Jump to content


Photo
- - - - -

hard drives & raid - benchmark and compare!


  • Please log in to reply
1540 replies to this topic

#1537 catkicker

catkicker

    Certified Geek NOT

  • Members
  • PipPipPipPipPip
  • 3316 posts
  • Gender:Male

Posted 03 January 2008 - 04:00 PM

Would be great to see some benchmarks with matrix raid setup.
As soon as I have my new/used intel build up and going. i'm going to post benchmarks on the 50gig Raid) slice and the Raid10 storage slice.
One thing that I've noticed with all the talk about matrix raid and it's stripe sizes no one has mentioned cluster size.
I'm wondering if ExRoadie's formula for cluster size would apply to the matrix storage arrays.

#1538 supershanks

supershanks

    Certified Geek

  • Members
  • PipPipPipPip
  • 1932 posts

Posted 03 January 2008 - 04:24 PM

Just a few pages back m8:lol:
Enter the Matrix There's a few assocated posts after the link i just gave.

luck:)

#1539 SpeedEuphoria

SpeedEuphoria

    Geek

  • Members
  • PipPipPip
  • 755 posts

Posted 01 February 2008 - 11:48 AM

hey just wanted to bump this great thread and give my 1 cent.

1st I have 2 questions, since I seemed to hear in this thread that the Raid controller is linked to the pci e bus has anyone tried up the pci freq from 100 and see if anything resulted?(I know I read hear a long time ago that it shouldn't be OC'd much because of problems although I've seen some running @105?) Or what about chipset voltage, have anything to do with the raid controller and helping it out? I've allways left mine default b/c raising it never helped me get more clocks, but I'm prob psu limited and my stepping sucks.

Any way I have 2 arrays, on 6.70 drivers(been running 2 arays for 2 years on the raid controller so it deff works)

1st I want to say that like some others have mentioned, I dont believe the benchmarks are an exact replica of real world performance but may be helpful and also the burst speed may not be realistic(like why dont they show that burst speed on the graph in HD tach if its actually being applied?). They are somewhat inconsistent, I have conflicting results between ATTO and HD Tach v3.0.4.0:mad:

1st On the Raptors 32/4(os), they are faster than the Hitachi's obivously but especially at write(they write more than read).

I seemed to do better overall with ncq off, 7.9 access times, otherwise 8.4-5 access

burst and reads were best w/ only read cache off and this had the smoothest graph(no dips at all just smooth) but 8.5 access and reads looked good, but on ATTO didnt look as good with just ncq off. So what to do?:confused:

Can someone please explain to me what ATTO is actually doing with the settings talked about in the thread?

The Hitachi's

disabling ncq or read cache didnt seem to affect anything except for burst(406max:eek2:)

Conclusion: I seem to like ncq off on the Raptors and default(keeping sataII on and spectrum off) on the Hitachi's for now.

So basically the tests are back and forth. This is seeming more like the pagefile debate:rolleyes:. So since there is some reasoning behind the 4:1 I will stick with that(an a pagefile also:)). SO 16/4 on the raptors for boot and 64/16 on the hitachi's for video stuff

#1540 Hemidare

Hemidare

    Certified Geek

  • Members
  • PipPipPipPip
  • 1398 posts
  • Gender:Male
  • Location:Sylvan Lake, AB

Posted 02 February 2008 - 08:09 AM

http://www.anandtech...doc.aspx?i=2922

That says that the NCQ was not available on the 74 gb raptors. Just something called Tagged command queing (TCQ). That is probably why NCQ doesnt really effect anything on your raid with 2 74 gb raptors.
Posted Image

#1541 SpeedEuphoria

SpeedEuphoria

    Geek

  • Members
  • PipPipPip
  • 755 posts

Posted 02 February 2008 - 01:16 PM

Correct it is TCQ mine are WD740GD FLCO's, I havent messed with these since I installed them 2yrs ago so I'm trying to reinform myself so bare with me as I also forgot what the jumpers were for but found that info here,
http://wdc.custhelp....ated=1071171477

So has anyone with Raptors enabled spread spectrum? I guess I can try it, but afaik all the spectrum stuff in bios and on the HD's is for emi emissions, and basically it makes the equipment work out of sync so its not pulsing/emiting at the same times so emi doesnt spike as high. Basically it should be a good idea for a HTPC or sound editing where emi can affect sound output/processsing.

What exactly is the Enable Command queuing for in windows? I read all of this thread and I believe that most of it was talking about turning it off or on, also I believe someone mentioned that it was more for a multiuser environment
Edit: I found a page on microsoft's website and it said basically when on it sends a message to flush the cache so that its safer yet slower for data, then on another OS(forgot which one 2000 I think) there is another option for Power Protection and if you have that on and enable cache it will use the cache on the HD's. I tried with both off and the results sucked on mine especially for writes that were the best numbers with in on.