How It Works:

[Home] [Products] [HowItWorks]
[HIW Topics] [CHSxlat] [DOSboot] [FAQ] [MBR] [OS2boot] [PartTable]


First the "legal" stuff...

  1. This FAQ is not intended to replace any other FAQ on this subject but is an attempt to provide historical and technical information about the ATA interface.
  2. This FAQ is not an endorsement of any vendor's product(s).
  3. This FAQ is not a recommendation to purchase any vendor's product(s).
  4. Every effort is made to insure that all of the information presented here is not copyrighted, not proprietary and unrestricted.
  5. When opinions are stated they are clearly identified, including the person's name and email address. Such opinions are offered as long as they contribute to the understanding of the subject being discussed. No "flames" allowed.

Table of Contents


Read and understand these terms. You will be lost and confused if you do not! Many of these are described in much greater detail in other parts of this FAQ.

ATA or AT Attachment

ATA is the proper and correct name for what most people call IDE. In this document, ATA refers to all forms of ATA (ATA-1, ATA-2, etc, IDE, EIDE, etc). The ATA interface uses a single 40-conductor cable in most desktop systems.


ATA-1 is the common name of the original ATA (IDE) specification. ATA-1 is not an official standard yet. Final approval is pending.

ATA-2 or ATA Extensions

ATA-2 is the common name of the new ATA specification. ATA-2 is still in early draft form and has not been submitted for approval as an official standard.


ATA-3 is the common name of a future version of the ATA specification. The ATA-3 working group has held several meeting but the only things adopted so far are a DMA version of the Identify command, a description of "device 1 only configurations" and a set of "security" commands.

There is much discussion going on concerning merging ATA-3 with ATAPI. This will require some kind of "command overlap" capability. The details of this are consumming much of the meeting time.

ATAPI or ATA Packet Interface

ATAPI is a proposed new interface specification. Initially it will probably be used for CD-ROM and tape devices. It uses the ATA hardware interface at the physical level but uses a subset of the SCSI command set at the logical level. The ATAPI specification work is currently being done in the SFF committee.

The ATAPI folks have delayed forwarding their CD-ROMspecification from SFF to X3T10 so the X3T10 ATAPI working group has nothing to work on yet and have held no meetings.

Block Mode

Block mode is the name given to the use of the ATA Read Multiple and Write Multiple commands. These commands generate a single interrupt to the host system for each block of sectors transfered. The traditional Read Sectors and Write Sectors commands generate an interrupt to the host for each sector transfered.

CAM (Common Access Method) Committee

The Common Access Method committee, now disbanded, worked on two specifications: the CAM SCSI and CAM ATA specifications. Both specifications were forwarded to the X3T9 committee for further work years ago.

CHS or Cylinder/Head/Sector

CHS is the old and traditional way to address data sectors on a hard disk. This style of addressing relates a sector's address to the position of the read/write heads. In today's ATA devices, all sector addresses used by the host are logical and have nothing to do with the actual physical position of the sector on the media or the actual position of the read/write heads.

Command Block
Control Block

These are names given to the I/O register interface used by ATA devices. It refers to a set of I/O registers, or I/O ports and I/O port addresses used to program the device. These names replace the older term Task File.

DMA or Direct Memory Access

DMA is a method of data transfer between two devices that does not use the system's main processor as part of the data path. DMA requires lots of hardware: a DMA arbitration unit, a DMA data transfer unit and host bus signals that enable the DMA controller to assume control of the host system's bus. When the DMA controller has control of the host system's bus, it moves data between the two devices by generating the appropriate bus read/write cycles. For the ATA READ DMA command this means generating an I/O read cycle and then a memory write cycle for each 16-bit word being transferred. For the ATA WRITE DMA command, a memory read cycle is followed by an I/O write cycle for each 16-bit word transferred.

EIDE or Enhanced IDE

EIDE is a marketing program started by Western Digital to promote certain ATA-2 features including ATAPI. WD has encouraged other product vendors to mark their products as "EIDE compatible" or "EIDE capable".


See MFM.

Fast ATA

Fast ATA is a Seagate marketing program used to promote certain ATA-2 features in newer ATA devices. Seagate has encouraged other product vendors to mark their products as "Fast ATA compatible" or "Fast ATA capable".

Host or Host System

The computer system that the ATA device is attached to.

