Processor comparisons

Discuss Demos / Games / Applications written by Atari coders

Moderators: Mug UK, Silver Surfer, Moderator Team

User avatar
EmpireAndrew
Captain Atari
Captain Atari
Posts: 397
Joined: Fri Jul 15, 2016 5:46 pm
Location: NYC, USA

Re: Processor comparisons

Postby EmpireAndrew » Wed Jul 19, 2017 10:19 am

I've heard two different stories. One around price and one around timescale. The latter sounds the most plausible although price is always a consideration.
1977 VCS Heavy Sixxer (Boxed)
1990 Atari 1040STE, 4MB, UltraSatan, TOS 2.06, TT Touch -> Atari SC1435 Colour CRT Monitor
1991 Atari TT030, 2/64MB, Int 8GB Gigafile SCSI2CF, TOS 3.06, CaTTamaran Accelerator -> Atari TTM195 19" Mono CRT Monitor
1993 Atari Falcon030, 14MB, Int 4GB IDE2SD, TOS 4.04 -> Atari PTC1426 Color CRT Monitor
Amiga, Mac, DOS, Newton, SGI, Sun, NeXTStation and more!

User avatar
calimero
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2064
Joined: Thu Sep 15, 2005 10:01 am
Location: STara Pazova, Serbia
Contact:

Re: Processor comparisons

Postby calimero » Wed Jul 19, 2017 10:55 am

leech wrote:
AtariZoll wrote:I kind of always thought during the 8088 days that the reason the 68000 wasn't chosen instead was that it was very expensive.

8088 was chosen because of CP/M.

CP/M was only proven "OS" for microcomputers back then. Right? Even Apple II need Z80 board to be able to run CP/M software... AutoCad was writen for CP/M in 1981. I think.
using Atari since 1986.http://wet.atari.orghttp://milan.kovac.cc/atari/software/ ・ Atari Falcon030/CT63/SV ・ Atari STe ・ Atari Mega4/MegaFile30/SM124 ・ Amiga 1200/PPC ・ Amiga 500 ・ C64 ・ ZX Spectrum ・ RPi ・ MagiC! ・ MiNT 1.18 ・ OS X

AtariZoll
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2945
Joined: Mon Feb 20, 2012 4:42 pm
Contact:

Re: Processor comparisons

Postby AtariZoll » Wed Jul 19, 2017 11:37 am

calimero wrote:8088 was chosen because of CP/M.
CP/M was only proven "OS" for microcomputers back then. Right? Even Apple II need Z80 board to be able to run CP/M software... AutoCad was writen for CP/M in 1981. I think.

I don't think that it is some big problem to port CP/M to any CPU.
And you are wrong, it is 8080, an 8-bit CPU, with which Z80 is compatible. 8088 is variant of 8086 with 8-bit data bus - like 68008.
Not to mention that PC came with DOS, what certainly has lot of common with CP/M, but is not it.
There is way to stop global warming.

Zarchos
Captain Atari
Captain Atari
Posts: 224
Joined: Wed May 09, 2012 7:38 pm
Location: FRANCE

Re: Processor comparisons

Postby Zarchos » Wed Jul 19, 2017 1:56 pm

I never really understood Motorola with its 68000 series and then PPC and why it all stopped.
Have you got a link to some good infos about that ?
OK the 68040 is the exact opposite of the RISC philosophy, but hey it was really powerful ... I remember reading an article at the time with this title in a French mag, possibly SVM : '68040 : the anti RISC chip' but it was clear the results were clearly there.
Atari 1040 STE+SATAN, 520ST, 800xl, xegs, Amiga 500, 2000 with 68020, Archimedes, RISC PCs + Iyonix, Omega, BBC B, Atom, Electron, ZX 81, Spectrum 48/128/+2/+3, Speccy2010, Russian clones, Sam Coupe, V6Z80P, QL with accelerators, Enterprise 128, Einstein inc 256, Oric Atmos, MSX 1, 2, Thomson MO5, Amstrads inc CPC+, C 16, 64, 128, VG5000, Apple IIGS and more ! Yes I want to create a museum when I retire.

User avatar
calimero
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2064
Joined: Thu Sep 15, 2005 10:01 am
Location: STara Pazova, Serbia
Contact:

Re: Processor comparisons

Postby calimero » Wed Jul 19, 2017 2:36 pm

AtariZoll wrote:
calimero wrote:8088 was chosen because of CP/M.
CP/M was only proven "OS" for microcomputers back then. Right? Even Apple II need Z80 board to be able to run CP/M software... AutoCad was writen for CP/M in 1981. I think.

I don't think that it is some big problem to port CP/M to any CPU.
And you are wrong, it is 8080, an 8-bit CPU, with which Z80 is compatible. 8088 is variant of 8086 with 8-bit data bus - like 68008.
Not to mention that PC came with DOS, what certainly has lot of common with CP/M, but is not it.


