Which benchmarks for a RAID card review?

[Computurd]

Well if you are going to be doing tests with dedicated hardware cards only two benchmarks are reliable and wil lgive you reproducable results, and that is i/o meter and everest. ATTO will give you so many borked readings you will be tearing yoiur hair out LOL it does not handle cards well AT ALL. The big key is being able to reproduce the results, one run to the other, if your benchmark will not do that then you know that it is having issues. I am speaking from my experiences with SSD and harware raid, but i have heard from many that they have the same issues with hardware hdd raids. Consider using Everest really for read only tests, requires blank disk for write testing, but i/o meter will give you awesome results, and the bootup profile has a mix of commonly used input/ouput patterns that are used in a normal everyday usage of an OS, lots of small files used randomly. There are some really good info HERE ALSO there are downloads for different profiles in the second post.

[pjkenned]

ATTO isn't too bad on any of the cards I have used with mechanical drives. Sure one run may be 730MB/s max another may be 718 and another 738, but that's a reasonable margin of error. The SSD's on the ICH10R can swing a few percent per run on almost anything I run.

Then again, SSD + HW Raid is going to be a bit different.

[Griff805]

I would talk with PC Per. and see what IOMeter test they run. That way you have an already established test that you could compare results with.

Are you going to be taking into account the whole fresh/degradation with the SSDs?

If you're going to do this review, I would seriously consider not half-assing anything (not saying you will). These SSD folk are picky and vicious and you'll be torn apart on the OCZ forums and these forums and all the others. So, I wouldn't rush any tests, or neglect to cover all your bases.

[Griff805]

Well if you are going to be doing tests with dedicated hardware cards only two benchmarks are reliable and wil lgive you reproducable results, and that is i/o meter and everest. ATTO will give you so many borked readings you will be tearing yoiur hair out LOL it does not handle cards well AT ALL. The big key is being able to reproduce the results, one run to the other, if your benchmark will not do that then you know that it is having issues.

How do you know for sure if the inconsistent results are not the drive or controller? You may be looking for consistent numbers, but what if that's not what the drive is doing? How do you know the other benchmarks aren't rounding everything?

I'm not saying you're wrong, as I think ATTO is a waste of time, but I don't see how you can make a statement like that unless you've written the benchmark software, or at least know at a basic level with the benchmark is doing and how it interprets the results?

[Computurd]

well i do have a bit of experience with this area, and i have personally tested just about every major benchmark with the following things
ICH10r/ICH9r
Areca 1220
Areca 1680-ix
Highpoint 3520
Highpoint 4320
LSI 9260-8i

and the above statements that i have made about this holds true, across all of those differnet controllers. I speak from personal testing. The differences in the benchmarks results, themselves, attest to the fact that they are borked when used for hardware raids with certain controllers. how many attos have i seen where it tells me that my drive is writing at 1000mb/s but reading at 150? i couldnt even begin to tell you. the results are sporadic enough, and have enough of a shift in realistic performance numbers as to be written off as unreliable. this very same issue has been talked about extensively on another forum that i frequent that is more storage oriented.

[F.E.A.R.]

@ Naplam

Not fast enough, because you can´t reckon on the cache all times.
Try this without MFT and Ramdisks...

http://www.hardwareluxx.de/community...&postcount=123

[NapalmV5]

why you drag me into this thread again ??

ramdisks ?? are you suggesting ive been "ramdisking" ?? FYI ive never done ramdisk.. never posted any ramdisk benchmarks.. never posted any real apps results/vids done on ramdisks.. ever! why?? why would i throw down the sheethole the systems ram performance when i need every drop of it

as for mft just the benches ive posted.. thats all.. i dont use mft @ real apps.. never posted any real apps on mft.. why?? its actually slower than just raw 1231 performance

i guess i have to repeat once again.. all the vids on youtube were done @ 512mb cache.. this very moment i type this on the 512mb cache.. 512mb cache is what i use 24/7.. the few apps that are faster on the 2gb cache i dont use or have use for

the performance you see in the vids is my daily norm pc performance raw and pure just the way i like it

i dont need mft.. i dont need ramdisks.. you guys do ramdisks just to get closer to my performance.. all your ramdisks cant match my 1231.. fool yourselves all you want

sorry F.E.A.R your acards are no closer despite your 1261.. the acards just arent what iometer cracksem up to be.. dont fool yourself


PS.. you guys want to challenge me.. again? want to suggest smtg about smtg about me? start a new thread or continue per Rhys: http://www.xtremesystems.org/forums/...d.php?t=229628

[F.E.A.R.]

But i saw your perf. without cache (comparable with full cache) http://img355.imageshack.us/img355/4111/1231bc8.jpg
I say it again and again. Your setup with Flash-SSDs can´t beat a setup with Ram-SSDs at long distance and very heavy loading. Sorry napalm.

[NapalmV5]

once again.. what have i claimed/posted @ post #8 ???

once again.. wtf have i claimed/posted here @ my own thread: http://www.xtremesystems.org/forums/...d.php?t=207557 ???

heavy loading?? 106 app loadup light loading ??

if your acards would be all that.. you should be faster than 18 seconds.. the whole world would know about it from you

sorry fear wishful thinking your ram based ssd is faster

[F.E.A.R.]

heavy loading?? 106 app loadup light loading ??

i would say - yes, because the apps are too small (serial loading - not parallel).
E.g. the first 10 apps are ready with loading, the next apps begins to load.

Try it with virusscan and/or IOmeter (e.g. workstation or database) in background.

[NapalmV5]

sorry to say this fear but you dont have a clue what that kind of load is on the system.. the system fires on all cylinders.. every yota of the system in motion

pause the vid @ 1:40 at the 106 app loadup and youll see 144 processes running before it settles @ 129 look how much ram is used and then i loadup cod4.. light load??

dude wtf!? you make it sound like i launched 106 notepads + another notepad.exe

dont worry ill show you what a light (hundreds of apps) loadup looks like

dude what is up with you? tell me what you have attempted and your acards failed to satisfy you?

i know.. you dont have to say it.. i know you cant match my normal/cached/mft benches @ pcmark.. neither @ 100app loadups.. i know what your acards are capable of.. im sorry it hurts.. get another acard.. 2x acards whatever it takes..

if you got beef with me.. take it on me with your system not with senseless statements

i expect more from you

[Computurd]

Napalm this is awesome!

[Griff805]

What's with the Napalm hate- The other thread was started by Rhys and Napalm obliged. He spent a lot of time posting videos and pics of his benchmarks that he did not fake. Why would he? He's got a great setup and now everyone is busting his balls for being proud of it. What gives?