blgardne@esunix.UUCP (Blaine Gardner) (11/23/88)
I've got one of the first A2090A controllers with a Seagate ST4096
hooked up to the ST506 interface. Diskperf consistantly returns maximum
read speeds of 200K to 220K.
This seems awfully slow for a DMA controller, especially since a
couple of friends with the old, non-DMA, CLtd SCSI controller are
getting 300K to 350K. One has an ST506 (really, a 5M ST506) with a SCSI-
ST506 adapter, and the other has a 65M Seagate SCSI (I don't know the
model number).
Why would a DMA interface with a 28 mS drive be so much slower than a
non-DMA interface with 40 and 85 mS drives?
One possiblity is that I need to change the interleave from 1:1 to
another value. But I was under the impression that the 2090A could handle
data at full speed from the drive.
If I do need to change the interleave, is it possible to just reformat
one partition with a different interleave, or do I have to start from
scratch with Prep? If interleave is the problem, I'd like to make trying
different interleave values as painless as possible.
What about buffered seek? This is a feature of the ST4096, and I
understand that it's supposed to improve drive speed. Does the 2090A
support this feature?
On a slightly different subject, has anyone successfully formatted an
ST506 drive that has more than 8 heads? My ST4096 is a 9 head drive, but
I can only format with 8 heads, so I've got 9 meg going to waste. :-(
I got some e-mail from Dale Luck saying that if the 2090A didn't work
with more than 8 heads, a PROM swap would fix it. Dale offered to get
the PROM to me, but I've been unable to make e-mail contact with him.
Any help with getting the 9th head working would be greatly appreciated!
For the record, this is an A2000, 3M RAM, A2090A, ST4096, WB & KS 1.3.
The 2090A is in the slot nearest the CPU slot, and the 2M ASDG card is
in the second slot.
--
Blaine Gardner @ Evans & Sutherland 580 Arapeen Drive, SLC, Utah 84108
Here: utah-cs!esunix!blgardne {ucbvax,allegra,decvax}!decwrl!esunix!blgardne
There: uunet!iconsys!caeco!pedro!worsel!blaine (under construction)
"Nobody will ever need more than 64K." "Nobody needs multitasking on a PC."ejkst@cisunx.UUCP (Eric J. Kennedy) (11/30/88)
In article <1106@esunix.UUCP> blgardne@esunix.UUCP (Blaine Gardner) writes: >This seems awfully slow for a DMA controller, especially since a >couple of friends with the old, non-DMA, CLtd SCSI controller are >getting 300K to 350K. One has an ST506 (really, a 5M ST506) with a SCSI- ^^^^^^^^^^^^ >ST506 adapter, and the other has a 65M Seagate SCSI (I don't know the Hey, I've got an old, non-DMA, CLtd SCSI controller, with an ST506-type drive and SCSI-ST506 adapter, and I can get 70K on a good day! That's *with* the fast file system! (30-40K with the old file system.) I didn't think speeds in the 300K range were possible. What can I do to speed it up? This is a stock CLtd. 33Meg drive system for the A1000. Thanks, -- --- Eric Kennedy ejkst@cisunx.UUCP
blgardne@esunix.UUCP (Blaine Gardner) (12/05/88)
From article <14015@cisunx.UUCP>, by ejkst@cisunx.UUCP (Eric J. Kennedy): > In article <1106@esunix.UUCP> blgardne@esunix.UUCP (Blaine Gardner) writes: >>This seems awfully slow for a DMA controller, especially since a >>couple of friends with the old, non-DMA, CLtd SCSI controller are >>getting 300K to 350K. One has an ST506 (really, a 5M ST506) with a SCSI- > ^^^^^^^^^^^^ >>ST506 adapter, and the other has a 65M Seagate SCSI (I don't know the > > Hey, I've got an old, non-DMA, CLtd SCSI controller, with an ST506-type > drive and SCSI-ST506 adapter, and I can get 70K on a good day! That's > *with* the fast file system! (30-40K with the old file system.) I > didn't think speeds in the 300K range were possible. What can I do to > speed it up? This is a stock CLtd. 33Meg drive system for the A1000. Here's what I remember from talking with my friends: They are using CLtd's SCSI-DOS 3.0 (a fairly new upgrade, and freely distributable to CLtd owners as I recall), and an interleave of 2:1 (that's a mountlist entry of Interleave = 1). Since I crashed my hard drive a couple of days ago, I took the opportunity (!) to play with interleave. The results were underwhelming. I preped the drive and formatted a 10 cylinder (~700K) FFS partition. I did