HBA or Host Bus Adapter or Host Adapter

The hardware that converts host bus signals to/from ATA interface signals. An ATA-1 host adapter is generally a very simple piece of hardware. An ATA-2 host adapter can be simple or complex.


IDE can mean any number of things: Imbedded Device (or Drive) Electronics (yes, you can spell embedded with an "i"), Intelligent Device (or Drive) Electronics, etc. The term IDE is the trademark of someone (Western Digital does not claim IDE as theirs but they do claim EIDE). Many hard disk vendors do not use IDE to describe their products to avoid any trademark conflicts.

LBA or Logical Block Addressing

LBA is a newer (for ATA it is newer) way to address data sectors on a hard disc. This style of addressing uses a 28-bit binary number to address a sector. LBA numbers start at zero. In today's ATA devices, all sector addresses used by the host are logical and have nothing to do with the actual physical location of the sector on the media.

Local Bus

Usually this refers to the processor's local bus in a high performance computer system. Usually the processor, the external processor instruction/data cache, the main memory controller and the bridge controller for the next low speed system bus, for example, a PCI bus, are located on the local bus. Lower speed local buses may have connectors that allow the attachment of other devices. For example, the VL-Bus is a local bus that can allow attachment of video, SCSI or ATA controllers. It is very difficult to attach other devices to high speed (say faster than 100MHz) local buses due to electrical restrictions that come into play at those higher speeds.


ATA device 0. Device 0, the master, is the "master" of nothing. See Slave.

Megabyte or MB

Megabyte or MB is 1,000,000 bytes or 10^6 bytes. IT IS NOT 1,048,576 bytes or 2^20 bytes, repeat NOT!


In this document MFM refers to any of the older hard disk controller interfaces, MFM, RLL and ESDI. It is used to describe any hard disk controller that uses the Task File interface on the host side and the ST506/ST412 interface on the drive side.


Operating System.


We can thank the Personal Computer Memory Card International Association for the PC Card specification. The PCMCIA is a nonprofit industry association. The PC Card ATA specification is another form of the ATA interface used by PCMCIA compatible ATA devices. This interface uses the PCMCIA 68-pin connector. Most 68-pin ATA devices are dual mode - they can operate as either a PCMCIA PC Card ATA device or as a 68-pin ATA device.


We can thank Intel and the other members of the PCI committee for the PCI bus specification. PCI is intended to be the next high performance computer bus. PCI is not generally described as a processor local bus.

PIO or Programmed Input/Output

PIO is a method of data transfer between two devices that uses the system's main processor as part of the data path. On x86 systems, the REP INS and REP OUT instructions implement this data transfer method. INS reads and I/O port and writes the data into memory. OUTS reads data from memory and writes the data to an I/O port. Each time an INS or OUTS instruction is executed, the memory address is updated. The REP prefix causes the instructions to be repeated until a counter reaches zero.


See MFM.


ATA device 1. Device 1, the slave, is a "slave" to nothing.

See Master.

Task File

This is the name given to the I/O register interface used by MFM controllers. It refers to a set of I/O registers, or I/O ports and I/O port addresses used to program the controller. In ATA, this name has been replaced by the terms Command Block and Control Block.


See the SCSI FAQ.

SFF or Small Form Factor

The SFF committee is an ad hoc committee formed by most of the major storage device and system vendors to set standards for the physical layout of hard disk and other devices. SFF has published many specifications that describe the physical mounting and connector specifications for hard disk devices, including ATA devices. During a brief period of time when the X3T9 committee was not doing much work on the ATA-1 interface, the SFF committee published several specifications that were not really part of the original SFF charter. Most, if not all, of these nonphysical specifications have now been incorporated into the latest X3T9 or X3T10 ATA specifications. ATAPI is currently an SFF specification.

ST506 and ST412

This is the common name for the hard disk controller to hard disk drive interface used by MFM, RLL and ESDI controllers and disk drives. ST stands for Seagate Technology. The ST506 and ST412 were the Seagate products that set the de facto standards for this interface many years ago. This interface is composed of two cables: a 34-conductor and a 20-conductor cable.

VESA and VL-Bus

We can thank the Video Electronics Standards Association for the VESA Local Bus or VL-Bus specification. The VL-Bus is one example of a local bus. VESA is a nonprofit industry association like the PCMCIA.

