ICHR5 Raid/IDE performance

[Shayd]

Hey Folks,

I have moved my 2 74GB raptors over from the promise controller on my P4C800-E Dlx (1018) to the ICHR5 sata connections. I replaced the sata drive I had on the ICHR5 sata with a PATA 200GB drive to free up the slot for a RAID0 setup.

Performance seems to be low though with the PATA running at 43MB/s and the two 10krpm raptors only putting out 106MB/s

106 seems really low :cry: as I think I was getting (forgot to write it down) 113 or so on the promise.

I have updated to the lastest IAAR and 875 chipset drivers.

Any suggestions? Thanks

[Shayd]

BTW, I was suprised to find out that the Sata Raid controller is on IRQ 18 with the network controller, as well as the USB controller if you have 6 ports enabled :yikes:

[Shayd]

Well, Tweektown's review was getting 113MB/s on Sandra.

Since noone had any comments I'm going to swap back to the promise and see what the performance is again. I was sure that someone said that ICHR5 was faster for RAID0. We will see I suppose barring my own induced problems ;)

[Snafu]

Hmmm. I thought I had posted. Try Atto to judge the performance of your raid. Here's a link (http://www.bleedinedge.com/forum/showpost.php?p=28577&postcount=16) to follow from pointreyes. It was in the thread Higher Speed or More Space? (http://www.bleedinedge.com/forum/showthread.php?t=3076&highlight=atto).

I tried Sandra and I get abissmal scores even worse than yours. I seem to recall that they were 68MB/s and my non-raid sata was 48MB/s. I started a thread about my raid set-up on promiseRaid configured right? (http://www.bleedinedge.com/forum/showthread.php?t=3104&highlight=raid+configured)

Now I wonder where my post went to :scratch: . Maybe I never hit send :bonk: .

[Shayd]

hehe...the sock eater from the dryer must have wanted desert and ate your post :lol:

[Snafu]

Yeah, that's it. :lol:

[Freddy]

Hello Shayd,

that really surprises me, but... I am not really a Sandra fan. What are your HD Tach 2.70 (http://www.simplisoftware.com/Public/index.php?request=HdTach2.7) scores? For my two 74 GB Raptors on ICH5R in RAID0 the scores are:
RAT= 7.8, RBS=133.2, Read=112.8/53.6/82.7, CPU=4.0.
By the way: what PATA drive is that? The scores for my 160 GB Hitachi 7K250 are:
RAT=12.1, RBS= 90.3, Read= 65.6/31.1/49.6, CPU=3.3.
And I would not rely on Sandra for making these decisions...

Freddy

[Shayd]

Ok, DL'd HDTach and here is what I got for PATA (Maxtor 200GB Diamond Max Plus 9.3 UATA133 7200rpm)
RAT=14.1, RBS 110.4MB, Read 62.6/31.7/47.5 CPU 3.0

SATA 10krpm Raptors ICHR5
RAT=7.9, RBS=115.5, Read 106.2, 41.9, 70.3 CPU 2.7

I'll check promise next...

[Shayd]

Ok, here are the Promise numbers for the 2 74GB Raptors

RAT=7.8, RBS=99.0MB, Read=85.9, 50.1, 69.6 CPU 3.9

Interesting, Promise uses 1.2 percent more CPU thats ~23% more then ICHR5

[Shayd]

HAHA....just when you think you know what is going on....:lol:

Here is the ICHR5 numbers after moving the drives back over from the other mobo connections.

RAT=7.9, RBS=115.9, Read=103.9, 50.7, 70.5 CPU 2.5

Go figure...:scratch:

OOPS, I guess that is within the error margin after all... :o

[Snafu]

For coparison sake using HD Tach 3.0 and just noticed you are using 2.70 :bonk: not sure how comparable these will be.

My set-up are 36GB Raptors in Raid with 64kb stripe and about 10GB used up. The following numbers were generated by running the long bench (32MB zones) rathe rthan the quick bench (8mb zones)
RAT = 8.8ms
CPU Utilisation 3%
Ave. Read: 67.9MB/z
Burst Speed 111.0MB/s

Looking at the graph for Sequential Read Speed it never drops below 50MB/s and peaks at 95MB/s

[Freddy]

Hello Shayd,

those scores for your PATA look fine! Just one question though: RBS=110.4 MB/s on a UDMA-100 connection!? :scratch: How can that be? It is connected to Primary/Secondary IDE, not to the PATA Promise connector?
The Raptor scores are not too bad either. Am I correct that you are using a 128 kB stripe size, considering the low CPU usage and the little low RBS and Read speeds? I am using a 64 kB stripe size.
Compared to the Promise there is some difference, although it is not that big. But next to the slightly higher scores and the lower CPU usage, the ICH5R does not burden the PCI bus. So the ICH5R is still the best option!

Freddy

[Shayd]

The 200GB is connected to secondary IDE and a DVD RW on primary (It was the originally the only PATA drive) Nothing is connected to the Promise.

Sandra sayes its UDMA-6. The interface is UATA-133

Yes, the Raid0 is 128k stripes as well

[ThugsRook]

My set-up are 36GB Raptors in Raid with 64kb stripe and about 10GB used up. The following numbers were generated by running the long bench (32MB zones) rathe rthan the quick bench (8mb zones)
RAT = 8.8ms
CPU Utilisation 3%
Ave. Read: 67.9MB/z
Burst Speed 111.0MB/s

my single SATA WD36gb:
burst = 105.6 mb/s
avg = 49.7 mb/s
seek = 8.4ms

...and thats w/o using a sata driver (so its only in ata100 mode)

interesting :scratch:

[Snafu]

Interesting :scratch: ? I should :hide: :o . Something seems Snafu'd with these results - not a heck of a lot of gain running raid 0. Curious if you results were using version 3.0 and the 32MB zones. Sometimes versions report different results (I can only hope :lol ).

BTW the temps really jump when running the test with 32MB zones

[ThugsRook]

v3.010
32mb test

you might wanna try a 64k or 32k stripe

:beer:

[Snafu]

Hmmm. I ran it at 64k stripe. Do you think 32k will perform better or would it only help with this bench (I heard larger 128k would be better for real use).

[ThugsRook]

prolly only help with this bench.
try a search on ABX for other ppls results at certain stripes.

[pointreyes]

If you are using more smaller files then larger files and prefer to defrag your drive more ofter then use 32k stripe. 128k stripe is good for large files and should be used for things like imaging, video encoding, etc. 64k stripe strikes a middle ground. Just note that you will have to defrag a drive more often the smaller the stripe is.

I prefer 128k.
Here's my SATA controller results. Not a miserable Promise so-called controller and not ICH5R, it's not even RAID0. (http://members.dsl-only.net/~pointreyes/ga/RC4852.jpg)
And just for fun. Er, this is a single drive. (http://members.dsl-only.net/~pointreyes/ga/Seagate.jpg)