I am not sure in which part I am wrong?

Let me explain a little bit more:

in late 1970 and 80s all useful software for microcomputers was available only on CP/M.
When IBM decide to enter microcomputer business two major things was priority: to have software, to have basic
Microsoft could provide both (in dubious way): clone of CP/M that will work on 16bit 8088 (provided by Tim Paterson) coz IBM wanted 16bit CPU for its microcomputer and CP/M have greatest software library for micros.

btw
Porting CP/M to other CPU probably was not problem, but Digital Research did not hurry to port it.
using Atari since 1986.http://wet.atari.orghttp://milan.kovac.cc/atari/software/ ・ Atari Falcon030/CT63/SV ・ Atari STe ・ Atari Mega4/MegaFile30/SM124 ・ Amiga 1200/PPC ・ Amiga 500 ・ C64 ・ ZX Spectrum ・ RPi ・ MagiC! ・ MiNT 1.18 ・ OS X

User avatar
EmpireAndrew
Captain Atari
Captain Atari
Posts: 397
Joined: Fri Jul 15, 2016 5:46 pm
Location: NYC, USA

Re: Processor comparisons

Postby EmpireAndrew » Wed Jul 19, 2017 4:27 pm

Zarchos wrote:I never really understood Motorola with its 68000 series and then PPC and why it all stopped.
Have you got a link to some good infos about that ?
OK the 68040 is the exact opposite of the RISC philosophy, but hey it was really powerful ... I remember reading an article at the time with this title in a French mag, possibly SVM : '68040 : the anti RISC chip' but it was clear the results were clearly there.


The problem with the 040 is that it ran too hot.
They couldn't get the clock speed up on it.
I think the fastest was 40MHz?
1977 VCS Heavy Sixxer (Boxed)
1990 Atari 1040STE, 4MB, UltraSatan, TOS 2.06, TT Touch -> Atari SC1435 Colour CRT Monitor
1991 Atari TT030, 2/64MB, Int 8GB Gigafile SCSI2CF, TOS 3.06, CaTTamaran Accelerator -> Atari TTM195 19" Mono CRT Monitor
1993 Atari Falcon030, 14MB, Int 4GB IDE2SD, TOS 4.04 -> Atari PTC1426 Color CRT Monitor
Amiga, Mac, DOS, Newton, SGI, Sun, NeXTStation and more!

AtariZoll
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2945
Joined: Mon Feb 20, 2012 4:42 pm
Contact:

Re: Processor comparisons

Postby AtariZoll » Wed Jul 19, 2017 4:51 pm

calimero wrote:I am not sure in which part I am wrong?
Let me explain a little bit more:
in late 1970 and 80s all useful software for microcomputers was available only on CP/M.
When IBM decide to enter microcomputer business two major things was priority: to have software, to have basic
Microsoft could provide both (in dubious way): clone of CP/M that will work on 16bit 8088 (provided by Tim Paterson) coz IBM wanted 16bit CPU for its microcomputer and CP/M have greatest software library for micros.
btw
Porting CP/M to other CPU probably was not problem, but Digital Research did not hurry to port it.

Your writing suggested that 8088 was compatible with 8080 - for which was CP/M written originally, and with which Z80 is compatible. But 8088 is not compatible with 8080. Only more similar than 68000 - since same, Intel made both.
I think that you overestimating CP/M - it may be first serious OS, for usage with disks, but such things last not long. As DOS did not last long.
Apple used Z80 extension probably just because CP/M would be too slow with 6502 (as Commodore did with C128), + then was no need to recompile, rewrite all CP/M SW. Since PC used complete new architecture, new CPU, CP/M support was not that big deal - you could not run old CP/M SW on it. They went ahead instead it. And by such decisions like CPU choice never only 1 factor decides, they considered lot of it.
There is way to stop global warming.

User avatar
Hazzardus
Atari freak
Atari freak
Posts: 53
Joined: Sun Jun 21, 2009 10:43 pm
Location: Hove Beach

Re: Processor comparisons

Postby Hazzardus » Wed Jul 19, 2017 5:54 pm

Atarieterno wrote:
EmpireAndrew wrote:
So yes, a Falcon with an 030 (especially with a DSP) could do things a 386 or 486 could not. But try something that takes raw CPU power and watch a 486 or even 386DX40 stomp a Falcon into oblivion...


So everything is very relative, it is not equal to compare processors thinking of Doom or Quake, than in Cubase and Logic.
With very poor microprocessors, rockets have been brought into space, but Intel 386 has not been able to sustain a music recording studio ever.
If someone has succeeded, please: tell me how he did it.


