Talk:Parallel ATA/Archives/2013/November

ATA is a Command Set not a Transport Protocol

This page appears to be significantly misnamed.

ATA is the name of the command set. ATA is an analog of SCSI. Like SCSI's Parallel SCSI, SAS and iSCSI, ATA can be transmitted over a varity of physical protocols including SATA, PATA and AoE. In fact the current proposed standard ACS-2[1] does not include the transport protocols. And the abstract (second page) makes no mention of them.

A more logical layout would be to split PATA into its own page, leaving the ATA page to discuss the command set and contrast with SCSI.

ATA standards versions, transfer rates, and features

The table says on ATA/ATAPI-7 (ATA-7, Ultra ATA/133) that there is SATA/150. But where is SATA/300 ? It is not listed. -- Frap 16:03, 01 October 2006 (UTC)

It's certainly in the Serial ATA article. I'm not sure why SATA/150 is in the table in this article at all. Jeh (talk) 09:36, 22 June 2008 (UTC)

AT/ATAPI doesn't describe physical interface anymore - it could be Parallel ATA, Serial ATA or eSATA (and FireWire or USB for that matter), and any ATA/ATAPI drive will still respond to commands defined by the ATA/ATAPI standard (what's more, Serial ATA can also be used to interconnect SCSI devices as in Serial Attached SCSI). --Dmitry _{(talk •contibs )} 21:13, 5 October 2006 (UTC)

But there is a standard describing the physical connection. As a practical matter a "parallel ATA device" still follows the physical connection standards described in ATA/ATAPI 6. Jeh (talk) 09:36, 22 June 2008 (UTC)

Suggested fix for picture of cables

On my monitor I can't see any difference between the "80 conductor cable" and the "40 conductor cable" (second picture), except that the connectors are colored. I think we should find a picture that better shows the different conductor spacing. All of you others, get right on that. ;) Seriously, I'll see if I can't take one myself. Jeh 09:19, 21 October 2006 (UTC)

Yeah that pic is horrible, note to editors black objects do not photograph well. Plugwash (talk) 21:09, 22 October 2008 (UTC)

Apparent error in the "cable select" section

I think this passage is confusing or inaccurate:

"With the 40-wire cable it was very common to implement cable select by simply cutting the pin 28 wire between the two device connectors. This puts the slave device at the end of the cable, and the master on the "middle" connector. This arrangement eventually was standardized in later versions of the specification. If there is just one device on the cable, this results in an unused "stub" of cable. This is undesirable, both for physical convenience and electrical reasons: The stub causes signal reflections, particularly at higher transfer rates."

I would've fixed it myself, but I'm not sure I understand it. If early users were hacking ribbon cables to have an open pin 28, that would only work by cutting between the second and third connection. This would put the master to the middle connector, as the passage says. It says this arrangement was standardized in later versions, but all the ATA ribbon cables I've seen put the black master connector at the end, and the gray slave in the middle. I think what the passage means to say is that these early hacks had the opposite configuration of modern cables, which just leave gray pin 28 with no wire contact. Someone who understands what this paragraph is trying to say should probably fix it so someone doesn't put their cables on backward. --Loqi T. 02:51, 7 June 2007 (UTC)

I read it as saying

• The most common way of implementing cable select prior to the 80 conductor era (though cable select back then was bloody rare anyway) was to cut one wire between the two device connectors (that is between the master device and the slave device). This put the slave connector in the middle which was undesirable.
• This was standardised in some later version of the ATA spec (not having read the spec I can neither confirm nor deny this), it would also be usefull to know which version.

The next paragraph then goes on to say that this was changed with the introduction of 80 conductor cables which do indeed put the master device at the end.

Personally I have never seen a 40 wire cable that supported cable select.

I've seen a few that indeed had a "hole" punched out of the pin 28 wire between the second and third connetions, just like used to be done on floppy cables. But not many. Jeh (talk) 13:22, 7 July 2008 (UTC)

Never mind "seeing", I HAVE some like that... And plenty of pre-ATA-6 (and so, pre-80-conductor-cable) hard drives and motherboards that claim to support the standard also. 193.63.174.211 (talk) 18:02, 10 May 2012 (UTC)

windows limitations

according to http://discountechnology.com/Seagate-160GB-IDE-ATA-100-Hard-Drive:

• windows 2000 up to sp3 have a limitation of 137 GB
• windows xp up to sp1 have a limitation of 137 GB

so we could add it asomewhere
GNUtoo 12:50, 13 July 2007 (UTC)

Found some further info at seagate d17 sata product manual google cache in section 3.8.1:
W2000 Sp3 + XP Sp1 needs "Big Drive Enabler"
Maxtor Knowledge Base Answer ID 960
MS KB 303013
Electron9 23:59, 25 July 2007 (UTC)

Size limits - Win98 64GB

Good article, but too bad it does not list all the size boundaries. This article seems excellent, mentions the Win98 64GB boundary: [2] -69.87.199.99 19:53, 18 September 2007 (UTC)

I'm not sure that restrictions imposed by operating systems belong in an article about the ATA interface and standards, since the ATA interface and standards have nothing to do with these restrictions. Jeh (talk) 22:26, 22 June 2008 (UTC)

Maybe we need a page summarising the common PC storage size limits and what part of the system (hardware, bios, OS etc) imposed them. Then linking to the appropriate articles for further details. Plugwash (talk) 22:42, 23 June 2008 (UTC)

I never knew Win98 had a 64GB limit... in fact, I came onto the talk page to discuss/dispute the alleged 128GB one, as I've successfully used a 250GB drive with it (albeit partitioned into two equal halves...) 193.63.174.211 (talk) 16:45, 10 May 2012 (UTC)

Clue: 125 < 128. 86.166.70.84 (talk) 12:02, 23 September 2012 (UTC)

...Pardon? Sorry, if you're referring to how I split it into two 125GB (actually 116GiB) parts, then yes, well done, you figured out my method. However, my confusion is over how the Dickens this actually worked. After all, what's effectively happening is that Windows is able to read beyond 128GiB... so long as that data isn't used to, say, denote the start of a partition. Partition (and FAT, maybe)? Oh nononono. Two-gigabyte video file? Why of course! ... But data is data, surely, so how come it can handle one but not the other? Also, to reiterate: WHAT "64GB" limit?! 193.63.174.211 (talk) 11:51, 1 August 2013 (UTC)

comparing IDE/ATA speed with others

The speed presented in this article is in MB/s, generally accepted as the abbreviation of Megabyte per second. In the SATA and USB aritcles the speed is presented in Mbits/s (Megabits per second). Could someone check the speed of ATA in Mbits/s please? it is possible it is alredy in megabits but somone did not write it properly. --Iamcon (talk) 07:01, 22 January 2008 (UTC)

The speeds quoted here are the correct numbers for megabytes/second (where mega = 1,000,000). For example, UDMA 5 runs at 100,000,000 bytes/second. I hope this is unambiguous enough.

