#amiga 1200 & hard disks
18 messages in this thread
Fred: I cannot see any reason to attribute your slow disk access to MBX 1200 or
RDPrep, can you? What is your "mask value" (see RDPrep after loading your RDB).
It should be 0x7fffffff. And your bufmemtype should be set to "1" for fast ram.
Sorry I didn't get immediately back to you –but I did see both of your
messages, not just the one. –Redmond
I certainly didn't mean to imply any problem with any Microbotic program or
hardware. I was hopeing you would have some ideas of how I can use RDPrep to
speed this thing up.
Mask comes up as 0x7ffffffe. I've tried several values including what you just
recommended with no help. (0x7ffffff#) where # = c or d results is worse
results. btw…I do reboot after making changes.
I noticed something that might be strange…RDPrep shows RDBLow=0 and RDBHi=509
???
I have reselect off but tried with it on and no different.
HILun, HIDrive and HiId are all YES (this is the only HD in my 1200)
I have hypercache but a cache is really only going to help when you need to
read a file again. In fact a 1 meg hypercache does not help if I copy a 1 meg
file from dh2: to ram: 2 or more times. (a 2 meg or bigger cache does help but
I wwuld think a 1 meg cache should help some)
I have 2 partitions 20,80,20 megs each. Using addbuffers of 128 for each, only
dh1: shows any improvements with DiskSpeed tests. The other 2 need at least
twice as much before DiskSpeed shows some improvements.
what's strange is dh1: is 99% full and dh2: is 99% FREE yet DiskSpeed reports
dh1: as faster than dh2: when both have 128 buffers!
thanks!
This is getting a bit strange… I have exactly the same symptoms as you,
except my config is a bit different.
I have the M1230XA 50Mhz with '882 & 8M ram, and I have 4 partitions on my
ST9144A ~14,86,11,11, and my second partition is also the only partiition
that gets any kind of benefit from changing any of the partition settings.
(I wonder if it's the size of the partition, or the positition… Hmmmm,
may have to get me set of backup disks out, and repartition everthing.)
I use hypercache also, but that doesn't help writes at all <130Ks.
Redmond, I'm not saying anything is wrong with either RDPrep, or the
M1230XA, I just wnat to know how to get this disk to go faster! I know it
can, because I can trick it into going about 900Ks the same way Fred does
(by issuing a SCSI enquire) for the length of a series of disk
accesses…
btw, I can't change my buffers to fast ram only 🙂 I tried that, and I
thought I killed my Ami! After a while, it dawned on me that fast ram
isn't mapped, under ADos 3.0, until after the system reads the RDB and
mounts the disk. And the 1200 does not recover to well from that
condition!
Bob Comer — Flying high and fast with my A1200HD'030 on Autopilot!
From Cheyenne, Wyoming U.S.A.
Yesterday afternoon my 80 meg dh1: (99% full) (128 buffers) Diskspeed reported
500k+/sec reads (Writes about 150k/sec) while my 20 meg DH2: (99% FREE) (128
buffers) remained at 180k/sec reads and 140k/writes!!
But last night for no apparent reason, my DH1: was only getting 180k/sec reads
and Writes were now up to 400k/sec!!! I even increased buffers (addbuffers)
from 128 to 500 with no changes!
I wonder if we just don't have a batch of bad 1200's???
later!
I don't think it's my 1200, as I got about 400Ks with my original 40M
Seagate.
Maybe, the 120M Seagate is the problem… (I hope not, and it works
perfectly in every other way.)
Bob Comer — Flying high and fast with my A1200HD'030 on Autopilot!
From Cheyenne, Wyoming U.S.A.
Bob: if you set BufMemType to "1" it'll take whatever ram type is available and
thus using AutoXA, the fastram will be available to your buffers. Give that a
try. Also: dump a mountfile and upload it for us to look at. Best, –Redmond
I forgot about the old Mountlist BufMemType, sorry.
(I've gotten spoiled by things like RDPrep, and HDToolBox!)
