SWEET Raid Benchmark Result

Judas

Playing around with AIDA32... came across there sweet Disk Benchmarking utility....

Ran the Linear run... and i'm not sure if the results are right... but wouldn't these state i have a pretty nice... setup?



To bad i couldn't do a write....

Judas

BTW.. considering i'm running 2 30gb 2mb cache maxtor HD's In raid 0 AND 256kb block/stripe length

easyrider

So what's the best raid to setup? I have 2 120gig seagates in raid0, not sure what the KB is on it.
What's the best way to set the array at?
Also, if I adjust the array will I have to format to do it?

Judas

well... lately... i've tried 64kb/128kb/and 256kb now.. (i can adjust from 4kb all the way to 2048kb)..... each step up has actually given me a higher preformance.... Depending on what your doing.. with your computer... you should adjust it acordingly... I'm using a Highpoint Technology 374 onboard chipset... which is pretty simple to navigate .... I'm not sure about you... you might be using Promise ..... you could look at your manual on how to setup..... And to adjust the block size.. you must reformat the drives computer... even remove the partitions.... this will all be/should be done by the controller itself... if you play games.. the latest.... i think 256kb is right were it's good..... i think i'll push up to 512kb my next reformat... reinstall.. and see what that does for a speed increase....

easyrider

Hrmm, odd, I was told that setting it lower would result in faster speeds, and setting windows cluster size the same as the raid. Mine were set at 16k in the raid, so I jsut did partition magic, changed the cluster size from 4k to 16k, and ran hdtach, and I get the EXACT same benchmarks.

Judas

... saposedly... but then again.... you have any idea how old those are... we are now dealling with files that are anywere from 40mb to .PK files at well over 400mb to even the montrous 1gb files.... I let windows set the default cluster size.... lower was saposed to be faster..... but i'm thinking it's not the case anymore enless we get choices of 8mb all the way up to 512mb block sizes... (or larger).... see with 16kb or even 64kb for that matter.... deviding a file that's lets say is 5120kb (16kb=2x160 chunks) (64kb=2x40 chunks) ... mine setup with 256kb is (256kb=2x10) ...theres a preformance difference... i think when you start cutting the file into more chunks.... it actually start degrading preformance..... you basically want to pull one file off both hardrives at the same time through one block size if you can..... that'll give you the full bandwidth... but having to pull 320 block...gets to be a little teadious... However, this i my understanding...

Judas

BTW... changing the cluster size won't impact preformance if i understand clusters correctly....only the RAID block size (or stipe length.. which ever you want to call it)

BBThumper

......for your set-up 16k or 32k stripe is the best for performance. You will lose all data on the drives when you change the stripe size so only do it if your ready and willing for a new format.......

Judas

you sure?

I've yet to see a some recent facts that say this as every time i pump up the stipe length..... (block size)... i get a bandwidth jump (although small).. and a average is jumped up a good amount.....

16k or 32k is just to small for nowadays files and whatnot......

BBThumper

........I was told this in AMDForums so I thought I would give it a try. For my AMD set-up with the Sil Valley controller 16k worked better than 32k-64k-128k. I know with Intels it's a little different. No real performance lose between the stripes.......If I can find the article explaining this I'll post the link........I also tested different clusters and just like you stated, the default 4k with XP was no different than 8k-16k-32-64k when changed with partition magic 8.........

BBThumper

......here's one and I'll look for more if you like:

http://www.pcguide.com/ref/hdd/perf/...erf_Stripe.htm

......and here is a Intel forum that helped me with stripes:

http://www.abxzone.com/forums/showth...threadid=66205

Judas

yep... seen those raid links before... the one first on being rather old... 97 it was produce... even though the site itself my have been updated since last 2001.... still old...

BBThumper

.....from all I have read if you are moving around large files like DVD/Video then the larger stripe is better. I myself use only small files so the small stripe is better for me....

Judas

how small though.... if your a gamer... even if soe of the games are over 3-4 years old (i've even some well over that and some if not majority are well over 20mb in size)..... the large stripe size is better.. you lose speed and preformance when positioning the files for best speed starts to counter what your trying to do... it has to split the file into more junks.. place the equally.......... like i mentioned before... a 5120kb file

Judas

BTW... most system files are much larger now as well..... they don't sit between 32kb in size and 256kb.. they are either well under.. .being 1kb to 5kb... average... and those that really well over 256kb.....

BBThumper

.......thks for the info. I will try larger on my next format. I will be going offline at the end of the month until the end of the year so I am sure I'll be formatting alot just trying to keep busy ...........

Judas

sorry to see you go..... do try multiple things...

Zardon

judas, im just looking at the image you posted. I cant see that bench being too accurate. your scores are all the same, for average, maximum etc.

Judas

i figured as well.. but then i ran some other test or programs... they are reading nearly identical....

Judas

I tried HD Tach again... i've never had success with it... it always takes my readings wrong... i either get way way way wrong results for my burst speeds and average/min...max transfers.... stuff like that... it did get my transfer rate of max 87mb/s and average of 72mb/s and a min of 42mb/s..... but the benchmark runs for about 2 seconds and

Access time is all screwed... first it's 12ms... then it's 30+... then... who knows.. my CPU utilization is the only things that remains somewhat constant.... over then 40% mark... which is wrong for sure cause i was running hd tach at the same time as a few other CPU utilization programs..... and the faster speeds.... and not one time... did my % go over 5....maybe a jump to 7-10% once a inawhile.. that's it

BBThumper

......just do not have the money right now and this is one thing that isn't a necessity . I need a break from the internet anyway. Thks for asking ........

.......I have found H/DTach to be more consistent for me than Atto:

Judas

nice read burst.... wow... what'cha running?.... i know when i run the program on other computers using the standard controllers and stuff.. it works great... and is consistent.. but not on my stuff... i just can't get it to read it right....

BBThumper

.....sorry i thought it was in my sig, 2 x Raptors 36gig in Raid 0. I know that feelin when certain programs work good for everyone else but as soon as I install, crash ..............

Judas

it doesn't crash.. it just incapable of reading my stuff correctly... the highpoint controller.. it's way of testing... just isn't reading right... for all i know... it might be guaging the cpu usage by the controllers own control usage.... or something of the sort... it's not getting correct readings...

ChoGGi

any ideas why my cpu utillization is 0.0? shouldnt there be at least a little bit of it
i tried runnning it a few times but still no change
asus k8v deluxe, two maxtor 80gb diamondmax (on the via controller, it also has a promise controller)
think it might have something to do with cool & quiet?



edit:those aida tests are cool

Judas

i think HD Tach has a few problems with some controllers.... i'm not sure what it does...but i just recently reran it and it said my CPU ussage was 98% and i was watching my cpu using Taskmanager... and it never once went over 7%