The reason SATA and USB quote in bits per second is that they are serial protocols: one bit is sent at a time. Parallel ATA is not like that; it sends 16 data bits at a time. In each case these are the "natural" units, according to the respective technologies. They are also the same units and numbers quoted in industry standard specifications and sales information for these interfaces.

You can't just take the PATA number and multiply by eight to get a number comparable with serial protocols, either. SATA uses an encoding involving 10 bits on the wire for eight bits on the disk. The "1.5 Gbit/sec" figure for the original SATA is the bit rate on the wire, not on the disk, and translates to 120 megabytes/second of actual data read or written. USB uses a "bit-stuffing" protocol in which the ratio is not even constant. Jeh (talk) 07:32, 22 January 2008 (UTC)

150 (decimal) megabytes/sec, shurely? (1500 / 10 = 150, not 120) ... Hence the secondary designations sometimes seen, of SATA-I as "150", SATA-II as "300" and SATA-III as "600"... 193.63.174.211 (talk) 11:52, 1 August 2013 (UTC)

Bits may be transferred by different means, it still boils down to effective transmission capacity per second. And associated latency. It's much harder to make comparisions when the same thing is noted in different units. Electron9 (talk) 15:02, 15 February 2008 (UTC)

The point I'm making is that "effective transmission capacity per second" is, at least, more accurate for PATA's "133 MB/sec", than for SATA's "1.5 Gb/sec": All issues of latency, inter-block delays, etc., aside, it is possible that there could be periods of time during which PATA would be transferring 133 MB/sec of end user data. This can't be said of SATA's "1.5 Gb/sec". Yes, switching units makes things even more confusing, but it would also be confusing to cite specs for these buses using units other than those commonly quoted. Adding corrections for e.g. 10 physical-layer bits to 8 transport-layer bits would simply compound the confusion, and would likely result in frequent "corrections" by new wiki editors who didn't stop to read the explanations.

I do think it would be worthwhile to "rationalize" all of these specs so that useful comparisons could be made (perhaps in a separate article pointed to by the PATA, SATA, USB, etc., articles), but this should be done in addition to, not instead of, cites of the "official" numbers using the usual units for each. Also it should be done with a LOT of explanation. Jeh (talk) 00:59, 16 February 2008 (UTC)

I suggest then to add a seperate value "effective transfer rate". Electron9 (talk) 07:12, 16 February 2008 (UTC)

apparent conflict between article page and other reference

From this article:

In Maximum security mode, you cannot unlock the disk! The only way to get the disk back to a usable state is to issue the SECURITY ERASE PREPARE command, immediately followed by SECURITY ERASE UNIT. The SECURITY ERASE UNIT command requires the Master password and will completely erase all data on the disk. The operation is rather slow, expect half an hour or more for big disks. (Word 89 in the IDENTIFY response indicates how long the operation will take.) [3] [4]

From the article in c't:

When setting his or her password the user can choose between the security levels "High" and "Maximum." When the level "High" is chosen the disk will accept either the user or the master password to unlock the disk or disable the protection function. When "Maximum" is the choice only the user password will provide access to the data. Should it get lost then the administrator with his or her master password will only be able to unlock the disk after forfeiting all the data stored upon it. Which step is accomplished by the command Security Erase: It erases all the information by writing zeros onto all sectors of the hard disk before again allowing access to it.

These seem contradictory.

Ealex292 (talk) 02:44, 28 April 2008 (UTC)

request for clarification on maximum security mode

OK...so this means once this mode is entered, the only way to use this other than a heater and doorstop is to erase the media and start again? I don't know the specification...do you mean instead "if the password is provided incorrectly too many times?

(comment text added by Rchandra to the article on 24 May 2008. Copied here by Jeh (talk) 19:15, 27 June 2008 (UTC)

Request seconded. Is maximum security mode a "brick mode" intended to prevent any future access of the data by any person whatsoever? No unlocking possible except for the purpose of erasure? Or am I misunderstanding this paragraph? The wording is not very clear to say the least. Note that I'm only talking about what's officially the case, without taking vendor supplied back-doors into account. -- 77.7.188.166 (talk) 10:17, 12 March 2011 (UTC)

According to the Rockbox reference, maximum security mode is a "brick mode." I'm hesitant to include that in the article, though, without a proper look at the spec. --Joe Sewell (talk) 19:55, 11 October 2011 (UTC)

Place Word DMA as stub

I propose we should integrated Word DMA as a stub article to Advanced Technology Attachment, anyone agrees? WDMA (computer) --Ramu50 (talk) 02:23, 7 July 2008 (UTC)

You mean merge it with this article? Maybe, but then the section containing that info should also explain the ATA PIO modes (merged from the article on PIO, which really needs to talk much more about PIO in general (eg to serial and parallel ports), right now it is too specific to ATA) and also the UDMA modes. Then you will have doubled the size of this article. Maybe instead, put all three of those into a separate article on "AT Attachment transfer modes". Otherwise I think the existing articles on PIO and WDMA should stay as they are, and an article on UDMA added. That can be an easy next step. Adding an article is always easier than merging. Jeh (talk) 13:09, 7 July 2008 (UTC)

"Technical criticisms" section

Upon comment by Ramu50 I agree, the "technical criticisms" section from the article counts as WP:OR as per WP:CRITICISM. I have moved it here until someone finds reliable sources that state these points as criticisms. Quoting the spec is not a source for a criticism as the conclusion that the spec is describing a problem is that of the WP editor. A blog is not even a reliable source and the blog entry cited does not support the contention that the ATA specification is at fault. It does criticize the default drive behavior of "write cache enabled" and the refusal of some OSs to allow this to be overridden, but neither is a problem of the ATA spec, nor specific to ATA drives. Jeh (talk) 03:51, 17 July 2008 (UTC)

Technical criticisms

Criticisms of current versions

The write cache of ATA disk drives is enabled by default to increase performance. If a power failure occurs before data in the write cache has been flushed to the disk, data will be lost. There is a "flush cache" command in ATA protocol, which will write the entire cache to medium before returning. However, the protocol does not allow a way to inquire a drive if a particular sector has been written or not.

