gallersd@andorfer.fmi.uni-passau.de (Rainer Gallersdoerfer) (03/28/91)
Hello world. Last week I visited a friend who has got a 600MB SIEMENS hard disk. We installed it with different controllers ( WD7000FAAST2 and AHA1542B ). Surprisingly we measured different seek times with some standard test programs! I don't exactly remember but i think the 'track-to-track' ( average ) times were 3.3ms ( 13.3ms ) for the WD and 6.7ms ( 16.7 ) for the AHA. HOW CAN THIS BE ??? Note: The only things we've changed were the controllers! Thanx in advance, RAINER. ########################################################################## # # # Rainer Gallersdoerfer, Siegfriedweg 6, 8480 Weiden i. d. Opf., Germany # # # # gallersd@andorfer.fmi.uni-passau.de, gallersd@unipas.uucp # # # ##########################################################################
csbrod@immd4.informatik.uni-erlangen.de (Claus Brod) (03/28/91)
gallersd@andorfer.fmi.uni-passau.de (Rainer Gallersdoerfer) writes: >Last week I visited a friend who has got a 600MB SIEMENS hard disk. >We installed it with different controllers ( WD7000FAAST2 and AHA1542B ). >Surprisingly we measured different seek times with some standard test programs! >I don't exactly remember but i think the 'track-to-track' ( average ) >times were >3.3ms ( 13.3ms ) for the WD and 6.7ms ( 16.7 ) for the AHA. >HOW CAN THIS BE ??? Very well possible. I wrote a hard disk test program that is quite something like a standard for the ATARI ST line. In an early version of it, I tested seek times like this: - send the controller a SEEK command - isssue TEST UNIT READY commands to monitor if the seek is complete - issue another SEEK and so on... With older Seagate SCSI drives (ST157N, ST296N), this scheme failed totally. I turned out that these drives are very slow in processing the TEST UNIT READY command and thereby cluttered up the measured results. So it seems very likely that the controllers used have an influence on seek times measured by benchmark programs. Naturally, this says nothing about the real performance of the hard disk, especially because TEST UNIT READY commands are not used in standard driver operations. ---------------------------------------------------------------------- Claus Brod, Am Felsenkeller 2, Things. Take. Time. D-8772 Marktheidenfeld, West Germany (Piet Hein) csbrod@medusa.informatik.uni-erlangen.de Claus Brod@wue.maus.de ----------------------------------------------------------------------
bking@nro.cs.athabascau.ca (Barry King) (04/06/91)
It makes sense to me that the speed of the HBA would affect the seeks times. The processor on the HBA must interpret the SCSI command and fire it off to the drive. If the processor is very fast, theSCSI overhead will be reduced and effective seek times will drop and throughput will increase. I used to run my Conner CP3200F with an Adaptec 1540B in a 386/33MHz ISA box. I now use a DPT PM2001/95 (SmartConnex). The difference in seek times between the two HBAs is almost 1 ms in favour of the DPT. The DPT has a full 16 bit path and uses a 10 MHz 68000. The Adaptec uses an 8 bit Intel processor of some sort. Throughput is also several percentage points better using the DPT. I attribute this to the processor speed and maybe the BIOS on the HBA(???). Barry King ersys!bking@nro.cs.athabascau.ca Edmonton Remote Systems: Serving Northern Alberta since 1982
det@hawkmoon.MN.ORG (Derek E. Terveer) (04/16/91)
ersys!bking@nro.cs.athabascau.ca (Barry King) writes: >I used to run my Conner CP3200F with an Adaptec 1540B in a 386/33MHz ISA >box. I now use a DPT PM2001/95 (SmartConnex). The difference in seek >times between the two HBAs is almost 1 ms in favour of the DPT. [..] >has a full 16 bit path and uses a 10 MHz 68000. The Adaptec uses an 8 > Throughput is also several percentage >points better using the DPT. I attribute this to the processor speed and >maybe the BIOS on the HBA(???). That is a pretty small margin to attribute to anything unless you have a pretty exacting benchmark and a very controlled environment. What kind of test are you using to determine this? The "several percentage points" that you are observing may very well be lower than the resolution of your test... -- Derek "Tigger" Terveer det@hawkmoon.MN.ORG -- U of MN Women's Lax I am the way and the truth and the light, I know all the answers; don't need your advice. -- "I am the way and the truth and the light" -- The Legendary Pink Dots