Overview Features Instructions Performance Forum Downloads Products OrderV4 Reseller Contact

Welcome to the Apollo Forum

This forum is for people interested in the APOLLO CPU.
Please read the forum usage manual.



All TopicsNewsPerformanceGamesDemosApolloVampireAROSWorkbenchATARIReleases
Performance and Benchmark Results!

68080 (680xx Vs X86)page  1 2 

Örjan Olofsson

Posts 9
13 Oct 2020 20:42


Hi all, my first post here. First I'd like to say this is an
exciting project, have been lurking here for a while.

Is it correct to say that the 68080 equals the performance of a
Pentium 1 running at 166 mzh?

How much faster than today is it possible for the 68080 to become?

My next question is a bit historical: When comparing older 680xx
cpu's to x86's, is it fair to say that:

68000 = 286,
68020 = 386,
68030 = almost the best 486,
68040 = better than all 486,
68060 = worse than Pentium 1?




Michael Farrell

Posts 18
13 Oct 2020 21:58


Well, if you *just* base it off the mips,
 
  a 68040 is about 2/3 the speed of a 486DX2/100.
 
  and an '060 is 3/4 of the speed of a Pentium 100.
 
  So, the AC68080 is around the same speed as a P100.
 
  EXTERNAL LINK 
  It's not a fair comparison though. Just a straight "millions of instructions per second"


Gunnar von Boehn
(Apollo Team Member)
Posts 4964
14 Oct 2020 06:21


Hello Örjan,

Örjan Olofsson wrote:

  68000 = 286,
  68020 = 386,
  68030 = almost the best 486,
  68040 = better than all 486,
  68060 = worse than Pentium 1?

Comparing CPUs is very complex as a CPU can do so many things and has so many different attributes.

If you look at CPUs then the you can look at many different aspects.

a) Strength and Power of the instructions
E.g Can it do difficult instructions like "divide"?

b) Size and number of the Registers
E.g. Does it has enough registers to do most algorithms nicely?

c) Can it work directly on memory?
Or does it need to use extra instructions to load values to registers first?

d) Does it support complex and nice to use address modes?

e) Does it support Immediate operants in instructions?

f) Does it support Vector operations to accelerate e.g. graphic and video?

g) Does it has good Instruction Caches and Data caches?

h) Does it has smart branch prediction?

i) How fast can it execute a single instruction?
Does the "average" instruction take 1 cycle or more?

j) Does it support executing more than 1 instruction per cycle?

k) Can it operate well on misaligned memory data?

l) How smart is it on operating on memory streams?
Does it support "prefetch" instructions or does it even prefetch automatically?

m) How wide are the registers?

If you look in this in details then you will always get a "complex" answer...
If you compare all 68K and all x86 from 8080 to Pentium 1 then
68080 is by far the most advanced of all of them.


Gunnar von Boehn
(Apollo Team Member)
Posts 4964
14 Oct 2020 06:51


Michael Farrell wrote:

So, the AC68080 is around the same speed as a P100.

The 68080 is a lot stronger than the Pentium P100.
The 68080 in the V4 can peak do 340 MIPS.

But in my experience the MIPS is not the most realistic comparison.
A lot more real world are in my opinion "STREAM" and "LMBENCH".

Lets look at some results:

The VAMPIRE scores here about 500 MB/sec which is comparable to an AMD ATHLON runnning at 800 MHz!




Örjan Olofsson

Posts 9
14 Oct 2020 11:08


Gunnar von Boehn wrote:

 
  The 68080 is a lot stronger than the Pentium P100.
  The 68080 in the V4 can peak do 340 MIPS.
   
  But in my experience the MIPS is not the most realistic comparison.
  A lot more real world are in my opinion "STREAM" and "LMBENCH".
 
 

 
  Thanks for replying. Yes I know it's complex to compare different CPU's and that MIPS is not always relevant.
 
  So, which Pentium is a realistic comparison to the current 68080 at full peak?
 
  How much further is it possible to increase the speed of the 68080?
 
  :)


Gunnar von Boehn
(Apollo Team Member)
Posts 4964
14 Oct 2020 11:30


Örjan Olofsson wrote:

So, which Pentium is a realistic comparison to the current 68080 at full peak?

   
Apollo-Core == 68080 has many modern features which only very modern  INTEL CORE chips have.
   