Here's my mountlists,
/* RigidDiskBlock. */
ST9144A 04.0: disk = scsi.device
Unit = 0
BytesperBlock = 512
Cylinders = 980
Heads = 15
BlocksPerTrack = 17
CylinderBlocks = 255
RDBlow = 0 ; RDBhi = 509
MinCyl = 2 ; MaxCyl = 979
Interleave = 1
HiLun = TRUE
HiID = TRUE
HiDrive = TRUE
Reselect = FALSE
#
/* Partition. */
DH0: device = scsi.device
Unit = 0
Flags = 0x0
Surfaces = 1
BlockSize = 512
BlocksPerTrack = 255
Reserved = 2
Interleave = 0
LowCyl = 2 ; HighCyl = 114
PreAlloc = 0
Buffers = 20
BufMemType = 1
DOSType = 0x444f5301
MaxTransfer = 130560
Mask = 0x7fffffff
Mount = 1
/*! Bootable = 1 */
BootPri = 1
#
/* Partition. */
DH1: device = scsi.device
Unit = 0
Flags = 0x0
Surfaces = 1
BlockSize = 512
BlocksPerTrack = 255
Reserved = 2
Interleave = 0
LowCyl = 115 ; HighCyl = 810
PreAlloc = 0
Buffers = 20
BufMemType = 1
DOSType = 0x444f5301
MaxTransfer = 130560
Mask = 0x7fffffff
Mount = 1
/*! Bootable = 0 */
BootPri = -128
#
/* Partition. */
DH2: device = scsi.device
Unit = 0
Flags = 0x0
Surfaces = 15
BlockSize = 512
BlocksPerTrack = 17
Reserved = 2
Interleave = 0
LowCyl = 811 ; HighCyl = 898
PreAlloc = 0
Buffers = 20
BufMemType = 1
DOSType = 0x52455356
MaxTransfer = 130560
Mask = 0x7fffffff
Mount = 1
/*! Bootable = 0 */
BootPri = -128
#
/* Partition. */
DH3: device = scsi.device
Unit = 0
Flags = 0x0
Surfaces = 1
BlockSize = 512
BlocksPerTrack = 255
Reserved = 2
Interleave = 100
LowCyl = 899 ; HighCyl = 979
PreAlloc = 0
Buffers = 20
BufMemType = 1
DOSType = 0x444f5301
MaxTransfer = 130560
Mask = 0x7fffffff
Mount = 1
/*! Bootable = 0 */
BootPri = -128
#
Bob Comer — Flying high and fast with my A1200HD'030 on Autopilot!
From Cheyenne, Wyoming U.S.A.
Bob: Your Maxtransfer setting is too high for an IDE drive; it should be 65024
or less. Give it a whirl and let me know what happens. Note that this limit is
one that results from IDE firmware, not anything MicroBotics imposese on you.
–Redmond
Well, I tried your maxtransfer settings, and I have something for you to
ponder… (I know Microbotics is not at fault, but any insight from
anyone is valuable, and I think I'm getting close to the answer!
I'll post my scsi speed and diskspeed at the end of this message, as it
might help confirm what I'm saying, and maybe someone will have the
answer!
When I set my maxtransfer to the suggested 65024, writes and creates in
Diskspeed jumped dramatically for me in all my partitions, (Though DH1: is
still the fastest for some reason.) However, reads, are still abysmal,
still only 180Ks. Could the IDE driver in the 1200 be requesting a bigger
read block than the above size, and could the seagate drive bump into s
slower mode when it get's a request like this? (I have none of the file
corruption problems as on the Maxtor's, so my drive handles bigger blocks,
just not so fast…)
It seems to me, when looking at the diskspeed numbers and comparing them
to older ones (reads and creates were a lot about 120Ks with the
maxtransfer higher,) that maxtransfer only covers transfers one way, ie.
writes to the drive.
If this is so, is there a way to limit the amount of the read request?
The SCSISPeed results are much better all the way around now, but I don't
the results to real world like when copying files.
MKSoft DiskSpeed 4.2 Copyright 1989-92 MKSoft Development
——————————————————-
—–
CPU: 68030 AmigaOS Version: 39.106 Normal Video DMA
Device: DH1: Buffers: 140
Comments: DiskSpeed 4.2 Started from Project Icon
CPU Speed Rating: 2722
Testing with a 262144 byte, MEMF_FAST, LONG-aligned buffer.
Create file: 196191 bytes/sec | CPU Available: 84%
Write to file: 421505 bytes/sec | CPU Available: 75%
Read from file: 189821 bytes/sec | CPU Available: 89%
Testing with a 262144 byte, MEMF_CHIP, LONG-aligned buffer.
Create file: 194878 bytes/sec | CPU Available: 74%
Write to file: 404735 bytes/sec | CPU Available: 55%
Read from file: 190650 bytes/sec | CPU Available: 84%
Average CPU Available: 77% | CPU Availability index: 2096
MKSoft ScsiSpeed 4.2 Copyright 1989-92 MKSoft Development
——————————————————-
—–
CPU: 68030 AmigaOS Version: 39.106 Normal Video DMA
Device: scsi.device:0
Comments: ScsiSpeed 4.2
CPU Speed Rating: 2718
Testing with a 512 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 316006 bytes/sec | CPU Available: 26%
Testing with a 4096 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 196470 bytes/sec | CPU Available: 86%
Testing with a 32768 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 728359 bytes/sec | CPU Available: 65%
Testing with a 262144 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 1025216 bytes/sec | CPU Available: 54%
Testing with a 512 byte, MEMF_CHIP, LONG-aligned buffer.
Read from SCSI: 292275 bytes/sec | CPU Available: 23%
Testing with a 4096 byte, MEMF_CHIP, LONG-aligned buffer.
Read from SCSI: 196645 bytes/sec | CPU Available: 82%
Testing with a 32768 byte, MEMF_CHIP, LONG-aligned buffer.
