Settle a Cable Select Debate

[djnes]

Okay, forget whether it's useful or not. I personal never do because I don't see the point, when I can just as easily set the hard drives as a master and a slave.

Okay, here's the question that needs an answer. Another member, who will remain nameless to avoid embarassment, says to me that if your using an older 40 pin IDE cable...ATA33 and down...that you cannot use cable select for hard drives. Now I find this to be absolutely wrong, as I've used cable select back in those days to get to misbehaving HDDs from two different brands to work together. So tell me honestly...can you use an ATA33 40 pin cable and set your drives to cable select? I've done it...but that's not good enough for this guy. Please post if it's possible, and if you've done it.

 

[Tigerbiten]

You can but...........
Master is the middle drive and Slave is the end one.

Its the reason Cable Select got a bad name as the drives are reversed.
Ps. I don't use it.

Luck........

 

[djnes]

That's what I cared about...that it was possible. I didn't realize the drives were reversed. Add another reason why I hate it. I've always had 2 identical drives in my systems back in those days....never noticed that it switched them.

 

[Dark Ember]

Yeah, definitely possible. Whats interesting about CS switching master/slave positions from normal is this: back in those days, the master drive wasn't required to be on the end of the cable to work when setting manually. So maybe that has something to do with it switching them?

 

[TLS2000]

Again, I stated that a standard 40 wire cable that you picked up from the computer store for $.50, didn't support it. I stated that for cable select to work on a 40 wire cable, you needed a cable that supported it. I never said that 40 wire cables did not support cable select.

Please at least use what I said instead of opening the debate by paraphrasing what I said incorrectly. This makes it look like I was wrong, when I wasn't.

Again: NOT ALL 40 WIRE CABLES SUPPORTED CABLE SELECT, ALTHOUGH SOME DID!

 

[phatrabt]

I never use(d) it. To me it was always easier to set the drives manually. It's just a matter of opinion...

 

[Doctor X]

I found out with my drives, if i set them master/slave. It would be slower than cable select. I verified this with benchmarks before. What I cannot figure or find out is why.... I was able to reproduce this with any drive that was ata66 or later... but I noticed my maxtors being really pronounced. Any one have comments on this? When I bought my samsungs a few weeks ago, I tried the same thing and it is there also. I just wish I knew why. Could it depend on the controller?

Doctor X

 

[tdg]

The cable itself has nothing to do with supporting or not supporting cable select, it's up to the BIOS and the hard drives for it to work. The only difference between ATA33 and ATA66+ cables is that the ATA33's have 40 pins/40 conductors, and ATA66+ have 40 pins/80 conductors, there is no other logical difference in them that would have any effect on cable select.

Most cable select issues came from WD hard drives and their hatred for being set on cable select, master or slave instead of master stand-alone when used in a single drive setup.

 

[TLS2000]

The difference is that wire on pin 28 isn't hooked up to the middle connector of an Ultra ATA cable. A lot of older ATA cables had it hooked up through both connectors, which would make both drives try to act as master. This was especially true with cheap cables.

edit: My last post on the subject. There's no point in arguing with you any more. I have my point of view, you have yours. Let's leave it at that.

 

[Ice Czar]

The only difference between ATA33 and ATA66+ cables is that the ATA33's have 40 pins/40 conductors, and ATA66+ have 40 pins/80 conductors,

more specifically
40 pins/40 data wires, interspersed with 40 ground wires to cutdown cross-talk

and of course that Cable Select is different between the two
(if its an ATA33 CS capable cable)
an 80 wire (ATA66) cable has the Master at the end of the cable with a Cable Select jumpered drive
....

Configuration Using Cable Select @ PC Guide

An alternative to the standard master/slave jumpering system used in the vast majority of PCs is the use of the cable select system. As the name implies, with this system the cable--or more correctly, which connector on the cable a device is attached to--determines which device is master and which is slave. The goal of cable select is to eliminate having to set master and slave jumpers, allowing simpler configuration.

To use cable select, both devices on the channel are set to the "cable select" (CS) setting, usually by a special jumper. Then, a special cable is used. This cable is very similar in most respects to the regular IDE/ATA cable, except for the CSEL signal. CSEL is carried on wire #28 of the standard IDE/ATA cable, and is grounded at the host's connector (the one that attaches to the motherboard or controller). On a cable select cable, one of the connectors (the "master connector") has pin #28 connected through to the cable, but the other (the "slave connector") has an open circuit on that pin (no connection). When both drives on the channel are set cable select, here's what happens:

* Master: The device that is attached to the "master connector" sees the CSEL signal as grounded, because its connector has pin #28 attached to the cable, and the host's connector has that signal grounded. Seeing the "zero value" (grounded), the device sets itself to operate as master (device 0).
* Slave: The drive that is attached to the "slave connector" does not see the CSEL signal as grounded, because its connector is not attached to the CSEL signal on the cable. Seeing this "no connection", the device configures itself as a slave (device 1).

If you switch the devices between the two connectors, they swap configuration, the master becoming the slave and vice-versa. Not a very complicated arrangement, and a good idea, it would seem. In fact, if cable select had actually caught on, it would have been great. The problem is that it has never been widely used, and this lack of universality has made cable select unattractive, which is a bit of a chicken and egg situation. Since cable select was never accepted in the industry, most drives come, by default, with the drive jumpered as a master or single drive. This means that to enable cable select, you have to change a jumper anyway, which obviously negates some of the advantage.

