Can one IDE have 2 hard disks capable of different speeds ?

[vinayag]

I have a older Seagate 80GB 7200RPM diak capable of ATA 100. I plan to get a
new Maxtor 80GB disk capable of ATA 133. If I put them both onto the same
IDE as primary and secondary devices, will they both run at ATA 100 or will
they run at indivijal speeds with no probems ?

Any inputs are appreciated.

 

[Broadkipa]

They will run at the speed of the slower drive as far as I know

 

[AndyHui]

All modern chipsets implement independent device timing for each IDE device.

The speed of one device on the cable will NOT affect the speed of the other.

 

[Pariah]

The speed of one device on the cable will NOT affect the speed of the other.

I thought this was true myself, but it isn't. I know Intel's boards will drop the IDE channel to the lowest UDMA speed and since I haven't researched it on other boards it may be the case with some of them as well. Check the faq's on the website of your MB manufacturer. In this case I wouldn't worry about it, dropping 133 to 100 isn't any big loss, certainly nothing you would notice.

 

[Lord Evermore]

Even the HX, VX and TX chipsets from Intel had independent device timing. The last Intel chip that didn't was the 430FX. One wonders how many years it will take before everyone will know about this.

whoa how come Intel can't get on the bandwagon? Aside from making it easier to deal with signal strength and integrity, why would they not have implemented independent device timing on their own boards? Given the much wider range of available device speeds now than back in the days of the HX chipset, this is a severe limitation of your upgrade possibilities.

 

[Pariah]

nm, I was just about to post that link.

I just checked the faq of another board, and Intel says it will drop to the lowest depending on the capabilities of the native OS IDE driver you are using. If you use Intel's Application Accelerator the drives will run at optimal speeds.

 

[Lord Evermore]

That's kinda frigging dumb. If the chipset supports them running at independent speeds, then the drivers should automatically do that. I can't believe Microsoft would include default drivers that won't do that, but only when using certain chipsets. Default MS drivers work fine at independent speeds on several non-Intel systems I have.

 

[Brian48]

I really don't think this is a problem so much for the chipset, but rather a limitation of the ATA standard that it forced to use. Here's another write up on the issue. I generally mix HDD and optical drives so I can do "on-the-fly" burning without having to make an image on the HDD first.

 

[Davegod]

I dunno about intels; but generally drives will be limited to the speed of the master on that cable, unless you have intependent timings in which case it makies no difference -it looks like the intel's dont have this, looking in here. In general it usually makes most sense to order drives, going fastest first, on PriMaster SecMaster SecSlave PriSlave - getting more complicated if you like onthefly burning. If you have a raid controller you basically have a tertiary and maybe quaterniary line, and are unlikely to have any problem anyway. Generally ensure primary drive is on the end connector and slave on the middle one, with jumpers set accordingly or to "Cable Select" (not older mobos).

edit: oops, thx Pariah

 

[Pariah]

Generally ensure primary drive is on the middle connector and slave on the end nearest it, with jumpers set accordingly or to "Cable Select" (not older mobos).

With 80 wire cables, the master always goes on the end, with the slave on the middle connector when using cable select.

 

[Lord Evermore]

Cable select is stupid since master/slave jumpers are easy to set, and you have to set a jumper for the drive to acknowledge cable select anyway, since most drives come configured as master (and with cable select, the middle connector is slave anyway), and most instruction manuals for ATA/66/100 drives say to put the master on the end connector, and 80-wire cables with labels are labelled that way. In practice it doesn't matter where you put the drives on the cable, do it however is easiest for you to arrange the cables in your case.

It is no longer the case that "generally" drives will be limited by the slowest device on the cable, nor will they be or have they ever been limited specifically by the master device. Only Intel boards seem to have any issues with not using independent timings, and only if you don't use the right drivers. I haven't had any boards in years that didn't run drives automatically at the proper speeds independent of the slowest devices, with chipsets from all the makers but nvidia and ATi.

