[Search for users]
[Overall Top Noters]
[List of all Conferences]
[Download this site]
| Title: | AMIGA NOTES |
| Notice: | Join us in the *NEW* conference - HYDRA::AMIGA_V2 |
| Moderator: | HYDRA::MOORE |
|
| Created: | Sat Apr 26 1986 |
| Last Modified: | Wed Feb 05 1992 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 5378 |
| Total number of notes: | 38326 |
472.0. "Hard Disk Benchmarks" by LEDS::ACCIARDI () Tue Apr 21 1987 21:57
Got this off of Plink... Haven't studied it with a magnifying glass yet,
but it looks like the Pal Jr. leaves the others for dead. Interestingly
enough, the Pal Jr. is supposedly using the SCSI/DMA controller designed
by Commodore. Also in the rumormill, ASDG and Comspec are both planning
to market Amiga hard drives. Comspec claims that their product
will be able to boot Kickstart and Workbench off the hard disk under
1.2. ASDG claims that their combo will be 10 x faster than PAL.
Where will this all end?? On the bright side, six months ago, all
there was for the Amiga was slow Microforges and discontinued Tecmars.
Now it looks like things are heating up, with five vendors shipping
products, and two more announcing. Will CBM join the fray? Stay
tuned.
Ed
-------------------------------------------------------------------------
The Amazing Computing Hard Disk Review
April 16, 1987
This document describes the hard disks tests performed by John Foust and
Sheldon Leemon for the upcoming hard disk comparison in Amazing Computing
magazine. Copyright (c) 1987 John Foust and Sheldon Leemon. This document
is freely distributable in electronic form, but permission must be granted
for paper publishing.
All drives tested were 20 megabyte versions. They are the C Ltd., the
Microbotics MAS-20, the Byte-by-Byte PAL Jr., the Supra 4x4, and the Xebec
9720H.
THIS IS A PRELIMINARY VERSION OF THE RESULTS, SUBJECT TO PROOFREADING. IT
IS OPEN TO PUBLIC AND PRIVATE DISCUSSION. THIS IS NOT THE FINAL VERSION OF
THE REVIEW, AND DOES NOT CONTAIN THE TEXT OF THE REVIEW. MANUFACTURERS
PLEASE CALL JOHN FOUST ANYTIME AT (608) 257-3954 IF NECESSARY.
Test Outline:
-------------
1. Record the time of low level format, if possible.
2. Record the time of an AmigaDOS 'format'.
3. Copy 'diskperfa' to the hard drive, then run it on the freshly
formatted drive, and record the results. 'diskperf' may not be an accurate
measure of Amiga hard disk performance, but has emerged as a common
benchmark.
Only the first run on each machine was recorded, but testing revealed that
the non-read/write 'diskperf' numbers would vary by as much as 3 each, and
the numbers for read/writes would vary as much as 300 bytes per second from
run to run, so take them with a grain of salt.
4. Copy 'run', 'time', and 'arc' to the blank drive. First 'cd' to the
hard disk, and 'assign' c: to dh0: so 'time' can find 'run' on the hard
disk. Then 'time arc' and take the average of ten runs, with the printed
output going to the screen. Divert the 'arc' output to NIL:, and record
the times of another ten runs. Take the average of the ten runs and record
that number.
The public domain 'time' program by Scott Everndon is used to accurately
record the execution time of a program. It is available on the networks.
In general, the numbers recorded here are far too precise for most
purposes, so only consider the first two decimal places of the times here.
In other words, 2.209894 becomes 2.21 for all practical purposes.
The proper 'arc' for this test is version 0.07, which should be 62208 bytes
in length, found on AMICUS #8. The official AMICUS #8 has a representative
set of files and several levels of subdirectories. It has 1544 blocks
used, 214 free. The master was re-copied under AmigaDOS 1.2, so it carries
the improved file structure. (Due to a duplication error, you might have a
bad AMICUS #8, so check these numbers.)
5. Add 'copy' to the hard disk, and 'makedir' to the RAM: disk, then:
ram:makedir dh0:1
time copy df1: 1 all quiet
Then 'makedir' 19 new directories named '2' to '20'. Then
time copy dh0:1 2 all quiet
copying from directory '1' to each other directory in succession, for all
19 directories. Record the times for each copy.
