Crucial C300 256g 8xR0 lsi 9211/9260

**mbreslin** · 03-13-2010, 03:32 PM

My vantage score was 7k or whatever off mike's wr, I don't remember and I'm too lazy to go look at the record, it's the nearest 10 thousandth to the wr, 60k or 70k one of those I don't recall. I will do a hardcore push for vantage whenever I get my loops finished and the 980 finally shows up, 60/70k was with 920@4ghz/5870/9260. I will have 980 at hopefully 4.5ghz or close/crossfire with 5970/5870. I should be able to put up quite a good number.

This is windows dynamic striping on the 9211..

I'm pretty happy with this though as I whined about earlier 4k randoms don't scale at all.

I'm sure I read I can make this bootable but does anyone know of any way to winsat this configuration?

Edit: actually looked at vantage scores and I don't know why I thought 60/70k, 20k was my score, I'll throw it up again later.

**SteveRo** · 03-13-2010, 04:02 PM

[QUOTE=mbreslin;4285548]My vantage score was 7k or whatever off mike's wr, I don't remember and I'm too lazy to go look at the record, it's the nearest 10 thousandth to the wr, 60k or 70k one of those I don't recall. I will do a hardcore push for vantage whenever I get my loops finished and the 980 finally shows up, 60/70k was with 920@4ghz/5870/9260. I will have 980 at hopefully 4.5ghz or close/crossfire with 5970/5870. QUOTE]

yep me to on the 980 - hopefull for a 4.5.

**SteveRo** · 03-13-2010, 04:06 PM

To run winsat against a dynamic disk - you should be able to just format the d drive via windows disk mgt and identify the drive letter in the command line like any other disk.

**Computurd** · 03-13-2010, 04:24 PM

jesus that is impressive. WOW !! that 9211 softraid is UBER
did you do the full vantage run, or just the HDD suite?
you should at least be able to be in the top twenty.

**mbreslin** · 03-13-2010, 04:34 PM

Oh I did full vantage, I'll have to buy it to do the hdd test, I will do this later this evening.

@steve, winsat fails and says this drive involves more than one physical.. :/

**Computurd** · 03-13-2010, 05:49 PM

i was also curious if you had placed just one device on the controller and benched it to see if it is performing above 3g? does the controller see it as a 6gb/s device and treat it as such? your results are great keep up the good work~!

**felix_w** · 03-13-2010, 06:08 PM

Originally Posted by Computurd

i was also curious if you had placed just one device on the controller and benched it to see if it is performing above 3g? does the controller see it as a 6gb/s device and treat it as such? your results are great keep up the good work~!

Isn't the one below with one single drive ? I'm A bit Confused with this result...

Originally Posted by mbreslin

That's 1 drive with your usual 9260 r0 settings from earlier only thing I turned on cache, I bet that skewed it, I have to go out for an hour I'll fix it when I get home if you need something different. (64k stripe)

**Metroid** · 03-13-2010, 06:12 PM

Originally Posted by felix_w

....

Cache is on.

**felix_w** · 03-13-2010, 06:23 PM

Originally Posted by Metroid

Cache is on.

Originally Posted by mbreslin

That's 1 drive with your usual 9260 r0 settings from earlier only thing I turned on cache, I bet that skewed it, I have to go out for an hour I'll fix it when I get home if you need something different. (64k stripe)

This is the point where i fell asleep probably...gonna rest myself now...i need it...

ty metroid

**mbreslin** · 03-13-2010, 06:37 PM

Originally Posted by mbreslin

That's 1 drive with your usual 9260 r0 settings from earlier only thing I turned on cache, I bet that skewed it, I have to go out for an hour I'll fix it when I get home if you need something different. (64k stripe)

Originally Posted by felix_w

Isn't the one below with one single drive ? I'm A bit Confused with this result...

This is single drive 9260, I actually tested each one because I thought one was bad, seems fine now though. "only thing different is cache, I bet that skewed it" I said this then because until then I hadn't turned cache on with the 9260 (I didn't think you were supposed to with ssds, obviously I see now that the 9260's cache shines) The 9260 sees all of my drives as sata 6gb devices, (and it's obvious in this bench from well over sata ii speeds in reads).

