Jump to content
Compatible Support Forums
Sign in to follow this  
slaveOne

RAID Performance Problems

Recommended Posts

I've been experiencing extremely low performance with my on-board HighPoint HPT372 RAID controller. Below are my system specs:

 

Windows XP Professional (non-SP1)

Abit KD7-RAID

AMD Athlon XP 2600+ (333)

512MB PC2100 DDR RAM

nVidia GeForce3 Ti200

 

Master on IDE 3 (RAID): Maxtor 20GB HDD (92049U6) ATA/66

Slave on IDE 3 (RAID): Western Digital 120GB HDD (WD1200JB) ATA/100

Master on IDE 4 (RAID): Western Digital 40GB HDD (WD400BB) ATA/100

 

I have tested each HDD on PC Mark 2002, and have achieved an average score of 400 on the HDD benchmark. I have the latest VIA 4in1 drivers installed, the latest HPT372 drivers, and my BIOS is flashed to the most current revision (CY). I am using the IDE cables that came with the motherboard. That's about all I can think of that would pertain to this problem. When I connected the drives to IDE 1 and 2, they were much faster, but I did not benchmark them in that situation. Though the speed difference was definitely noticable.

 

Any help on this matter is greatly appreciated. Thanks.

Share this post


Link to post

Didn't mention what RAID mode, but none the less, the problem lay in the hard drives themselves. Mixing a 20 GB with a 120 GB with a 40 GB with different brands and different ATA capabilities is really a "no-no" when you do RAIDs of any kind in my opinion.

 

If you want performance, try to stay not so much brands, but try to stay within the same HD drive limits (2 x 120 GB drives for example)

Share this post


Link to post

Well, I'm really just trying to use them as extra IDE channels (not striping or mirroring). So if I were to connect a single drive (say, the 120GB drive) to the RAID controller, I would get much higher performance?

Share this post


Link to post

Depends on the capabilities. It would defintely not bog the resources as much and you would most likely get a better performance. But if you are using them as extra IDE channels, you have to remember it comes at a price considering those "extra" channels were meant to be used as RAID devices.

 

Good luck to you.

Share this post


Link to post

Well, thanks for the info jdulmage. Any idea if I can connect ATAPI devices to this RAID controller? I'm guessing the anwer is no, but please humor me =]

Share this post


Link to post

look is there is a jumper to switch the controller from raid to ata function.

Share this post


Link to post
Quote:
look is there is a jumper to switch the controller from raid to ata function.


No.. it's an on-board controller. It acts as an ATA controller if you don't set up any RAID arrays. From my motherboard manual: "KD7-RAID's built-in HighPoint HPT 372 chipset gives you the capability to support Ultra DMA 133. It provides two IDE channels (IDE3 & IDE4) that also support Ultra DMA 133 specifications, and it allows for four additional IDE devices in your computer system."

Share this post


Link to post

First, PCMark itself is not good enough for benchmarking harddisk (actually a crap benchmark for everything). Use HDTach for best results.

 

The 400 mark provided by pcmark doesn't indicate where the problem originates.

 

Run the benchmark on all drives, and take a note on what the beginning/end speed is (for the graph above), and the burst speed. Please report the scores that you're getting and i could tell u what's causing the slow speed.

 

What device do you have on the VIA IDE controller?

 

Remember only one device can be active on a channel at a given moment. So from the HD info you provided, it's best to remove the maxtor drive, since it's the slowest drive that'll bog down the channel that it's on.

 

A very important thing that you need to do is to make sure both highpoint driver and bios version is the same, otherwise it'll easily cause problems.

 

Just in case the highpoint controller BIOS imbedded inside the abit bios files is always outdated, even with newest revisions. So what you need to do is find a tool called CBROM, which could be used for up[censored] the highpoint bios imbedded inside system bios file.

 

If you can't find that tool on the net, i can send u the file and the instruction of how to use it.

Share this post


Link to post
Quote:
Well, thanks for the info jdulmage. Any idea if I can connect ATAPI devices to this RAID controller? I'm guessing the anwer is no, but please humor me =]


I can not humor you in that way laugh

Share this post


Link to post

Hey Newumbrella, I am at work right now so I cannot benchmark my drives with HDTach for another 5 hours or so. I have a burner and a dvd drive attached to IDE 1 and 2. Actually, I did not benchmark the 120GB as I thought I did, and I tried it last night and got 800 in PCMark 2002 on it. So, the other hard drives are probably just a bit slower. Although, like you said the Maxtor is probably slowing down the WD 40GB drive because it is on the same channel.