Criticisms of obsolete versions

ATA1 (section 9.22), ATA2, ATA3 specifies 'Set Features' command allows the user to enable it if the drive shall 'Enable write cache,' but does not specify the command to flush the cache.

ATA4 (section 8.10) specifies 'Flush Cache,' however should an error occur while executing the command, the disk will return the failing sector address and the sector is removed from cache. An alternative way to initiate flush cache does exist from ATA4 (section 8.37.9) and onwards: "When the subcommand disable write cache is issued, the device shall initiate the sequence to flush cache to non-volatile memory before command (see 8.10)."

Page moved

I moved the page, to agree with content.Bettering the Wiki (talk) 05:13, 17 September 2008 (UTC)

Thank you! The page was originally "AT Attachment", renamed to "Advanced Technology Attachment" by someone who did not realize that that is not the correct name. Since the "AT Attachment" page was now a redirect with more than one edit in its history this change could not be simply reverted. (I also tried going through "requested moves", arguing for "speedy revert of undiscussed move", to no avail.) So, uh... how did you do it? Thanks, regardless! Jeh (talk) 08:51, 19 September 2008 (UTC)

Hot-pluggability

To the editor who keeps putting "hot-pluggable: yes, with software" back into the article:

Sure, you can put a filter driver on top of just about any device to make it "hot pluggable at the software level." However the article describes ATA as described in the ATA standards docs. The specifications describe, among other things, the hardware interface - the electrical characteristics of the signals and the mechanical characteristics of the connectors. And these aspects of ATA simply do not permit ATA devices to be safely hot-plugged.

A hot-pluggable device interface has many well known characteristics. Typically the contacts are staggered and the connector designed so that upon connection ground mates first, then power, then the data signals. ATA is not like that and as a result you stand a good chance of electrically damaging either the device or the host interface if you try it. PATA does not even have power and signals on the same connector (except in the laptop 44 pin version), or on mechanically linked connectors as far as the cable side is concerned, so cannot make any guarantees about make or break sequencing.

There is nothing that any software can do about this.

Yes, there are of course "mobile racks" and similar things that work around this, mostly by mechanically requiring that the drive be powered off during insertion or removal. But they are not covered or allowed for by the ATA spec; in fact, they violate it, as the ATA spec does not allow for any "intermediate" connectors between the host controller and the device. The fact that these devices sometimes do work (and, in my direct experience, sometimes not) is beside the point; the article is referenced to the PATA specs.

I hope this clarifies the issues involved. Jeh (talk) 08:57, 19 September 2008 (UTC)

Laptop 44 pin ATA connector

There should he pinout and picture of laptop 2.5" 44 pin connector —Preceding unsigned comment added by Valent (talk • contribs) 13:53, 10 October 2008 (UTC)

Afaict (from looking at an adaptor) the connector is a 2mm pitch with 43 (a 2x22 grid with a pin missing as a key) pins 39 of which are the same as the standard ATA connector and the remaining pins carrying power. Plugwash (talk) 01:39, 20 October 2008 (UTC)

I'd like to see a pinout picture added similar to this one: http://www.unitechelectronics.com/ide44pinout.gif which really helped me to visualize and understand the pinouts between the 3.5" 40 pin connector and 2.5" 44 pin connector. Furthermore, it helped me to see where pin 1 is actually located on the hard drive itself. Kpr00 (talk) 19:32, 1 August 2009 (UTC)

Something about the 1.8in connecters - 50pin, zif and lif would be nice too. HughesJohn (talk) 11:22, 25 November 2009 (UTC)

The 50-pin connector mentioned above (usually used for laptop optical drives) is, I believe called a JAE connector, for Japan Aviation Electronics. Possibly the KX15-50K3DE connector? — Preceding unsigned comment added by 209.143.195.65 (talk) 21:34, 2 November 2011 (UTC)

Merger proposal

I am proposing that the ARMD-HDD page be merged here. ARMD-HHD is essentially an orphan page, very few things link to it, it is most definitely a subtopic of "AT Attachment", and requires the background info from this page to understand. Jeh (talk) 00:10, 16 December 2008 (UTC)

Go ahead.--Anss123 (talk) 09:03, 16 December 2008 (UTC)

This page is getting quite long. It might be suitable to keep them seperate. Btw, is ARM-HDD PATA or SATA? Electron9 (talk) 00:59, 17 December 2008 (UTC)

It long predates SATA. Whether it also now applies to SATA, I don't know. I imagine that SATA covers booting, etc., from these devices "natively" rather than with an add-on spec. Jeh (talk) 02:29, 17 December 2008 (UTC)

Hearing no other objections I went ahead and merged. Page length is still well within WP guidelines. Jeh (talk) 05:43, 28 December 2008 (UTC)

Eliminating slavery!

Following Mmj's (Talk) worthy edits to the article my attention was drawn to this sentence in the Master and slave clarification section:

Both are technically equal slaves to the driver in the host OS.

IIRC I wrote this sentence, other than the word "equal" and the wikilink, but now that I read it, it grates. When I write a device driver I usually don't think of the subject device as a "slave."