This part of the test hopes to describe 'real world' hard disk use. It
means a lot of thrashing disk access, and a lot of interfacing with
AmigaDOS, which is widely known to be slow. It means copying a lot of
small files with the unusually small buffers of the 'copy' command.
However, it could be interpreted as a fair benchmark. The resulting twenty
numbers reflect the performance of the drive on 'real world' tasks over
time, as the drive approaches fullness. Again, these numbers would vary by
as much as five seconds each in time trials.
6. Again run 'diskperfa' on the roughly seventy percent full drive, and
record the results.
7. Again record the times of loading 'arc', as above in #4.
Diskperf results, sorted
------------------------
Number of files created per second
Blank Full
Pal Jr. 7 2
C Ltd 7 2
Supra 6 2
MAS-20 4 2
Xebec 2 1
Number of files deleted per second
Blank Full
Pal Jr. 25 25
C Ltd 12 6
Supra 12 6
MAS-20 9 6
Xebec 3 2
Number of directory entries scanned per second
Blank Full
Supra 51 52
C Ltd 51 51
MAS-20 49 49
Xebec 48 46
Pal Jr. 44 42
Number of seeks and reads of one byte per second
Blank Full
Pal Jr. 76 72
Supra 55 48
Xebec 49 42
C Ltd 49 40
MAS-20 36 35
Read/write averages of tables below
-----------------------------------
Blank Full
Read Write Read Write
Pal Jr. 50776 14105 48169 13475
Xebec 36313 13163 33826 12052
C Ltd 24387 16455 22848 14487
Supra 22896 18469 22029 17333
MAS-20 16802 10516 15908 9438
Pal Jr.
Buffer Blank Full
Size Read Write Read Write
---- ------------- -------------
512 39125 13443 37449 12725
4096 52428 14169 51400 13653
8192 55775 14324 52428 13797
32768 55775 14483 51400 13724
Xebec
Buffer Blank Full
Size Read Write Read Write
---- ------------- -------------
512 32363 12663 29454 11107
4096 37449 13306 34492 12136
8192 37449 13443 34952 12423
32768 37991 13239 36408 12542
C Ltd
Buffer Blank Full
Size Read Write Read Write
---- ------------- -------------
512 24730 13041 22405 11299
4096 24499 17593 22795 15065
8192 24272 17712 22995 15697
32768 24049 17476 23198 15887
Supra
Buffer Blank Full
Size Read Write Read Write
---- ------------- -------------
512 22995 16192 21845 15420
4096 22995 19275 22028 17712
8192 22995 19275 22028 17955
32768 22598 19134 22215 18204
MAS-20
Buffer Blank Full
Size Read Write Read Write
---- ------------- -------------
512 15240 10485 14169 9362
4096 17246 10527 16384 9429
8192 17246 10527 16487 9497
32768 17476 10527 16591 9463
Specific comments about each drive:
-----------------------------------
C Ltd
-----
1. The low-level format program took about 15 minutes. C Ltd supplies a
quick-format program that will erase the directory information for a
partition in a matter of seconds. The drive has 648 cylinders. This
format left 12 used, 44118 free, for 22 megabytes formatted capacity,
slightly more than the other 20 megabyte drives.
2. AmigaDOS format took 21:40.
3. -above-
4. 'arc' with print: 6.549920
'arc' to NIL: 4.347965
5. Results:
1 187.485907
2 117.251790
3 119.751828
4 123.485196
5 125.418550
6 127.668579
7 128.801907
8 131.585267
9 133.401971
10 136.151996
11 136.985335
12 140.485383
13 141.968722
14 144.785416
15 146.468762
16 148.918797
17 150.502140
18 152.735502
19 151.652155
20 148.302111
Copy left 31039 used, 13091 free, or 70 percent full.
6. -above-
7. 'arc' with print: 6.788964
'arc' to NIL: 4.570886
Microbotics MAS-20
------------------
1. ???
2. AmigaDOS formatting of 612 cylinders took 1574 seconds, or 26:23.
This left 16 blocks used, 41666 free.
3. -above-
4. 'arc' with print: 8.128445
'arc' to NIL: 5.918
5. Results:
1 221.553081
2 167.752382
3 173.452434
4 172.619069
... (not recorded, "went up smoothly," according to Sheldon.)
20 188.635902
6. -above-
7. 'arc' with print: 11.700130
'arc' to NIL: 9.490139
Copy left 30583 used, 11099 free, or 73 percent full.