 

The thing that confuses me is that I used to get much better performance when I had the same drives on the main IDE channels (1 and 2). I never benchmarked them with that set up, but it was noticably faster. I would think that the HPT372, acting as an ATA controller, would offer better performance.

 

Also, I am using the HPT 2.32 driver because the BIOS only has the HPT372 controller flashed to 2.32. So... no problems there. I found that tool, CBROM, but I'm a little afraid to use it. If you could give me instructions on how to properly update the Abit BIOS image to contain the new HPT372 2.34 BIOS, it would be much appreciated.

 

I'll try HDTach when I get home and let you know how that goes. Thanks for the help.

Share this post


Link to post

Hey AlecStaar, thanks for the input. I am planning on buying another WD1200JB (120GB) and ditching the smaller two drives (or I'll just put them on my Promise ATA/133 PCI controller), so that should speed things up. If I were to do that, should I run the two 120GB drives separate or should I stripe them? Which would be faster?

 

Edit: on a side note, earlier I mentioned that my WD 40GB (ATA100) was probably going slow because it was paired with the Maxtor 20GB (ATA66), but I just realized that the WD 120GB is paired with the Maxtor! So, does that mean I will get better performance on the WD 120GB if I switched it with the WD 40GB (so that it's on its own IDE channel)?

Share this post


Link to post

Well, I ran HDTach on all three of my drives (in the same configuration), and I got some pretty interesting results.

 

Maxtor 20GB: http://www.zbattle.net/maxtor.gif

WD 40GB: http://www.zbattle.net/wd400bb.gif

WD 120GB: http://www.zbattle.net/wd1200jb.gif

 

It appears that the WD 120GB got some pretty decent speeds, though it lagged quite a bit giving it a really low minimum result. I suspect that this is because this drive is paired with the slow Maxtor.

 

Something that really bothered me in these tests was the CPU Utilization. Out of all three drives it was an average of 53%, which is insanely high. Though, I suspect this is because I am using HPT372 BIOS version 2.32, which I hear is the culprit for this behaviour.

 

Newumbrella mentioned that I could update my HPT372 BIOS to version 2.34 by embedding the update into the Abit BIOS image, but I don't really know how this is done (except that you need to use CBROM). If anyone knows how to properly perform this patch, please e-mail me or reply to this thread.

 

Quote:
* Now, I don't know if the HPT-372 is more "modern" & can address disks of diff. rpm, &/or ATA-66/100 rating or not, on SAME CABLE... but that's ANOTHER ISSUE TO CONSIDER!

 

I.E.-> Since I heard more modern EIDE controllers can now do this & address disks @ their TRUE speed even if on same cable (& are indeed, diff. speed drives in either ATA-66/100 rating, OR IDE vs. EIDE OR even 5400rpm vs. 7200rpm ratings)!

 

This I only heard recently in fact here @ this site on these forums only days ago...

 

Truthfully/QUITE HONESTLY, hearing that was a SURPRISE to me in fact, that a weakness of single-threaded I/O controllers in the IDE/EIDE world have overcome this weakness fairly recently in more modern controllers for them! It's good news, & seemed to be supported well by those that spoke of it to me, & they had documentation & support as well.

 

Plus, were I looking @ it programmatically? It wouldn't be impossible to tag each channel's devices in their datastreams with a (on EIDE channel #1) "I am 100 ATA & hit me @ this speed" & on EIDE channel # 2 if need be, & it is slower "I am 66 ATA & hit me @ this speed ONLY"...

 

(Now, as far as MY HPT-370 RAID controller circuit here? Like yourself, I did not see it doing that here, being able to address diff. speed disks on the same cables on it @ their TRUE max speeds, but instead saw them operating only as fast as what seemed like slowest devices on their cables!)

 

Now.. & from the sounds of it there, the same issue exists still for you that the SLOWEST DISKS on a controller chain on these HighPoint RAID controllers lag you... & the fastest disks do NOT seem to be addressed @ their TRUE MAX SPEEDS!)

