mockingbird wrote:

Western Digital drives had a DOS utility to permanently (That is, it is retained after a power cycle) set the maximum UDMA rate... I don't know if this is true for their last IDE drives, but this was at least the case for their IDE drives of the late 90's and early 00's.

Do you happen to have saved a copy? Or a name of such a tool?

I found my ALI card so I thought I'd post a picture of it:

(Click on attachment for bigger picture)

Mau1wurf1977 wrote:

mockingbird wrote:

Western Digital drives had a DOS utility to permanently (That is, it is retained after a power cycle) set the maximum UDMA rate... I don't know if this is true for their last IDE drives, but this was at least the case for their IDE drives of the late 90's and early 00's.

Do you happen to have saved a copy? Or a name of such a tool?

The file is called DLGUDMA.EXE

IIRC, there were different versions of this file that worked for different drives.

It also occurs to me that if you want to limit the drive to UDMA2, you can just use a 40-conductor ribbon cable instead of an 80-conductor cable.

That's the odd thing, I tried a 40 pin cable but during POST the drive still shows up as UDMA 5.

But maybe I should test this in DOS?

Absolutely. DLGUDMA should help you then.

I recently got an ALi M1489 / M1487 board running. I used the ALi driver that you provided for Windows 95c and ran some HDD benchmark tests before and after the installation of the driver. The results are confusing. Did you acquire any benchmarks before and after?

Sandra 99
before driver = 4417
after driver = 2375
[larger score is better]

Winbench96 HDD Winmark
before driver = 994
after driver = 925
[larger score is better]

WinTune98
before driver = 11 MB/s cached
before driver = 1.6 MB/s uncached
after driver = 11 MB/s cached
after driver = 3.3 MB/s uncached

For the most part, the results were the same or worse than before using the ALi driver. I am using an Am5x86-133 with a 133x CF card. I generally felt that CF card access was slow before and after the ALi driver, however I am used to using UDMA100 or Ultra2 SCSI 80 on my 486's. The perception of slow could also be because I am used to hearing some sort of access sounds from the HDD, with the lack of sounds implying to my mind that something has "stalled". Anyway, this doesn't change my feeling of HDD access being the same before and after using the driver.

I wonder if anyone else has attempted this?

These are my results in Windows 98 SE with aliide driver and with the standard one, and a Maxtor 82560A4 hard drive:

These are my results in Windows 98 SE with aliide driver and with the standard one, and a Quantum Fireball 1280AT:

What have you concluded from these results?

Absolutely nothing yet. I want to make some real world tests but I am thinking of how to design the experiment so that the results are confident.

Edited: By the way, there is no DMA tickbox in the disk drives with the aliide driver (neither with the standard driver).

Interesting. On my UMC 8881-based system, when using a LS120 disk drive, there is a DMA tick box, but after I tick it and reboot, the tick is removed.

I have timed file transfers between both hard disks using a 123MB file. I have defragged the hard disks first, and I have only taken one time per session restarting the computer after each test.

With aliide driver:
- From Maxtor Partition 1 To Maxtor Partition 2 (primary master): 123 s
- From Maxtor Partition 1 (primary master) To Quantum (secondary master): 71 s
- From Quantum (secondary master) To Maxtor Partition 1 (primary master): 80 s
- From Maxtor Partition 1 (primary master) To Quantum (primary slave): 75 s
- From Quantum (primary slave) To Maxtor Partition 1 (primary master): 86 s

With standard driver:
- From Maxtor Partition 1 To Maxtor Partition 2 (primary master): 117 s
- From Maxtor Partition 1 (primary master) To Quantum (secondary master): 64 s
- From Quantum (secondary master) To Maxtor Partition 1 (primary master): 67 s
- From Maxtor Partition 1 (primary master) To Quantum (primary slave): 63 s
- From Quantum (primary slave) To Maxtor Partition 1 (primary master): 66 s

From these results, the standard driver appears faster at sequential read/writes. But from your previous benchmarks though, the ALi driver was faster at writes and slower at reads. Are you able to setup a RAM drive and re-run the timed tests with the hard drive writing to the RAM drive and the RAM drive writing to the hard drive?

feipoa wrote:

From these results, the standard driver appears faster at sequential read/writes. But from your previous benchmarks though, the ALi driver was faster at writes and slower at reads. Are you able to setup a RAM drive and re-run the timed tests with the hard drive writing to the RAM drive and the RAM drive writing to the hard drive?

I have created a 33 MB RAM drive by inserting this line in the autoexec.bat file:

xmsdsk.exe 33792 R: /t /y

I have copied a 32 MB file...

With aliide driver:
- From Maxtor to RAM drive: 20 s
- From RAM drive to Maxtor: 20 s
- From Quantum to RAM drive: 26 s
- From RAM drive to Quantum: 22 s

With standard driver:
- From Maxtor to RAM drive: 17 s
- From RAM drive to Maxtor: 17 s
- From Quantum to RAM drive: 22 s
- From RAM drive to Quantum: 21 s

Huh. Anyone have any idea why the sustained transfers are slower when using the ALi DMA driver?