As the Intel link shows, if you're going to end up having the devices limited by the slowest device on a single cable, it's best to put the two fastest drives on one cable, and two slower devices on a second cable. If you arrange them Primary Master ATA100, Secondary Master ATA100, Primary Slave ATA33, Secondary Slave ATA33, then every drive will run at ATA33. If you do manage to run the devices at their proper speeds with independent timings, then you can in some cases get better performance by having the two fastest drives as masters, and the slower drives as slaves, so that both the fast drives can operate at the same time (since two devices on one cable can't do that).

However it depends entirely on what you're doing. If you're only ever accessing one hard drive at a time, then you probably won't even notice any difference. If you're burning data to a CDR, then having the CDRW drive on a separate cable than the hard drive the data is coming from is a better arrangement, since the hard drive won't have to stop and wait for the slower CD burner to relinquish the bus.

 

[KF]

Originally posted by: Brian48
I really don't think this is a problem so much for the chipset, but rather a limitation of the ATA standard that it forced to use. Here's another write up on the issue. I generally mix HDD and optical drives so I can do "on-the-fly" burning without having to make an image on the HDD first.

From the relevent section of that write up:

next page
Independent Master/Slave Device Timing: Hard disk controllers on modern systems support running the master and slave device at different speeds, if one supports faster transfer modes than the other. Some systems, however, especially older ones, do not.

I didn't see them saying the ATA standard prevents independent timing. However there was something about not being able to send commands to another drive on the same cable until after completion on the first drive.

"Master/Slave Channel Sharing: By its very nature, each IDE/ATA channel can only deal with one request, to one device, at a time. You cannot even begin a second request, even to a different drive, until the first request is completed. " Maybe he knows; maybe he doesn't.
That would be extremely unfortunate if you can't do seeks simultaineously on two drives. Maybe they should fix that in one of those ATA specs they are forever revising, if that is the case.

In general the info content of the IDE FAQs there is low. Very low. It is mostly quibbling with the terminology the modes go by than anything else. What exactly is in these ATA standards, for instance?

As for the Intel's FAQ about their mobo not using independent timing, I don't see why it should be any more correct than the other FAQs everybody reads. The authors probably read the same FAQs as everbody else. It would depend on the BIOS Intel bought, wouldn't it, and I doubt these guys know any more about what's in there than anyone else. I know the modes are posted on the boot up screen for the (non-Intel) mobos I have. (Maybe they don't actually use those modes? You can also set modes in the BIOS. Maybe they don't use those modes either?) And Windows redoes everything anyway once it takes over.

 

[corkyg]

Originally posted by: Lord Evermore
Cable select is stupid since master/slave jumpers are easy to set, and you have to set a jumper for the drive to acknowledge cable select anyway, since most drives come configured as master (and with cable select, the middle connector is slave anyway), and most instruction manuals for ATA/66/100 drives say to put the master on the end connector, and 80-wire cables with labels are labelled that way.

Not stupid at all! If it were not for CS settings I would have to open up my machine and change jumpers every time I switched drives with mobile rack switches. Also - the Trios II drive selector uses CS as it electronically lets you switch between three drives.

 

[Lord Evermore]

KF: I would rather think an Intel FAQ about an Intel motherboard using Intel chipsets would be consider "authoritative" on the capabilities of the board. Other FAQs written on random websites might be suspect because they're tracking down information about things themselves, but one would expect a manufacturer to be able to provide accurate information about their own product.

The ATA standard has ALWAYS had the issue of only one device being active at a time on each channel, that's why there's so much discussion over what arrangement of drives in a system will yield best performance. That's nothing new, and it's not going to change (at least not for the foreseeable future). Being able to send commands to more than one drive at a time requires things that IDE never had, and SCSI does have, which is why SCSI still has some advantages over IDE despite the higher bandwidth of newer IDE specs, compared to "common" SCSI specs, and the serious price differences. If they start adding in features like that, it increases costs (they'd have to develop a way to integrate it into the ATA specs).