It gets the point across but... is "slaves" really the best word here? Would it be reasonable to say "Both are under control of"? Considering that the preceding sentences have this point covered, do we really need this sentence at all? Am I worrying too much about minutaie? Jeh (talk) 08:12, 23 December 2008 (UTC)

If you're happy, I would support removing that sentence entirely. It doesn't add much more that the previous sentence didn't already explain (the drivers in the host operating system perform the necessary arbitration and serialization, and each drive's controller operates independently.) and, as you rightly point out, using the term "slaves" again is misleading so it's a pretty ineffective analogy. mmj (talk) 01:01, 24 December 2008 (UTC)

... I've gone ahead and removed it as it looks like we're in agreement, though feel free to modify and reinstate it. mmj (talk) 01:05, 24 December 2008 (UTC)

We're in agreement. Thanks! Jeh (talk) 05:58, 24 December 2008 (UTC)

Cheers for the clarification of the relationship between device 0 and 1. But as to the politicizing of the terms master/slave in control applications, get over it. That terminology refers to the relationship of electro-mechanical devices, not the impressed servitude of Messenians in Sparta, Africans in the southern states 150 years ago, or child "carpet slaves" in India today. —Preceding unsigned comment added by 192.101.136.224 (talk) 19:07, 1 July 2010 (UTC)

Significant figures

Article says:

Parallel ATA only allows cable lengths up to 457.20 mm (18.00 in)

Does anyone truly believe that increasing the cable length to 457.21 mm would cause communication failure? Be realistic! —DIV (128.250.247.158 (talk) 05:48, 14 April 2009 (UTC))

While what you say is undoubtedly true, the T13's standard does give that exact number, hence a 457.21mm cable would be out of spec. Feel free to reword it if you want. -- KelleyCook (talk) 19:57, 14 April 2009 (UTC)

Cable lengths up to 457.21 mm are guaranteed to work. Longer cable setups may fail! This is related to reflections, impedance mismatches, etc.. P-ATA electrical interface is also hodgepodge so it's wise to adhere to these limits. Electron9 (talk) 11:03, 23 May 2009 (UTC)

I guess the figure comes from it being 1.5 ft rather than a metric size? Sounds like the problems of metric wood sized when people needed a 4x2 and had to say it in mm. --Gibnews (talk)

"457.20 mm" is false precision. The specification was originally written in inch units and later converted from inches to millimeters. The dimension should be shown as "18 in (457 mm)" in the article. —Quicksilver^{T @} 20:18, 14 January 2010 (UTC)

It's not false precision, the specification states "18.00 inches". -- KelleyCook (talk) 20:51, 14 January 2010 (UTC)

How about "needless", then? Or did the original author really mean that anything up to 18.004999(rec.) inches would be fine, but 18.005 inches and above out of the question, presuming you could accurately measure such lengths? More likely they were just working to a particular prescribed number of sig figs. I suppose if you wanted to be really pedantic you could imagine a corner-case where someone who was told "a maximum of 18 inches" rather than "18.0" or "18.00" would then make an 18.49" cable that was technically within-spec but only-just failed to work (as it's almost a full 12.7mm too long), but most people involved in the production of such cables should hopefully realise that if you told them 18 inches (or, if you like, 18.0 inches), then it would be wisest not to be a smart-alec and just manufacture them in lengths no longer than 457mm, preferably rounded-down a little-finger nail's width to 450mm (45cm). Or even 457.2mm, which is beyond the apparent precision of the average bulk cable cutting tool, crimper and solderer anyway. There's a reason even the 2.5" interface still uses 2mm pin spacings...

And besides all that, running 457.2mm to a further sig fig of accuracy then makes your converted output figure more than 25x as precise as the input. Surely 2.5x would be sufficient? You could spec it as 457.1~457.3mm and still be as (still needlessly) accurate, as it puts your range within the 457.073 ~ 457.327mm implied by 18.00 inches... 193.63.174.211 (talk) 16:59, 10 May 2012 (UTC)

Parallel ATA - why the redundant "parallel" describer?

Whilst amateur enthusiasts may refer to any "non-SATA, ATA interfaces" as "PATA" (whatever that is) and "parallel ATA", shouldn't the article use correct terms? Thus, I propose that reference to "PATA" is replaced with ATA, and "parallel" only used as a comparitor where ATA is being compared with SATA. No need to invent a term that doesn't exist when there is a perfectly useful and valid term already (i.e. ATA). —Preceding unsigned comment added by 81.107.183.201 (talk) 23:15, 7 May 2009 (UTC)

I'm just going to go ahead and move the page. Also there is clearly a confliction with already the third sentence stating there is no such term and further on in the article again. --Nestea Zen (talk) 08:03, 13 May 2009 (UTC)

Ok, so after having ran across Retronym - citing no source - and then checked t13.org I decided to undo the change for now. They are using the term in nearly every document.--Nestea Zen (talk) 08:44, 13 May 2009 (UTC)

It's been "renamed for clarity", similar to how various other devices, systems, etc have been in the past. The ISA bus was never originally or officially called that, either in the PC, XT or AT days, but whoops, there's an article about it, as it's the term that was coined for the connector when more modern competitors came along. I'm no fan of it either, but it does at least make it clear what type you're talking about when it matters... 193.63.174.211 (talk) 17:05, 10 May 2012 (UTC)

Meaning of IDE

In this article IDE is defined as 'Integrated Drive Electronics" however the original acronym from Western Digital was "Intelligent Drive Electronics" This is missleading. --Gibnews (talk) 18:09, 29 May 2009 (UTC)

Everyone knows it as "Integrated..." so I think a source would be needed for this. Jeh (talk) 19:17, 6 October 2010 (UTC)

Need to block IP 119.111.65.80

This guy is continuosly vandalizing this page. Is there some administrator who can do something? —Preceding unsigned comment added by Michele.alessandrini (talk • contribs) 07:09, 15 July 2009 (UTC)

UDMA should be new article

I think there should be new article about UDMA, because SATA uses UDMA too. Also comparing of different UDMA levels 0-5 should be described. —Preceding unsigned comment added by 62.16.33.43 (talk) 04:12, 16 October 2009 (UTC)

Also UDMA redirected here when I was looking for information on UDMA Compact Flash cards. 86.176.156.189 (talk) 16:42, 10 April 2010 (UTC)

XT Attachment?

Wasn't there originally also a sibling technology called XT Attachment, which used an 8-bit interface for easy bridging to the old XT Bus interface, the same way that ATA is essentially a 16-bit ISA slot in disguise? If I'm not mistaken here, it should be mentioned, at least in passing. -- 85.179.162.155 (talk) 04:19, 28 June 2010 (UTC)

Don't think so. I worked on genuine IBM PC-XTs. Hard drives on the XT were interfaced through ST506-compatible things like the WD 1003, each different from all the rest, so there was not a single "XT attachment" standard. Jeh (talk) 19:15, 6 October 2010 (UTC)

Well I was talking not so much about real IBM XTs but more about the late-1980s el-cheapo XT clones that lasted into the early 1990s, until Windows swept them away. Did they use ST-506-like hard drives with separate controllers until the end? -- 77.187.134.125 (talk) 17:02, 11 March 2011 (UTC)

For example http://www.cknow.com/cms/ckinfo/xta---xt-attachment.html says it was a "Rarely used implementation of IDE with an integrated 8 bit XT controller". -- 77.187.134.125 (talk) 17:04, 11 March 2011 (UTC)

Interesting. A little more Googling turned up a bit more info here: http://datarecoveryspecialist.com/glossaryofterms.htm - from a company that appears to be in a position to know about such things. This also appears to be a timeline, and it implies that XTA was developed after ATA, before ATA2. Jeh (talk) 22:00, 11 March 2011 (UTC)

I do remember XT-bus drives being used in Commodore A590 sidecars and de:Schneider Euro PC HDD add-ons, both using the same WD 93028-X drives. (We sold both types of drives back then.) The 40-pin connector looks the same as PATA but pinout isn't compatible. I guess the technology used is very similar and the software side looks like ST-506 as well, but don't know for sure. Zac67 (talk) 20:46, 11 October 2011 (UTC)

PERFECT confusion.

