timothym@tekigm2.UUCP (03/24/87)
Hello net.landians; What follows is a list by performance of various computer/controller/harddisk combinations. It is intended for reference only, and does not intend to shape ones mental attitude for or against any one controller or hard disk. Read the numbers, and it should be obvious to you whether your system is performing at the norm, or if it could be improved with a simple modification to the system interleave. A rough showing says that interleave settings are largely depend- ent on controller and processor clock speeds, but more-so the controller. ****************************************************************************** DISKTIME benchmark results sorted by System Performance: Page 1 ****************************************************************************** Computer uP Clk Controller Int Disk Time BPS ?? ------------------------------------------------------------------------------ IBM PC-AT/Intel 80386 16.0 OMTI/8620 1 MAXTOR/4380 29.0 352,678 4 IBM PC-AT/Intel 80386 16.0 WD-1003-WAH 2 MAXTOR/1140 46.1 212,334 Compaq 386 80386 16.0 WD-1003 ?? 2 ST4051 ?? 49.7 205,995 Eltech PC-AT 80286 10.0 Everex 2 ST4038 54.8 186,743 IBM PC-AT/Intel 80386 ?? IBM/WD-1003 2 CMI ?? 55.2 185,339 2 IBM PC-AT 80286 8.0 IBM/WD-1003 2 CMI 56.2 182,206 Compaq 386 80386 8.0 WD-1003 ?? 2 ST4051 ?? 59.3 172,798 IBM PC-AT 80286 8.0 IBM/WD-1003 2 ST4038 60.1 170,269 IBM PC-AT 80286 8.0 IBM/WD-1003 2 ST225 74.2 137,997 IBM PC-AT 80286 8.0 IBM/WD-1003 3 CMI ?? 74.6 137,173 * Compaq Portable 8088 4.8 Adaptec 2010a 2 Micrpls 1302 74.6 137,200 Compaq Deskpro 8086 7.2 Adaptec 2010a 2 Micrpls 1302 76.3 134,128 IBM PC-AT 80286 8.0 IBM/WD-1003 3 CMI ?? 77.1 132,797 * Compaq Deskpro 8086 7.2 Adaptec 2010a 2 Priam V170 77.5 132,129 IBM PC-AT/Intel 80386 ?? IBM/WD-1003 3 Priam 60meg ?? 124,956 * HP Vectra 80286 8.0 WD-1003 ?? 3 ST225 ?? 88.3 116,008 * Compaq Deskpro 8086 7.2 Adaptec 2010a 2 ST225 88.4 115,863 Epson Equity 3 80286 8.0 WD-1003 ?? 3 ST225 ?? 88.9 115,224 * Compaq Deskpro 8086 7.2 Adaptec 2010a 2 ST225 94.7 108,364 AT&T PC6300 8086 8.0 DTC 5120 3 ST225 119.4 85,721 IBM PC-XT 80286 10.0 WD-1002 ?? 3 ST225 119.6 85,597 3 Zenith Z158 8088 8.0 WD-1002 ?? 3 ST225 ?? 77,201 Tandy 1000 8088 4.8 Everex 4 ST225 ?? 158.0 64,802 AT&T PC6300 8086 8.0 WD-1002-WX2 6 Tulin TL240 158.4 64,646 Compaq 386 80386 6.0 WD-1003 ?? 2 ST4051 ?? 173.0 59,204 Compaq Deskpro 8086 4.8 Adaptec 2010a 2 ST225 176.9 57,899 IBM PC-XT 8088 4.8 IBM/WD-1002 7 ST225 ?? 204.5 50,076 Compaq Deskpro 8086 7.2 Adaptec 2010a 2 Priam V170 223.9 45,731 1 AT&T PC6300 8086 8.0 WD-1002-WX2 5 Tulin TL240 343.0 29,819 * Notes: *) Performances shown may be improved by altering interleave. 1) HDC BIOS ROM incompatible, correct ROM improved throughput to 132,000 bps. The Adaptec 2010a will not perform at Int=2 in a Compaq Deskpro, unless the BIOS ROM is modified. 2) Intel iSBC 386AT 80386 motherboard installed. 3) Breakthru 286 Accelerator Card installed. 4) Accomplished by full track buffering on the controller. Other things to consider when using this chart, is that although most of the results are repeatable to a large degree, DOS as written makes it impossible to ensure a 100% repeatability. Therefore gauge results with a 5% margin of uncertainty. Another example of variations like this is the Compaq Deskpro, and the Compaq Portable who's backplane run at the exact same frequency, and should yield an identical transfer rate, yet they show a 2% variation. In this case, both of the machines were tested just after formatting the harddisk with DOS 3.1. An explantion for the variation was found by running the CORE International disk access time analyzer program on both machines. The faster of the systems had the faster hard disk. This variation between disks was unexpected, but indeed affected the overall system throughput. Both disks were within the advertised specifications also. The percentage of used diskspace also affects throughput in that longer seeks are needed during the transfer. Although the benchmark was coded in a manner attempting to minimize this, DOS has a way of hampering the best efforts. If you would like to add your particular system to the list, please E-Mail a list of the machine details as listed above, your name, and the results. The program source is available as well, and may soon be available in 'C' as well as the original Turbo Pascal. -- Tim Margeson (206)253-5240 PO Box 3500 d/s C1-937 @@ 'Who said that?' Vancouver, WA. 98668 {amd..hplabs}cae780!tektronix!tekigm2!timothym (this changes daily)