space@nadia.UUCP (Lars Soltau) (04/12/89)
Tales of Mystery and Imagination ================================ Personae: - Lucky (?) A2000 Owner - His helpful friend - Commodore A2090 SCSI controller - 44MB Rodime hard disk - 82MB Seagate hard disk - Prep First Act. ---------- The protagonist, our lucky A2000 owner, still hoping he will see AMIX one day, feels the need to expand his mass storage from 40MB to 80MB. Enter stage left: Seagate ST296N, 80MB SCSI hard disk, for an amazingly low price. (Let's just say, less than for an ST4096, which is full height.) The lucky A2000 owner rejoiceth and greedily lays his hands on the Seagate hd. All data is copied from the old to the new drive and the old drive is dismounted. Exit stage right: Rodime 44MB SCSI hard disk. With trembling hands the lucky A2000 owner runs his very own hd performance test. This test just opens a big file and reads 512K blocks. The Rodime hd made a quite satisfactory 500KB/s, so the new, bigger drive should be faster. lAo: What! 179KB/s! There must be something wrong! (Derisive laughter in the background...) (Curtain) Second Act. ----------- The Hannover fair. Seagate's booth. lAo: (stubbornly) But 179KB/s is absolutely ridiculous for an SCSI drive! Seagate employee: Well, I've never heard of this controller. What did you say was the name of the manufacturer? Commodore? Hmmmm... The only thing I can think of is that you are using a wrong interleave. lAo: (very astonished) Interleave? I always thought that interleave is not a question with SCSI drives. Seagate employee: Quit thinking. Interleave is configurable with our SCSI drives. Later, in the evening. The lAo is sitting at his desk, desperately prepping and reprepping his shiny new hard disk, with different interleaves. Not the slightest effect on the transfer rate. Finally he picks up the phone and calls a friend. lAo: Please, you must help me! I can't stand it any more. If life's like this, I don't want to live any longer. Helpful friend: Hmmm, I think I know the man that can help you. Bring your HD to me, I'll call him and then we'll see what we can do. Third Act. ---------- Three people are sitting on the floor and staring at a monitor. The hd is connected via the host adapter that was sold together with it to a Commodore (again that ominous name...) PC10. The host adapter's rom is being disassembled, the screen shows ugly things like "mov ax,[byte PTR]ds". After 15 minutes, the mystery of the lost interleave is solved. The rest is routine. The hd is lowlevel formatted on the PC with different interleaves, the data rate on the PC changes. But there: again black clouds loom on the horizon. Prep seems to reformat the drive with a different interleave. Our lAo breaks down, sobbing and crying. Helpful friend: Hey, don't prep the f*ing drive. Just mount your FFS partition and format it. lAo: You think that will work? A quick examination reveals that it does work, and it doesn't. A quick format works, a slow format doesn't. lAo: What the heck, I don't need to slow-format it. But there, he is wrong. Quick-formatted, the drive produces strange r/w errors that vanish when CANCELled. Several executables don't run anymore. Fourth and Final Act. --------------------- Showdown. On one side Prep, armed with a dreadful hddisk.device. On the other side the small and helpless lAo, his only weapon a tiny Aztec db. Prep: Go ahead, punk. Make my day. I'll format your drive MY WAY! lAo: Und bist Du nicht willig, so brauch' ich Gewalt. (Goethe, Erlkoenig) (And if you don't agree, I'll use force.) Let's not be witnesses of this ugly scene. Several deafening detonations later, we dare again look upon the opponents. Prep is lying dead on the ground, shot down with his own hddisk.device twisted out of its hand with the tiny Aztec db. The lAo holds in his hand the Scroll of Wisdom, on which fiery letters state: "SCSI Format Command: 04 00 00 00 XX 00, where XX is the desired interleave." After a lowlevel format with interleave 3, the Seagate hd makes nearly 400KB/s. (Curtain.) -- Lars Soltau UUCP: ...uunet!unido!pfm!nadia!space BIX: -- no bucks --
hugh@censor.UUCP (Hugh D. Gamble) (04/13/89)
In article <345@nadia.UUCP>, space@nadia.UUCP (Lars Soltau) writes: ... > With trembling hands the lucky A2000 owner runs his very own hd performance > test. This test just opens a big file and reads 512K blocks. The Rodime > hd made a quite satisfactory 500KB/s, so the new, bigger drive should be > faster. > > lAo: What! 179KB/s! There must be something wrong! ... > which fiery letters state: "SCSI Format Command: 04 00 00 00 XX 00, where > XX is the desired interleave." After a lowlevel format with interleave 3, > the Seagate hd makes nearly 400KB/s. > > (Curtain.) > -- > Lars Soltau UUCP: ...uunet!unido!pfm!nadia!space BIX: -- no bucks -- The ST296N is an RLL (I think) drive that spins faster than the lower end Seagates (sorry, the spec's at home). With a vanilla B2000 and a Comspec SA2000 SCSI controller (non DMA) the optimal interleave is 5:1. With an A2620, and a new PROM for the SA2000 that takes advantage of the '020, empirical evidence from diskperf (the recent version from FISH 18? [Thanks Fred, and Joanne, of course]) showed that a 4:1 interleave was optimal. 3:1 gave slightly faster reads, but writes slowed way down. With SetCPU FASTROM (it makes a difference [thanks Dave]) this configuration with 4:1 gives a little over 200kB/s. Needless to say, this is a little dissapointing. The upside is that this performance is rock steady no matter how much hi-rez, interlace, overscan screen DMA you've got going. I was assuming that if I put in a 2090A (maybe so I can run AMIX, but forget I said so because it's irrelevent & I'm talking about AmigaDOS/WB here) I could get the drive down to 1:1 and expect performance in the 700kB/s range. The 400kB/s and 3:1 interleave mentioned above seem low to me. Where's the bottleneck? I'm assuming (possibly incorrectly) that the performance rating used by the poster should be in the same ballpark as diskperf. What is the theoretical (for you head game types) and actual (for you hackers) max rate for a 2500 with a ST296N? -- Hugh D. Gamble (416) 581-4354 (wk), 267-6159 (hm) (Std. Disclaimers) hugh@censor, kink!hugh@censor # It may be true that no man is an island, # but I make a darn good peninsula.