Drives Won't Play Nice Together!!

[Tomteriffic]

Or, How I Spent My 4th of July Weekend.....

OK, the short story:

I've been working on a machine that will be used for multitrack audio, and HDD sustained transfer rate is real important in this situation. My boot drive was a WD 400BB 2 meg cache job, while the files drive is a WD 400JB 8 meg cache (in a removable drive sled for quick swaps between projects)

Mobo is a Gigabyte GA-7DXE, which has a AMD 761 Northbridge and a VIA 686 southbridge.

The boot drive was running at an unusually slow transfer rate (12 megs/sec or so), while the files drive was humming along happily in the mid 40's. I decided to see if I could improve the boot drive's transfer rate, wondering if it might be the hotswap sled (it wasn't).

By the end of last night, my transfer rate was down to 2.4 megs/sec on BOTH drives, with the drives running in "real" mode, the motherboard appeared to have lost it's primary drive controller, and I was convinced that somebody up there did NOT want me messing with audio on a computer (this has been a six year running battle).

By this afternoon, a new mobo and second 8 meg cache hard drive, and I discover, by trial and error, that if you put an 8 mag cache WD drive and a 2 megger together on the same chain, this mobo just freaks out, and proceeds to put itself into survival mode and crawl along as far as the drives are concerned. The original mobo is fine, they both do the same thing.

I don't know if it's a drives issue, a mobo issue or a HDD controller drivers issue. It's zipping right along now with two 8 meg cache drives in it, but it sure has me wondering about whether WD or anybody has ever said anything about this.

BTW, the drivers are just the ones that came with the mobo, part VIA, part AMD 761.

I'd like to buy a clue, please.....

Which Windows??

Windows XP will downgrade your DMA mode after 6 CRC errors. It will drop from DMA5 to DMA4, right on down, unless your IDE drivers don't support it. Then it will go straight to PIO. Is all your cabling alright? Good, sound, 80 wire? (I know it is) Everything jumpered properly? (I'm sure it is) I know you had DMA enabled for both drives. Sometimes you need to toggle the DMA if available/PIO only drop down a couple of times to get Windows to use DMA.

You have the drive set up for IDE Block mode in BIOS? WD drives support it. Block mode will increase your throughput.

Mixed drivers, eh? Your soundcard gonna accept that?

[Tomteriffic]

Prexapes, good questions, all. I forgot to mention that, for now, the OS is Win98SE and the cabling and jumpering was straight.

Well, OK, I goofed on the jumpering during one of the many, many drive swaps

Reason for the Win98 is that my audio interface is one that's outta business and no drivers were ever written for any NT kernel OS'es (I've heard rumors of some Win2K betas, but never have found them.)

As far as the mixed drivers, that's pretty much standard-issue AMD 761 chpset stuff. The onboard audio codec is an AC97 thang on the VIA southbridge, which I'll disable and remove the drivers once I actually get the big dawg in and ready to go.

As an aside, every other audio machine I've built has had persistent mystery gremlins of one sort or another. I finally decided that what it was was the spot on the floor where the machine sat. It MUST have been a Vortex Of Evil. So this one's going in a rack about three feet away.

Still has me baffled as to why the primary drive controller would suddenly decide that it didn't have it's drivers or was otherwise SNAFU'ed, just because of two differing cache sizes on the drives. Couldn't find anything on the WD site about not mixing the two types.

http://support.microsoft.com/default...NoWebContent=1

I know, you don't need a MS KB article right now, but this kinda has me stumped. I'd probably point at the IDE driver for the board. Seems like there'd be no reason at all for the drives to act like that, especially considering they're the same manufacturer.

[Tomteriffic]

I'm gonna have to read that article while I'm at the machine in question, but, from what I can tell, the proper driver is/was loading when the machine was acting up. Gotta check a few other details, though, to see what's what.

Good find, man. I appreciate it!

[sabwafare2001]

If you want to use two hd on the same channel try puting them both on cable selet or try them with the jumpers off. I have a Maxtor and a WD800jb both 80gb with 8mb cache and they would work well with the jumpers on slave and master, so I put them on cable select and they woker fine until. But now I have no jumpers on them which I assume means the same as cable select. If you do this, with it set to cable select or no jumpers, make sure the end opposite of the cable, thats not connected to the mb, has the master drive on it and the middle your slave drive.

[Tomteriffic]

Well, I have both JB's in there humming along happily, but I just know that this will come up again, so I'll try the cable select thing. WD drives are very jumper-fussy, but that just might work.

[Tomteriffic]

Just to blow the cobwebs off an aging thread, I found the answer, courtesy of my local "cutting edge of obsolescence" computer shop.

The reason for the freakout, which can take a number of forms, is that the 2 meg cache memory on the WD 400BB (and most others) uses a 16 bit word length. The 8 meg cache on the JB (and most other drives with 8 megs of cache) uses a 32 bit word length for it's memory. Thus the freakout due to the incompatibility of the two cache types.

Seems like a pretty dumb oops on WD's part, but, since the prices of the two drive types are so close, I'll live with it.

See? Ya learn something new every day.....

[Peter M]

If you ask me, that's just technobabble to cover up a freaky firmware screwup. The drives don't even see each other's cache architecture let alone having to manage each other's. After all you can mix and match completely different drives, even different vendors', on one IDE cable. As long as they're about the same generation of IDE.

[Tomteriffic]

Originally posted by Peter M
......a freaky firmware screwup...

It certainly is that, Peter, at the very least. Doesn't make sense to me either. But, then again, I saw a new high-endy Maxtor and an older one doing exactly the same thing on an AOpen board with a VIA chipset last night.