LSI 9211-8i

[Ao1]

@ gulllars ----> http://www.xtremesystems.org/forums/...d.php?t=243725

Hi Tiltevros. Could you please post the iometer config file that you used for the random reads?

[Computurd]

it is simple
4k 100 percent random 100 percent read. then adjust outstanding i/o , one worker .
we run them in messenger together all the time to test results so i know how he formats it, same as me

[Computurd]

NEW WINDOWS DRIVERS RELEASED TODAy

[Computurd]

new FW and driver for win7 has got me up and running raid 0 is scaling MUCH better, hit 1300 in sequential with quick test, havent got over 1050 previously. will post with more later today WOOT!!
Triple post FTW!!

[GullLars]

I see tiltevros, thanks for the numbers. At 4KB random you hit the IOPS roof at QD=24, at wich point you get 72928 IOPS @ 0,3001ms and 0,7904ms max. Those are really impressive numbers!
What also impresses me about this controller is from QD=1 to QD=10 you only gain 0,0233ms average latency.
It's also really nice you are able to exceed 1000MB/s @ < 1ms for 16-64KB random.

Now i would just like the numbers of a single x25-M on 9211 for comparing the scaling of multiple devices (the integrated RAID overhead). From what i can tell by just looking at the numbers, you have great scaling untill you hit the IOPS roof. For the larger packet sizes it seems the scaling reaches diminishing returns around QD = ([Devices]*[10(channels)])/([block size]/[4KB]), wich is where you start to see channel saturation.


Computurd: The reason for benching a single SSD of the same type as the ones in the RAID is to compare the scaling of a single drive by QD to the scaling of the RAID by QD, and do [RAID latency]/[singledrive latency] for the same QD/#drives. This directly translates into RAID overhead. If you take the number you get, subtract 1, and multiply by 100 you get the % overhead.


EDIT: BTW, the site dump.no has been taken down, so the files uploaded there are no longer accessible. The site was shut down due to lack of funding.
So stevero, if you have hosted the spreadsheets anywhere else i would love a link (i recently installed W7 on both my laptop and gaming rigg, and lost the benching files in the process)

[SteveRo]

Gul, I think I still have the files but I do not have a place to post them.

[SteveRo]

new FW and driver for win7 has got me up and running raid 0 is scaling MUCH better, hit 1300 in sequential with quick test, havent got over 1050 previously. will post with more later today WOOT!!
Triple post FTW!!

waiting, impatiently waiting!

[fhj52]

NEW WINDOWS DRIVERS RELEASED TODAy

Thanks!

NEW LINUX (PH4) DRIVERS RELEASED too... .

Sneaky LSI guys did not change the dates, yet. (ED: that's cuz the dates are correct this time: November 2009. => not "new", just new to public|website!) )
9211-8i Link: http://lsi.com/storage_home/products...-8i/index.html

I hope this resolves some issues w/ compatibility.

...they need new|updated firmware for SAS dual-port usage too. The adapter sees 'em but does not know it is same drive. It allows me to try make a 1TB array out of four 147GB(136GB real) SAS2 HDD. (... I wish, )

[GullLars]

Stevero, how about adding the raw .csv files to a zip archive and uploading it to this site as an attachment to a post? It's next to the smiley face when you post full replies.

Offtopic:
Gul = yellow in norwegian.
My nick is GullLars (by looking at caps, Gull Lars).
Gull = gold (also used as slang for luck) in norwegian.
Lars is my first name.
The nick means either Gold Lars (as in gold winner) or Lucky Lars, it's also a joke reference to a TV character named Gul-Lars (yellow Lars, a yellow human-sized bird in childrens-TV). Usually it's meant as lucky Lars, as it's my gaming nick, and people always accuse me of "just being lucky" when i have a streak.

[SteveRo]

Stevero, how about adding the raw .csv files to a zip archive and uploading it to this site as an attachment to a post? It's next to the smiley face when you post full replies.

Offtopic:
Gul = yellow in norwegian.
My nick is GullLars (by looking at caps, Gull Lars).
Gull = gold (also used as slang for luck) in norwegian.
Lars is my first name.
The nick means either Gold Lars (as in gold winner) or Lucky Lars, it's also a joke reference to a TV character named Gul-Lars (yellow Lars, a yellow human-sized bird in childrens-TV). Usually it's meant as lucky Lars, as it's my gaming nick, and people always accuse me of "just being lucky" when i have a streak.

Very good - didn't know I could just attach files! - ich10R ziped files attached!

[SteveRo]

1231ML-2G files attached.

[jcool]

Computurd (or anyone else using the 9211), could you tell me the controller init time at bootup (ie how many seconds it adds to the booting process?)
Now I normally don't reboot at all (24/7 cruncher after all), but I test new CPUs every so often which means a LOT of reboots. Now if this card adds 10-15s I can live with that, however if it behaves like an Adaptec and adds 45s it would be a no go for my mainrig.