While the clockrate of the 68080 is limited by the FPGA chip - the performance of the CPU is extremely high for its clock.
The Apollo CPU has much higher performance than a 800 MHz 68030 CPU in mostly all cases.
 
The performance of the APOLLO CPU can range between "Pentium 100" to "Pentium 3@1000 MHz" - depending on what enhancements and features you trigger with your program.
So its just impossible to give you a "simple" number and be correct.

Apollo 68080 is clearly the fastest 68K Amiga ever - and even outruns PowerPC machines at 1000MHz in some benchmarks.




A1200 Coder

Posts 69
14 Oct 2020 13:57


Regarding the slower CPUs, 68030/50MHz is about the same as 486/25 MHz in integer performance. 68040 beats the 486 significantly, both in integer and FPU performance. 68060 is slightly faster in integer performance than a Pentium 1, but loses in FPU performance.

Pentium 1 can execute max 2 instructions per clock, 68060 2-3 instructions per clock (yes, Motorola manuals actually mention somewhere 3 instructions per clock), while 68080 can execute 4 instructions per clock.

The 68080 is better than Pentium Pro, P2,P3,P4 in some aspects, but P3 introduced the SSE instruction set, which is better than MMX in Pentium and AMMX in Vampire, as these are integer-based, while SSE is floating point.

As already pointed out, FPGA-based designs, like 68080 are limited to clock speeds of around 100 MHz with current used FPGA technology, while those Pentiums could easily run at clock speeds from 200 MHz- over 1 GHz in ASIC.


Gunnar von Boehn
(Apollo Team Member)
Posts 4964
14 Oct 2020 15:34


A1200 coder wrote:

68060 2-3 instructions per clock (yes, Motorola manuals actually mention somewhere 3 instructions per clock)

You can answer this very easily.

68060 Icache does provides at maximum 32bit per cycle.
32bit is enough for maximal 2 of the shortest 68k instructions.
68k instructions have a size of 16bit or longer.

As many 68k instructions are 32bit or 48bit or longer the 68060 is often limited by this. Motorola was aware of this bottleneck and the 68060-dev-team wanted actually to double the Icache fetch ... but this never happened.

A1200 coder wrote:

The 68080 is better than Pentium Pro, P2,P3,P4 in some aspects, but P3 introduced the SSE instruction set, which is better than MMX in Pentium and AMMX in Vampire, as these are integer-based, while SSE is floating point.

Please note that MMX and AMMX have not much in common except 3 letters. The instructions are different and AMMX has a huge number of features that MMX not has.

You are fully correct that AMMX is optimized for integer and pixel operations.

But this is not bad.
For GFX operations and video operations the integer math is most important.
And for 2D game coding the AMMX special instructions give AMMX a huge advantage over SSE.

For 2D games AMMX does beat SSE hands down.
 
 


Saladriel Amrael

Posts 148
15 Oct 2020 09:38


Gunnar von Boehn wrote:

Michael Farrell wrote:

  So, the AC68080 is around the same speed as a P100.
 

  The 68080 is a lot stronger than the Pentium P100.
  The 68080 in the V4 can peak do 340 MIPS.
 
 
  But in my experience the MIPS is not the most realistic comparison.
  A lot more real world are in my opinion "STREAM" and "LMBENCH".
 
  Lets look at some results:
 
 
  The VAMPIRE scores here about 500 MB/sec which is comparable to an AMD ATHLON runnning at 800 MHz!
 
 
 
 

So, long story short, Memory access speed is often an ignored bottleneck in CPU/MoBo designs, while Apollo/Vampire avoids this bottleneck thus letting the CPU/Chipset express their true potential under heavy workload situations?

That is a thing I'd wish to see more often even in modern PC/Console production




Örjan Olofsson

Posts 9
15 Oct 2020 14:07


Gunnar von Boehn wrote:

Örjan Olofsson wrote:

  So, which Pentium is a realistic comparison to the current 68080 at full peak?
 

   
  Apollo-Core == 68080 has many modern features which only very modern  INTEL CORE chips have.
   
  While the clockrate of the 68080 is limited by the FPGA chip - the performance of the CPU is extremely high for its clock.
  The Apollo CPU has much higher performance than a 800 MHz 68030 CPU in mostly all cases.
 
  The performance of the APOLLO CPU can range between "Pentium 100" to "Pentium 3@1000 MHz" - depending on what enhancements and features you trigger with your program.
  So its just impossible to give you a "simple" number and be correct.
 
