Processor comparisons

[catmando]

Saw this on another forum thought it might be worth something

68000 7 = 8086 5
68020 14 A1200 no FAST = 286 8
68020 14 w/fastmem = 386DX 20
68030 16 = 386DX 25
68030 25 = 386DX 40
68030 40 = 486SX 20
68030 50 = 486DX 25
68040 25 = 486DX 33
68040 40 = pentium 66
68040 50 overclocked = pentium 90
68060 50 = pentium 233 (no MMX type cpu)

Amiga 3000 16 mhz 3.97 BogoMIPS
Atari Falcon 16 mhz 3.95 BogoMIPS

[wongck]

Wow... I remember playing a high colour 3D space simulator on what I think is a Pentium 100+ MHZ machine. I was like WOW this is so much better than Elite.
So my 68060 can also do that ?

[Moulinaie]

Here are two sources...

One that gives the result of SuperPI for his AMD 586-75:
http://www.vogons.org/viewtopic.php?t=25288#p191584

And the program I made for Atari :
http://gtello.pagesperso-orange.fr/kronos_soft.htm

You can see that, for the 16K calculation, the AMD requires 22 seconds, the same for a Falcon CT60 at 95MHz.
My Atari program is in assembler, and I hope, it is optimized and don't use the FPU.

So: 68060 50MHz = Pentium 233... I don't think so!

Guillaume.

[AtariZoll]

Saw this on another forum thought it might be worth something

68000 7 = 8086 5
68020 14 A1200 no FAST = 286 8
68020 14 w/fastmem = 386DX 20
68030 16 = 386DX 25
68030 25 = 386DX 40
68030 40 = 486SX 20
68030 50 = 486DX 25
68040 25 = 486DX 33
68040 40 = pentium 66
68040 50 overclocked = pentium 90
68060 50 = pentium 233 (no MMX type cpu)

Amiga 3000 16 mhz 3.97 BogoMIPS
Atari Falcon 16 mhz 3.95 BogoMIPS

This list is a mess, and far from accurate. First, what sense has comparing 68020 in Amiga with slow RAM with some 286 at full speed ?
And in next line is same CPU with fast RAM faster approx 5x ???? No way.
We know that 68030 was not faster than 386DX at same clock.
68040 faster than Pentium - which has 64-bit data bus ? Etc ...
Why Apple needed to go on Power PC CPU if 680xx line was so fast ?

[CheshireCat]

The entire comparison table in the OP is quite a bit off for different reasons.

At same clock speed, 68060 is faster than Pentium for integer but usually slower for float.
In any case, a Pentium@233 is several times faster than a 060@50, and that is not even considering MMX code !

[Zarchos]

This kind of comparisons just makes no sense.

MIPS are just not even the adequate tool to evaluate a system's speed.
And that is an Acorn nutter with ARM2 8 Mhz boasting its 4.5 MIPS telling you.
Different CPUs do different operations per instruction, so the idea of 'number of instructions executed per second' to evaluate speed is just plain stupid from the very beginning : it just means nothing from one type of CPU to another one.(and even when it is the same type of CPU it means nothing from one generation to the next).

Consider this too : the architecture your CPU is inserted in, the sources of interrupts, the memory bandwith your memory bus can provide, the various DMAs ... all this must be taken into account.
A CPU doesn't work all alone ...

I still believe specific algorithms coded in assembly language, for let's say a hundred or a few hundreds of typical cases, should be timed and these results would give you the basis to appreciate this or that CPU, when facing this or that task to complete, with this or that architecture.
But then it wouldn't be an answer with 10 words, and people are intellectualy lazy and prefer short, false answers, most of the time.
Reality is complex, sorry.

[EmpireAndrew]

I 100% agree with Zarchos.
Recently I've been going through my collection and doing tests on how long it takes to download and display a simple webpage, which becomes unmeasureable on faster machines but is interesting at the more vintage end.
In some cases I've tried to use the same browser across platforms when possible so the differences are the OS or the hardware at least.
No I know it doesn't take into account the FPU and so on, its not a full CPU comparison which is difficult to pull off and meaningless anyway, it's just a real world use case that excercises the whole platform although there is very little disk involved.
For example:

386DX 20 12MB 640x480x64k win3 IE3 = 144 seconds
386DX 20 12MB 640x480x64k os2v3 webex 1.1 = 91 seconds
486DX 33 12MB 640x480x64k win3 IE3 = 22 seconds
486DX2 66 12MB 640x480x64k win3 IE3 = 12 seconds
P233MMX 128MB 800x600x16m win95 IE3 = 3 seconds

From memory (I need to double check and write down):
TT030@48MHz 2/16MB 1920x960x2 Cab2.8 = 57 seconds?