Yeah, it's all relative. There is no way a 68040 is even close to a 386, let alone the P66, for pure compute power, but show me a 386 or even a 486 that came close to what the 68030 did in the Falcon, or even the 68020 in the Macintosh 2, or even the 68000 in the original ST line when it came to music.

The DSP in the Falc and the FPU in the Mac 2, and the direct MIDI in the ST (which the Mac 2 "borrowed") made a huge difference for certain things, but lets be honest, there was no way any of the x68000 processors were similar or even competitive to the x86 architecture for pure computing power.

If they were, The power Mac beige box and the G3 would have been running a motorola chip.

User avatar
calimero
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2064
Joined: Thu Sep 15, 2005 10:01 am
Location: STara Pazova, Serbia
Contact:

Re: Processor comparisons

Postby calimero » Wed Jul 19, 2017 11:42 pm

AtariZoll wrote:Your writing suggested that 8088 was compatible with 8080 - for which was CP/M written originally, and with which Z80 is compatible. But 8088 is not compatible with 8080. Only more similar than 68000 - since same, Intel made both.
I think that you overestimating CP/M - it may be first serious OS, for usage with disks, but such things last not long. As DOS did not last long.


ah, I see.
My point was that IBM choose Intel 8088 because of CP/M and existing software.
In 1980. all useful software for micros was on:
1) Apple DOS (Apple II - 6502)
2) Commodore DOS (Commodore PET - 6502)
3) CP/M (8080)
For IBM it was no brainer: first two was propriety system by Apple and Commodore, third was, hm, "available".
There was no other software library in 1980. for microcomputers!

AtariZoll wrote:Apple used Z80 extension probably just because CP/M would be too slow with 6502 (as Commodore did with C128), + then was no need to recompile, rewrite all CP/M SW.

There was no CP/M for 6502 in 1980.

AtariZoll wrote:Since PC used complete new architecture, new CPU, CP/M support was not that big deal - you could not run old CP/M SW on it. They went ahead instead it. And by such decisions like CPU choice never only 1 factor decides, they considered lot of it.

You could easily port CP/M software to PC-DOS! This was main goal: to have usable software on new IBM microcomputer from day one.
using Atari since 1986.http://wet.atari.orghttp://milan.kovac.cc/atari/software/ ・ Atari Falcon030/CT63/SV ・ Atari STe ・ Atari Mega4/MegaFile30/SM124 ・ Amiga 1200/PPC ・ Amiga 500 ・ C64 ・ ZX Spectrum ・ RPi ・ MagiC! ・ MiNT 1.18 ・ OS X

AtariZoll
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2945
Joined: Mon Feb 20, 2012 4:42 pm
Contact:

Re: Processor comparisons

Postby AtariZoll » Fri Jul 21, 2017 7:57 am

calimero wrote:...
You could easily port CP/M software to PC-DOS! This was main goal: to have usable software on new IBM microcomputer from day one.

I don't see what it has with chosen CPU for PC. CP/M SW runs not on Intel 8086/8088 . It needs to be recompiled.
Then saying that usable CP/M SW was main goal with some new computer ... If that would be main goal, nobody would talk today about IBM PC more than about some Dragon home computer. You can not like IBM and Microsoft, but they made something what market actually missed in those times: architecture and OS, what is future proof. Unlike others did - every new model was incompatible with older ones from same house. IBM showed the right direction, what was followed then in oncoming 16-bit micros more or less - but here need to add that 8-bit CPUs were simply not for that task. Only newer 16 and mostly 32-bit CPUs could be expanded while keeping them compatible with original instruction set, SW.
And again, you can not love Intel (myself not loving that company), but it seems that is is only CPU, which still uses original instruction set, established almost 40 years ago (of course, with plenty of added). Motorola changed it when went on PowerPC. That may be just because 68000 was designed too good, to say so - instruction set was excellent for that time, but later is was limit in further progress, so they needed to go on new model.
RISC was great thing in it's time, when complexity of CPU chips was much lower, and best way to improve speed. But again, it is now not relevant, because new CISC can do same speed as RISC, so execution in minimum cycle counts of complexer instructions.
Technology went in direction of multi core CPUs - clearly visible in case of smart phones. And most of sold in it is not Intel, but ARM based.
Anyone remember Transputer ? It was claimed as next big thing in CPU architecture, but never was success. As I know, only expensive Next workstation was sold with it, and Atari sold couple hundreds of Atari Transputer Workstation - ATW-800.
There is way to stop global warming.

User avatar
Cyprian
Atari God
Atari God
Posts: 1405
Joined: Fri Oct 04, 2002 11:23 am
Location: Warsaw, Poland

Re: Processor comparisons

Postby Cyprian » Fri Jul 21, 2017 9:33 am

Zarchos wrote:
Cyprian wrote:Zarchos, yep, thats obvious.
CPU in Archimedes was 4 times faster than in ST and more than in amiga


It isnt ;-)

can you elaborate that?