WG or Working Group

The actual work on various specifications and standards documents within the X3T9, X3T10 and SFF committees happens in working group meetings. Most WG meetings are held monthly.

X3T9 and X3T10

These are the names of the official standards committees that have worked on the ATA-1 and ATA-2 specifications. X3T9 was responsible for the SCSI and ATA-1 specifications and standards. X3T10 has replaced X3T9 and is now responsible for the current SCSI and ATA specifications and standards work.


This term is used in this document to describe the capacity boundary that exists in most x86 system software. This boundary limits the size of an ATA disk drive to 528MB. For cylinder/head/sector style addressing of disk data sectors, this number is computed as follows:

  1. the number of cylinders are limited to 1024, numbered 0-1023.
  2. the number of heads (per cylinder) are limited to 16, numbered 0-15,
  3. the number of sectors (per track) are limited to 63, numbered 1-63.
  4. a sector is 512 bytes,
  5. 528MB means the following values:
    ( 1024 * 16 * 63 ) or 1,032,192 data sectors
    ( 1024 * 16 * 63 * 512 ) or 528,482,304 bytes.

68-pin ATA

This refers to a variation of the ATA interface that uses the PCMCIA 68-pin physical interface but does not use the PCMCIA electrical or logical interface. Most 68-pin ATA devices are dual mode - they can operate as either a PCMCIA PC Card ATA device or as a 68-pin ATA device. This interface was developed within the SFF committee and is now included in ATA-2.

Basic Questions

Where did ATA come from?

What we now call the ATA-1 interface was developed for Compaq many years ago by Imprimis (then part of CDC, now part of Seagate) and Western Digital. The first ATA-1 hard disk drives were made by Imprimis but it was Conner that made the interface so popular.

How is ATA different from MFM?

From the host software standpoint, ATA is very much like the Task File interface used by MFM controllers. A properly written host software driver should not notice any difference between the MFM Task File interface and the ATA Command Block interface while doing basic commands such as Read/Write Sectors.

At the hardware level, ATA uses a single cable between a host bus adapter and the ATA device, where the MFM controller interface uses two cables between the controller and the drive.

In the MFM environment, the controller is one piece of hardware and the drive another piece of hardware. Most likely these two pieces of hardware are from different vendors. The MFM controller is dependent on the design of both the host bus and on the drive.

In the ATA environment, the host adapter is the one piece of hardware that is dependent on the host system bus design. The ATA interface is (mostly) system independent. All of the hard disk controller and drive logic is contained in the ATA device hardware. This gives the hard disk designer complete control over both the controller and drive functions.

Why is ATA so popular?

Two basic things make ATA so popular today: cost and hard disk drive technology. An ATA-1 host adapter is cheap, usually much less than $25US and it uses only one cable. On the technology side, current hard disk features, such as, defect handling, error recovery, zone recording, cache management and power management require that the controller be fully integrated with the read/write channel, the servo system and spindle hardware of the disk drive.

What are the basics of the ATA interface?

The ATA interface is a very simple interface based on an ISA bus I/O device architecture. The interface consists of two sets of I/O registers, mostly 8-bit, for passing command and status information. The registers are like a set of mail boxes with a door on front and back connected such that both doors can not be open at the same time. The front door is open when the Busy bit in the Status register is zero and the host can read and write the registers. The back door is open when the Busy bit in the Status register is one and the ATA device can read or write the registers.

The physical interface contains just enough signals for a 16 bit data bus, five register address bits, and a few control signals like read register, write register and reset.

ATA devices look like traditional hard disk drives even though some are not really a hard disc with rotating platters. User data is recorded in 512 byte sectors. Each sector has a sector address. There are two ways to express sector addresses: by cylinder/head/sector (CHS) or by logical block address (LBA). CHS is standard, LBA is optional.

What is EIDE or Fast ATA?

Both are marketing programs used to promote various ATA-2 features, mostly the faster data transfer rates defined by ATA-2.

WD defines EIDE as:

Seagate defines Fast ATA as:

What does all of this mean to us?

Support for the ATA-2 high speed PIO and DMA data transfer modes is both a hardware and software issue.

Support for more than one hard disc controller (or ATA host adapter) requires the BIOS and/or the operating system to support more than one Task File or Command/Control Block register set on the host bus.

