All values are in minutes:seconds [hundreds]. Smaller time values
indicate higher speed.
This document and the included charts are copyright 1996 Bahman Moallem.
Any form of redistribution without written permission is prohibited.
Have a look at the screenshots of the 68060 - CyberVision64 Amiga system that was used for these tests. You can view the original 1024x768 images by clicking at the thumbnails of them.
Can you resist? ;)
Follow this link to see additional information about the Amiga that was used for the following tests.
In the fall of 1977 in Austin, Texas the Motorola engineers were
gathered to design a microprocessor to succeed the companie's 8-bit
6800 chip. Some pushed the idea of a 16-bit chip, double the
bandwidth of its predecessor while the more ambitious argued for a
chip with a 32-bit internal architecture. The go-getters group won,
and in July of 1979 the Motorola 68000 CPU was introduced.
Faster, more robust versions of the CPU, including the not so common
68010, 68020 (Amiga 1200), 68030 (Amiga 3000) and 68040 (Amiga 4000)
gave the power hungry user a boost, but the big leap for the Amiga
users come in the form of the 68060.
With its 90 MIPS (millions of instructions per second) and 89286
Dhrystones per second, the 68060 50 MHZ is 260 times faster than the
original 0.35 MIPS 68000 that powered millions of Amiga 500s, 1000s
and 2000s.
Motorola's 060 is a 32-bit processor with a superscalar, pipelined
architecture. It has more than 2 million transitors, using a 0.5-
micron, triple-layer metal, 3.3 volt process in a static design.
Separated in the 060 are instruction caches and data caches;
instruction fetch pipelines (with four-stage instruction processor,
a physically-mapped 8Kbyte instruction cache with four-way self
associative, a virtually mapped, 256-entry branch cache and a FIFO
instruction buffer), and operand execution pipelines (with four-
stage operand processor, a physically mapped 8Kbyte data cache with
4-way interleaving and a 4-entry, 32-bit write buffer).
68060 has the ability to execute more than one instruction per
clock cycle due to two parallel instruction pipelines for integer
instruction and also a third pipeline for floating point
instructions. The chip also has the ability to do branch prediction
to increase the speed even more.
View the complete Motorola 68060 facts sheet here.
The 680x0 family and 68060:
Do you know the word "fast"?CyberStorm MK-II:
CyberStorm MK-II for the A3000(T) A4000(T) from the German company
Phase5 features:
Phase5 releases PowerUp uppgrade specifications.
If you want even more then you need Phase5's Super 'Amiga', A\Box.
Review of the MK-II:
And what about the future:
Phase5 PowerUp uppgrade allows your Amiga to run 680x0 and PPC
native programs at the same time. This was demonstrated at Computer 96
at Köln - Germany.
More information soon.
68030 25 MHZ vs 68060 50 MHZ | ||
---|---|---|
68030 25 MHZ | 68060 50 MHZ | |
Boot [network OFF] | 00:54 | 00:22 |
Scenery Animator, 24bit IFF 1024x768 frame, Yosemite.land, all details ON |
31:32 | 04:24 |
LHA, compressing a text file of 1.6 MB | 00:51 | 00:10 |
LZX, compressing a text file of 1.6 MB | 00:58 | 00:14 |
LHA, uncompressing the above archive | 00:08 | 00:01 [50] |
LZX, uncompressing the above archive | 00:06 | 00:01 [35] |
LHA, complete backup of XDH0: my boot partition, 47 MB, 4771 files, 443 dirs |
54:08 | 10:31 |
Image cataloguing of 49 mainly 800x600 JPG files to one JPG file using ImageFX v1.51 Cover Disk version |
35:19 | 03:39 |
CygnusEditor, CED, replacing 3795 "and"s to "Amiga"s |
00:05 [17] | 00:01 [46] |
DMS, compressing PARBENCH disk in RAD: | 00:50 | 00:12 |
DMS, umcompressing the above file | 00:09 [34] | 00:02 [10] |
MagicNews analyzing and filtering a 2.4 MB UseNet news packet |
03:40 | 00:38 |
IBrowse showing a WEB page containing several JPG and GIF files on a 16bit screen |
00:25 | 00:02 |
Complete rebuild of NetInstall, the internet installer for Amiga |
01:21 | 00:10 |
Notes:
This stops it using some later improvments to the instruction set. ;(
View the AIBB benchmarks, 20 640x200 GIF pictures with a sum of 285 KB.