May I quote from the table:

ATA/ATAPI-7

ATA-7, Ultra ATA/133

UDMA 6 aka UDMA/133 SATA/150

SATA 1.0, Streaming feature set, long logical/physical sector feature set for non-packet devices

NCITS 397-2005 (vol 1) NCITS 397-2005 (vol 2) NCITS 397-2005 (vol 3)

"SATA/150"? "SATA 1.0"? I thought until now that this is a Parallel ATA article. Am I the only one here to think that the SATA stuff in the columns to the right is VERY confusing for people that want to get themselves informed about this stuff yet unknown/obscure for them? -andy 77.7.111.144 (talk) 11:39, 4 September 2010 (UTC)

The table shows the evolution of the standard. ATA/ATAPI-7 does include UDMA 6 for Parallel ATA, but it is also significant that that version of the standard includes Serial ATA. Just because the article is about Parallel ATA does not mean that related standards can't be mentioned. Jeh (talk) 19:22, 6 October 2010 (UTC)

okay, that makes sense. thank you. 174.56.57.138 (talk) 13:58, 24 August 2012 (UTC)

Channel Listings aka Master/Slave

Jeh (talk | contribs) m (47,912 bytes) (Undid revision 392226961 by Lostinlodos (talk) rv. They're not the official terms according to the specs.

My post lifted directly from Jeh talk page

RE Parallel ATA Maybe we're not hitting on the same use of 'official'. I was referring to the understanding that 'master', and 'slave', are printed on all major-manufacturer drives since ATA r2. I was looking at the history (where IDE is commonly confused to be ATA) as that few people would immediately recognise the channels without the more generic printed-on-device terms. Would you consider leaving both terms for the benefit of further understanding. Lostinlodos (talk) 12:21, 25 October 2010 (UTC)

Out of either ignorance or deference to popular usage, there are already 30 occurrences of "master" and 23 of "slave" in this article, most of them in the section on cabling and connectors. The two sentences in question here are (in a no doubt futile effort to set the record straight) noting the actual terms as reflected in the spec. The specs do not say "Device 0 (master)" or "Device 1 (slave)". They simply say "device 0" and "device 1", and have done so since ATA-2; even ATA-1 notes that "master" and "slave" are "industry terms." The sentence as it now stands furthermore says "device 0 and device 1, respectively," the "respectively" referring to the occurrence of "master" and "slave" in the sentence immediately preceding. This should be completely sufficient to establish which is which. It is unnecessary, redundant, and more generative of confusion to state "The two devices are correctly referred to as device 0 (master) and device 1 (slave), respectively." Jeh (talk) 20:15, 25 October 2010 (UTC)

Point taken.Lostinlodos (talk) 20:22, 25 October 2010 (UTC)

btw, back in ATA-1 there was provision for spindle sync. This used to be important for RAID configurations. If the spindle sync feature was used, device 0 was the source of the spindle sync signal (hence the "master") and device 1's rotation was (wait for it) "slaved" to device 0's. Spindle sync was an optional feature, never implemented on any actual drives that I know of, and gone from the spec by ATA-3 at the latest. But this was the only way in which one device was ever truly a "master" over the other that I've ever been able to find. I conclude that this is the true origin of the terms "master" and "slave" here... but that conclusion is of course OR. I'll try to find a non-OR way of expressing this in the article. Jeh (talk) 05:25, 17 November 2010 (UTC)

Hmm. I can only add to this my own experience of setting up various machines from the early 90s onwards. Until the point that BIOSes (and disks) became clever enough that they would report their name, size and serial number as an identifier that could be used to e.g. choose the boot device, you had to set a lot of things manually with knowledge of the physical layout, including the head/sector/cylinder counts in the very earliest ones. The main disk, on the end connector of the cable, was always - always! - called the master, and was the one which the system would boot from. The "second" disk, on the middle connector, was always referred to as the slave, and was meant as additional storage only, especially in the earlier systems where booting from it even via a "master/slave swap" (similar to "floppy A/B swap") simply wasn't allowed. Later BIOSes eventually allowed you to choose the boot device (primary IDE channel master, primary slave, secondary master, secondary slave...), then choose to boot from a non-floppy removable, then did away with the terms altogether and just referred to them by name. I'm not sure if I ever saw any use of the "Device 0" and "Device 1" terms.

It's all a case of oversensitivity anyway. Might as well demand that people start calling floppy disks "flexible disks" (for 8") / "flexible mini-diskettes" (for 5.25") / "micro-diskettes" (for 3.5" and smaller). Once a term enters common enough use, and starts hitting the dictionaries under these definitions (OK, I haven't actually checked that one, but I'd put a small stakes bet on it), can't we use it to make things clear and understandable? The cases of people installing a solo disk as "slave" or "device 1", and or on the secondary rather than primary IDE channel (also terms which you could describe as misleading, rather than channel 0 and channel 1) and using it as their main boot and storage drive are pretty minimal. Boot disk is primary master. CD drive is secondary slave. Any additional devices fill the other slots. Or you could say 0/0, 1/1, and confuse everyone like you would calling every road by number instead of using at least a few names... It's not like we're literally saying that the one disk owns the other and forces it to work for free... 193.63.174.211 (talk) 17:20, 10 May 2012 (UTC)

