IBM SSD Hard Disk Drive Frequently Asked Questions ================================================== . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What do I need to do to use an IDE HDD with more than 528 MB capacity? Upgrading to an IDE HDD with capacity greater than 528 MB may require an alternative to BIOS INT13 to address the drive. The combination of bit assignments in INT13 and in the CHS fields of IDE produce the 528 MB maximum limit. Starting in 1993, the BIOS in some PCs has a translation mode that extends the maximum addressable capacity above 528 MB. For these BIOS versions, using larger HDDs requires no BIOS changes. For older BIOS versions, or those without a translation mode, one alternative is to use DiskManager from OnTrack Computer Systems. Another is to use OS/2 and HPFS format instead of FAT format. Note: Windows 32-bit Disk Access is not intended to address the 528 MB capacity limit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What is '32-bit access'? The drivers that allow Windows to do disk or file access with the microprocessor in protected (32-bit) mode instead of real (16-bit) mode are collectively referred to as '32-bit access'. OS/2 has used protected (32-bit) mode for all accesses in all releases, so you won't hear about an upgrade of this nature for OS/2. As an example, 32-bit Disk Access provides a method of doing HDD I/O operations that does not use BIOS INT13 and leaves the processor in protected mode. Use of BIOS INT13 requires switching the processor into real mode, which is relatively time-consuming. Windows performance is improved with 32-bit access, as outlined in many articles (a good example is by John Woram in the June 1994 issue of Windows Magazine). Other testing reveals that the performance gains from 32-bit File Access are significant, while the gains from 32-bit Disk Access are more modest. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Why can't I select 32-bit Disk Access with my new IBM IDE HDD? On some IBM HDDs, Windows v3.1 will not recognize that the drive supports the WD1003 standard. In these cases, the following lines need to be added into the 386enh section of the SYSTEM.INI file in the WINDOWS subdirectory: device=*wdctrl device=*int13 This allows the system to offer the 32-bit disk access option. It can be enabled by clicking on the 32-bit box under the Virtual Memory option, or by entering 32BitDiskAccess=on in the 386enh section of the SYSTEM.INI file. Note: Windows 32-bit Disk Access is not intended to address the 528 MB capacity limit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Which models of PS/1 use the IDE interface for their HDDs? Any PS/1 using a 386 or later microprocessor uses the IDE interface. The PS/1 models which use a 286 microprocessor employ a non-standard HDD interface. Also, PS/1 models announced after May 1994 have the added feature of a 'translating BIOS' which allows HDDs with more than 528 MB capacity to be fully used. Some particular examples of models using the IDE interface include 2121, 2133, 2144, 2155, and 2168. An example of a model which does *not* use the IDE interface is 2011. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Why does a WRITE BUFFER command take longer to complete when it is the first accepted command after power is applied to my Ultrastar? This delay occurs only at power up, and is the result of the execution of unit diagnostics. At power-up, DFHS/DFMS does a thorough test of the unit data buffer. This test requires 6-7 seconds to complete, and during that time no other buffer operations are allowed. All SCSI commands are accepted after the 2 seconds it takes to initialize the interface, and unit attention (resulting from POR) is cleared. However, any commands involving buffer transfers are not executed until after the buffer test is completed. Commands which do not require use of the buffer will execute (e.g. START UNIT, INQUIRY, TEST UNIT READY). Ultrastar will disconnect from the bus, if allowed, when receiving a READ or WRITE BUFFER command while the buffer test is in progress. If not allowed to disconnect, the bus will stay in command phase until the buffer test is complete. The command is accepted, but execution is delayed, and no error condition is generated as the result of delayed execution. Note that the recommended minimum time-out limit for a WRITE BUFFER command is 30 seconds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What do media transfer rate, cylinder transfer rate, buffer transfer rate, interface data rate, and bus rate mean? Which one is the 'speed' of my HDD? These rates measure the speed at which bytes of data are moved, but at different points of the complex system that moves your computer data from RAM to disk and back. The 'effective' speed of your HDD is the average speed of all these parts working together. How your application software, operating system, and HDD adapter work together greatly affects the 'effective' speed for a given HDD performance. _____ | | - rotation speed, recording density, |HDD | read/write channel electronics - - - - -> media rate | | - format efficiency, latency - - - - - - -> cylinder rate | | - access time, error recovery - - - - - - -> throughput | | . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . |_____| - controller/buffer/interface electronics -> buffer rate : : : : - interface protocol (IDE, SCSI, etc.)