I seem to recall the TT being under a minute but I did so many tests on it and it was a few months ago so I could be mistaken.
I'll re-test at some point.
I'lol also be doing this on other Ataris, Amiga, Mac and other platforms.
It's for my own interest but I'll share as it might be interesting if not definitive proof of anything...

[Atarieterno]

The point is that Intel 356 is not capable of recording 16 audio tracks even if a miracle is requested.
That comparison is as relative to the utilities that they have had in each epoch.

[wongck]

waaah... hoping to have a fast 3D space simulator on my 68060...

[mikro]

68060 50 = pentium 233 (no MMX type cpu)

Yeah, sure. Do a simple mind experiment, imagine Quake, or even better Quake II, running on the Pentium 233 and on the 68060, I'll even give you the liberty to overclock it to 100 MHz and use SuperVidel's 8-bit chunky graphics.

No freaking way.

[AtariZoll]

The point is that Intel 356 is not capable of recording 16 audio tracks even if a miracle is requested.
That comparison is as relative to the utilities that they have had in each epoch.

Sure, Intel 356 is not capable. Something what exists not is not capable
And what some CPU is capable depends a lot of how SW is written, + peripherals.
To add, CPU does not recording - storage does it - hard disk for instance

[Zarchos]

68060 50 = pentium 233 (no MMX type cpu)

Yeah, sure. Do a simple mind experiment, imagine Quake, or even better Quake II, running on the Pentium 233 and on the 68060, I'll even give you the liberty to overclock it to 100 MHz and use SuperVidel's 8-bit chunky graphics.

No freaking way.

For Doom which is simpler it will be ok, though.

[mikro]

For Doom which is simpler it will be ok, though.

My point exactly -- for Quake II P@233 is totally OK, for 68060@100 MHz it isn't even close.

[Atarieterno]

The point is that Intel 356 is not capable of recording 16 audio tracks even if a miracle is requested.
That comparison is as relative to the utilities that they have had in each epoch.

Sure, Intel 356 is not capable. Something what exists not is not capable
And what some CPU is capable depends a lot of how SW is written, + peripherals.
To add, CPU does not recording - storage does it - hard disk for instance

I am sure that anyone has seen that it is a simple errata and I wanted to indicate "386".
Even a simple musician like myself know that many more factors intervene in the operation of a computer besides its CPU, but an Intel 386 with all the software of the world and all the peripherals was never able to work professionally in a music recording studio ; That is what I was trying to say and that is why I find the comparison with Atari absurd.
However, even with the idiomatic difficulty, I will try to be more precise to avoid more responses of this type.
Greetings.

[AtariZoll]

..
I am sure that anyone has seen that it is a simple errata and I wanted to indicate "386".
Even a simple musician like myself know that many more factors intervene in the operation of a computer besides its CPU, but an Intel 386 with all the software of the world and all the peripherals was never able to work professionally in a music recording studio ; That is what I was trying to say and that is why I find the comparison with Atari absurd.
However, even with the idiomatic difficulty, I will try to be more precise to avoid more responses of this type.
Greetings.

I was cynical not because that typo. Your claim about 386 is just totally biased. 386DX at 40 MHz surely can do it. CPUs does not differ so much as some may think.

[Atarieterno]

..
I am sure that anyone has seen that it is a simple errata and I wanted to indicate "386".
Even a simple musician like myself know that many more factors intervene in the operation of a computer besides its CPU, but an Intel 386 with all the software of the world and all the peripherals was never able to work professionally in a music recording studio ; That is what I was trying to say and that is why I find the comparison with Atari absurd.
However, even with the idiomatic difficulty, I will try to be more precise to avoid more responses of this type.
Greetings.

I was cynical not because that typo. Your claim about 386 is just totally biased. 386DX at 40 MHz surely can do it. CPUs does not differ so much as some may think.