Suggestion: section on cables

This article alludes to two basic types of cables: 1) a 40 conductor version with uniformly colored connectors (all black) and 2) an 80 conductor version with three differently colored connectors. It would be useful to have a table in the "Parallel ATA interface" section showing 1) the associated controller standards for each cable and 2) how each cable handles the pin 20/28/34 options. I find the verbal description hard to follow. For example, the section does not address how pins 20/28/34 are wired on the 40 conductor cable. Also, there seems to be a contradiction with which controller standard was the first to use the 80 pin cable; did it appear with Ultra DMA/33 or UDMA/66, as shown in the subsequent features table?. 3dimen (talk) 07:42, 2 November 2010 (UTC)

Suggestion: backward compatibility

In the table "Features introduced with each ATA revision" could a column be added showing backward compatibility? For a given hard drive, which of the previous system revisions can it be used in? In other words, can a ATA/ATAPI-6 drive (which normally uses an 80 conductor cable) be used in a ATA/ATAPI-4 system with a 40 conductor cable? The drive would obviously not perform at rated speed, but would it function? I think this information would be useful. 3dimen (talk) 07:42, 2 November 2010 (UTC)

"medium" vs "media"

In the sentence

For example, any removable media device needs a "media eject" command, and a way for the host to determine whether the medium is present,[...]

I believe the word "medium" is actually correct. "Medium" is the singular of "media", and since only one medium can be present at a time, "media" would not be correct. See e.g. http://grammar.about.com/od/alightersideofwriting/a/mediagloss.htm

Of course, strictly speaking, it should also be "removable medium device" (just like we say "cassette deck", not "cassettes deck"), and "medium eject", but here "media" is used as a collective noun (like "any medium"), and at any rate these two terms have become established as phrases.

The correction "media" -> "medium" was made, then reverted (incorrectly, I believe) on January 9th, 2011. I'd like to come to a consensus here before changing it again. Sebastian (talk) 22:29, 11 January 2011 (UTC)

The ATAPI command set calls it "media." "Media change requested" and "media changed" status bits, "ATAPI Removable Media Device", etc. More, it is extremely rare in the industry to heard the term "medium" instead of "media" in this or similar contexts. Maybe it's a reflection of the fact that even though it's packaged in discrete lumps like disks or tapes, the actual media is uncountable quantities of magnetic domains or dye spots? Whatever the reason for the usage, I think the "principle of least astonishment" applies - the industry (not excepting the ATA/ATAPI standard) doesn't use the term "medium," nor should anyone come away from this article thinking that we do. Jeh (talk) 23:04, 11 January 2011 (UTC)

Difference between old DMA and UDMA?

What's the difference between WDMA and UDMA signaling apart from the speed? Or is it ONLY the speed? -- 77.7.188.166 (talk) 10:28, 12 March 2011 (UTC)

From looking at the spec table, I would presume the bit width increased with each generation, and possibly UDMA either double-pumped it's "bus" (the cable) - probably with a switch to "edge" rather than "level" signalling - or co-opted some of the conductors that wouldn't otherwise have normally carried data in order to do so, given that it's the third-gen standard and the second one already doubled the transfer speed:minimum cycle time ratio before UDMA did similar. 193.63.174.211 (talk) 17:25, 10 May 2012 (UTC)

UDMA is double-pumped, take a look at the table with cycle times. Apart from that, these modes are more or less just an agreement of how fast the cycles may be clocked (with respect to what the controller and the drive negotiated and what the cable supports). Zac67 (talk) 18:16, 10 May 2012 (UTC)

BIOS size limitations of C/H/S ?

I am tired and I must sleep but I get a nagging feeling looking at C/H/S values of 1024, 16, and 63. Was the sector count a maximum of 63 or 64 per cylinder?

The full math: 1024 * 16 * 63 = 1032192 sectors * 512 bytes = 528482304 bytes / 1024 (per kb) / 1024 (per mb) = 504 megabytes

If the C/H/S is increased to 1024 / 16 / 64, then the total is 512 megabytes, and I am certain 504 was the limit, not 512.

Odd.. DMahalko (talk) 07:16, 24 May 2011 (UTC)

Nope, you were right second time. I can't account for it, it's just one of those things. Maybe setting the sector count to 64 denoted some kind of "reserved" status? (EG "use pre-programmed hard drive type X" (where X is from 0 to 16384...), "use alternate device type", "switch to LBA mode", "assume size is 8-times-X megabytes", or whatever) ... but, why choose the Sector field for that, instead of the Cylinder one? No-one would much miss 1/1024th of the capacity, vs 1/64th, and none of the examples I can think of would need more than 64x16 (=1024!) sub-options... That, or it was a simple way of setting the device as "disabled", but it can't have been much harder to check that all 20 bits of the CHS field were zeroes, vs checking that all six of that one were ones, would it?

Wait! I see now. Cylinders would be 1-1024. Heads would be 1-16. SECTORS would be ZERO to 63. They still should have made it so sectors were 1-64 and cylinders were 0-1023, with 0 cylinders = device disable, but maybe the bits had to be arranged a certain way in memory and the easiest method of quickly checking if a disc was set as installed or not would be to check the high nibble of it's Sector count (read the first byte (from the probably quite slow CMOS memory), only pay attention to those bits). No realistic fixed disc would have 3 or less sectors, so if the bits encoding "32", "16", "8" and "4" were all zero, it could be assumed there was no disk configured at that position and the POST could move immediately to the next stage. (And, it would be limited to 20 bits by the PC architecture's real-mode addressing, even on a 286? Just with 512-byte sectors times 1M (or more accurately 1008 K) addresses rather than 1-byte memory cells...)193.63.174.211 (talk) 17:52, 10 May 2012 (UTC)

An open letter to 50.12.132.24 (talk · contribs)

...who just joined the legions of "ready, fire, aim" editors who have changed "AT Attachment" here to "Advanced Technology Attachment", without, apparently, reading so much as the first graf of the article body (that's the one titled "History and terminology").

First, thanks for self-reverting your error. Now, please note: In the wikitext there is both a big block comment above the lede, and use of the "not a typo' template, both stating clearly that "AT Attachment" is correct as written. The comment further explains that it is not to be expanded to "Advanced Technology Attachment" as that is not the name of the standard. Furthermore that point is explained in the visible article text, first paragraph after the lede. Despite all that, you, 50.12.132.24 (talk · contribs) proceeded to make that very change. Yes, you self-reverted... but I'm just wondering... what does it take? Flaming letters in the sky? Heck, why do we even need the comment or the template? Is it really asking too much that people read an article before proceeding to edit it? Please see WP:CIR. Jeh (talk) 02:27, 1 March 2012 (UTC)

Split ATAPI into new article?