68060/50 MHZ CyberVision64 vs Amiga 4000 68040/25 MHZ AGA | Compare | 060/50 CV64 | 040/25 AGA | Difference | Put Pixels | 2586883 | 205774 | 12.57 | Draw Lines | 43341 | 1123 | 38.59 | Draw Hor/Ver | 139824 | 1536 | 91.03 | Draw Circles | 112795 | 1155 | 97.66 | Draw Ellipse | 87148 | 1040 | 83.80 | Draw Boxes | 14653 | 138 | 106.18 | Scroll X | 3232 | 44 | 73.45 | Scroll Y | 3235 | 47 | 68.83 | Print Texts | 63715 | 2636 | 24.17 | CON: Output | 1739 | 371 | 4.69 | Open Windows | 539 | 75 | 7.19 | Size Windows | 1022 | 98 | 10.42 | Move Windows | 205 | 28 | 7.32 | Swap Screens | 34 | 480 | 0.07 | Areafill | 34624 | 180 | 192.36 |
---|---|---|---|
68060/50 MHZ CyberVision64 vs 68040/28 MHZ PICASSO II | Compare | 060/50 CV64 | 040/28 Picasso II | Difference | Put Pixels | 2586883 | 541182 | 4.78 | Draw Lines | 43341 | 14364 | 3.02 | Draw Hor/Ver | 139824 | 34128 | 4.10 | Draw Circles | 112795 | 34742 | 3.25 | Draw Ellipse | 87148 | 3046 | 28.61 | Draw Boxes | 14653 | 2148 | 6.82 | Scroll X | 3232 | 442 | 7.31 | Scroll Y | 3235 | 446 | 7.25 | Print Texts | 63715 | 10027 | 6.35 | CON: Output | 1739 | 436 | 3.99 | Open Windows | 539 | 171 | 3.15 | Size Windows | 1022 | 212 | 4.82 | Move Windows | 205 | 79 | 2.59 | Swap Screens | 34 | 24 | 1.42 | Areafill | 34624 | 427 | 81.09 |
---|---|---|---|
Notes:
Larger values indicate higher speed.
Tests are performed on 256 colors screens.
MK-II/CyberVision64 speed tests were performed on 1024x768 screens, whereas 640x480x8 screens were used for testing other Amigas.
MK-II/CyberVision64 performs a tad faster on smaller screens so
for the fun of it I performed the tests on a 640x480x8 screen
on the 68060 Amiga.
Here comes WSpeed's results:
As you can see the combination of MK-II and the excellent CyberVision64 boost the Amiga into an ultra-fast workstation. The Picasso II graphic board is many times faster than the Amiga's native graphic system.
For the fun of it, I ran SystemInfo and grabbed a snapshot of its screen. Have a look at it and observe the comment that it outputed ;)
View the SystemInfo results, one 640x200 GIF pictures.
Real-life performance of a P90/Win95 vs A3000 68060/50:
Pentium 90 MHZ vs 68060 50 MHZ - The booting time | ||
---|---|---|
Pentium 90 MHZ | 68060 50 MHZ | |
Boot [network OFF] | 00:53 | 00:22 |
Notes:
The 68060 Amiga loads 32 user specific programs while it boots resluting in a sum of 57 tasks in the system.
The P90 PC loads 3 very small user specific programs.
Pentium 90 MHZ vs 68060 50 MHZ - Communication programs | ||
---|---|---|
Pentium 90 MHZ | 68060 50 MHZ | |
Non-stop scrolling through the entire MessageBase of a message area in a BBS - 520.415 KB of text |
02:18 | 01:48 |
Notes:
The 68060 Amiga loads 33 user specific programs including term
The P90 PC loads 3 very small user specific programs and Win95's HyperTerminal.
Pentium 90 MHZ vs 68060 50 MHZ - Editors | ||
---|---|---|
Pentium 90 MHZ | 68060 50 MHZ | |
Loading. | 00:02:52 | 00:01:47 |
Saving. | 00:04:37 | 00:00:84 |
Replacing 1654 "Amiga"s with "ABC"s. | 00:02:30 | 00:00:80 |
Selecting All. | 00:01:47 | 00:00:32 |
Scrolling through the entire document using the Page-Down key. |
01:22:30 | 00:51:22 |
Alpha-sorting the entire document. | 00:21:18 | 00:12:98 |
Notes:
All values are in minutes:seconds:hundreds. Smaller time values indicate higher speed.
