HD Controller Question

[midwestfisherman]

I have a MSI K7N2 Delta-L mb. I currently have a Hitachi 7200 rpm HD on it. I am planning on adding an older 17 gb HD that I have to this box. This will be used to run linux and to store files. No big performance demands needed for thesecond HD.

My question is if I add the older HD (ATA 66 - 5400 rpm) to the same controller as the current HD (ULTRA ATA100 - 7200 rpm), setting the older HD as the slave, will the controller run both at the slower speed or will it recognize the speed of each drive and run them accordingly?

Thanks.

 

[tcsenter]

EIDE/ATA has had independent device timing since...well...a long time. Outside of some firmware conflict between the two drives or a BIOS flaw, each drive will operate at its respective mode to the extent it is supported in the chipset and BIOS.

It is wise to keep your two most utilized drives on different channels. Other than that, there are no performance issues with pairing drives that operate at different modes on modern systems.

 

[midwestfisherman]

Thanks for the reply.

 

[Peter]

There are still limitations in mixing drives. One shouldn't put drives w/ UDMA-66 or higher on the same cable as drives w/ UDMA-33 or lower, since that tends to go wrong electrically.

 

[DieHardware]

Is that just for HDD's or ODD's too?

 

[Peter]

Anything. The problem is that in master/slave setups, both devices are always "live" on the bus even when not talked to. This means that an old/slow device's bus interface circuitry might mess up the higher frequency signalling taking place when the machine is talking to the other, faster drive. I've seen it happen.

Some UDMA-66 or higher capable HDDs detect the situation at powerup, and fall back to reporting a maximum speed of UDMA-33 themselves, such that fast data transfers don't happen. Of course this is no good to performance at all.

 

[tcsenter]

Pairing an old first generation UDMA/33 or older drive with a much newer UDMA66/100/133 drive can give one drive or the other fits, but that would generally not be an inherent limitation of the ATA subsystem as opposed to an issue originating from one of the drives themselves being intolerant of the other.

I had a first generation Fujitsu UDMA/33 HDD that refused to permit any other paired drive to POST higher than PIO Mode 4, no matter which was the slave or master. I tested it in numerous configurations of slave/master, primary/secondary, and whenever that G-D Fujitsu was connected, the other drive on the same channel would not POST (or be initialized) higher than PIO-4. It otherwise worked fine when nothing was paired with it.

Assuming the hardware is not faulty or flawed, and the IDE cables meet ATA recommendations for the drive modes used (cables longer than 18" were not recommended for UDMA mode operation in previous revisions of the ATA standard), there *should* be no problem.

 

[Peter]

tcsenter, that'd be a drive firmware problem then. Fujitsu has the worst track record in that regard, at least in my book.

We've seen the actual electrical signal integrity problems I've been talking about above - both the symptoms (data loss while accessing the faster drive) and, in the lab, the absolutely screwed signal shape on the cable. This happens e.g. with a UDMA-100 HDD and an MWDMA2 CDROM on the same cable.
Recent Maxtor HDDs declare themselves UDMA-33 when paired with such a CDROM drive, which prevents bad things from happening, but also prevents performance from happening.

 

[tcsenter]

Some of these signaling errata caused by disparity in timings can be resolved by using a shorter IDE cable (18" instead of 24") and swapping connector positions on the cable (in addition to using an Ultra-ATA cable from a good vendor). One of the connectors on an IDE cable, either the end or middle connector, I forget which now, has tighter signaling margins than the other and this can be exploited to compensate for the disparity in timing between two drives.

I discussed this at some length with Quantum's Kent Pryor and Eric Kvamme (both major contributors to T13) a few years ago now. Their take on this issue was that in the majority of cases they have seen these problems rear their ugly head, given all relevant components of the system are billed Ultra-ATA compliant, it is usually traced to a vendor's design that results in unforgiving electrical or signaling properties.

Both were quite adamant that, although things do get tight on the upper end of UDMA, Quantum has proven there is adequate margins to accomodate for different drive modes from UDMA-5 (highest at the time of our discussion) down to MWDMA-2, if only every interested party made compatibility as much of a priority as others.

It does not help that so many ATAPI drive vendors refused for so long to adopt the UDMA/33 interface over MWDMA-2, since their drives didn't require more than 16.6MB/sec. Even today, there are a few ATAPI vendors who limit their interface to MWDMA-2.

But you are correct that issues can arise and some vendors have simply resorted to negotiating a safer mode through their driver or firmware when certain criteria are present such as a drive that reports no higher than MWDMA-2. Of course, all bets are off when attaching ancient PIO mode devices.

 

[Peter]

You are telling me no news. In the slow+fast drive combination, we've seen signal integrity go bad on 4" cables in the lab. We have tracked it down to poorly designed bus attachment circuitry on those slow drives, and since switching vendors, it hasn't come back.

It happens with cheapo CDROM/DVD/burner drives, and particularly often with FlashROM drives. The vendor's excuse usually being exactly what you said - "our drive isn't fast enough to need fast interface support, so why use better (and more expensive) interface circuitry than what OUR drive needs?" Because else your drive <BEEP>s up the data transmissions that go to the other drive, you ignorant <BEEP>! Duh!