Xebec 9720H
-----------
1. The low level format and verify took 8:45.
2. AmigaDOS formatting took 613 seconds, or 8:26, leaving 12 blocks used,
41739 free.
3. -above-
4. 'arc' with print: 6.635088
'arc' to NIL: 4.433391
5. Results:
1 221.286478
2 227.136551
3 235.436636
4 239.886699
5 240.153373
6 239.120036
7 242.970070
8 242.553090
9 245.620095
10 245.920105
11 248.686803
12 248.553461
13 251.470163
14 253.603543
15 260.720291
16 263.020309
17 263.086988
18 261.353629
19 259.253605
20 257.203577
6. -above-
7. 'arc' with print: 7.016767
'arc' to NIL: 4.500046
Copy left 31040 used, 10711 free, or 74 percent full.
Supra SupraDrive 4x4
--------------------
1. The low-level format and testing program took 25 minutes total, with
the first two minutes being the low-level format itself, followed by the
verification of the surfaces of the drive.
2. AmigaDOS format took about nine minutes for 599 cylinders, leaving 12
blocks used, 40786 free.
3. -above-
4. 'arc' with print: 6.645076
'arc' to NIL: 4.455040
5. Results:
1 180.019150
2 118.118445
3 120.801840
4 123.051850
5 125.051891
6 124.301869
7 127.035226
8 128.018571
9 129.951923
10 130.685279
11 132.635299
12 134.601988
13 135.551966
14 137.068670
15 138.368691
16 139.818715
17 141.052052
18 141.102047
19 139.518688
20 138.452007
Copy left 31041 used, 9757 free, or 76 percent full.
6. -above-
7. 'arc' with print: 6.645076
'arc' to NIL: 4.455040
Byte-By-Byte PAL Jr.
--------------------
1. The low-level format and testing program
2. AmigaDOS format took 5:07 for 615 cylinders, leaving 12 blocks used,
41670 (last was 40786?) free.
3. -above-
4. 'arc' with print: 5.386714
'arc' to NIL: 3.195864
5. Results:
1 180.069149
2 98.468235
3 102.501633
4 101.001631
5 102.018285
6 101.818298
7 104.601659
8 105.084986
9 107.868346
10 107.885014
11 110.285048
12 110.051707
13 112.601757
14 112.251735
15 114.952146
16 111.885056
17 113.901746
18 111.785054
19 109.585045
20 107.701714
Copy left 31054 used, 10628 free, or 74 percent full.
6. -above-
7. 'arc' with print: 5.533394
'arc' to NIL: 3.333365
| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 472.1 | New Life | WHAMMY::SPODARYK | Binary Throttle | Sat Dec 30 1989 17:51 | 41 |
| Well, this note is quite old, but no sense in creating a new note.
With all the discussion lately about HD/Controller performance, how
about using this note to keep track of the various possibilities
around.
I have used the new DiskSpeed, that Ed uploaded. The archive is
found at tape""::user2:[upload]diskspeed2.lzh. There is a small bug
in the 'save results' section. Very minor, and just a repetition
of "Bytes/s Create", rather than "Create", "Write" and "Read".
Other than that, the program is pretty slick.
It might be a good idea to describe your setup, and post results while
displaying a non-interlaced screen, an overscanned/interlaced screen,
and a overscanned/interlaced/4plane screen. Whatever.
For kicks, here is a test of my RAM: disk. I'm using a A2620 w/2Meg.
The test was done using medium intensity. For fast devices the
difference between med/high is very small ( I've found .5 - 1% ).
DiskSpeed 2.0 - Copyright (c) 1989 by MKSoft Development
Device: ram:
Test Intensity: Med
21 Files/s Create
41 Files/s Open/Close
813 Files/s Scan
43 Files/s Delete
142 Seek/Read
Buffer Size 512 4096 32768 262144
--------- --------- --------- --------- ---------
Bytes/s Create 168298 250874 266834 268973
Bytes/s Write 400949 1431655 2097152 2209345
Bytes/s Read 400849 1410343 2075531 2200290
~Steve - like those 2M/s transfers!
|
| 472.2 | HardFrame/ST277N-1 | WHAMMY::SPODARYK | Binary Throttle | Sat Dec 30 1989 18:09 | 38 |
| This test was performed on a 31M partition of my Seagate ST277N-1.
