did fix the issue. wonder what they did with the drives? very disappointing. but i positively love the 3520 so it is a toss-up here.
but all past dealings with highpoint aside, i am still interested in this controller.

http://www.tweaktown.com/reviews/330...er/index9.html

[Computurd]

and back to serious fastpath stuff i am refreshing my drives right now and imaging them, then i will run the detailed tests...which post number has them again? wanna make sure i get the right ones!

[mbreslin]

For 4 ports I'd just stick with ich.

[Computurd]

yeah i can see that view...but ICH cant do 2gb/s! this device can.
just wish they had a reviewer who knew what he is doing. this guy who does that review has no clue. i mean where is the straight up numbers? lol he is benching series 1 series 2 series 3...what the hell does that mean>?