ASRock939Dual RAID - 22% CPU utilization?

[CFster]

Check this post out.

Basically this guy is saying that in order to perform RAID functions on the SATA1 ports requires 22% CPU utilization.

I ran my own test:

First, 2 Maxtor 80GB SATA150 in RAID 0 configuration on the SATA1 ports:

Test 1

Second, a single WD 120GB by itself on IDE2

Test 2

While the speeds are faster on the RAID config, CPU utilization is 27.3% vs. 2.9%.

Is this normal for RAID? Or is the controller on the ASRock board crappy.

 

[CFster]

Nobody?

Is this normal for a RAID controller?

 

[Greenman]

I'll test it on my setup and let you know.
cpu usage is 11.1% on my raid setup.

 

[CFster]

What's your setup?

Do you have an IDE drive too?

 

[Greenman]

A7N8X delux
AXP Barton @ 2200mhz
2 36 Raptors Raid 0 and an old 20 gig ide I use for backup.

Edit: I was going to pick up an ASRock board and a 146 opteron, but if the raid controller sucks I might have to find something else.

 

[CFster]

Did you by any chance test your IDE drive?

 

[Greenman]

No, but I will.

Edit: The ide drive is 3% cpu usage.
Since the raid array has an avg throughput of 88.8 mb/second as opposed to 29.2 for the ide drive, I'm not going to make a fuss over the cpu usage.

 

[CFster]

Thanks for your help.

This is a deal breaker for me. If I can't get an aftermarket RAID card working properly this mobo is getting yanked.

Glad it was cheap.

 

[Greenman]

I hear ya. Everyone says raid 0 isn't worth it, but 88 megs a second is pretty good. I don't think I want something half that speed.
It sure seems like a brand new board with a high end cpu should be able to do a little better than my 3 year old rig.

Edit: It would be nice if someone with an nforce 4 board could test the raid controller, perhaps high cpu usage is the norm on newer boards?

 

[Greenman]

bump:
Anyone else have anything to add to this? Is the raid controller on this board junk?

 

[Greenman]

Last bump.
Are we the only two people on earth that use raid 0? Thow me a bone here.

 

[CFster]

Yeah, wow.

Must be not a lot of people using the ASRock board.

 

[CFster]

Well, I just RMA'ed the board - too many problems.

- Lousy RAID performance
- Flakey drivers - I had to reinstall XP three times to get it working right, and even now it's not 100%. XP reports AGP 4x even though it's set to 8x in the BIOS.
- 1T memory setting is very unstable, even though my memory officially supports it
- Onboard audio sucks.
- No hardware monitoring app comes with the mobo - and I don't think MBM supports this chipset.
- Max memory voltage is 2.7v
- The power LED header on the board is 2-pin, but every case I've ever had has 3 pin (1 blank in the middle). I can't figure out why they did this.

 

[Greenman]

I'm sorry to hear that. I had really hoped to use that board for my next upgrade. Since I flat out refuse to toss my AGP video card, it looks like I'm stuck going with an NF3 board. Or perhaps I'll just say forget it and live with my old AXP for another year.

 

[CFster]

I have an K8N nForce3 Ultra coming. I wanted the Platinum, but it seems that it isn't readily available anymore.

People say the MSI boards are buggy, and if you get a good one then you've got a great one. I've traditionally used ABIT and MSI over the years and have had very few problems - this ASRock was a departure for me.

In any event, I'll post some before and after benches (graphics, cpu, memory etc..) to see the difference, as everything else will be identical.

 

[designit]

NO no no,
You dont understand. the 27% you're getting is during the test. try your windows task manager and you will see cpu usage =0.
Those other problems you described is only in your case, I dont have any.
MBN works on this mobo
Good Raid performance
download drivers from ULi website and all are working perfectly. You might have drivers conflict if did not do a fresh install
1T has to w/ your mem settings and brand of mem, nothing to do w/ mobo. You cant overclock w/ cheapo mem and expect top performance. My patriot XBLK(TCCD) runs @ 300 mhz fsb 1:1
and got lower 32M Spi score than NF4 DFI UD using same ram, both were @ 2.7 ghz
TCCD only needs 2.6v up to pc4000. and 2.7 pc4200. anything above that needs 2.9v
a simple voltmod will take care of it. It is Ironic no one complains about voltmods on other boards that cost twice as much. All you have to do is solder a wire from vdimm to 3.2v plate next to it and you will have 3.2v. Even a child can do it.
OK so what power led header is 2 prong, it works.

 

[CFster]

Originally posted by: designit
NO no no,
You dont understand. the 27% you're getting is during the test. try your windows task manager and you will see cpu usage =0.

Why do I get 3% with a regular IDE drive?

Those other problems you described is only in your case, I dont have any.

You're one of the lucky ones. I've heard of problems with this board on multiple forums. IMO, in order to have both AGP and PCIe, and another CPU socket on this board they had to cut some corners and there are conflicts.

Good Raid performance

Not according to my benches.

download drivers from ULi website and all are working perfectly. You might have drivers conflict if did not do a fresh install

I did a fresh install about 6 freekin times.