And how many cycles takes on Archimedes:
1) move 4 bytes? Best case (source and destination address are word aligned) for 68000 is 20 cycles - "move.l (A0),(A1)"
2) set one pixel on? In case of ST/amiga it is 16 cycles for one bitplane (multiply it by 4, for 16 color mode) - "bset #1,(A0)"
Jaugar / TT030 / Mega STe / 800 XL / 1040 STe / Falcon030 / 65 XE / 520 STm / SM124 / SC1435
SDrive / PAK68/3 / CosmosEx / SatanDisk / UltraSatan / USB Floppy Drive Emulator / Eiffel / SIO2PC / Crazy Dots / PAM Net
Hatari / Aranym / Steem / Saint
http://260ste.appspot.com/

Zarchos
Captain Atari
Captain Atari
Posts: 224
Joined: Wed May 09, 2012 7:38 pm
Location: FRANCE

Re: Processor comparisons

Postby Zarchos » Fri Jul 21, 2017 9:37 am

Cyprian wrote:
Zarchos wrote:
Cyprian wrote:Zarchos, yep, thats obvious.
CPU in Archimedes was 4 times faster than in ST and more than in amiga


It isnt ;-)

can you elaborate that?

And how many cycles takes on Archimedes:
1) move 4 bytes? Best case (source and destination address are word aligned) for 68000 is 20 cycles - "move.l (A0),(A1)"
2) set one pixel on? In case of ST/amiga it is 16 cycles for one bitplane (multiply it by 4, for 16 color mode) - "bset #1,(A0)"


It is about this MIPS idea to give the power of a machine : you can't use that, it doesn't reveal the truth.
Or the assertion is not enough without further details.
If you don't mind 1st reading the comments I wrote in this YT video, please
https://youtu.be/hU23fEL154k
where I explain it all (almost)
and then come back for further questions.

Setting one pixel on (in 8 bit but pixel mode) is done with the STRB instruction.
It takes 4 cycles.
Loading one pixel (in 8 bit but pixel mode) is done with the LDRB instruction.
It takes 4 cycles too.

In fact the real formulaes are 2N and 1S+1N+1I respectively.
It is also the reason why you can avoid losing 1 cycle when there is DMA video reloading (occuring every 16 bytes) because this will be merged with the execution of this instruction.
All this is still complex to me and has been explained on the stardot foum.
I have enhanced some of my routines thanks to the explanations given by Sarah Walker who wrote the Arculator emulator and has an in-depth knowledge of the Archimedes architecture clearly I do not have.
I'll find back the posts and will post them here.
Thanks for your interest.
Xavier.

PS : Are you sure of your figures for the 68000 timings ?
I'd be happy to correct my comments with more exact infos than those I had read from an Amiga fanboy ...
Last edited by Zarchos on Fri Jul 21, 2017 10:08 am, edited 1 time in total.
Atari 1040 STE+SATAN, 520ST, 800xl, xegs, Amiga 500, 2000 with 68020, Archimedes, RISC PCs + Iyonix, Omega, BBC B, Atom, Electron, ZX 81, Spectrum 48/128/+2/+3, Speccy2010, Russian clones, Sam Coupe, V6Z80P, QL with accelerators, Enterprise 128, Einstein inc 256, Oric Atmos, MSX 1, 2, Thomson MO5, Amstrads inc CPC+, C 16, 64, 128, VG5000, Apple IIGS and more ! Yes I want to create a museum when I retire.

User avatar
Cyprian
Atari God
Atari God
Posts: 1405
Joined: Fri Oct 04, 2002 11:23 am
Location: Warsaw, Poland

Re: Processor comparisons

Postby Cyprian » Fri Jul 21, 2017 9:47 am

unfortunately there is no info about CPU cycles pixel manipulation instruction.

There https://docs.google.com/spreadsheets/d/ ... edit#gid=0 I see code:

Code: Select all

; Apply sprite data to screen      
ORR R4,R4,R9,LSL #8      1


does it meant 1 CPU cycle per two instructions OR and LSL?