Well, I'm not so sure he can.
Intel 386 appeared in 1986 and until 1994 Steinberg did not entrust its first Cubase Audio to PCs ... operating in a rather "dubious" way.
I've seen the evolution of computing in recording studios, I've seen huge sums of money relying on a PC that did not work properly, I've known the Turtle Beach Multisound and the arrival of SoundScape.
No, I'm sorry, but with a 386 connected on a PC it was not possible to record 16 channels of audio for very fast that it was the hard disk and very expensive that is the card used.
It is my opinion, but if someone has evidence that proves otherwise I will be happy to see them.

[Faucon2001]

Even on a P150, Cubase VST a turtle beach sound card and win95 audio was lagging compared to Falcon.
How come on a 386 ?

[EmpireAndrew]

One of the benefits of Ataris and Amigas was how the internal components were wired up.
For example on the ST MIDI ports and other components are accessed in turn on a regular interval, so timing is precise.
On an Amiga with a serial to MIDI adapter that's not the case and the timing can be off.
Same on a Mac or PC where access to the expansion boards is interrupt driven. Their processors could be much faster but the way it has to talk to the other components is more flexible, but unpredictable in terms of timing.
However the tight integration of the architecture of an Atari or Amiga is a barrier to bringing out improved, faster models. You get all sorts of incompatabilities by the necessary re-designs.
But back then, you could do more with less on a system like an ST or Amiga, and that's before you consider other benefits like DMA access for things like driving laser printers, or blitter and other custom chips.
But long term, the more flexible but less efficient designs of the PC and to a lesser extent, the Mac, were the right way to go.

Comparing just the processor, however it's done, whether calculating pi or whatever, doesn't tell the full story.

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.

All this is what made computers back then far more interesting imho...

[Atarieterno]

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.

[arf]

Saw this on another forum thought it might be worth something

68000 7 = 8086 5
68020 14 A1200 no FAST = 286 8
68020 14 w/fastmem = 386DX 20
68030 16 = 386DX 25
68030 25 = 386DX 40
68030 40 = 486SX 20
68030 50 = 486DX 25
68040 25 = 486DX 33
68040 40 = pentium 66
68040 50 overclocked = pentium 90
68060 50 = pentium 233 (no MMX type cpu)

Can’t agree really on this data. The 68k stuff is slower than shown here, IMHO, way slower. My experience is that until the 386DX the speed is more or less the same with the same MHz, with the exception of the 68000, which is faster than the 8086.

But from the 486DX2-66 or 486DX50 ongoing, the 68k stuff can’t keep up the pace.

[AtariZoll]

Some, who made huge claims here should correct self - it is not CPU what is not capable for something, but some concrete computer architecture.

[Atarieterno]

Some, who made huge claims here should correct self - it is not CPU what is not capable for something, but some concrete computer architecture.

Nobody omits the complete architecture or hardware and software accessory set, you also said it in an earlier post. The point is that none computer with the Intel 386 microprocessor has served professionally for multitrack audio recording.
I could have been a doctor, but I have not studied medicine ... the end result is that I am not a doctor; The 386 has power to manage recording systems ?, I do not know, but the story does not prove it. Everything else is supposition.

[Zarchos]

I don t agree with your idea of 'supposition'.
There is enough knowledge about hardware to know if yes or not some applications would work on it or not.
It is an issue I have very often with the Amiga owners, as an Archimedes owner : they don t want to admit that with the knowledge of the hardware, the timings of some specific pieces of codes, eventually yes it is possible with a 90% accuracy to say wether this or that Amiga game could have been converted to the Archie and would have run at 50 fps or 25, how much memory would have been necessary ...
At the time systems were simple enough to foresee with great precision where and when the limits would be reached.
No need to have the actual software developed to know it.

[Cyprian]

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

[Zarchos]

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

It isnt
Oh dear I'm going to oppose everybody in this thread
Graphics manipulations are a bottleneck if you don t use tailored routines ... But once you have them, you can compute your CPU load and from that quite easily foresee the feasibility in 50 or 25 fps ...
Do not forget until recently the Archie had no precise Hsync, so everything using this feature on the Amiga or St had to be rethought using the CPU ...