this with interleave values of 0, 1 and 2. There was no practical difference between any of the interleaves. An interleave of 0 gave max read speeds of 180K to 230K, interleave = 1 produced 200K to 230K, and interleave = 2 yielded 218K and 238K. From this I have to conclude that interleave on the A2090A has virtually no effect on performance. Surprising! The high speeds my friends are getting with the Cltd interface make my nasty, suspicious mind wonder if CLtd might be distributing a (shall we say) "optimized" version of diskperf. Despite the disappointing performance of the ST506 interface on the A2090A, it does seem pretty immune to overscan. I haven't tried diskperf with a HAM picture yet, but switching between my usual 704x470 Workbench screen and the stock 640x200 made no real difference in the speeds. (PS If any hints on speeding up the A2090A and this ST4096 were posted around Thanksgiving, I'd appreciate an e-mail copy. My system lost all news that should have come in during the 4 day weekend.) -- Blaine Gardner @ Evans & Sutherland 580 Arapeen Drive, SLC, Utah 84108 Here: utah-cs!esunix!blgardne {ucbvax,allegra,decvax}!decwrl!esunix!blgardne There: uunet!iconsys!caeco!pedro!worsel!blaine (under construction) "Nobody will ever need more than 64K." "Nobody needs multitasking on a PC."
jdow@gryphon.COM (J. Dow) (12/10/88)
In article <1135@esunix.UUCP> blgardne@esunix.UUCP (Blaine Gardner) writes: >From article <14015@cisunx.UUCP>, by ejkst@cisunx.UUCP (Eric J. Kennedy): >> In article <1106@esunix.UUCP> blgardne@esunix.UUCP (Blaine Gardner) writes: >>>This seems awfully slow for a DMA controller, especially since a >>>couple of friends with the old, non-DMA, CLtd SCSI controller are >>>getting 300K to 350K. One has an ST506 (really, a 5M ST506) with a SCSI- >> ^^^^^^^^^^^^ >>>ST506 adapter, and the other has a 65M Seagate SCSI (I don't know the >> >> Hey, I've got an old, non-DMA, CLtd SCSI controller, with an ST506-type >> drive and SCSI-ST506 adapter, and I can get 70K on a good day! That's >> *with* the fast file system! (30-40K with the old file system.) I >> didn't think speeds in the 300K range were possible. What can I do to >> speed it up? This is a stock CLtd. 33Meg drive system for the A1000. > >Here's what I remember from talking with my friends: They are using >CLtd's SCSI-DOS 3.0 (a fairly new upgrade, and freely distributable to >CLtd owners as I recall), and an interleave of 2:1 (that's a mountlist >entry of Interleave = 1). > >Since I crashed my hard drive a couple of days ago, I took the >opportunity (!) to play with interleave. The results were underwhelming. >I preped the drive and formatted a 10 cylinder (~700K) FFS partition. I >did this with interleave values of 0, 1 and 2. There was no practical >difference between any of the interleaves. An interleave of 0 gave max >read speeds of 180K to 230K, interleave = 1 produced 200K to 230K, and >interleave = 2 yielded 218K and 238K. > >From this I have to conclude that interleave on the A2090A has virtually >no effect on performance. Surprising! > I can think of two possible reasons for this lack of whelming performance changes while diddling interleave. Note that the "mountlist" interleave will make no difference on FFS disks. To effect a change you must hard format the drive. A soft format will do zippo. The other possible reason is continued use of OLDFS. It will show little improvement with interleave changes and may actually slow down as you get to smaller interleaves. This is because it makes only 512 byte calls to the device driver. (FFS is smart enough to make BIG reads where appropriate.) WIth the HardFrame (should be same with 2090(a)) there is a very dramatic performance change going frlom 2:1 to 1:1 on a SCSI drive. (ST277N, Mini 8245S, or Adaptec controllers on ST506 disks. I;ve tried 'em all.) -- Sometimes a bird in the hand leaves a sticky deposit. Perhaps it were best it remain there in the bush with the other one. {@_@} jdow@bix (where else?) Sometimes the dragon wins. Sometimes jdow@gryphon.CTS.COM the knight. Does the fair maiden ever {backbone}!gryphon!jdow win? Surely both the knight and dragon stink. Maybe the maiden should suicide? Better yet - she should get an Amiga and quit playing with dragons and knights.