Zarchos wrote:PS : Are you sure of your figures for the 68000 timings ?
I'd be happy to correct my comments with more exact infos than those I had read from an Amiga fanboy ...

yes, they are 100% correct, and those figures are the best case.
I've just re-verified that in Steem Developer emulator.
Jaugar / TT030 / Mega STe / 800 XL / 1040 STe / Falcon030 / 65 XE / 520 STm / SM124 / SC1435
SDrive / PAK68/3 / CosmosEx / SatanDisk / UltraSatan / USB Floppy Drive Emulator / Eiffel / SIO2PC / Crazy Dots / PAM Net
Hatari / Aranym / Steem / Saint
http://260ste.appspot.com/

Zarchos
Captain Atari
Captain Atari
Posts: 224
Joined: Wed May 09, 2012 7:38 pm
Location: FRANCE

Re: Processor comparisons

Postby Zarchos » Fri Jul 21, 2017 10:12 am

Shifts cost no additional cycles on any instructions, as long as they are a constant.
I make heavy usage of this feature in my routines.
For the story it is Hermann Hauser, Acorn's boss, who asked to have this feature for the ARM, and as he said in a recent interview, he is still very happy he wanted that, because of course it gives a big plus to the ARM, be it for maths or graphical manipulations.

Thanks for the figures : I am sure Galaghan/FT on eab will love my amending my comments of my YT video.
And as he pointed it, not being happy of what I wrote, I'll be happy to stress he kept from telling me I was wrong and the timings on the Amoeba are in fact even worse.
I do not like thieves, in case I didn't tell you, and I told him when he was touting piracy didn't hurt sales on the Amoeba.
Atari 1040 STE+SATAN, 520ST, 800xl, xegs, Amiga 500, 2000 with 68020, Archimedes, RISC PCs + Iyonix, Omega, BBC B, Atom, Electron, ZX 81, Spectrum 48/128/+2/+3, Speccy2010, Russian clones, Sam Coupe, V6Z80P, QL with accelerators, Enterprise 128, Einstein inc 256, Oric Atmos, MSX 1, 2, Thomson MO5, Amstrads inc CPC+, C 16, 64, 128, VG5000, Apple IIGS and more ! Yes I want to create a museum when I retire.

Zarchos
Captain Atari
Captain Atari
Posts: 224
Joined: Wed May 09, 2012 7:38 pm
Location: FRANCE

Re: Processor comparisons

Postby Zarchos » Fri Jul 21, 2017 10:22 am

Now there is something to take into account it is the cost of DMA video refill on the Archie : it is not free, it takes some cycles.
But there is a way to save some cycles.(I mean : to avoid the 'penalty' cycles due to DMA video refill).

There is a also a way to save some cycles by a special rearranging of your ARM instructions .. again a reason why I had to rewrite 8 times my plotting routines, but also my fast filling routines.
See here the comparison 'normal code' vs 'optimised' code :
https://youtu.be/Pjev7O7fKjA
And it is not what a real speed test should be : there are too many cases to test, between the position 0 to 15 from a quadword boundary, and the length of the segment you want to plot.

I have not yet reached my litre of coffee drunk so without my notes I can't even give you an example as I simply do not remember what to do (which I did !) exactly ;-)

And I can guarantee you that nobody never ever used that on the Archie.
It has been discovered by somebody coding for the 3DO.
It is not in the documentation of the ARM by VLSI (If I am wrong, please tell me where it is, I own the 2 editions).
Last edited by Zarchos on Fri Jul 21, 2017 10:32 am, edited 1 time in total.
Atari 1040 STE+SATAN, 520ST, 800xl, xegs, Amiga 500, 2000 with 68020, Archimedes, RISC PCs + Iyonix, Omega, BBC B, Atom, Electron, ZX 81, Spectrum 48/128/+2/+3, Speccy2010, Russian clones, Sam Coupe, V6Z80P, QL with accelerators, Enterprise 128, Einstein inc 256, Oric Atmos, MSX 1, 2, Thomson MO5, Amstrads inc CPC+, C 16, 64, 128, VG5000, Apple IIGS and more ! Yes I want to create a museum when I retire.

AtariZoll
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2945
Joined: Mon Feb 20, 2012 4:42 pm
Contact:

Re: Processor comparisons

Postby AtariZoll » Fri Jul 21, 2017 10:32 am

Cyprian wrote:And how many cycles takes on Archimedes:
1) move 4 bytes? Best case (source and destination address are word aligned) for 68000 is 20 cycles - "move.l (A0),(A1)"
2) set one pixel on? In case of ST/amiga it is 16 cycles for one bitplane (multiply it by 4, for 16 color mode) - "bset #1,(A0)"


You can move 4 bytes faster with movem - of course when moving not only 4 bytes, but much more at once.

68000 CPU is old design, from time when memory speeds were low. That's why RAM access is 500nS by 8 MHz clock. With less than 4 cycles for RAM access, for instance 250nS in 2 cycles, 68000 at 8 MHz would be pretty much faster, some 50% . And RAM in ST is 250nS (together with all delays in logic).
I don't see that we need to compare 68000 with much later design ARM.
My point is that in many cases it is RAM (access) speed what determines speed of some OP (as in case of moving (better word is copy) 4 bytes) .
There is way to stop global warming.