Also, just to be clear, I'm not done with benches or anything and I'll be doing different array sizes to see scaling, maybe if steve or someone will give me a standard idea of what benches to do so he can easily arrange them I'll start them soon, I just only have so much time and I've tried to cram everything in.

My verdict so far is that the drives are pretty awesome..

4k random doesn't seem to scale at all, this is dissapointing.

On the 2 controllers I have and ich10r latency seems to be about the same, I knew this going in, I don't see what can be done possibly crucial updating firmware but this seems doubtful, I believe I read a mod post saying it was unlikely there would be a fw update anytime soon. Is my array worth what I paid in performance? Probably not, certainly not looking at steve's acards, but they sure are fast and it doesn't look like something much faster is coming anytime this year so that's good enough for me.

Minor funny thing about the drives, when you take them out of the bag within minutes those pretty crucial stickers on the drives peel up and don't stop peeling until they litterally become unstuck from the drives, it's hilarious, this is not a heat issue or anything like that as it's just ambient temps drives aren't plugged in yet or anything, just cheapo branding sticker, or cheap adhesive or whatever I just thought it was funny.

**mbreslin** · 03-13-2010, 06:41 PM

windows dynamic striping on 9211, quite a good score I think.

Edit: dynamic disk is silly but according to tilt I can get these kind of iops without dynamic disk, I'll try in a little bit, lsi didn't lie, the 9211 can push 260k

(this is 4k random read)

**mbreslin** · 03-13-2010, 10:33 PM

Vantage HDD scores go here..

9211: 64640

9211 w7 dynamic striping: 70857 (I haven't seen anyone else's HDD suite scores but I bet this score is uber, too bad this drive isn't bootable/wei-able)

**SteveRo** · 03-13-2010, 10:37 PM

Mr. mbreslin, great scores!
Maybe you can't do winsat on dynamic disk - Anvil?
Tomorrow I will try to post some iometer setups that I like - the others you already have - AS SSD, CDM, ATTO, winsat, ... - maybe Everest.
CompuT - what bench program did you say was best for latency test?
Personally - I like iometer - it seems to be the most trustworthy.
Iometer is so configurable; the trick is getting consistent test setups for comparisons - from one system to another.
That is why sharing iometer setup files makes sense for most of us - but you probably already knew all this.

**mbreslin** · 03-13-2010, 10:55 PM

@steve: I did and that's why I asked anvil for iometer config files, no you can't winsat dynamic disks, or boot from striped dynamic disks, so even though I get awesome results I will never use it for 24/7. My original goal was to spend a month testing/configuiring/building and then settle in to a 24/7 set and forget. The 980 will come and then I'll wait and see how the 1880 is then after that I'm done.

**Nizzen** · 03-14-2010, 12:19 AM

Originally Posted by mbreslin

@steve: I did and that's why I asked anvil for iometer config files, no you can't winsat dynamic disks, or boot from striped dynamic disks, so even though I get awesome results I will never use it for 24/7. My original goal was to spend a month testing/configuiring/building and then settle in to a 24/7 set and forget. The 980 will come and then I'll wait and see how the 1880 is then after that I'm done.

Here is the 3 IOMETER configs we use for testing in the norwegian forum

http://www.diskusjon.no/index.php?sh...ost&p=15357931 (IOMeter.zip)

Random read, random write and Workstation.

The benchresults are in the first post

http://www.diskusjon.no/index.php?sh...c=1149834&st=0

**SteveRo** · 03-14-2010, 02:35 AM

Originally Posted by mbreslin

@steve: I did and that's why I asked anvil for iometer config files, no you can't winsat dynamic disks, or boot from striped dynamic disks, so even though I get awesome results I will never use it for 24/7. My original goal was to spend a month testing/configuiring/building and then settle in to a 24/7 set and forget. The 980 will come and then I'll wait and see how the 1880 is then after that I'm done.

Exactly my plan also - 980 now and 1880 when it comes out.
However, I will never be done testing - testing is a favorite pasttime for me.
Also - I learn so much in the process - not only from the testing itself but also from the comparison and analysis when collaborating online as we do here.

Which 1880 will you wait for?
I'm holding out for the 24 port 4GB version this time.