This drive is rated at 62M formatted, 28ms, and costs approximately
$450 mail-order. It's a half-height, 5 1/2" drive. The drive is
pretty new, so fragmentation shouldn't be a factor.
The controller was a MicroBotics HardFrame. This is a DMA controller,
on a 'hard card' type frame, with room for mounting a 3 1/2" drive.
This is approximately $250, perhaps lower mail-order.
Does CPU make any difference? I'm using a A2620.
I've only included one timing. This is using a overscanned/interlaced
Workbench screen, and displaying (ala superview) a _severely_
overscanned/interlaced/4bitplane picture. The timing for this was
within a fraction of a % of my timings for a non-interlaced workbench
screen, and a overscanned/interlaced workbench.
~Steve
-----------------------------------------------------------------------
DiskSpeed 2.0 - Copyright (c) 1989 by MKSoft Development
Device: dh1:
Test Intensity: Med
8 Files/s Create
23 Files/s Open/Close
181 Files/s Scan
11 Files/s Delete
375 Seek/Read ( lo-res did manage 400 seek/read, only real difference )
Buffer Size 512 4096 32768 262144
--------- --------- --------- --------- ---------
Bytes/s Create 28238 133020 242051 321095
Bytes/s Write 29641 175792 461229 556920
Bytes/s Read 102639 177732 476794 575629
|
| 472.3 | ST-157R-1 and ST-251-0 with OMTI 5528 RLL | FRA04::LUKAS | Dieter, SWAS-Advisory Net&Comms, @FRS | Wed Jan 03 1990 08:20 | 55 |
| I have tested my two Seagate drives using DiskSpeed 2.0.
This was my configuration:
Amiga 500, 68000 CPU, KS 1.3
3 MByte Memory
OMTI 5528 RLL Controller with 2 drives connected (47 MByte + 62 MByte)
ALF2 V2.3 driver software (ALF.device, SoftSCSI_XTC0FEF.device)
Buffer Size: 26
Unit 1: Seagate ST-157R-1, 28 msec, 6 heads, 26 sect./track, 615 cyl.
---------------------------------------------------------------------
DiskSpeed 2.0 - Copyright (c) by MKSoft Development
Device: DH2:
Test Intensity: Med
11 Files/s Create
22 Files/s Open/Close
57 Files/s Scan
26 Files/s Delete
192 Seek/Read
Buffer Size 512 4096 32768 262144
----------- --------- --------- --------- ---------
Bytes/s 28526 142481 203360 232009
Bytes/s 29571 177126 326762 369237
Bytes/s 29576 177146 338364 372827
Unit 2: Seagate ST-251-0, 40 msec, 6 heads, 26 sect./track, 820 cyl.
--------------------------------------------------------------------
DiskSpeed 2.0 - Copyright (c) by MKSoft Development
Device: DH3:
Test Intensity: Med
10 Files/s Create
20 Files/s Open/Close
58 Files/s Scan
12 Files/s Delete
190 Seek/Read
Buffer Size 512 4096 32768 262144
----------- --------- --------- --------- ---------
Bytes/s 27785 124564 170869 227102
Bytes/s 29504 172516 317863 356724
Bytes/s 29498 173035 307838 349525
|
| 472.4 | Fragmentation makes a difference | FSDEV3::JBERNARD | John Bernard 292-2591 YWO/E3 | Mon Jan 08 1990 18:35 | 173 |
| The following is a sample of the disk performances on my system. The system
configuration was a 2000 w/Midget Racer (68020/88881 @12mhz), ASDG Dual Serial
Board, with MACHII V2.5 and two spoolers running in the background. 3MB 16 bit mem.
2 Disk controllers in the system, CLTD KRONOS 2000 and Microbotics Hardframe.
Please note the difference in performance in the two tests made on DH0:.
The first run was on the drive after normal use for a week. Using FREE, the
disk free space was somewhat fragmented, although not severely. I quick formatted
(AmigaDOS) the drive, and restored the drive from the backup drive. The second
performance run was on this newly restored drive. FREE showed the drive to have
contiguous free space and I assume any fragmented files were restored in one piece.
This exercise gained an increase in measured performance as shown below. As purely
subjective input, when I do this to the system drives (DH0,1,2,3) once a month,
the system does appear "faster". Since the degradation in performance takes place
over a month, I really don't notice it.
The following INFO shows the relative disk usage on the drives tested...
