LSI 9260-8i Versus Areca 1231ML-2G

[Anvil]

I apologize for not getting back to you sooner -
The settings I used in iometer are as you see them listed with everything else on default settings with each measurement run for 1 minute each.
One minute for 4MB 100% sequential 100% read, queue set to 64
One minute for 4MB 100% sequential 100% write, queue set to 64
One minute for 4KB 100% random 100% read, queue set to 64
One minute for 4KB 100% random 100% write, queue set to 64
...

No problem, I received the last drive yesterday so I haven't had time to do the tests yet.

What filelength did you use and how many workers?

[XS Janus]

For that kind of price... I don't know What they can offer that will be worth it over the new LSi card...

[SteveRo]

No problem, I received the last drive yesterday so I haven't had time to do the tests yet.

What filelength did you use and how many workers?

file length was 8-10GB, workers was 1.

[F.E.A.R.]

@ SteveRo

Nice comparison.
The Copy-Test and seq. r/w with LSI and Acard is disappointing me.
LSI only a paper tiger?

I want to see more tests with real apps.
I think i can stay with Areca

[SteveRo]

@ SteveRo

Nice comparison.
The Copy-Test and seq. r/w with LSI and Acard is disappointing me.
LSI only a paper tiger?

I want to see more tests with real apps.
I think i can stay with Areca

Yes, the Areca is faster at random IO than the LSI no questions.
When using acards it makes a big difference.
That said, the LSI is faster at sequential reads - even with acards.

[Computurd]

i would like to see some bootup runs with the 1231 on I/O meter, and the random file runs on I/o Meter as well. I guess that i remain unconvinced because these benchmarks are not reliable. How in the world, for instance, can the areca be getting 1635 mb/s seq read speed in the crystal disk benchmark??? that is physically impossible for that controller, and that is a widely accepted fact. The atto benchmark screens are totally unreliable as well with the controllers,l look at the 128.0 results..you expect me to believe that the 1231 is reading at 100 mb/s but writing at 850 at that block size? on other forums that i frequent there has been a agreement that the only real benchmarks that work with hardware raid is Everest and I/O meter. I would really really like to see runs with those if you have time stevero, i could provide you linkage to i/o meter if you need it.

[NeedMoMegaHurtZ]

I guess that i remain unconvinced because these benchmarks are not reliable. How in the world, for instance, can the areca be getting 1635 mb/s seq read speed in the crystal disk benchmark??? that is physically impossible for that controller, and that is a widely accepted fact.

2GB cache. The bench is running in cache.

Just like some benchmarks posted for the LSI are unreliable, unless you use a larger test size. Do some runs at 50MB and then 1000MB on CDM and I suspect you will see the difference on a 512MB cache controller. You see people posting 1500MB/s using only 2 drives on some benchmarks because the bench is run in cache.

[Computurd]

I am aware of that, and also that lots of benchmarks only measure the peak performance of the device, they do not give an accurate representation of the whole, that is why i only really trust everest and I/O meter personally. However, this also does not account for the atto readings, atto is a borked bench when it comes to hardware raid. AS SSD is also not accurate with hardware cards.

[SteveRo]

I am aware of that, and also that lots of benchmarks only measure the peak performance of the device, they do not give an accurate representation of the whole, that is why i only really trust everest and I/O meter personally. However, this also does not account for the atto readings, atto is a borked bench when it comes to hardware raid. AS SSD is also not accurate with hardware cards.

I agree with you on iometer, it is the one I trust the most also.
I did a complete iometer compare of the LSI vs the Areca controller in the first posting of this thread, see the first table that I posted -

http://www.xtremesystems.org/forums/...d.php?t=234374

- iometer 2008.06.18-RC2 was used. The table (in the first post again) is the excel output from iometer.
All of the conclusions that I draw from the comparison are based on these iometer results.

[Computurd]

Well that makes me feel dumber than usual, shoulda noticed it said i/o meter in the upper right corner there... for some reason i was thinking that was a compilation of the otyher benchmark scores...my fault!

[dinos22]

im testing the Areca 1231ML
how are you getting those huge crystalmark readings with SSDs?!?

[SteveRo]

@ Mr dinos - good morning and congrats on your acquisition!
If the CDM runs you are looking at are unbelievably high they are probably being run out of the areca controller cache.
When will we see your pcmv score - can't wait!

[dinos22]

i only have stock cache 512MB on them
i hacked up one of my corsairs thats 2GB and killed the stick so lol have to test with 512MB
no wonder....you are running the bench on cache and im on actual drives hence the performance is ...same as adaptec

[dinos22]

ok 2GB ram and i can see similar figures now..........but vantage is aweful

have u done any PCMV testing with SSDs?

[Nizzen]

LOL I got 24k with ONE ssd now

[Computurd]

awesome nizzen! which drive, and do you have it on your controller?