The 528MB problem is due to the original design of the x86 BIOS which limits cylinders to 1024 and sectors to 63. The ATA interface allows up to 65,535 cylinders, 16 heads and 255 sectors - that is about 136GB (137GB if is LBA is used). Support for devices over 528MB requires the BIOS and/or operating system to support some form of CHS translation. Note that LBA alone does not solve this problem (in fact, LBA may make things more complex).

Support for CD-ROM and tape will probably be done via the ATAPI interface which uses a different command structure than ATA. That makes ATAPI another host software issue.

What does an ATA-1 host adapter do?

An ATA-1 host adapter is a simple piece of logic whose main purpose is to reduce the system bus address lines from 12 (or more) down to 5. It may also buffer some signals giving some degree of electrical isolation between the host bus (usually an ISA or EISA bus) and the ATA bus. In ATA-1, the ATA interface is controlled directly by the host bus so that all timings are controlled by the host bus timing.

What does an ATA-2 host adapter do?

This answer is complex because it depends on how smart your ATA-2 host adapter is. First, an ATA-2 host adapter supports the ATA-2 higher speed data transfer rates. That requires that the host adapter is attached to something other than an ISA or EISA bus. Second, an ATA-2 host adapter may perform 32-bit wide transfers on the host bus. This requires FIFO logic and data buffers in the host adapter. Third, an ATA-2 host adapter may use a different data transfer protocol on the host side than is used on the ATA device side.

Can I put an ATA-2 device on an ATA-1 host adapter?
Can I put an ATA-1 device on an ATA-2 host adapter?

The answer to both questions is yes, as long as the electrical timing specifications of the device are not violated. In general it is impossible for an ATA-1 host adapter to violate the specifications of an ATA-2 device. It is possible for an ATA-2 host adapter to violate the timing specifications of an ATA-1 device but this is not common. However, host adapter hardware design errors or software driver bugs can cause such a problem. The result will be corrupted data read or written to the ATA-1 device.

I have an ATA-2 host adapter with an ATA-2 device. I want to add an ATA-1 device to this host adapter. Can I run the ATA-2 device in its ATA-2 data transfer modes?

Sorry, NO you can NOT run the new drive in its faster data transfer modes. Be very careful, most of the ATA-2 host adapter vendors do not have anything in their setup documentation or software to prevent this sort of thing.

When you run the new drive at a data transfer speed that is faster than the older drive can support, you are violating the electrical interface setup and hold times on the older drive. There is no telling what the older drive will do about this, but you are asking for data corruption and other nasty problems on your older drive.

How many disk controllers and/or ATA host adapters and/or SCSI host adapters can I put in my system?

From a hardware standpoint - as many as you want as long as there are no I/O port address, memory address or interrupt request signal conflicts. From a software standpoint it is a whole different story.

First the simple x86 system hard disk controller configurations...

  1. 1 ATA with 1 or 2 drives at I/O port addresses 1Fxh/3Fxh using interrupt request 14 (IRQ14).
  2. 1 ATA with 1 drive at I/O port addresses 1Fxh/3Fxh using interrupt request 14 (IRQ14) plus a SCSI with 1 drive.
  3. 1 SCSI with 1 or 2 drives.

Other configurations are possible but are most likely not supported in the system or SCSI host adapter BIOS. And if its not supported at the BIOS level, it is unlikely to be supported by an operating system, especially DOS. The primary reason the above configurations are so restrictive is that the original IBM x86 BIOS supported only one MFM controller with a maximum of 2 drives. This restriction was then coded into much x86 software including many early version of DOS. The configurations above work because they do not break this old rule.

Just remember this - most systems will always boot from the first drive on the first controller. In a) that is ATA drive 0, in b) that is ATA drive 0, in c) that is SCSI drive 0.

And now the more complex configurations...

Once you go beyond the three configurations above all bets are off. Most likely you will need operating system device drivers in order to access any drives beyond the first two. And now your real problems start especially if you like to run more than one operating system!

If you do run more than one OS, then you need equivalent drivers for each system if you would like to access all the drives. Plus it would be nice if all the drivers configured the drives in the same manner and supported all the possible partitioning schemes and partition sizes. It would be especially nice if a driver would not destroy the data in a partition just because it did not understand the file system format in that partition.