In this edit summary Ruud Koot says

ATAPI should have its own article with a summary here

I don't disagree that ATAPI should have its own article—there is far more to it than is appropriate to include here, especially since ATAPI also exists and works under SATA—but I wouldn't like to see the ATAPI section here reduced much from its present state. Of course, the ATA protocol should have its own article with Parallel ATA and Serial ATA articles covering the transport layer. Jeh (talk) 19:35, 7 March 2012 (UTC)

I agree that the section shouldn't be shortened much. I implied that a separate article on ATAPI would have much potential for growth beyond the current size of this section. —Ruud 13:21, 9 March 2012 (UTC)

Someone has re-instated the split tag without bothering to leave a comment on the talk page. I do not intend to spend my life trawling through the article history in the vain hope that an explanation has been given. If I understand the request correctly then it is without merit for the following reasons:

The content to split off is too small to make a viable article.

The notability of the content to be split off has not been established.

If I were to carry out the split as requested then the split article is liable to be deleted for the above reasons and my time would be wasted.

I will acknowlege the input of Jeh above and will not summarily remove the split tag in April as I would have done otherwise. However, by May 2012, I would hope to see some progress towards getting this issue resolved i.e. 1 of the following has happened:

1) The split has been carried out.

2) The problems stated above have been rectified.

3) A comment has been placed on this talk page by the person intending to do the above informing us of the proposed timescale.

I am sorry, but if it is left to me to do the split with the article in this state then I will have no choice but to remove the split tag because to do different would:

waste my time if I did the split

just add to the maintenance backlog and eventually be deleted by the next volunteer that comes along to clean up (after I have given up and done something else).

Op47 (talk) 11:22, 9 March 2012 (UTC)

If you don't want to invest the time to perform the split properly, then please just leave the tags alone. That would be even less work for you. —Ruud 13:21, 9 March 2012 (UTC)

So the "split" was done, but no significant content has been added to the resulting separate article in months. Jeh (talk) 20:17, 24 August 2012 (UTC)

Split history

Should we split history into a separate article? It's very long, and the history section is mainly interesting for, well, historical reasons, in distinction to the description of the current standard. --Colapeninsula (talk) 17:19, 8 March 2012 (UTC)

I don't think "History of Parallel ATA" is of sufficient general interest to warrant an article by itself. It's not as if it's tough to skip it if you're not interested. Jeh (talk) 17:59, 8 March 2012 (UTC)

I agree with Jeh. Especially history sections should remain in the main article if size allows. —Ruud 13:23, 9 March 2012 (UTC)

No split, I agree with Jeh and Ruud. Additionally, the technical and historical details overlap somewhat (due to the long history of use and development) which would make it hard to split off cleanly. Zac67 (talk) 12:59, 15 April 2012 (UTC)

Split tag has been removed per discussion above. Op47 (talk) 17:36, 15 April 2012 (UTC)

How important are the published standards, particularly timewise?

I'm just wondering, as I'm seeing things written down here that I'm sure were not the case in practice.

For a start, I have a 1994-vintage 486 board which has at least upto PIO 3 on it and maybe PIO 4 unless I'm conflating it with a successor (along with some kind of early DMA mode which seemed faster, rather than slower), even though that supposedly wasn't put in the spec for another few years. In fact most of the specified modes - PIO, UDMA, and so forth - appear to have gone live in the marketplace well before the date on the linked specification PDFs.

Were those documents simply setting previously-informal evolved standards in stone so they could be used as more definite references and enhance compatibility (even though they were somewhat "after the fact" for anything but embedded systems and held-over legacy kit by then), or is the idea that nothing should have been available with those facilities until after their publication?

Also, I have a puzzling conundrum for you:

The alleged max disk size limit of Win98 SE (plus early Win2K and XP) is supposedly 128 gig (roughly), up from 64 in the original release... ((though I could be sure the latter one is nothing but a display issue in the DOS-based FDisk and even Win95 OSR2 was fine with >64GB disks if they were already formatted; Win95 original balked at anything over... hmm, it was either 8 or 32GB, I forget...)

However. I have a "250gb" drive (actually 233GiB thanks to the binary/decimal disparity) kicking around in my spares box, from waybackwhen (about 2002?). It is partitioned with two 116GiB drives, that were once C:\ and D:\... in a Win98 SE system. Last time I plugged it in, both of them were quite well filled with files dating from no later than the time I last powered up my creaky old W98 box in order to transfer stuff to a shiny new XP one.

How, exactly, did I get away with that? I don't think I installed any patches, and definitely didn't use any bootsector overlay things. Certainly the patch mentioned in the article text is completely alien to me, and probably post-dates my retirement of that computer. 250 was about as large as disks came at the time (there were *maybe* some 320s about), and I'd gone out on quite a limb to add it. The only thing I can tell you is that it was partitioned like that because C:\ couldn't be made to work when any larger than 127gb, nor could D:\ be accessed at all if it started at or beyond C:\'s maximum size (or even a certain amount before it; 116 each merely gave a nice amount of wiggle-room).

Oh yeah. And there's also the "185gb" one that was installed alongside it, installed about a year prior, so it wasn't even the first one to exceed the 127gb "limit"... (again, subpartitioned, but less evenly...) plus the machine's by-then "original" 80gb, both these on an add-in ATA-133 card (is THAT even within the published timescale? I retired it sometime around '05/'06)... plus the DVDROM and CDRW on the secondary IDE, and an ATAPI Zip as primary-slave ;) 193.63.174.211 (talk) 18:36, 10 May 2012 (UTC)

Removed Master/Slave Controversy

I removed the following because it is better suited for the Master/slave (technology) article.

The terms "master" and "slave" have not been without controversy. In 2003, the County of Los Angeles, California, US requested that, when possible, suppliers stop using the terms because the county found them unacceptable in light of its "cultural diversity and sensitivity".