Wouldn't it be great if MS could figure out how to make dynamic stripped disks bootable?

**Anvil** · 03-14-2010, 04:25 AM

Dynamic disks behave differently in benchmarks, some work and some don't.
I did a few tests a while a go but I need to review the results to make sure what worked and not.

mbreslin,
According to AS SSD your array is not aligned but the scores are awesome

, whether a correct alignment would up the results I'm not sure but they probably would be.

I wanted to create a benchmark thread in cooperation with GullLars but he's been busy lately so I might do it myself.
(with iometer patterns, charts etc)

Here's a link to the iometer config file we've been using a lot lately. Link
Save the results to a separate file and post the results.
The test is read only (except for creating the test file) and performs QD tests from 1 to 128 on 512B to 64KB block sizes.
(total runtime is about 34minuttes)

You need to select the drive to test and set the size of the testfile. (4x the size of the cache on caching controllers, otherwise just use the default 1GB filesize)

**~~johnw~~** · 03-14-2010, 10:51 AM

Originally Posted by mbreslin

4k random doesn't seem to scale at all, this is dissapointing.

I guess you mean the synchronous 4K random (QD=1)? If I understand that test correctly, it synchronously reads random 4K blocks. Since 4K is much less than your stripe (chunk) size, I think it will only be reading from one drive at a time (unlike, say, 512K which could read from 8 drives in parallel if you have 64K chunk size). If 4K random QD=1 is only reading from one drive at a time, then it cannot scale. Have you seen any results posted where 4K QD=1 did scale with additional drives in the array?

On the other hand, 4K random QD=32 can and does scale, since the controller and drives are more than capable of reading multiple 4K blocks in parallel, and moreso as you add drives to the array. Interestingly, I've noticed that with single SSDs, even the 4K random does not scale with QD (i.e., QD=1 and QD=32 are the same) with Indilinx, Toshiba and Samsung controllers, but it does scale with Intel, Micron, and Sandforce controllers.

**Metroid** · 03-14-2010, 11:32 AM

Quite right at some extent johnw.

The 4K random write does scale, the 4K random read does not scale.

As fas as I have seen the raid0 decreases the 4k random read performance by 3 to 10% and in some rare cases %1.

http://www.xtremesystems.org/forums/...30&postcount=1

http://www.xtremesystems.org/forums/...34&postcount=1

http://www.xtremesystems.org/forums/...61&postcount=1

I have never seen a controller that scales the random read speed. The only thing that can scale 4k random read speeds are special controllers with cache like the Areca 1231ML-4G.

http://www.xtremesystems.org/forums/...0&postcount=46

Edit: http://www.newegg.com/Product/Produc...&nm_mc=OTC-RSS

**Anvil** · 03-14-2010, 02:39 PM

It is not possible to run the winsat disk assessment Tool on dynamic disks.

WINSAT_2R0_DYNAMIC.PNG

"Unable to open the drive or device.
This drive may be allocated across more than one physical disk. Before re-running the Disk assessment, check that your drive is not in a software RAID configuration and is allocated on only one physical disk."

**~~onex~~** · 03-14-2010, 03:12 PM

I guess you mean the synchronous 4K random (QD=1)? If I understand that test correctly, it synchronously reads random 4K blocks. Since 4K is much less than your stripe (chunk) size, I think it will only be reading from one drive at a time (unlike, say, 512K which could read from 8 drives in parallel if you have 64K chunk size). If 4K random QD=1 is only reading from one drive at a time, then it cannot scale.

sounds like a very good point

..!!

mberslin - wow! 264k 4k random (1GBps) is one freaking outstanding results!

**mbreslin** · 03-14-2010, 03:25 PM

pcmark vantage: 20814

puts me like 14th or something in top 20 (not sure why it doesn't show up in that list?)

http://service.futuremark.com/compare?pcmv=293661

**Computurd** · 03-14-2010, 06:15 PM

it will take time to show up. it will.

Excellent job bro! keep pushing

i think the FastPath that is coming will help you tremendously.

@stevero---Everest is the benchmark i refer to. of the 'easy' GUI based benchmarks it is one of the best imo. also, it measures down lower on latency than other benchmarks (hdtune etc.)