One of the things EIDE promotes is BIOS support for up to four ATA devices -- 2 ATA host adapters each with 1 or 2 drives. The first would be at I/O port addresses 1Fxh/3Fxh using interrupt request 14 (IRQ14) and the second at I/O port addresses 17xh/37xh using interrupt request 15 (IRQ15). Acceptance of this configuration has not been spreading like wild fire through the BIOS world.

Lets look at a two complex configurations...

  1. ATA with 2 drives and 1 SCSI with 1 or more drives.
    Nice configuration. The ATA drives would be supported by the system BIOS and the SCSI drives may be, could be, should be, supported by the SCSI host adapter BIOS but probably not. So in order to use the 2 SCSI drives you probably have to disable the BIOS on the SCSI card and then load a device driver in CONFIG.SYS. And because the SCSI BIOS is disabled, you then need a SCSI driver for that other OS you run.
  2. ATA with 1 or 2 drives on each.
    Also a nice configuration. But because the system BIOS probably only supports the first controller address, you will need a DOS device driver loaded in CONFIG.SYS in order to access the drives on the second controller. You will need that driver even if there is only one drive on the first controller. You also need a similar driver to support the second controller in your other OS.

Note: I understand that OS/2 does support both MFM/ATA controller addresses and does allow up to four drives - I have not confirmed this for myself.

Are disk drives the only ATA devices?

No. Over the years there have been ATA tape drives, ATA CD-ROMS and other strange devices. Most of these are expected to be added to an existing ATA host adapter as the second device (device 1) with an existing ATA disk drive (device 0). In general these require software drivers to operate with your OS.

Now, we have ATAPI CD-ROM and tape devices that can be placed on an ATA host adapter. And soon we should see system motherboard BIOS support for booting from an ATAPI CD-ROM device. The general idea is that an ATAPI device can coexist with an ATA device on the same cable.

What can be done to improve ATA device performance?

A difficult question. But the first step is usually to reduce the number of interrupts that the host sees during a read or write command. ATA disk drives have three types of read/write commands:

Now see the next question...

What else can be done to improve ATA device performance?
What is PIO mode "x" ?

An even more difficult question. The second step is usually to increase the rate at which the host transfers data.

(Ahh... I can see the funny look on your face from here. You are saying to yourself: "the rate at which the host transfers data? Read on...)

The rate at which data is transferred to or from an ATA device is determined by only one thing: the PIO or DMA cycle time the host uses. No, the drive does not have much to do with this! The only requirement is that the host not exceed the minimum PIO or DMA cycle times that the device supports. For example, during a PIO read command when the device signals an interrupt to the host this means that the device is waiting for the host to read the next sector or block of sectors from the drive. The host must execute a REP INSW instruction to do transfer the data. The rate at which the host executes this instruction determines the PIO cycle time. Technically, for a read command, the cycle time is the time from the host assertion of the I/O Read signal to the next time the host asserts the I/O Read signal.

Be careful when looking at the table below - the data rate is the data transfer rate achieved while transfering the sector or block or sectors. It is an "instantanous" data rate. The overall data transfer rate for a command includes many time consuming events such as the amount of time the host requires to process an interrupt. Note that on many fast ATA drives today, the time it takes the host to process an interrupt is frequently greater than the time required to transfer the sector of block of sectors for that interrupt! It is not uncommon for the host overhead to reduce the data rate to or 1/3 of the instantanous rate shown here.

The ATA PIO modes are defined as follows:

PIO mode min cycle time data rate comment
0 600ns 3MB the rate at which a system running at 4.77MHZ could execute the REP INSW.
1 383ns 5MB the rate at which a system running at 6MHz could execute the REP INSW.
2 240ns 8MB the rate at which a system running at 8MHz could execute the REP INSW.
3 180ns 11MB requires an ATA-2 host adapter.
4 120ns 16MB requires an ATA-2 host adapter.

The complete description of the PIO (and DMA modes is much more complex and will be cover in more detail later in this FAQ.

Do I need BIOS or OS drivers to support more than 528MB?

Warning: Read the previous question before reading this one.

Maybe, probably, yes. The answer to this very complex and will be discussed in detail in Part 2. Here is the brief answer...