Thank you

[One_Hertz]

Computurd (or anyone else using the 9211), could you tell me the controller init time at bootup (ie how many seconds it adds to the booting process?)
Now I normally don't reboot at all (24/7 cruncher after all), but I test new CPUs every so often which means a LOT of reboots. Now if this card adds 10-15s I can live with that, however if it behaves like an Adaptec and adds 45s it would be a no go for my mainrig.

Thank you

Off the top of my head, it is just under 20 seconds. But then it also ninja adds another 7-8 seconds when it hangs going "verifying DMI pool data". So I guess roughly 25seconds longer boot time with the 9211 installed.

[jcool]

Damn.. just over my "limit" then I guess. Is that verify DMI pool data hang specific to a certain motherboard maybe?
This is why I loved the Highpoint 3510/3520.. just under 8 seconds total

[Anvil]

jcool,

Why the limit?

I reboot once a month or whenever Windows Update needs a reboot.

I've got 2 raid controllers on my main rig, LSI 9260-8i and a PERC 6/i, sure it takes some time but then again I don't boot that often.

[jcool]

Read
Ordinarily I never reboot (like once a month for updates), but if I am testing new CPUs I do 100+ reboots in a single day. Waiting more than 10-15s for a Raid card can get REALLY annoying then.

[Anvil]

A simple solution to you problem would be removing the raid card while you test the new cpu's
It is an option, it might not work for you though.

[jcool]

Yeah, it would require the OS drives NOT to be on the Raid card though, which kinda defeats the purpose.
Or, I set up an SSD/HDD specifically for CPU testing. Might be better considering I risk tanking my main OS all the time

[Anvil]

Get the Intel X25-V 40GB it is a great boot drive.

I've been buying a few of those lately and it would fit the bill.

[jcool]

Yeah, I think I'm gonna wait for the new SSD gen though.. I'm thinking 4x Micron C300 @ LSI 9211-4i or something to replace my single CSX Indilinx MLC SSD. I'd use that one for CPU testing then

[Anvil]

Good thinking,

The C300 is first in line for me as well but the SandForce sounds tempting too.

[fhj52]

Computurd (or anyone else using the 9211), could you tell me the controller init time at bootup (ie how many seconds it adds to the booting process?)
Now I normally don't reboot at all (24/7 cruncher after all), but I test new CPUs every so often which means a LOT of reboots. Now if this card adds 10-15s I can live with that, however if it behaves like an Adaptec and adds 45s it would be a no go for my mainrig.

Thank you

I'd say ~12 sec's here.*
It depends on how the adapter is configured and a great deal on the system BIOS as well as system config. My AMI BIOS adds an additional ten second + wait before loading the boot menu after the adapters' BIOSes are finished. I too do not normally (re)boot but have been a lot recently. That extra 10+ sec drives me crazy ... I am not a patient man, .
Newer mobo and BIOS should, ihope, fair better.


*After timing it I have to say that's a best time. A 15 to 30 sec. wait is more typical and then there's another BIOS wait of ~10-15 secs while the system BIOS and booltloader spin their wheels. End result on this Supermicro H8DCi (dual opteron) with only four SAS drives in one LUN on the 9211 is 25 - 45 secs.

[jcool]

12 seconds sounds fine.. hmmm. I'm curently using a DFI mobo (Award Bios) but will be switching back to my E760 classified soon (Award as well, but a slow booter...)
I really like the bootup post speed of the DFIs, can hardly read anything before gettng to the Windows Boot Logo
Configuration would be 4x SSD in Raid 0 ofc, single volume.

[Spoiler]

If I remember right, my 9211 took close to 30 seconds. It's still better than my adaptec card, but nothing like booting a PC without a raid card.

Off topic, I agree with the Intel 25-v 40G as a boot drive. My pc boots to the Win 7 desktop in 14 seconds from power up.

[fhj52]

Yea well the LSI 9211-8i part of the boot will take longer if the system is being crashed, gets locked and requires hard reset. Then is prbly closer to Spoiler's value ... 20-25 secs than to 10-15. Both the system BIOS and the adapter(s) BIOS will run longer. I know this because, among other reasons, I have been doing that for the last few days due to a borked Linux OS boot, .
For crashing systems, whether with CPU tests or perhaps OS or other software modifications that will cause such, a BBU as with the 9260-* would be advantageous. Otherwise, prbly should use a separate system dedicated to such tests and do not keep any valued data on it.
In any event, configure the adapter for the target. E.g., if only LUN 0 is being used, no need to scan for all the rest of the stuff that the adapter can handle, so configure it appropriately and it will knock-off some time.