In order to test how the most popular 32bit text editor for Windows95 stands vs CygnusEditor, the popular editor for the Amiga, I grabbed the shareware copy of UltraEdit-32 version 4.10 from www.windows95.com and installed it in the Pentium-PC.
A 1.9 MB ASCII file was used for the tests.
CygnusEditor lacks an internal sort-block funtion. To alpha-sort a block of text, I use an AREXX script that cuts the marked block, saves it to a temp file in T:, calls my own (old, dirty and not so optimized) sort program that sorts the temp file and then loads it back into CygnusEditor. Still Amiga sorts the file 50% faster than the PC.
As you can see, CygnusEditor is clearly faster than UltraEdit when performing the above operations.
CygnusEditor for Amiga and UltraEdit for Windows95 are 2 very powerful and feature-laden editors but CygnusEditor, like many other Amiga programs, sports an AREXX port. With a few AREXX commands and CygnusEditor you can do things that are virtually impossible to do on a PC.
Even for these tests, the 68060 Amiga was loaded with 33 user specific programs, whereas CygnusEditor was one of them while the P90 PC was loaded with 3 very small user specific programs and of course UltraEdit.
More real-life tests become available soon.
How can a 50 MHZ 68060 be faster than a Pentium 90 MHZ?
To mention a few reasons:
The 68060 executes two instructions at the same time, just like the Pentium and the Pentium Pro. The drawback of the Pentium series is its instruction dependency as they can not execute two instructions per clock all of the time. The instruction dependency of these chips causes it to backtrack and change its execution sequence when one instruction depends on another to finish before it can execute. Motorola 68060 suffers from this a lot less because of its dual pipeline, instruction cache, data cache and branch cache.
The Pentium has 8 32bit integer registers and 8 64bit floating point registers. The 68060 has 16 32bit integer registers and 16 64bit floating point registers.
The 68060 has dual 8K cache, 0.5 micron static cmos allowing for lower heat operation and the ability for the chip to run from a 3.3 volt power supply. The Intel Pentium is based on older 0.8 micron and does not have these features. Also the Pentium requires more transistors on chip to remain compatible with earlier Intel chips which adds to the heat problem. The Pentium has something like 3.5 million transistors with larger 0.8 micron spacing and the 68060 gets the same preformance with 2 million transistors and tighter 0.5 micron spacing. The closer you get the tansistors together and the fewer you have the lower the operating temperature of the chip and the higher the frequency you can run the clock at.
The 68060 is faster even counting in MIPS (not that it matters that much when you compare CPUs with different architecture) and it needs a lot less instructions for the same amount of work. An Intel Pentium 90 MHZ does 72427 Dhrystones/second while the Motorola 68060 50 MHZ sports 89286 Dhrystones per second.
The main strength of the Amiga is in its ultra-fast Operating System , AmigaOS. The 68060 50 MHZ is slightly faster than the Pentium P90 in raw number-crunching operations as it will be proven later in this document but because the Amiga does not have a ball-and-chain OS like those popular ones available for IBM-PCs to deal with, the 68060 50 MHZ Amigas feels almost 2 times faster than the P90 PC in normal real-life operations. I have also worked with a very expensive Win/NT system running on a PentiumPro 166 MHZ - 128 MB 60 ns RAM - Matrox M 4 MB (VRAM) and that system is pretty much comparable to the 68060 - CyberVision64 Amiga, at least if you do real-life multitasking instead of using them as pure rendering engines.
Compressing/uncompressing tests, Pentium 90 MHZ vs 68060 50 MHZ | Packer | Archive size | Pentium 90 MHZ | 68060 50 MHZ | LHA - compress | 6.809.315 | 02:22 | LZX - compress | 5.372.314 | 02:12 | ARJ - compress | 6.736.215 | 02:02 | Zip - compress | 6.803.286 | 02:11 | LHA - uncompress | 6.809.315 | 00:59 | LZX - uncompress | 5.372.314 | 00:55 | ARJ - uncompress | 6.736.215 | 01:04 | Zip - uncompress | 6.803.286 | 01:02 |
---|---|---|---|
Notes:
As you can see in the above tests, Amiga-LZX packs 21% better than latest versions of WinZip and ARJ at almost same time.