A traditional x86 system BIOS supports only CHS mode addressing with cylinders limited to 1024, heads limited to 16 and sectors limited to 63. This allows addressing of drives up to 528MB. These limitations come from the INT 13 read/write calls that combine a 10-bit cylinder number with a 6-bit sector number into a 16-bit register.

Note that this is entirely a software problem: the ATA interface supports up to 65,535 cylinders, 16 heads and 255 sectors.

While the head number usually requires only 4-bits, up to 6 or 8 bits are available in the INT 13 interface. This fact has allowed the SCSI folks to support big drives by increasing the number of heads above 16. The SCSI host adapter BIOS converts this "fake" CHS address to a different CHS or an LBA when it issues a read/write command to the drive. The following table shows some approximate drives sizes and the "fake" CHS parameters that you may see from a SCSI BIOS:

cyl head sector size
1024 16 63 512MB
1024 32 63 1GB
512 64 63 1GB
1024 64 63 2GB
1024 256 63 8GB

The last entry represents the largest possible drive that a traditional INT 13 BIOS can support.

The system BIOS folks must start supporting drives over 528MB in their BIOS by implementing some type of CHS translation. To date, few systems have such BIOS. And here is the bad part: Microsoft says that the BIOS must support it in order to use it in their OS. The algorithm is simple (but warning: this is not the complete algorithm!):

INT 13 input action ATA interface
cyl number "multiply" by n modified cyl number
head number "divide" by n modified head number
sector number nothing sector number

The value of n must be selected such that the modified head number is less than 16.

LBA addressing at the hard disk drive level or at the BIOS or driver level is another solution. This solution will probably not be popular for several more years. It requires that the BIOS people implement a new INT 13 interface, called the Microsoft/IBM Extensions and that the OS people start using this new BIOS interface. Few system BIOS support this alternative interface today. Without this new interface, LBA support at the hard disk drive level is not required.

So most of us have older systems without much possibility of getting a BIOS upgrade, so what do we do? Well we must obtain one of the many driver products that are on the market that live in one of the disk boot sectors and "take over" the system BIOS INT 13 with an INT 13 that supports the translation. The biggest problem with this is that the replacement INT 13 BIOS must live someplace in memory. For DOS based systems, it can usually live at the top of the 640K of memory and DOS is made to think that that part of memory, usually around 8K bytes, does not exist. But the protected mode OS's do not like this and usually wipe out the driver when they load their kernel. So if you plan to run multiple OS's on your system, buyer beware!

Then there is the Windows problem: the standard FastDisk driver in Windows does not support such translation schemes and can not be enabled. So make sure the driver you purchase also comes with a Windows FastDisk replacement.

Buyer beware!

Do I need BIOS or OS drivers to support the ATA-2 data transfer rates?

Warning: Read the previous two questions before reading this one.

Maybe, probably, yes. The answer to this very complex and will be discussed in detail in Part 2. Here is the brief answer...

If you have a new ATA drive that supports the advanced PIO or DMA data transfer rates (ATA-2 PIO Mode 3 or 4, or, ATA-2 DMA Mode 1 or 2) then you also must have a new ATA host adapter that attaches to the VL-Bus or PCI bus or some other high speed bus (probably a 32-bit bus) in your system. That host adapter has I/O registers of its own that are used to control its advanced features. Controlling those advanced features requires software - either in the system INT 13 BIOS or in a INT 13 BIOS on the host adapter card or in a driver loaded via the boot record or later by your OS.

Depending on how that host adapter works you may also need a Windows FastDisk replacement in order to use the high speed data transfer modes in Windows.

Buyer beware!

I just purchased a new high speed host adapter for my VL-Bus (or PCI bus) system and a new 540MB hard disk. How do I get full use out of all this new hardware?

Did you read the previous three questions?

You need BIOS or driver software and a Windows FastDisk replacement. These must support both CHS translation (because your drive is over 528MB) and the host adapter hardware (to use the high speed data transfer rates).

Some drivers on the market today use LBA addressing on the ATA interface to get over 528MB. This may make your disk partition(s) unreadable by another OS.

Check the hardware and software specifications of the product before you buy it! Ask lots of questions - you probably get lots of incorrect or misleading answers - be prepared for that! If you plan to run something other than DOS and Windows, especially if you plan a "dual boot" or "boot manager" environment, be real careful.

Page updated 25 Oct 2005.