But the biggest reason why cable select never caught on was the cable itself. From the very beginning, all 40-conductor IDE/ATA cables should have been made so that they would work with cable select. There's actually no need to have different cable types, because if you set a drive to "master" or "slave" explicitly, it just ignores the CSEL setting. So a cable select cable can be used either way: regular jumpering or cable select.

Unfortunately, regular 40-conductor IDE/ATA cables don't support cable select. (Why this came about I do not know, but I suspect that some bean counter determined they could save five cents on each PC by doing this.) So to use cable select you need a special cable, and these are of course non-standard, making them a special purchase. Also, many people don't understand cable select, nor do they realize it needs a special cable. If you set both drives to "CS" and then use them on a regular (non-cable-select) IDE cable, both drives will configure themselves as "master", causing a configuration conflict.

Making matters worse, the 40-conductor IDE/ATA cable select cables have the "master connector" as the middle device and the "slave connector" as the device at the end of the cable, farthest from the host. For signaling reasons, it's best to put a single drive at the end of a cable, not put it in the middle leaving a "stub" of wire hanging off the end of the channel. But if you do this, that single drive sets itself as a slave with no master, a technically illegal configuration. Worse, suppose you do this, and your hard disk sets itself as a slave, and the system boots from it without problem, as most would. Then, you decide to add a new hard disk. You set it to cable select and attach it to the middle connector. The new drive then becomes the master, and thus moves ahead of the old drive in precedence! The system will try to boot from it instead of your old drive (which some people might want, but many do not.)

To get around this problem, a second type of 40-wire cable select cable was created, the so-called "Y-shaped" cable. On this one, the connector to the system is in the middle, and the slave and master connectors are on the two opposite ends of the cable. This certainly makes things less confusing, but has its own difficulties. For starters, IDE/ATA cables are very limited in length, which means this "Y-shaped" cable was hard to use in large tower systems. All your drives had to be mounted very close to the motherboard or controller card so the cable would reach. And again, the cable was a special item.

As you can see, the traditional way of doing cable select was a total mess, which was why it was never widely adopted. The key reason for this mess was--once again--lack of standardization. I rather expected cable select to eventually wither away. However, when the 80-conductor Ultra DMA cable was introduced, the cable select feature was much improved, changing the potential of this feature. The two key changes were:

* Drive Position: Unlike the old cables, with the 80-conductor cable, the master connector is at the end of the cable, and the slave is in the middle. As I explained above, this is a much more sensible arrangement, since a single drive placed at the end of the cable will be a master, and a second drive added in the middle a slave.
* Universality: All 80-conductor IDE/ATA cables support cable select (or at least, all of the ones that are built to meet the ATA standards). This means there's no confusion over what cables support the feature, and no need for strange "Y-cables" and other non-standard solutions.

These two changes mean a world of difference for the future of cable select. Since these cables will eventually completely replace all of the 40-conductor cables, all systems will be capable of running cable select without any special hardware being needed. As I mentioned before, you can still explicitly set drives to master or slave if you want to, and the CSEL signal will be ignored by the drives. So the bottom line is that these cables work either way, cable select or not. What will finally make cable select catch on? If drive manufacturers and systems integrators widely agree to use it, and the manufacturers start shipping drives with the "CS" jumpers on by default. We'll have to see if this happens.

Warning: 80-conductor IDE/ATA cables are often said to be compatible with 40-conductor cables. That's true of normal 40-conductor cables with drives jumpered as master and slave, but not cable select cables. If you swap a regular (non-"Y-shaped") 40-conductor cable select cable with an 80-conductor IDE cable, the master and slave drives will swap logical positions. If you don't that to happen, you'll need to change the order that the devices connect to the cable.

that should settle the debate
and educate a few members to boot

 

[Doctor X]

sup ICE Czar? Thanks for the info!

I have always wondered why I have better performance with cable select than master/slave. After reading what you posted, I am wondering if the controller I used made a difference. (i have run highpoint for years now). Or could it depend on the cable? I am leaning towards the cable now.

Anyone else verify that this is true for them also?

l8r!
Doctor X

 

[Ice Czar]

I am leaning towards the cable now.

Corruption 101 > Cables Section
specifically > ATA not so Frequently Asked Questions
Or: Why Ribbon Cables are unsuitable for RF transmission of data
(but the whole cable section has even more info)

cable, routing, controller, drives, HBA if applicable, the synergy of all those

which is why when you buy an external array,
an EE has gone through the multiple interfaces to balance and cleanup the signal quality
anything that employs CRC is prone to performance loss from mangled signals
its the overhead of resending the data (and more so for ATA which is unterminated )

which also explains why you can drop a small fortune on SCSI cables
for ATA cables, there are now a few supershielded makes available,
http://www.ioss.com.tw/web/English/RD3XPGladiator.html
http://www.ioss.com.tw/web/English/RD3Gladiator/TestReport.html
and while enthusiast reviews are often less than stringently scientific
they have shown promising and largely unexpected performance advantages
http://reviews.pimprig.com/miscellaneous/rd3xp_gladiator_ata133_cables.php

as to whether the average user would be able to percieve those benefits is an entirely different matter
most access patterns wouldnt require that

however corrrpution is never an issue till it actually happens to you