Unit Size Used Free Full Errs Status Name
---- ---- ---- ---- ---- ---- ------ ----
DF0: No disk present
DF1: No disk present
DH0: 51M 24390 77776 23% 0 Read/Write SYS_Disk
DH0a: 52M 24390 81083 23% 0 Read/Write SYS B\U
DH1: 51M 53401 49029 52% 0 Read/Write Util
DH1a: 52M 53401 52261 50% 0 Read/Write Util B\U
DH2: 51M 20612 81554 20% 0 Read/Write Work
DH2a: 52M 20612 84842 19% 0 Read/Write Work B\U
DH3: 51M 46398 56032 45% 0 Read/Write Graphics
DH3a: 52M 46398 59264 43% 0 Read/Write Gra B\U
DI0: 20M 13712 28038 32% 0 Read/Write Boot
RAM: 4K 7 0 100% 0 Read/Write RAM DISK
DiskSpeed 2.0 - Copyright (c) 1989 by MKSoft Development
Device: DH0:
Comment Conner 3100 - Cltd KRONOS Controller 51MB Partition 102MB Drive
Test Intensity: Med
7 Files/s Create
16 Files/s Open/Close
136 Files/s Scan
14 Files/s Delete
166 Seek/Read
Buffer Size 512 4096 32768 262144
--------- --------- --------- --------- ---------
Bytes/s Create 61922 123361 181866 227295
Bytes/s Create 68909 156157 320328 398272
Bytes/s Create 84599 226910 386052 448388
DiskSpeed 2.0 - Copyright (c) 1989 by MKSoft Development
Device: DH0:
Comment Conner 3100 - Cltd KRONOS Controller Restored Disk/Unfragmented Files
Test Intensity: Med
9 Files/s Create
18 Files/s Open/Close
146 Files/s Scan
14 Files/s Delete
169 Seek/Read
Buffer Size 512 4096 32768 262144
--------- --------- --------- --------- ---------
Bytes/s Create 60214 120320 182361 292732
Bytes/s Create 69291 164080 376751 462819
Bytes/s Create 85289 260364 499569 595640
DiskSpeed 2.0 - Copyright (c) 1989 by MKSoft Development
Device: DH1:
Comment Conner 3100 - Cltd KRONOS Controller 51MB Partition
Test Intensity: Med
7 Files/s Create
18 Files/s Open/Close
139 Files/s Scan
15 Files/s Delete
166 Seek/Read
Buffer Size 512 4096 32768 262144
--------- --------- --------- --------- ---------
Bytes/s Create 59865 116306 180279 272523
Bytes/s Create 68688 162392 366965 449640
Bytes/s Create 85136 247710 506163 615207
DiskSpeed 2.0 - Copyright (c) 1989 by MKSoft Development
Device: DH2:
Comment Conner 3100 - Cltd KRONOS Controller 51MB Partition
Test Intensity: Med
9 Files/s Create
20 Files/s Open/Close
144 Files/s Scan
15 Files/s Delete
167 Seek/Read
Buffer Size 512 4096 32768 262144
--------- --------- --------- --------- ---------
Bytes/s Create 60677 121464 186802 295633
Bytes/s Create 68759 166008 398272 488064
Bytes/s Create 84715 261123 511630 631612
DiskSpeed 2.0 - Copyright (c) 1989 by MKSoft Development
Device: DH3:
Comment Conner 3100 - Cltd KRONOS Controller 51MB Partition
Test Intensity: Med
9 Files/s Create
19 Files/s Open/Close
143 Files/s Scan
13 Files/s Delete
167 Seek/Read
Buffer Size 512 4096 32768 262144
--------- --------- --------- --------- ---------
Bytes/s Create 59758 116761 175333 287404
Bytes/s Create 69796 163033 374648 476794
Bytes/s Create 84078 264381 522926 638119
DiskSpeed 2.0 - Copyright (c) 1989 by MKSoft Development
Device: DH0a:
Comment Micropolis 1325/Omti 3527 - Microbotics Hardframe - 104MB Drive 52MB Partition
Test Intensity: Med
11 Files/s Create
19 Files/s Open/Close
134 Files/s Scan
16 Files/s Delete
173 Seek/Read
Buffer Size 512 4096 32768 262144
--------- --------- --------- --------- ---------
Bytes/s Create 27486 122685 210537 267899
Bytes/s Create 28926 172627 389601 515726
Bytes/s Create 68912 174308 409616 475668
|