Cant mix ATA-66 with slower devices.

4 replies [Last post]
pcy
Offline
Joined: Jul 6 1999

As far as I can establish, you cannot put an ATA-66 disk and any non ATA-66 device on the same EDIE controller.

Has anybody else noticed this effect?

It dosn't actually crash, but it does stop the system from working properly.

I have an Athlon 500 on a SD-11 motherboard. If you put (for instance a CD-ROM) on the same EDIDE contoller as an ATA-66 disk, and then measure the disk performance, two nasty things happen:
1. The disk speed drops by 65%
2. The CPU utilization goes up from 1 or 2% to 99%

Ouch.

As far as I can make out the lack of grounding for the extra 40 screening wires on the non ATA-66 device means it makes noise that interferes with the ATA-66 drivers. If so, I would expect this to happen on all chipsets/motherboards and lead to data corruption.

Nasty.

Peter

Peter

bcrabtree
Offline
Joined: Mar 7 1999

I think you are jumping to conclusions, possibly false ones.

First - it has long been the case that if you put a CD-ROM drive on the same cable as an ATA33 hard disk, the hard disk would slow down if the CD-ROM drive were an older model that did not support DMA.

This will be unchanged with ATA66 hard disks. So, if you've got an old CD-ROM, either connect it on a cable that doesn't have any fast HDDs or, if that is not possible, replace it with a DMA-capable CD-ROM drive or a SCSI CD-ROM drive.

If I put a real ATA33 drive as the slave to an ATA66 drive that has been set up to report itself as ATA33, then both drives can be run in DMA mode and both will run fast - the ATA66, typically, will be only a percent or so slower than if it were run at ATA66 on an ATA66-capable motherboard.

The genuine ATA33 drive does not cause data corruption problems with the ATA66 drive running in ATA33 mode.

I think the problem is that the CD-ROM is old, and non-DMA capable.

Bob C

bcrabtree
Offline
Joined: Mar 7 1999

Timing test - message sent at about 13:34

bcrabtree
Offline
Joined: Mar 7 1999

Timing test - message sent at about 13:37

pcy
Offline
Joined: Jul 6 1999

Bob,

I hear what you say, and I will go back and re-run all my tests in consequence. However:

1. The CD-ROM is a 6 week old A-Open x48, and DMA certainly WAS enabled. Indeed I ran the test with DMA enabled, disabled, and with the CD-ROM connected to the EIDE cable but no power. All the tests gave identical results.

2. The increase in CPU utilization from 2% to 99% is not like anything I have ever seen before. The old problem you refer to of CD-ROMs slowing down hard disks never produced an this scale of effect in my experience.

3. At one point I produced this problem with an ATA-66 disk and an ATA-33 disk on the same controller.

So maybe I had a setting wrong when I ran the tests, or maybe it only affects VIA chipsets, or maybe... which is why I posted the question.

Peter

Peter