blgardne@esunix.UUCP (Blaine Gardner) (12/12/88)
From article <9472@gryphon.COM>, by jdow@gryphon.COM (J. Dow):
< In article <1135@esunix.UUCP> blgardne@esunix.UUCP (Blaine Gardner) writes:
<>Since I crashed my hard drive a couple of days ago, I took the
<>opportunity (!) to play with interleave. The results were underwhelming.
<>I preped the drive and formatted a 10 cylinder (~700K) FFS partition. I
<>did this with interleave values of 0, 1 and 2. There was no practical
<>difference between any of the interleaves. An interleave of 0 gave max
<>read speeds of 180K to 230K, interleave = 1 produced 200K to 230K, and
<>interleave = 2 yielded 218K and 238K.
<>
<>From this I have to conclude that interleave on the A2090A has virtually
<>no effect on performance. Surprising!
<>
<
< I can think of two possible reasons for this lack of whelming performance
< changes while diddling interleave. Note that the "mountlist" interleave
< will make no difference on FFS disks. To effect a change you must hard format
< the drive. A soft format will do zippo.
The low-level or hard format on the A2090A is the Prep command, right?
I did redo the Prep (res0: mountlist entry), and FFS partitions both
with the different (matching!) interleave values.
The Diskperfa test were all done on the FFS partition.
If Prep is not sufficient to do the "hard" format, how do I go about it
with the A2090A and a ST506 drive?
--
Blaine Gardner @ Evans & Sutherland 580 Arapeen Drive, SLC, Utah 84108
Here: utah-cs!esunix!blgardne {ucbvax,allegra,decvax}!decwrl!esunix!blgardne
There: uunet!iconsys!caeco!pedro!worsel!blaine
"Nobody will ever need more than 64K." "Nobody needs multitasking on a PC."space@sns.UUCP (Lars Soltau) (12/12/88)
In article <9472@gryphon.COM> jdow@gryphon.COM (J. Dow) writes: >I can think of two possible reasons for this lack of whelming performance >changes while diddling interleave. Note that the "mountlist" interleave >will make no difference on FFS disks. To effect a change you must hard format >the drive. A soft format will do zippo. Could you explain the difference between hard and soft format? I mean, how do I hard format an SCSI device? Via prep, which does take much longer with an SCSI HD than with an ST506 HD? -- Lars Soltau UUCP: ...uunet!unido!sns!space BIX: -- no bucks -- Here's looking at you, kid! -- the Medusa
dak@ut-emx.UUCP (Donald A Kassebaum) (12/13/88)
Ten cylinders is a small partition. Running Diskperfa will fragment that pretty fast. The test goes from smallest to largest block size leaving the partition fragrmented. The 2090 really shows its speed when it can use DMA to read/write multiple blocks to disk. With 30k bytes records and a fresh clean partition, it can lay 17 block on a track in one operation with the Fast File System. When the partition get fragmented, FFS can't find contiguous block, it is forced to do more operations. With 700k, you might be doing single block read/writes with will be about the smae speed as the old file system. Give your partitions some space (10 or 20 meg) and try again. Dak
kallaus@leadsv.UUCP (Jerry Kallaus) (12/14/88)
In article <1154@esunix.UUCP>, blgardne@esunix.UUCP (Blaine Gardner) writes: ... >From this I have to conclude that interleave on the A2090A has virtually >no effect on performance. Surprising! > The Amiga 2090 User's Guide has this to say about interleave: (on page 22) Interleave ....should be "0" for ST506 drives.... (and on page 38, under Specification/Performance/HardDisk ST506) Sector Interleave 1:1 I suspect that the mountlist interleave is ignored for an ST506 device. -- Jerry Kallaus {pyramid.arpa,ucbvax!sun!suncal}leadsv!kallaus (408)742-4569 "Funny, how just when you think life can't possibly get any worse, it suddenly does." - Douglas Adams