Argel1200 (talk) 19:36, 16 July 2012 (UTC)

Should add that the master/slave model should be mentioned a bit more formally with a link to the article covering it. Currently, the first use for "master" is in the middle of the "Parallel ATA interface" section, with no specific description of the model or a link to the page on it. Argel1200 (talk) 19:40, 16 July 2012 (UTC)

Maybe a quick explanation of how PATA ddiffers from SATA would also be useful. Argel1200 (talk) 19:41, 16 July 2012 (UTC)

It isn't mentioned much here because PATA doesn't use a "master/slave model". The so-called "master" drive isn't a "master" of anything in PATA. It was a common but bogus myth that the "master" drive somehow controlled, arbitrated access to, or otherwise was a "master" to the "slave" in any way. How "drive 0" and "drive 1" ended up being called "master" and "slave" I do not know, but the only thing about "master/slave model" that belongs in an article on PATA is a note that it's a widely used but wholly misleading term. Jeh (talk) 15:44, 3 August 2012 (UTC)

I don't have a reference to back me up but my memories of building PC's throughout the 1990s is that it was to do with line termination of the cables. A signal transmission line works best when it has a resistor at the end of the cable. The resistor stops the signal from being reflected back to the source, reflections tending to confuse the next signal coming down the cable. Early SCSI drives usually had an external terminator that was plugged into the end of the cable or into the daisy chain port of the last drive on the cable but early IDE/ATA drives had the resistor built-in. If the drive was jumped to be master only then the id was set to 0 and the resistor was enabled. If the drive was set to be master with slave then the ID was set to 1 and the resistor. If the drive was set to be slave then the id was set to 1 and the resistor was enabled. This had the side effect that the slave always had to be at the end of the cable if you wanted reliable transmission of data. But in the mid to late 1990s the drives were able to handle the resistor automatically, so the jumpers became merely id 0/1 and the drives could be in any order of the cable.  Stepho  talk  22:35, 3 August 2012 (UTC)

Yes, a lot of people concluded that by extrapolation from SCSI. But if you'll look at the ATA specs you will find that they don't do that. Both "device 0" and "device 1" present the same load on the cable and neither matches the cable impedance; it is a high impedance load, not unlike the "bridging" connections used in audio. This is the reason for the extremely short cable length spec! Also recall (this is in the specs too) that with the 40-wire cable the end connector was designated for drive 1, but with the 80 wire cable the end is for drive 0 (these are the assignments provided if you use Cable Select). So if the drives had terminated, or not, according to whether they were jumpered (or Cable Selected) as "master" or "slave", you could have easily had a situation with either both drives adding termination, or neither. In short... an understandable theory, but no. Jeh (talk) 05:54, 4 August 2012 (UTC)

I'm talking about before cable select came in and well before 80-wire (CS and 80-wire era cables not having this termination problem). I'm talking about the first few years only. If we took a working system with master and slave and then swapped the master and slave drives, then the system would usually stop working - matching the symptoms you mentioned above in your next to last sentence. Setting up drives in the early 1990s was a pain :) But, like I said, I don't have a reference to back me up, so I'm not going to create a big fuss about it. I'll look around and see if I can find something more concrete beyond my memories and personal theories.  Stepho  talk  07:44, 4 August 2012 (UTC)

The specs are the reference, and they show no difference in the load the drive is supposed to put on the cable between drive 0 or drive 1. Even back in ATA-1, long predating the 80-wire cables. It's of course possible that some drive makers did something nonstandard in this area. This might explain why WD had a special jumper setting for "only drive," and why some combinations of drives were known to not play well together. If a RS (not a blog or forum site) can be found stating that the M/S setting determined whether the drive applied additional termination to the cable, we could of course include that info, but in the meantime it's just speculation, not even OR (particularly as the specs say otherwise). Jeh (talk) 09:05, 4 August 2012 (UTC)

It was well known that the manufacturers deviated from the specs at the drop of a hat and yep, some drives never were able to coexist - IDE CD players being the worst. And I agree that if I stuck this in the main article it would well and truly be OR - and possibly totally wrong :) I'll hunt around for something better than my own fuzzy memory.  Stepho  talk  10:05, 4 August 2012 (UTC)

Damn, I hate it when I'm wrong. The difference between 'master only', 'master with slave' and 'slave' seems to be wrapped up in the DASP signal. Early drives needed some extra help handling the DASP signal correctly in each situation (ie they needed to know about the presence of the other drive) while later drives could figure it out. http://books.google.com.au/books?id=E1p2FDL7P5QC&pg=PA513 (p513-514)  Stepho  talk  12:38, 4 August 2012 (UTC)

Line breaks

Jeh, you made a comment in the edit summary that WP:MOSNUM says numbers followed by words (ie not units) are not allowed to use {{nowrap}}. I read MOSNUM and could not find this rule. I did find a rule that says units should use {{nowrap}} but no rule to say other uses are not allowed. To my mind, bytes, MB, GB, etc are being used here in the same manner as m, km, miles, etc.  Stepho  talk  01:28, 3 August 2012 (UTC)

The point is not e.g. bytes vs. miles. No, the rule is that unit symbols (like B for byte) following a number should use nowrap. But in the previous section it says that unit names (like byte or mile or kilogram) are common nouns. So you should indeed use nowrap around "512 B", but not around "512 bytes". Similarly you should use nowrap around "27 km" but not around "27 kilometers" (and the latter is generally preferred in prose)... any more than you would use nowrap around "350 cylinders" or "27 hard drives" or a number followed by any other common noun.

You may have noticed a trend in some places to omit the space entirely before a unit symbol, e.g. "a 2TB hard drive". It's a lazy way of forcing nowrap, and it doesn't look too awful. But I doubt you would ever see "a 2terabyte hard drive" or, for that matter, "512bytes per sector". See? If you wouldn't omit the space, don't use nowrap either.

As for "it's allowed because it isn't forbidden," that is not a reasonable rule in typography. (Nor is "to my mind".) "Byte" and "B" are not the same thing. Excessive use of nowrap (and the article definitely used it excessively) clutters the edit window (and so makes edits more error-prone) and can mess up the page rendering.

If you still disagree this should probably be discussed at the WP:MOSNUM talk page. Jeh (talk) 08:37, 3 August 2012 (UTC)

Thank you for fixing the "nonexistent template" I left behind. btw, that would have been the perfect sort of edit to tag as minor. Jeh (talk) 08:37, 3 August 2012 (UTC)

Ah, it's the symbol vs word part that I missed. "To my mind" was only meant to say that I don't understand the reasoning and was trying to explain what I do understand in order to reach a mutual agreement. I'm happy enough now. Thanks.  Stepho  talk  08:55, 3 August 2012 (UTC)