KRUYSBER@HNYKUN53.BITNET (08/01/89)
I've done some ST testing using the public domain program QINDEX15.PRG, and
came up with the following results:
N: 1040ST
NT: 1040ST with TurboST to speed up BIOS text
ND: 1040ST with TurboDos to speed up disk I/O
NN: 1040ST with multi purpose accessory N_SYSTEM 2.03
M: Mega ST 2 without
MB: Mega ST 2 with Blitter
MT: Mega ST 2 with Blitter and TurboST to speed up BIOS text
MD: Mega ST 2 with Blitter and TurboDos to speed up disk I/O
MN: Mega ST 2 with multi purpose accessory N_SYSTEM 2.03
1: CPU speed (memory/register/divide/shift)
2: DMA read (read a 64K file) floppy drive A | hard disk C | RAM disk G
3: Gemdos I/O (file creation test) floppy drive A | hard disk C | RAM disk G
4: Disk (RPM) (disk speed) floppy drive A | hard disk C | RAM disk G
5: BIOS text (character output)
6: BIOS string (string output)
7: BIOS scroll (text scrolling)
8: GEM draw (drawing a resource)
For 1, 3 (drive A), 5, 6, 7 and 8 a score of 100 means equal to TOS 1.0; a
higher scoren means faster than this specific TOS version.
Both systems used a Atari Megafile 20 harddisk and N_RAMDSK.PRG as RAM disk;
disk size was half of the available memory.
1 2 3 4 5 6 7 8
N 100 193|3201|11824 98| 781| 9073 300|3600|11612 100 100 100 100
NT 100 193|3201|11824 98| 781| 9073 300|3600|11612 314 730 179 126
ND 100 193|3201|11824 98|1735|12416 300|3600|11250 100 100 100 100
M 99 186|3201|11824 98| 781| 4718 300|3600|11612 106 104 135 122
MB 99 186|3201|11824 98| 781| 4718 300|3600|11612 117 110 181 162
MT 99 186|3201|11824 98| 781| 4718 300|3600|11612 314 728 179 125
MD 99 188|3201|11824 97|1735| 7863 300|3600|11250 117 110 181 162
NN 99 193|3004|11588 84| 698| 6376 300|3600|36735 94 92 98 98
MN 98 181|3002|11361 118| 696| 3867 300|3600|36735 108 101 179 118
Conclusions:
1: The Mega appears to be somewhat slower than the 1040 (99 vs 100).
2a: DMA read operations appear to be somewhat slower on the Mega (186 vs 193).
2b: A harddisk is 17 times faster than a floppy; a RAM disk is 3.7 times faster
than a harddisk, so a RAM disk is 63 times as fast as a floppy! (3201/186,
11824/3201, 11824/186).
3a: TurboDos improves Gemdos operations (that's where it is made for in the
first place!). It speeds up operations about 220% for a harddisk and 170%
for a RAM disk; it doesn't appear to have any effect with floppy's.
3b: A RAM disk on a 1040ST is much faster than a RAM disk on the Mega ST! Is a
RAM disk on a Mega ST4 even slower?
4: There is no significant effect of TurboDos or TurboST on drive speeds (which
they shouldn't have!).
5a: A Mega ST in its simplest form is slightly faster concerning BIOS character
output operations than a 1040ST.
5b: The Blitterchip improves BIOS character output about 10%.
5c: TurboST improves BIOS character output 314%!
6a: A Mega ST in its simplest form is slightly faster concerning BIOS string
output operations than a 1040ST.
6b: The Blitterchip improves BIOS string output about 5%.
6c: TurboST improves BIOS string output 730%!!
7a: A Mega ST in its simplest form is 35% faster concerning BIOS scroll
operations than a 1040ST.
7b: The Blitterchip improves BIOS scroll operations about 35%!
7c: TurboST improves BIOS scroll operations about 179% for the 1040ST, whereas
it doesn't seem to have any (additional) effect to the Mega with Blitter.
8a: A Mega ST in its simplest form is 22% faster in drawing GEM resources than
a 1040ST.
8b: The Blitterchip improves GEM resource drawing about 32%.
8c: TurboST improves GEM resource drawing 126% for the 1040ST, whereas it
has a negative (!) effect on the Mega ST with Blitter (125 vs 162).
Comparing the results from NN and MN with the other ones shows that
1: CPU activity is somewhat slowed down.
2: DMA read operations are somewhat slowed down.
3: Gemdos operations are somewhat slowed down (the are slowed down more when
the disk speed is higher).
4: Disk speeds remain the same, except for the case of a RAM disk: the speed
increases about 360%!
5,6,7: BIOS operations are (somewhat) slowed down.
8:GEM drawing is somewhat slowed down.
These results are not surprising, since N_SYSTEM.ACC is a accessory which
installs 11 new systemroutines. These routines take care of all kinds of
things. These measurements are merely an indication of how much such routines
slow a system down (according to these measurements!). The very high disk
speed indicates that one should be carefull about drawing conclusions about
the speed of a program: it's theway you test it that makes all the difference.
(It is known from N_SYSTEM.ACC that it in fact slowes the system down for about
0.1%! You'll get all kinds of information and utilities for this minor 'lack of
speed'. The installed Gemdos routine by N_SYSTEM only speeds up the Dfree()
command, so there is no reason why disk speed should increase!).
Main conclusions:
- QINDEX15.PRG gives the opportunity to compare systems. Only comparisons can
be made, because for instance my harddisk has other partitions than yours and I
use another RAM disk program.
- QINDEX15.PRG should in fact only be used to COMPARE systems.
- TurboST is a fine utility which has its greatest effect on a 1040ST; the way
I use an Atari the benefits for speeding up BIOS text routines does not equate
the slowing down of GEM drawing on the Mega ST.
- TurboDos has its greatest effect on harddisks. It has no effect when you're
using a floopy-based system.
I hope everything is clear, and if not, or your results are completely
different, let me know!
Noud van Kruysbergen
Nijmegen Institute for Cognition research and Information technology
Mail Box 9104
6500 HE Nijmegen
The Netherlands
kruysbergen@hnykun53f-leoe@IFI.UIO.NO (Lars-Erik 0sterud) (08/02/89)
Have you tried to use a Hard-Disk cache-buffer (I use DCACHE)... The DMA and GEMDOS read goes way up (10 to 20 times faster on HD) Lars-Erik 0sterud / Summer & Christmas: / leoe@ifi.uio.no / f-leoe@ifi.uio.no / ____________________/ _______________________/