jesup@cbmvax.UUCP (Randell Jesup) (12/15/88)
In article <183@sns.UUCP> space@sns.UUCP (Lars Soltau) writes: >Could you explain the difference between hard and soft format? I mean, how do >I hard format an SCSI device? Via prep, which does take much longer with an >SCSI HD than with an ST506 HD? "Hard" format (also known as "low-level" format) is what decides which sectors will be in what positions on a track (interleave). On a SCSI drive, hard format also maps out any bad sectors on the drive, so you'll never see them. This requires testing the entire disk, which can take a while. ST-506 disks don't do this, so a low-level format in prep only actually modifies the first 2 cylinders, where prep stores it's bad block info and first partition info. "Soft" format (or high-level format, or AmigaDos format) is what sets up a disk structure on a partition of a drive. It may also (if you don't use the "quick" option) test every sector of the partition by reading and writing it. You cannot change the interleave with a soft format, since interleave applies to the whole disk. -- You've heard of CATS? Well, I'm a member of DOGS: Developers Of Great Software. Randell Jesup, Commodore Engineering {uunet|rutgers|allegra}!cbmvax!jesup
jim@b11.INGR.COM. (Jim Levie ) (12/16/88)
In article <5507@cbmvax.UUCP> jesup@cbmvax.UUCP (Randell Jesup) writes: > > "Hard" format (also known as "low-level" format) is what decides which >sectors will be in what positions on a track (interleave). On a SCSI drive, >hard format also maps out any bad sectors on the drive, so you'll never see >them. This requires testing the entire disk, which can take a while. ST-506 >disks don't do this, so a low-level format in prep only actually modifies >the first 2 cylinders, where prep stores it's bad block info and first >partition info. > > "Soft" format (or high-level format, or AmigaDos format) is what >sets up a disk structure on a partition of a drive. It may also (if you don't >use the "quick" option) test every sector of the partition by reading and >writing it. You cannot change the interleave with a soft format, since >interleave applies to the whole disk. > Well lets see now Randell has essentially said: 1) SCSI maps out bad blocks when the low-level formatting is done and 'prep' does this. 2) 'prep' only sets up the first 2 cylinders "where bad block info ..." is stored for ST506 drives and prep wants the user to enter the bad blocks in terms of head, track, and BFI. 3) The DOS format can do a read and write of every sector. The question is does format add to the badblock information for SCSI or ST506 drives? It would seem to me that one of the things either format or some other utility should do is to find and report badblocks that are present on hard disks and also update the bad block data on the drive. When the bad blocks are reported they should be in a form suitable for input to "prep' to avoid having to scan for them in the future. -- =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+ Jim Levie REMTECH Inc Huntsville, Al The opinions expressed above are just that. Ph. (205) 536-8581 email: uunet!ingr!b11!jim
cs161agc@sdcc10.ucsd.EDU (John Schultz) (12/16/88)
In article <5379@leadsv.UUCP> kallaus@leadsv.UUCP (Jerry Kallaus) writes: >Interleave ....should be "0" for ST506 drives.... >(and on page 38, under Specification/Performance/HardDisk ST506) >Sector Interleave 1:1 >I suspect that the mountlist interleave is ignored for an ST506 device. >Jerry Kallaus {pyramid.arpa,ucbvax!sun!suncal}leadsv!kallaus Read further: interleave is specified as x:y = x-y, so 1:1 = 0, 2:1 = 1, 3:1 = 2, etc. Just installed a 251-1 today with a 2090A, cool reboot times, looks quite snappy so fa'. John Schultz ...now studying brain waves at Scripps with an Amiga...
