hard drives & raid - benchmark and compare!

[SaFrOuT]

the link for the ATTO is not working

 

any other link please since i can't fiond it anywhere else and their site is not that friendly

[seamuso]

the link for the ATTO is not working

 

any other link please since i can't fiond it anywhere else and their site is not that friendly

 

http://www.atto.org/register/getfile1.php?...hbautil_300.exe .. it's part of this package

[soundx98]

Thanks ExRoadie, The Raptors are part of the new Lian-Li V1000-5x-57-EVA 7800GTX KO build. Gonna do up some pics (very clean) in a day or two. Getting better camera

[LanceDiamond]

After 6 days of building, burning in, tweaking and benchmarking my new system, I'm stable! My system is working well too.

 

Right now I'm running a RAID0 array of 3 x Hitachi T7K250 (250gb SATAII drives). Default cluster size/stripe size - just booted from SP2 CD and installed XP SP2, then updates, drivers and assorted other benchmarking stuff. No defrag needed as I have yet to delete anything lol.

 

My HDTach looks good:

 

 

But in the ATTO bench, I don't understand the results - if I run it with default settings, I get this result:

 

 

If I change to 32mb "Total Length" and "Neither" then I get this:

 

 

I do think my 2nd result, where I change from the defaults, is more "real" because I don't beleive I'm getting 650mb/sec reads. Any idea why ATTO would show me that on the default settings?

 

Ahh, actually I may have my own answer...Caching - 32mb "Total Length" gets me results where caching doesn't skew things maybe?

[red930]

@LanceDiamond

 

"32mb "Total Length" and "Neither" then I get this:"

 

If you look at some of the previous submissions you will see the exact settings we use.

[LanceDiamond]

@LanceDiamond

 

"32mb "Total Length" and "Neither" then I get this:"

 

If you look at some of the previous submissions you will see the exact settings we use.

 

If you look at my post, you'll see I did that - I used the same settings others used and in fact what I mentioned, what you quoted me on, ARE those settings.

 

My question, which I think I answered myself anyway, was why with the default settings I got such insane read figures. Cache is what I am assuming.

 

Anyhoo, thanks!

[red930]

You are correct in your analysis - the cacheing is kicking in.

 

Did you run the default setting more than once on fresh boots? It is always a good idea to run several of the same bench marks and then average the results out.

[Gogar]

unfortunately it seems that Si3114 controller is going to be in the new boards...RG and I are asking if we cannot get the Si3115 at least in the top of the line boards

Is that a pci-e controller or still capped to 133MB/s shared?

I think the bad performance with >2 drives on the sil 3114 is almost entirely because of the 133MB/s cap of the PCI bus, and it's already shared with the marvell controller.

 

Look into the sil 3132 combined with a sil 3726, this would yield 6 SATA II ports on a pci-e 1x link. (266MB/s in both ways)

look at this link:

http://www.siimage.com/products/product.aspx?id=32&ptid=1

 

If that's too expensive for DFI, perhaps just the 3132.

Separate sil 3726 1to5 port multipliers are available, maybe dfi can make their own?

 

Here is one implementation of a 3726.

http://www.cooldrives.com/cosapomubrso.html

 

I was thinking it would also be nice if there could be a small piece of pcb that could plug into the 2 sata ports of the sil 3132, with 2 3726's on the bottom and 10 SATA II ports on top. Maybe the 2 sil 3132 sata ports should be routed out as similar connectors as the karajan module along with a few power traces. Then a small piece of pcb plugs into it just like the karajan module and sits over the sil 3132. Probably a good idea to move it to the far end of the board though, so it won't interfere with pci cards.

 

mainboard:

 

[: : : : : : :]

_______

|sil 3132|

|-------|

|______|

 

[: : : : : : :]

 

 

 

storage module top:

 

 

 

storage module bottom:

 

two male headers and two 3726's

crystal and some smd caps?

 

edit: i've got some small additional ideas, like running the power of the sil 3132 through the module, effectively disabling the 3132 when the module is not in place.

 

Cheaper versions of the boards could simply come with a module that has only 2 sata II ports (just the ones of the 3132 routed through)

 

This should make it affordable for DFI because they can make the same boards for all versions.

Having a 3114 on cheap versions and 3115 on more expensive boards probably requires the boards to be different.

[Trevor]

[Angry_Games]

is that a 5400RPM 2MB drive?

[Trevor]

Its 7200 8mb, sucks compared to a raptor :tooth:

[{TCC}puTTs]

Here's my aging Maxtor drive.

 

 

Not to shabby