Zarchos
Captain Atari
Captain Atari
Posts: 224
Joined: Wed May 09, 2012 7:38 pm
Location: FRANCE

Re: Processor comparisons

Postby Zarchos » Fri Jul 21, 2017 10:34 am

Yes : Sophie Wilson said that many times in her interviews : eventually the speed of a processor / system is its ability to maximize the usage of memory bandwith.
The team at Acorn had always that in mind when designing all the chips.
Atari 1040 STE+SATAN, 520ST, 800xl, xegs, Amiga 500, 2000 with 68020, Archimedes, RISC PCs + Iyonix, Omega, BBC B, Atom, Electron, ZX 81, Spectrum 48/128/+2/+3, Speccy2010, Russian clones, Sam Coupe, V6Z80P, QL with accelerators, Enterprise 128, Einstein inc 256, Oric Atmos, MSX 1, 2, Thomson MO5, Amstrads inc CPC+, C 16, 64, 128, VG5000, Apple IIGS and more ! Yes I want to create a museum when I retire.

User avatar
Cyprian
Atari God
Atari God
Posts: 1405
Joined: Fri Oct 04, 2002 11:23 am
Location: Warsaw, Poland

Re: Processor comparisons

Postby Cyprian » Fri Jul 21, 2017 10:52 am

Zarchos, ok, how fast Acorn ram is? How many memory slots ARM CPU has per one VBL?
Jaugar / TT030 / Mega STe / 800 XL / 1040 STe / Falcon030 / 65 XE / 520 STm / SM124 / SC1435
SDrive / PAK68/3 / CosmosEx / SatanDisk / UltraSatan / USB Floppy Drive Emulator / Eiffel / SIO2PC / Crazy Dots / PAM Net
Hatari / Aranym / Steem / Saint
http://260ste.appspot.com/

User avatar
calimero
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2064
Joined: Thu Sep 15, 2005 10:01 am
Location: STara Pazova, Serbia
Contact:

Re: Processor comparisons

Postby calimero » Fri Jul 21, 2017 11:24 am

AtariZoll wrote: You can not like IBM and Microsoft, but they made something what market actually missed in those times: architecture and OS, what is future proof. Unlike others did - every new model was incompatible with older ones from same house. IBM showed the right direction,

Just small correction: IBM did not do this (invent or start) but rather Gary Kildall with CP/M years before IBM even decide to enter microcomputer business!

AtariZoll wrote: And again, you can not love Intel (myself not loving that company), but it seems that is is only CPU, which still uses original instruction set, established almost 40 years ago (of course, with plenty of added). Motorola changed it when went on PowerPC.

Major problem today is that you do not have another CPU ISA so you can not know how good (or bad) intel is right now (as you can not compare it to something else).
Thanks to Apple, ARM survived while all others: MIPS, Motorola, Sparc, DEC Alpha, even Intel Itanium go to oblivion because mass production of x86 would crash any competition no matter how good competition is because wintel is zilion times larger.
Today only ARM and IBM Power exist. IBM Power present brand new technology in CPU and Intel usually copy them few years later...

AtariZoll wrote:RISC was great thing in it's time, when complexity of CPU chips was much lower, and best way to improve speed. But again, it is now not relevant, because new CISC can do same speed as RISC, so execution in minimum cycle counts of complexer instructions.

Starting with PentiumPro intel is also RISC. They put microops layer between "40 years old instruction" and implementation in hardware. This allow intel to take best from RISC world and implement in their traditional CISC CPU.

AtariZoll wrote:Technology went in direction of multi core CPUs - clearly visible in case of smart phones. And most of sold in it is not Intel, but ARM based.
Anyone remember Transputer ? It was claimed as next big thing in CPU architecture, but never was success. As I know, only expensive Next workstation was sold with it, and Atari sold couple hundreds of Atari Transputer Workstation - ATW-800.

as far as I know, NeXT never build any computer with INMOS Transputer. NeXT build only MC 680x0 computers.
using Atari since 1986.http://wet.atari.orghttp://milan.kovac.cc/atari/software/ ・ Atari Falcon030/CT63/SV ・ Atari STe ・ Atari Mega4/MegaFile30/SM124 ・ Amiga 1200/PPC ・ Amiga 500 ・ C64 ・ ZX Spectrum ・ RPi ・ MagiC! ・ MiNT 1.18 ・ OS X

AtariZoll
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2945
Joined: Mon Feb 20, 2012 4:42 pm
Contact:

Re: Processor comparisons

Postby AtariZoll » Fri Jul 21, 2017 2:47 pm

OK, I will stop here with this. Really no point for continuing this discussion, when you write some illogical things, and some claims, which contradict to known ones: Gary Kildal, DR, did not design computer. I was talking about computer architecture, design, concept. IBM refused to buy CP/M from him, so no, IBM PC was not intended to run CP/M SW. The truth is that DR lost war, and became insignificant factor in 80-es. Was that war fair ? That's another story, and I really don't want to even think about it.
There is way to stop global warming.