- - -> interface rate : : - negotiation/communication/control __:_:___ | | - adapter hardware/software | | - main bus - - - - - - - - - - - - - - - -> bus rate |Computer| - operating system |________| - RAM Media transfer rate indicates how fast bits are detected from the disk, decoded, and checked for errors. This number is often quoted in megabits per second, reflecting the serial (1-bit wide) nature of read channels. Note that IBM HDDs, like Ultrastar, quote media rates in megabytes per second to allow direct comparison with interface data rates, also quoted in megabytes per second. Quoting media transfer rates in megabytes/second also helps avoid any confusion about whether the rate is for encoded or decoded (customer) bits. For example, RLL(1,7) is a 2/3 rate code so there are 3 encoded bits for each customer bit. A 120 megabit per second media rate for *encoded* bits would be an 80 megabit per second media rate for *customer* bits. The Ultrastar media maximum transfer rate of 12.58 megabytes per second is the rate of *customer* byte transfer. Not all of the bits on a track are customer bits. Some track space is used for block identification, synchronization fields, ECC fields and timing pads. Space must also be set aside for servo signals. These non-data areas can consume a significant part (25% or so) of the available recording area. So, even though the read/write electronics may be processing 12.58 MB/second, they are not reading only the bytes of data that the computer user is interested in. The fraction of a track on which user data is stored is called the format efficiency. Other times during which no user data is transferred from the disk are latency (waiting for the desired block to rotate to a position under the read head), seeks, and error recovery. These times add to the overall time required to read the number of user data bytes in a cylinder, reducing the net transfer rate (megabytes per second). The net transfer rate is sometimes termed the throughput of the HDD. The HDD saves up data in its internal memory buffer. The buffer is filled at a rate which, at maximum, is the throughput rate. In Ultrastar drives, this throughput is about 8 MB/second at the maximum data rate band. The buffer is emptied when data is transferred through the interface from the HDD to the system. Fast/Wide SCSI allows for transferring at rates up to 20 MB/second. Obviously, the buffer can be emptied faster than it can be filled. The result is that the long-term average data transfer rate through the interface can be no greater than the HDD throughput rate. The maximum buffer rate is, by design, equal to the maximum interface transfer rate. However, this maximum rate only occurs in bursts. The length of the burst depends on how much data is requested and how much data is actually in the HDD's buffer. In addition to data bytes, the interface must also pass command and communication bytes. If the transfer is not synchronous, communication signals must be asserted and recognized between each data byte, further increasing the time required to transfer a given number of user data bytes. Finally, the user data must be transferred from the HDD interface adapter card to the computer system RAM via the computer system bus. This transfer rate is usually much faster than the HDD interface transfer rate and does not limit the effective transfer rate. However, most (but not all) adapters require the CPU to handle the data transfers, which uses some of the CPU capacity for each transfer from the HDD. The overall result is that the maximum advertised data rate is realized only if measured during the correct and short enough time interval. Longer-time average data transfer rates will be reduced by the multiple efficiencies involved. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How do I disable SCSI parity checking on my DSAS-3xxx? SATSUMA (DSAS) does *not* have a way to select 'disable SCSI parity check' either by jumper or ModePage bit. SCSI bus parity checking is always enabled. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The HDD in my system makes noise when I'm not using the system. Is the HDD doing thermal updates? Deskstar models employ an idle time random access feature. Leaving the read/write heads positioned over a single track for extended periods of time can reduce the long-term life of the HDD. Moving the heads while the drive is idle ensures that the disk lubricant remains evenly distributed across the disk surface for maximum device life. The sound of this feature has been confused with 'thermal updates'. Deskstar models employ embedded servo and therefore do not require periodic thermal compensation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Will thermal updates affect the performance of Ultrastar? No. Ultrastar utilizes embeded servo, so thermal updates are not necessary or applicable. Ultrastar does periodically perform self-diagnostics, which include some fine-tuning adjustments for the servo system. These self-initiated diagnostics take approximately 300-400 milliseconds to complete. However, the HDD will not self-initiate these diagnostics until the SCSI bus has been inactive for at least five seconds. Since none of the diagnostics is an absolute requirement, the HDD will never interrupt bus-initiated activity to perform them. In a continuous usage environment where the SCSI bus is never idle for five seconds or more, the only delays are those that result from error recovery and saving logging information from HDD RAM to disk. When these actions are necessary, they are performed as part of a read or write command, extending completion time for the affected commands. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What exactly are thermal updates? In dedicated servo designs, one disk surface and head are used for positioning the head stack, a process generally called to as servoing. Thermal updates are used to measure the position of the data read head relative to the servo read head, since this position can change with thermal expansion of the HDD parts. In embedded servo designs, the servo information is on each track along with the data, so the data read head is also the servo read head. There is no relative movement to measure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How is SCSI TermPower enabled on DPES models? For DPES models, installing the jumper to enable internal SCSI bus termination also enables the drive to provide termination power to the SCSI bus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How is SCSI TermPower enabled on DALS models? For DALS models, SCSI bus termination power is always enabled. Users should be aware that this poses a risk of excessive current on the SCSI TermPower lines if multiple DALS units are on the same SCSI bus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I have a SCSI 3.5" drive that worked in an AS/400, but will not work in my PS/2. What's wrong? HDDs for AS/400 are low-level formatted to a 520B blocksize. They must be reformatted to 512B blocksize before they can be used in PS/2s. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Why doesn't my system recognize the new SCSI HDD I installed? The SCSI address is unique and the drive spins, is there something wrong with my installation or cables? Both ends of SCSI bus must be terminated. An unterminated bus often results in a drive not being recognized as attached to the bus. A 'SCSI bus' is any single section of multi-wire cable in which the conductors are continuous over the entire length. When the end of the conductor(s) is reached, the end of the bus is reached and electrical termination must be supplied. Either an external plug-type terminator can be used, or device on-board termination can be selected. Only the ends of the SCSI bus should be terminated. Even though there may be multiple connectors on the bus (cable), termination should be applied only at the two ends. Mid-bus termination reduces bus signal-to-noise and can produce effects similar to lack of termination. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Are IBM SCSI HDDs compatible with MACINTOSH systems? In general, all IBM SCSI HDDs are compatible with MACINTOSH. MACINTOSH SCSI compatibility requires the following options to be selected: - Disable Target Initiated (T.I.) Synchronous Negotiations - Disable Unit Attention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How do I make my UNIX system recognize my new SCSI HDD? To be compatible with many UNIX operating systems, the jumper selecting Disable Target Initiated (T.I.) Synchronous Negotiations must be in place. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Why does my 4 GB 0664 suddenly quit spinning? 0664-CSH/ESH (aka 0664-S20, 4 GB SCSI 5.25") devices can behave erratically under heavy utilization (high operations per second). An example of heavy utilization is compilation of a large program on a high performance UNIX workstation. Multiple reports of the device resetting itself and shutting down the motor have been reported. Operation is always regained by powering down and powering back up. No data is lost. The problem is confined to electronics card EC 483194, which is unfortunately the last EC level used for this product. No corrective action has yet been identified. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Who do I call for warranty questions? Warranty services for the end user of IBM HDDs is the responsibility of the distributor, wholesaler, or retailer from which the unit was purchased. IBM does *not* use a pass-through warranty. IBM warrants the drives only to the party with whom IBM contracts to provide drives. This way, all warranty claims are resolved between the seller and buyer in a particular transaction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Where can I find more technical information? The Storage Products Company of IBM's Storage Systems Division maintains a FAX-back service at (415) 903-0955. Enter item 9999 for a menu of available documents. A dial-up BBS is also maintained with similar documentation and, in addition, current bulletins and SCSI utilities. The BBS phone number is (507) 286-5314 (300-28.8k 8-N-1) through July 17, 1995. After that date, the number will be (507) 253-4112. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . HINTS HINTS HINTS HINTS HINTS HINTS HINTS HINTS HINTS HINTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FAT disk format cannot support partitions larger than 2 GB. This limitation results from the maximum allocation unit of 32KB, and the maximum file allocation table size. A boot partition cannot be larger than 2 GB. PS/2 Models 90 and 95 read part of their system IML code from the system partition of the boot drive. The SCSI controller shipped with these systems is specially designed to be available during system power-up so that the IML code can be read from disk when is it needed. Other controllers, including FutureDomain and Adaptec, do not complete the controller Power On Self Test in time to be ready to get the IML code when the system requires it. These adapters will not function as boot drive controllers, but will work fine as non-boot drive controllers.