1T has to w/ your mem settings and brand of mem, nothing to do w/ mobo. You cant overclock w/ cheapo mem and expect top performance. My patriot XBLK(TCCD) runs @ 300 mhz fsb 1:1

Why do you assume I have cheapo memory? I have very good memory, and achieved 1T on another board.

 

[Bad Dude]

Originally posted by: designit
NO no no,
You dont understand. the 27% you're getting is during the test. try your windows task manager and you will see cpu usage =0.
Those other problems you described is only in your case, I dont have any.
MBN works on this mobo
Good Raid performance
download drivers from ULi website and all are working perfectly. You might have drivers conflict if did not do a fresh install
1T has to w/ your mem settings and brand of mem, nothing to do w/ mobo. You cant overclock w/ cheapo mem and expect top performance. My patriot XBLK(TCCD) runs @ 300 mhz fsb 1:1
and got lower 32M Spi score than NF4 DFI UD using same ram, both were @ 2.7 ghz
TCCD only needs 2.6v up to pc4000. and 2.7 pc4200. anything above that needs 2.9v
a simple voltmod will take care of it. It is Ironic no one complains about voltmods on other boards that cost twice as much. All you have to do is solder a wire from vdimm to 3.2v plate next to it and you will have 3.2v. Even a child can do it.
OK so what power led header is 2 prong, it works.

Do you have any pics for the mods other than the ones on this forums?

Thanks.

 

[CFster]

Well I ran some benchmarks comparing the ASRock to a K8N Neo2 nForce 3.

All components, drivers (except board drivers) and OS is identical (XP Pro SP2). Nothing is overclocked except the video card. The ASRock drivers are the ULi's - the drivers from ASRock resulted in even worse performance.

And before you ask, yes I did a fresh install of XP, then the mobo/SATA drivers, then my video card drivers.

Athlon 64 3800 X2
G.Skill Extreme Series F1-3200PHU2-2GBZX PC3200 2GB(1GBx2) 2-3-2-5 Dual Channel
Leadtek A6600 GT TDH 6600GT 128MB GDDR3 549Mhz/1.10Ghz - 81.94 drivers
(2)Maxtor DiamondMax Plus 9 6Y080M0 80GB 7200 ATA150
(1)WD WD1200JB 120GB 7200 EIDE

I couldn't figure out how to get the text to space in the reply box - but the second row of numbers are for the MSI board...


ASRock 939Dual K8N Neo2

AquaMark3

57,943 65,260

Sandra:

CPU Arith Dhrystone 17939 18027
Whetstone 8093 8148

CPU Multi Integer 37839 37870
Floating 41024 41245

Memory RAM Int 3211 5136
RAM Float 3210 5081

File Drive Index 77MB/s 79MB/s


Everest:

Read 3856MB/s 5154MB/s
Write 1318MB/s 2001MB/s
Latency 84.3ns 59.6ns


HDTune:

RAID 0

Minimum xfer 61.7MB/s 61.6MB/s
Maximum xfer 88.0MB/s 99.9MB/s
Average xfer 78.9MB/s 84.7MB/s
Access Time 13.5ms 13.5ms
Burst Rate 88.9MB/s 95.9MB/s
CPU Usage 27.3% 5.1%

IDE

Minimum xfer 26.0MB/s 26.0MB/s
Maximum xfer 47.1MB/s 47.1MB/s
Average xfer 39.3MB/s 39.2MB/s
Access Time 14.1ms 14.3ms
Burst Rate 73.9MB/s 86.8MB/s
CPU Usage 2.9% 1.4%


The "old" nForce 3 board smoked the ASRock in just about every category. I was also able to run my memory at 1T on the MSI - something that would result in frequent crashes on the ASRock. This alone was reason enough to ditch the board - memory performance took a HUGE hit because of it.

If anybody doesn't hold stock in benchmarks, I can tell you that things "feel" like they run a lot smoother, and more stable with the MSI board.

I tried some overclocking with the ASRock, but if you go too far getting to it to post again is a PITA. The switching the power supply switch on and off several times trick doesn't work. I had to clear the CMOS jumper a couple times.

They've come out with some BIOS updates - features appear and disappear depending from subsequent updates - some things don't make sense.

You get half an instruction manual.

The onboard audio is horrible.

The power LED header on the board is 2-pin. The one in my Antec case is 3-pin. Most cases that I've had are 3-pin. Why would ASRock go out of spec like this.

The post screen went by so fast that I couldn't see what key to press to enter the BIOS. Since they included half an instruction manual (there was no section on the BIOS), I had to keep rebooting and pressing different keys until I found F2 (if I remember correctly) was the one to enter setup.

Also, the placement of the drive connectors is...unfortunate.

The ULi 1695 chipset was tempting because it allows both an AGP and PCIe video card, has a SATA2 controller on board, and future Socket M2 support. But for $65 I think they tried to cram too much into a budget board, and failed. I had it for two weeks and RMA's it.

I'd be interested in seeing if another manufacturer utilizes this chipset in a more upscale board. But, the question remains, why haven't any of them done so yet?