User avatar
Frank B
Atari Super Hero
Atari Super Hero
Posts: 905
Joined: Wed Jan 04, 2006 1:28 am
Location: Boston

Re: Processor comparisons

Postby Frank B » Fri Jul 21, 2017 3:45 pm

Zarchos wrote:Shifts cost no additional cycles on any instructions, as long as they are a constant.
I make heavy usage of this feature in my routines.
For the story it is Hermann Hauser, Acorn's boss, who asked to have this feature for the ARM, and as he said in a recent interview, he is still very happy he wanted that, because of course it gives a big plus to the ARM, be it for maths or graphical manipulations.

Thanks for the figures : I am sure Galaghan/FT on eab will love my amending my comments of my YT video.
And as he pointed it, not being happy of what I wrote, I'll be happy to stress he kept from telling me I was wrong and the timings on the Amoeba are in fact even worse.
I do not like thieves, in case I didn't tell you, and I told him when he was touting piracy didn't hurt sales on the Amoeba.


IMHO the 68k ISA is more elegant than the 32 bit Arm. The quirky conditional execution for each instruction is gone on Arm64.
There are several things on the ARM (32) that make me say, "yuck". The constant loading limitations for instance.
The lack of a divide instruction is constraining too. Not so bad if you're using a constant but with a variable divide? Yuck!
The 68k has far more flexible addressing modes and could clobber the ARM32 CPU for code density. There are 13 addressing modes in the original 68000.
By the time you hit the 020, you're up to 18. They are very very powerful. The arm32 arch has something like 4 addressing modes. Compare the flexibility of the bitfield instructions on the 020 to anything in the ARM. There's no equivalent to bitfield insert or bitfield extract instructions on the ARM. How many 32 bit (!) instructions would be needed to do the equivalent on ARM?

I looked at the ARM and the only thing I liked was the three operand instructions. Everything else seemed like a step backwards compared to the 68k. Sure it's fast but the 040 was faster and the Arm ISA is bugly in comparison. :)

Arm64 renders the 32 bit ISA obsolete and the programmer's model is very different.
My 2 cents. I own 3 archimedes btw. I have an A5000, a Risc PC and an a3000. I used to own an a3020 years ago. I don't dislike the machines. I just prefer the ST, Falcon and the Amiga to it.
Last edited by Frank B on Fri Jul 21, 2017 4:09 pm, edited 1 time in total.

User avatar
simonsunnyboy
Moderator
Moderator
Posts: 4860
Joined: Wed Oct 23, 2002 4:36 pm
Location: Friedrichshafen, Germany
Contact:

Re: Processor comparisons

Postby simonsunnyboy » Fri Jul 21, 2017 4:08 pm

No horizontal blank interrupt? And I remember so many great demos on the A3000 with raster bars.....Brothers In Arm come to mind *oh the memories* :)
Simon Sunnyboy/Paradize - http://paradize.atari.org/ - STOT: http://www.npoi.de/stot/

Stay cool, stay Atari!

1x2600jr, 1x1040STFm, 1x1040STE 4MB+TOS2.06+SatanDisk, 1xF030 14MB+FPU+NetUS-Bee

Jabber: simonsunnyboy@atari-jabber.org

Zarchos
Captain Atari
Captain Atari
Posts: 224
Joined: Wed May 09, 2012 7:38 pm
Location: FRANCE

Re: Processor comparisons

Postby Zarchos » Sat Jul 22, 2017 7:17 am

Frank B wrote:
Zarchos wrote:Shifts cost no additional cycles on any instructions, as long as they are a constant.
I make heavy usage of this feature in my routines.
For the story it is Hermann Hauser, Acorn's boss, who asked to have this feature for the ARM, and as he said in a recent interview, he is still very happy he wanted that, because of course it gives a big plus to the ARM, be it for maths or graphical manipulations.

Thanks for the figures : I am sure Galaghan/FT on eab will love my amending my comments of my YT video.
And as he pointed it, not being happy of what I wrote, I'll be happy to stress he kept from telling me I was wrong and the timings on the Amoeba are in fact even worse.
I do not like thieves, in case I didn't tell you, and I told him when he was touting piracy didn't hurt sales on the Amoeba.