SerialATA is where the ATA spec is going, and it gets a few features similar to SCSI, but SerialATA by design only has one device per channel/cable.

 

[thomsbrain]

oh sweet jesus, don't plug in two different ATA drives! they will explode into blue flames and unleash a powerful trojan world wide internet mouse!

 

[Mday]

okay, it was true that with older controllers, this throttling to the lower spec occured. but with newer controllers, this is no longer an issue, though communications between the drives will be throttled (obviously).

 

[BigToque]

christ... who the fsck cares?!?

A drive that is ATA 133 capable is not gonna run any fscking faster that an ATA 100 drive!

 

[KF]

Since I brought it up when I quoted this from one FAQ:

>Master/Slave Channel Sharing: By its very nature, each IDE/ATA channel
>can only deal with one request, to one device, at a time. You cannot even
>begin a second request, even to a different drive, until the first request is completed. "

and it's wrong, I should correct it here.

As of 1998, 4 years ago:
ATA history
"ATA/ATAPI-4 is ANSI document number NCITS 317-1998.
...
There is a command overlapping and command queuing protocol for ATA and ATAPI devices.
..."

This is a direct contradiction. New commands do not have to wait for completion. You can "stack" commands. Evidently the inability to do so is not "By its very nature..." I'd like to know how they do this.

As some people pointed out, running an ATA133 drive at ATA100 is not likely to make any difference, but people who read these threads apply them to other cases. What about an ATA133 HD with an DMA33 CDROM?

 

[KF]

>...Other FAQs written on random websites might be suspect because
> they're tracking down information about things themselves, but one
> would expect a manufacturer to be able to provide accurate information
> about their own product.

Lord Evermore,

Would you say the same thing if you were reading a FAQ on a Chaintech site? I mean, assuming it was intelligible?

Or, suppose you called the IRS about some tax question. Presumably the US government, the supreme authority, ought to know what its laws mean, even if no accountant, lawyer, or judge is completely certain.

Let's consider the chain of command. There is some web site programmer putting this all there. It is done on the orders of some supervisor. Neither of these know diddley about motherboards or even electronics. If the FAQ writing was done in a professional manner, as it may well have been at Intel, then the actual words were composed by a technical writer, who may understand these things, but he has no specific knowledge of what the engineering was, or what the BIOS programming was. Did he get the information from the engineers or programmers? Does he even have access to these people? Maybe he has read a lot of FAQs, like everybody on this forum, and thinks he knows, and so does not have to spend hours tracking down the actual designers and engineers who did the specifics, if he can even find out who they were, and if he is permitted to spend as much time as he sees fit writing up the FAQ.

Here's an example from me. I have a couple of old motherboards which I have decided to max out on memory, since SDRAM is so cheap and the max is so low. So the manuals of these two completely disimilar mobos (Intel TX vs VIA MVP3) say they take maximum size 128M DIMM modules. So does the memory selector on the Crucial site. For the heck of it, I try 256M modules. (I didn't have a lot of 128s.) Both mobos work perfectly with 256M modules, recognizing and reporting every bit. Memtest says all is well. One mobo is maxxed at 256M, the other with 384M = 256M + 128M. Maybe later BIOS revisions put this capability in. (OTOH, the newest 128M I have is only recognized as 64M.) So much for the authorities. BTW, this not the first time -by far- in my experience that the authorities were not authoritative. Usually it works against me though.

 

[Pariah]

"Would you say the same thing if you were reading a FAQ on a Chaintech site?"

No, because Chaintech doesn't use a Chaintech chipset on their boards nor do they write the ATA drivers for the board.

"I have a couple of old motherboards .... Usually it works against me though."

This is very common with old boards and even with more recent boards. The reason they have the limit of 128MB is not a technical reason, but because at the time those boards were produced, 256MB DIMM's were basically non-existent and not available for validation. You can't claim compatibility for something that isn't actually available for testing, well you can, but you see what I am saying. The FAQ's on Intel's site are for currently produced boards.