jesup@cbmvax.UUCP (Randell Jesup) (12/17/88)
In article <328@b11.INGR.COM.> jim@b11.UUCP (Jim Levie ) writes: >1) SCSI maps out bad blocks when the low-level formatting is done and 'prep' > does this. Yes, and in fact you don't even normally have to enter them, since the SCSI controller should map them out all by itself (that's why it takes so long - it's testing every sector). >2) 'prep' only sets up the first 2 cylinders "where bad block info ..." > is stored for ST506 drives and prep wants the user to enter the bad > blocks in terms of head, track, and BFI. Quite true. It reserved 2 cylinders for SCSI drives as well. >3) The DOS format can do a read and write of every sector. Will do so, unless you use the 1.3 QUICK option. >The question is does format add to the badblock information for SCSI or ST506 >drives? It would seem to me that one of the things either format or some >other utility should do is to find and report badblocks that are present on >hard disks and also update the bad block data on the drive. When the bad blocks >are reported they should be in a form suitable for input to "prep' to avoid >having to scan for them in the future. Actually the idea is that when format is formatting and testing the tracks, if the driver notes an error it should automatically map out that sector, and add it to the list of bad blocks at the beginning of the drive. Similar things happen with SCSI. -- You've heard of CATS? Well, I'm a member of DOGS: Developers Of Great Software. Randell Jesup, Commodore Engineering {uunet|rutgers|allegra}!cbmvax!jesup
blgardne@esunix.UUCP (Blaine Gardner) (12/18/88)
From article <5522@cbmvax.UUCP>, by jesup@cbmvax.UUCP (Randell Jesup):
< In article <328@b11.INGR.COM.> jim@b11.UUCP (Jim Levie ) writes:
<
<>The question is does format add to the badblock information for SCSI or ST506
<>drives? It would seem to me that one of the things either format or some
<>other utility should do is to find and report badblocks that are present on
<>hard disks and also update the bad block data on the drive. When the bad blocks
<>are reported they should be in a form suitable for input to "prep' to avoid
<>having to scan for them in the future.
<
< Actually the idea is that when format is formatting and testing the
< tracks, if the driver notes an error it should automatically map out that
^^^^^^
< sector, and add it to the list of bad blocks at the beginning of the drive.
< Similar things happen with SCSI.
Ok, maybe I'm being a bit thick here, but DOES the A2090A actually add
bad blocks found in formatting to the list of bad blocks on ST506
drives?
A little bit off the track: I've made a little script to run Prep, and
enter all the bad block data. I used this when I was messing with
different interleaves on my drive.
There are no comments (I didn't try it, but I figured any comments would
be read by Prep. Was I right or wrong?), but here's a rundown of what's
going on.
0 Select "User Defined"
8 Number of heads (Not 9. Sob, whimper. :-)
1024 Number of cylinders
Default (17) sectors
Default Write Precomp
Default (No) autopark
3 High Cylinder for DH0:
Default (30) buffers
y Mark bad blocks (Yes)
857;1;4151 Cylinder; Head; Offset (BFI) of bad block.
y Correct? (Yes) [Repeat these two lines as needed.]
Last bad block
n No more bad blocks
y Proceed with Prep? (Yes)
Dunno why this is here, but it wouldn't work without it.
=============================Cut Here==================================
0
8
1024
3
y
857;1;4151
y
177;2;8334
y
178;2;8335
y
359;2;9132
y
360;2;9131
y
737;3;9352
y
882;3;8061
y
954;4;6055
y
968;4;2577
y
969;4;2576
y
23;6;8703
y
24;6;4097
y
24;6;6795
y
819;6;10222
y
961;7;581
y
n
y
===============================Cut Here========================
All the blank lines are just a LF, and there needs to be a LF after the
last "y" in the script.
With this script, playing with interleave was as easy as
Mount RES0:
Prep <BadBlocks RES0:
with the script being named BadBlocks.
I hope this is of some use.
--
Blaine Gardner @ Evans & Sutherland 580 Arapeen Drive, SLC, Utah 84108
Here: utah-cs!esunix!blgardne {ucbvax,allegra,decvax}!decwrl!esunix!blgardne
There: uunet!iconsys!caeco!pedro!worsel!blaine
"Nobody will ever need more than 64K." "Nobody needs multitasking on a PC."