IMHO the 68k ISA is more elegant than the 32 bit Arm. The quirky conditional execution for each instruction is gone on Arm64.
There are several things on the ARM (32) that make me say, "yuck". The constant loading limitations for instance.
The lack of a divide instruction is constraining too. Not so bad if you're using a constant but with a variable divide? Yuck!
The 68k has far more flexible addressing modes and could clobber the ARM32 CPU for code density. There are 13 addressing modes in the original 68000.
By the time you hit the 020, you're up to 18. They are very very powerful. The arm32 arch has something like 4 addressing modes. Compare the flexibility of the bitfield instructions on the 020 to anything in the ARM. There's no equivalent to bitfield insert or bitfield extract instructions on the ARM. How many 32 bit (!) instructions would be needed to do the equivalent on ARM?

I looked at the ARM and the only thing I liked was the three operand instructions. Everything else seemed like a step backwards compared to the 68k. Sure it's fast but the 040 was faster and the Arm ISA is bugly in comparison. :)

Arm64 renders the 32 bit ISA obsolete and the programmer's model is very different.
My 2 cents. I own 3 archimedes btw. I have an A5000, a Risc PC and an a3000. I used to own an a3020 years ago. I don't dislike the machines. I just prefer the ST, Falcon and the Amiga to it.


You know I agree : the RISC philosophy has its weaknesses / drawbacks.
Can you please give an example of the bits extracting instructions (and what they do exactly, I do not remember well enough the bits of 68000 ASM I had try to learn several decades ago), and the number of cycles used ?
Your quote :
' Compare the flexibility of the bitfield instructions on the 020 to anything in the ARM. There's no equivalent to bitfield insert or bitfield extract instructions on the ARM. How many 32 bit (!) instructions would be needed to do the equivalent on ARM?
'
Then I'll try to tell you how it could be performed on the ARM chip.
There is a hidden feature of the LDR (load register) instruction, not detailed in the 2 books about ARM programming, but clearly detailed in the ARM VLSI Data Manual.
I make heavy use of this feature in all my graphical routines.
And if all this is beginning to be off-topic, please complain gently and not with harsh phrasing : the teacher/student behaviour is sthing I do not stand anymore, and I mean it has been like that for several decades. I am not sbdy's inferior, in any way.
Atari 1040 STE+SATAN, 520ST, 800xl, xegs, Amiga 500, 2000 with 68020, Archimedes, RISC PCs + Iyonix, Omega, BBC B, Atom, Electron, ZX 81, Spectrum 48/128/+2/+3, Speccy2010, Russian clones, Sam Coupe, V6Z80P, QL with accelerators, Enterprise 128, Einstein inc 256, Oric Atmos, MSX 1, 2, Thomson MO5, Amstrads inc CPC+, C 16, 64, 128, VG5000, Apple IIGS and more ! Yes I want to create a museum when I retire.

Zarchos
Captain Atari
Captain Atari
Posts: 224
Joined: Wed May 09, 2012 7:38 pm
Location: FRANCE

Re: Processor comparisons

Postby Zarchos » Sat Jul 22, 2017 7:30 am

simonsunnyboy wrote:No horizontal blank interrupt? And I remember so many great demos on the A3000 with raster bars.....Brothers In Arm come to mind *oh the memories* :)


None hardware indeed. So everything has to be done by the CPU.
Now RasterMan changes everything, we have the software, stable and precise solution, but who will demonstrate what an Archie can do with this new feature available ?
For sure all registers defining the screen modes and all registers in the memory controller can be redefined at the end of a scanline now + it is possible to change enough registers on the scanline, 16 times, for palette change, and also all infos defining the hardware sprite.

PS : Not sure about BIA for the rasterbars, maybe you are thinking of the demos by Karl Morton or the Colours demo by James Alldridge :
https://youtu.be/aeFY_JZYwyo
https://youtu.be/NeY7yLqETCA
sorry for the poor quality of the recording.
Atari 1040 STE+SATAN, 520ST, 800xl, xegs, Amiga 500, 2000 with 68020, Archimedes, RISC PCs + Iyonix, Omega, BBC B, Atom, Electron, ZX 81, Spectrum 48/128/+2/+3, Speccy2010, Russian clones, Sam Coupe, V6Z80P, QL with accelerators, Enterprise 128, Einstein inc 256, Oric Atmos, MSX 1, 2, Thomson MO5, Amstrads inc CPC+, C 16, 64, 128, VG5000, Apple IIGS and more ! Yes I want to create a museum when I retire.

AtariZoll
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 2945
Joined: Mon Feb 20, 2012 4:42 pm
Contact:

Re: Processor comparisons

Postby AtariZoll » Sat Jul 22, 2017 7:53 am

One question: is that old basic architecture and command set of ARM what is base for nowadays, widely used, multi core CPUs - used in smartphones, Raspberry Pi, etc. I had some 600 MHz ARM, manufactured by Intel in palmtop, some 10 years ago.
There is way to stop global warming.


Social Media

     

Return to “Non Atari Platforms”

Who is online

Users browsing this forum: No registered users and 1 guest