 

From what I can tell the HPT372 does allow a drive to operate independently of the drive it is sharing a channel with, but it does cause some lag that ultimately reduces the overall performance of the drive. However, this could be due to the fact that I am using HPT372 BIOS version 2.32.

 

Quote:
* What I want to know though, on your end is, did you use EIDE cables with 80 pins on them, OR did you use std, 50 pin IDE cables that you said came with your mobo... again, this DOES matter!

 

I am quite sure I am using 80 pin cables, but I'll try to find out for sure and get back to you. Is there an easy way to tell? I noticed that I have some older IDE cables and its individual wires looks much bigger than the ones for the IDE cables I'm using now (there are a lot more individual wires on my new ones). So I would think that if there are only two types of IDE cables, I am using the 80-pin versions.

Share this post


Link to post

From the HDTach results,

- Burst Speed - normal for all drive

- Seq. read speed - normal for Maxtor 20GB and WD 120GB, low for WD 40GB

- CPU Utilization - it's kinda too high....(take note that highpoint/promise controller's CPU Utilization will usually be higher then chipset IDE controller's figure)

 

I have a KA7-100 mb with HPT-370, and different bios/driver combination will sometimes cause problems with low seq. read speed, low burst speed(interface speed), and/or high cpu utilization.

 

It's usually preferred to set a higher performance device as master and slower device as slave.

 

Before doing anything could u run hdtach with the following configuration:

1. Primary Master - WD 40GB

Primary Slave - Maxtor 20GB

 

2. Primary Master - WD 40GB

Primary Slave - empty (disconnect maxtor drive)

 

This step is to verify:

- If maxtor drive is slowing WD 40GB down

- If there's any compatibility problems

 

Some possible solutions for high cpu utilization:

1. install VIA Raid patch - this will disable pci read caching, which improves pci performance

 

2. install latest VIA 4-in-1 driver

 

3. optimize bios settings

 

4. update highpoint driver/bios

 

Here's the instruction for using cbrom

 

1. copy bios image and highpoint bios file to cbrom dir

 

2. run dos prompt, change dir to cbrom

 

3. type: "cbrom xxxxx.bin(mb bios) /d, this will display the individual components of the bios

 

4. note that highpoint bios is a pci device

 

5. type: "cbrom xxxxx.bin(mb bios) /pci xxxxx.xxx(highpoint bios image), this will update highpoint bios component

 

6. type: "cbrom xxxxx.bin(mb bios) /d, check to make sure the bios is updated

 

7. u can update with the this bios image file (the file name will be going to be the same) and u can delete the temp files created(bios.rom, original.tmp)

Share this post


Link to post

HD Tach says my CPU usage high but my bus mastering works fine. I have a feeling it doesn't read the CPU usage properly for some controllers or configurations.

Share this post


Link to post

Bus mastering on all new boards will very likely to work properly (except if there's any compaibility with the BIOS/hardware). And you can try to optimize BIOS settings first, cuz it could rather make some difference on burst speed and cpu usage.

 

Cuz on my old KA7-100 board i have to use WPCREDIT and WPCRSET from H-oda.com to set some of the chipset options. Those options are simply not available in the BIOS (it's very conservative). Before optimization, the burst speed and cpu usage is 63mb/s and 50% respectively. After optimization and highpoint bios and driver update, it's 83mb/s and 15-20%.....and with a lower cpu usage and higher burst speed, that system is definitely "snappier" then before.

 

Comparing to older board, newer board are much "optimized". For example if i use WPCREDIT to check my KT3-ultra, those setting that will speed up the machine is all enabled. The default setting is much more aggressive. So it's unlikely that you have to use those two programs to tune the board, just changing the setting in the bios would be ok. Just make sure BIOS/driver is the latest and optimize BIOS and you sohuld be ok.

Share this post


Link to post

Well, I just bought another WD1200JB (120GB) HDD. I'm going to stripe them together and see what kind of performance I can get. Also, I followed your instructions (Newumbrella) on up[censored] my HPT372 BIOS, and it worked perfectly. Now my CPU Utilization on HD Tach is approx. 10%. Thank you very much for that info. I'll report back my findings of the 2xWD1200JB array to this thread.

Share this post


Link to post

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
Sign in to follow this  

×