Setup and Benchmarks of Several ESDI Hard Disks and Controller Cards

The Result - 24 MHz System

Do you remember your first gigabyte hard disk? The first one I had in use was a 3.5 inch Seagate SCSI drive in 1994. As early as 1991 Hitachi manufactured 5.25 inch full high gigabyte drives.

According to the manufacturer, the Hitachi DK516 is a 24 MHz disk drive, but with 45,880 bytes per track and 3,600rpm, the data transfer rate to storage media is 22MHz only. However, 1,2 gigabytes (with 77 sectors per track) storage capacity was outstanding in the year 1991, when a standard desktop disk drive had a capacity of 40 megabytes.

The tested drive has numerous media defects due to longtime usage. Only the WD1009V controller card was able to handle this large amount of defect sectors and to low-level-format and verify the drive.
The following table shows the results for head- mapping-mode with the Hitachi DK516 disk drive

This is an outstanding performance in every respect.

With 24MHz transfer speed the ESDI development ended. In relation to SCSI, drive configuration and drive setup was to complicated. Additionally there are incompatible controller cards and disk drives due to the insufficient support of the interface's data transfer rate. And there are incompatible controller cards and disk drives, which simply refuse to work together for whatever reason. The succeeding storage interfaces put the controller card electronics on the hard disk's drive card. Ever since then the low-level-format and the media defect handling was done by the drive manufacturer. Technicians simply have to plug in the drive at it's interface, at the power cord, and eventually instruct the system to detect the drive when booting. That's it.

Conclusions

What is the most fast ESDI controller card / disk drive combination?

The answer depends on the disk drive's storage capacity. If the drive holds not more than 528,482,304 byte (528MB), I would choose a track buffering card over the cards with cache ram. The most capable track buffering card is the DTC 6282-24 from Datatechnology.

The answer depends on the operating system too. If you use Novell NetWare 3.x or a contemporary Unix-like operating system, like SCO UNIX, SCO XENIX, ISC Unix, ESIX, or Interactive UNIX, the whole storage space of the ESDI disk drive will be accessible in 63-sector-translation mode, even if the disk holds more than 528,482,304 byte (528MB). Again, the DTC 6282-24 would be the best choice.
If you use DOS or OS/2 and the drive has more than 528,482,304 byte, you need either head-mapping-mode or drive split mode to access the whole storage space. Both modes have their drawbacks. Head-mapping-mode slows down the performance and drive split mode may create problems if used with disk maintenance software. The most capable controller card in head-mapping-mode is Western Digital's WD1009V.

Data throughput depends highly on motherboard timing. For the test system (Gigabyte GA486-TA) the combination of DTC 6282-24 controller card and DK515-78 disk drive gets the best results with approximately 2 MB/s, both in burst data transfer mode and in sustained data transfer mode.

Data throughput depends on the system bus characteristics. Micro channel 16bit ESDI controller cards have better throughput than ISA 16bit cards, but overall performance in micro channel systems is inferior, due to the head-mapping translation mode used.
32bit EISA controller reach exceptional high data transfer rates up to approximately 5 MB/s in ISA compatibility mode and up to approximately 20 MB/s in EISA mode. EISA mode burst data transfer rates are measured with COREtest from CORE International, since SpinRite from Gibson Research does not support EISA driver software.

What about the expensive caching controller cards?

In general caching controller cards are slower than the track caching cards. But what makes them slower? Ultrastor controller cards allow to turn off the cache ram but maintain the track buffer at the same time. I've tried them out this way.

Oops! Turning off the cache increases the performance significantly. Even the Ultra 12C, a low performer otherwise, reaches top results. It seems, that the cache management requires time, which adds to the drive's access time, making these controller cards relative slow. In the case of the Ultra 12C at least, I would recommend to use the minimal cache configuration with two 256KB SIMM RAMS (0,5 MB cache total) und turn off the cache when in the drive setup menu.

The behavior of the EISA caching controller cards is hard to judge, since they use operating system dependent software drivers, which are not supported by SpinRite. However, ist is interesting to see, that the Ultra 22C, which uses VRAM like the DTC 6282-24, performs better than the Ultra 22CA, it's successor, which uses SRAM and DRAM, when the cache is turned off.

SCSI versus ESDI hard disk.

At the end of the eighties the first true SCSI hard disks appeared on the market. Here SCSI means the first SCSI standard with 40MHz or 5MB/s data transfer speed. True SCSI hard disks are drives with a SCSI interface on the drive card, unlike those drives connected to a a ST412 or ESDI to SCSI converter card (e.g. the Emulex MD21).

I will test a contemporary Control Data CDC 94191-766 SCSI drive against a Maxtor XT-8760E ESDI drive, 15MHz both. (I would prefer to test a CDC 94196-766, which is the same Control Data drive with ESDI interface, alas I don't have one.) Additionally the Maxtor drive will be tested with an Emulex ESDI to SCSI converter card.

I know, the significance oft this test is limited, but worth to have a look at. Both drives have the same geometry (1632 cylinder and 15 r/w-heads) and in both drives platters spin with 3600rpm. It seems that the Control Data drive has a slightly better head positioning system, thus achieving a shorter access time. However, the overall performance measured in 'sector access velocity' of the ESDI drive combined with a Datatechnology controller card is higher. As compared to the true SCSI and true ESDI drive, the performance of the Emulex drive card is poor.
If speed is the deciding factor, ESDI storage is more performant than SCSI.

Sector Access Velocity Measurements for Other ESDI Systems

In the hypertext help system for SpinRite there are measurements for several more ESDI systems, which show the performance bandwidth of a certain drive in different system environments. Unfortunately there is no information about sector per track count, interleave factor, partitioning of the drive, the operating system or the test system used.