Read from SCSI: 720896 bytes/sec | CPU Available: 50%
Testing with a 262144 byte, MEMF_CHIP, LONG-aligned buffer.
Read from SCSI: 970903 bytes/sec | CPU Available: 35%
Average CPU Available: 53% | CPU Availability index: 1441
Bob Comer — Flying high and fast with my A1200HD'030 on Autopilot!
From Cheyenne, Wyoming U.S.A.
Bob: have you eliminated that hypercache software from the test? If not, do so.
Do you have an external floppy connected? Disconnect it (you might have a low
power situation unless you have a better power supply than the one that came
with the 1200) Eliminate all background software thingies and create as simple
an environment as possible to test in. –Redmond
Bob: the scsispeed test is showing the sort of results one would expect from an
IDE drive. But what really counts is what sort of speed are you experiencing in
actual applications useage? –Redmond
Fred: Did you make sure your BufMemType was set at "1". And: have you tried
your system *without* hypercache active? At the console you can do an RDPrepx
-v to get a report of how your system is set up. Either do that to a file or go
to Rdprep and save the mountfile and upload it for us to look at and maybe
it'll suggest something to us. Regards, –Redmond
Bufmemtype is set at 1. Whats strange is addbuffers seems to improve WRITES to
DH1: but not DH2: !!!
I'll send that mountfile to you asap!
thanks!
I'm at work now so I don't have any Diskspeed data to report but…
I changed Maxtransfer to 65024 on all partitions and rebooted.
(btw…supposedly Seagates and Western Digital 2.5" IDE drives didn't need a
lower Maxtransfer value but I'll do anything!"
After reboot I ran Diskspeed but forgot to turn off Hypercache.
Upon running Diskspeed again, my DH1: was getting 600k+/sec reads!
BUT…after checking other partitions DH1: was again down to about
190k/sec reads!
Like Robert Comer reported…my Creates and Writes have jumped up pretty good
except for READS!!! (this was with 128 buffers)
I then tried with buffers set at 3 for each partition then again with buffers
set at 1000 for each partition.
DH0: and DH1: each got better in CREATE/WRITE/READS when buffers got larger
except for DH2: which stayed about the same!
Weird! anyway, I'm compiling some results and I post them asap!
btw…I know Sysinfo isn't a good disk benchmark but if I do a Sysinfo's
scsi_inquiry the following Sysinfo disk speed test reports
over 1meg/sec! What's really odd is I can SEE and HEAR a difference!
later
Maybe we have some sort of benchmarking problem here as well? You might want to
just create a 1 meg file and read and write it and check it "in the real
world". After all: what counts is not what the benchmark tells you but what you
actually experience in use. I can also ask the author of scsi.speed for an
opion once I get your results. –Redmond
Funny you should ask!
I have a 1 meg animation file (Watch.anim) that I use for testing….
I have Hypercache off, buffers at 128 on each partition and maxtransfer is at
65024 for each partition.
From DH1: to DH2: = 13.6 secs first try …. 14.5 secs 2nd try
From DH1: to RAM: = 7.7 secs first try ….. 7.8 secs 2nd try
From DH2: to DH1: = 13.3 secs first try ….. 9.7 secs 2nd try
From DH2: to RAM: = 4.5 secs first try! ….. 4.6 secs 2nd try!
From RAM: to DH1: = 6.8 secs first try ….. 6.8 secs 2nd try
From RAM: to DH2: = 4.3 secs first try ….. 4.3 secs 2nd try!!
I'm doing this from a WShell window which I happen to have a timer in my Wshell
title bar. For some reason DH2: _now_ appears to be doing better! I'm going
to check with Diskspeed now….
Using a 256k Word aligned buffer in fastram, dh2: creates and writes = 140k/sec
reads are 180k/sec
DH1: gets creates and writes about 155k/sec and reads are 185k/sec
later!
The partition differences may be due to fragmentation of the partition. How
full are the respective partitions? –Redmond
here are results from INFO:
Mounted disks: Unit Size Used Free Full Errs Status Name DH0:
20M 41371 1212 97% 0 Read/Write Boot DH1: 80M 158462 6011
96% 0 Read/Write Work DH2: 20M 11208 31120 26% 0 Read/Write
Data
normally dh2: isn't so free…I have moved stuff to the others in order to play
around with RDPREP setting on dh2: without too much risk!
Strange thing is Diskspeed usually reports dh1: as fastest! I'm beginning to
think even Diskspeed isn't so hot in judging disk performance
I ran viewtek in the background rendering a large GIF and then started
Diskspeed
.. DH2: reported about 250k/sec WHILE VT rendered and 160-180k/sec while NOT
rendering the GIF!
right now I copied that 1 meg file from dh2: to ram: and it took 7.5 secs (my
last message to you said it took 4.5 secs!)
opps made a mistake! the above result is with a 1 meg hypercache…. I repeated
it and got 6.2 secs the 1st try and 2.4 secs the 2nd try… now thats twice as
fast as the previous msg!
later!