PC-LHA is not mentioned in the chart because I can't see anyone using it. Nevermind, it took 03:18 for it to compress (6.829.539 bytes) and 01:15 seconds to uncompress the test directory/archive.
Amiga LHA does pack almost as good as the PC programs but is a tad slower. Amiga LHA is old and has not been updated for a while.
A 68060 version of the Amiga-LZX is available to the registred users (I am afraid that I am not one of them) and should boost the performance of LZX even further.
As you can see in the above chart, both of the Amiga packers uncompress slightly faster than the PC programs.
Performance of the SCSI/EIDE controller and the SCSI/IDE HD affect the test results. While Amiga 3000's SCSI controller was one of the fastest devices available at the time of its introduction, a much faster SCSI expansion is available for the MK-II. The harddrive in the Amiga is not so fast either due to its age.
The P90 PC is equipped with a high performance ASUS P55TP4N motherboard that boosts a very fast PCI IDE controller right on the main bus. The Seagate IDE harddrive in the PC is also much faster than the old Quantum that is sitting in the A3000.
Adding the optional FAST SCSI-II DMA controller with a synchronous transfer on the SCSI bus of up to 10 MB/Sec and a faster harddrive should boost the performance of the 68060 Amiga much more.
The SCSI expansion of the MK-II was *not* used for these tests.
The bottom line is that the above tests are not fair because around 20 MB of data has to be read/written to/from the harddrives and the PC is equipped with a faster controller and a MUCH faster harddrive.
For the fun of it, I copied the test directory to the RAM: unit of the Amiga 3000 and used LZX to pack/unpack data from/to RAM: The result follows:
Compress - 01:45 seconds
Uncompress - 00:19 seconds
Lightwave 3D rendering test, Pentium 90 MHZ vs 68060 50 MHZ | ||
---|---|---|
Pentium 90 MHZ | 68060 50 MHZ | |
Test performed in CeBIT 1996 | 01:30 | 00:56 |
Notes:
Amiga Lightwave rendered the test picture faster than the P90 even if it is not really optimized for the 68060 yet. ;)
From "Amiga Report Online Magazine #5.03 -- March 22, 1997"
[The canonical Lightwave test of a year or so ago put an 060/50 and a P90
on roughly equal footing. -Jason]
An old version 1.4 of PatchReq that I am using here, tends to crash
under 68060. To eliminate this problem I run it this way:
; Downgrade the 060 *temporarily*
C:CX/PatchReq >NIL:
; Back to the real thing ;)
The only version of TinyMeter that I find reliable under 68060 is
TinyMeter version 3.5. All of the other versions, including
the latest version 4.31, cause look ups and other kind of problems.
If you have some memory left on the motherboard of your A3000 and
are facing problems with your harddrive, you better prove this one.
Change the MASK for your partitions to 0xfffffc and boot your
computer. Most of your problems should be gone.
This operation does not require a reformat of the actual partition
but do it at your own risk. YOU HAVE BEEN WARNED.
Snaphshot of the HDToolBox showing information about my boot
partition.
Jay Miner (may he rest in peace), Gerald Carda, Dave Haynie,
Thomas Rudloff, Ralf Schmidt, Olaf Barthel, Thomas Sontowski,
Frank Mariak, Steve Krueger, Carl Sassenrath, Fred Fish ...
And the AmigaOS group:
And anyone else who has been participating in Amiga developmet.
Amiga is a registred trademark of Amiga International.Tips and tricks for 68060 users:
If you are having problems with a few older programs that crash on
your 68060 system, downgrade your 68060 *temporarily*. This should
eliminate the problems. Replace these kind of applications with
newer ones as soon as possible.
C:CPU060 NI ND NS NB
C:CPU060 I D S BOther 68060 resources:
Thanks:
I wish to thank, in no particular order, the following individuals/
companies for bringing this very exciting technology to reality:Copyrights:
This document and the included charts are copyright 1996 Bahman Moallem.
Any form of redistribution without written permission is prohibited.
CyberStorm and CyberVision are registred trademarks of Phase5.
Microsoft and Microsoft Windows 95 are registred tradeark of Microsoft Corp.
CTOOLS and CTOOLS logo are registred trademarks of CTOOLS.
All program names mentioned in this page are either Copyright or Trademark of respective owners.
Copyright © 1995-1997 CTOOLS. Last Modified 1-Jan-97.
For comments, suggestions, questions E-Mail webmaster@ctools.pp.se