  Apollo 68080 is clearly the fastest 68K Amiga ever - and even outruns PowerPC machines at 1000MHz in some benchmarks.
 
 

Thanks for your reply, sounds great :)

Does that mean the 68080 has reached its full potential?


Vojin Vidanovic

Posts 1731
15 Oct 2020 15:19


Örjan Olofsson wrote:

  Does that mean the 68080 has reached its full potential?
 

 
  To me, as design yes, as product no. Meaning even in V4 it reaches 110Mhz or so, abet with great results.
 
  Its full potential would be in real silicon, in real ATX board, on couple hundred Mhz.
 
  Also, as seen in V1200 and V4 with (even) faster gen DDR2/3 plus RAM results of great memory controller (that is that faster part then in PPC) improve. So coupled with some blaze memory and SSD would be a taste of "real time OS and computer" to the end. Or maybe OS in fast ROM already loaded, like old BASICs and DOSes used to in 8bit era? We will see what it will reach ...
 
  Now its great even as FPGA.


Örjan Olofsson

Posts 9
15 Oct 2020 16:00


Vojin Vidanovic wrote:

Örjan Olofsson wrote:

    Does that mean the 68080 has reached its full potential?
 

 
  To me, as design yes, as product no. Meaning even in V4 it reaches 110Mhz or so, abet with great results.
 
  Its full potential would be in real silicon, in real ATX board, on couple hundred Mhz.
 

How hard/expensive would it be to make a 68080 in real silison?


Eric Gus

Posts 433
15 Oct 2020 17:03


Örjan Olofsson wrote:

 
  How hard/expensive would it be to make a 68080 in real silison?

VERY EXPENSIVE .. hundreds of thousands ..


Vojin Vidanovic

Posts 1731
15 Oct 2020 17:21


Örjan Olofsson wrote:

      How hard/expensive would it be to make a 68080 in real silison?
     

     
      So much we need a real world sponsor.
     
      But Vampire FPGA sales and software development help create market that could be beneficial to attract sponsors in future. In time. As well as right direction of free ApolloOS, so its independent of AmigaOS "future" but linked to its past by working on it with driver pack installed.
     
      For now, bigger and fatter FPGAs with faster RAM should be enough to showcase its power.
     
    68000 = its faster and way more mature then XT
    68020 = its better design then most of 286, even higher freq
    68030 = with FPU and 40Mhz clock comparable to 386
    68040 = Beats normal 486 low clock, but not DX2 or DX4 plus
    68060 = Similar design as Pentium, but low clocked
    68080 - Catching up P1 P2 in FPGA, with many P1 MMX to P4 tech included, would do waay better in silicon
 
    So some point up to 030 we were kind of leading, and in XT days we had a major upper hand. Reason why 68 000 was so popular :)
   
   


Gunnar von Boehn
(Apollo Team Member)
Posts 4964
15 Oct 2020 20:29


Vojin Vidanovic wrote:

So some point up to 030 we were kind of leading, and in XT days we had a major upper hand. Reason why 68000 was so popular :)   

I personally think 68K was popular because its great to code.
And this is still true.
68000 is still leading here today - even against other 5 Ghz chips




Boban Krsmanovic

Posts 8
15 Oct 2020 21:33


Örjan Olofsson wrote:

  cpu's to x86's, is it fair to say that:
 
  68000 = 286,
  68020 = 386,
  68030 = almost the best 486,
  68040 = better than all 486,
  68060 = worse than Pentium 1?
 

I can only answer from an (average) user point, and not at all technical.
Back in the day, I have a friend that had 386DX/40 and a friend that got plain (1MB) A500.
I had only Commodore 64, so I guess, I was in the position to be objective.
In my opinion, Amiga 500 wipes floor with 386 on just about everything. We played (and compared) of course, mostly games, and 396 was horrible in that part comparing to the A500. But also, when I saw Deluxe Paint, I felt in love right that moment, and we had fun playing with Workbench. On the other side, Windows 3.11, looked dull and uninteresting.
Not to mention overall feeling... Amiga had always Joystick, where PC were very expensive, it could connect to the TV, where PC couldn't (or it costs more then PC), Amiga had sound, while PC had PC speaker.
The only thing I can think of that was A500 disadvantage is floppy drive. But then again, it wasn't really that bad as some people mentioned today. Lot's of games loaded fairly quickly, and only (some) adventure games was a pain to play.
And considering the price difference, with price of the 386, you could have not A500, but A1200 with HD and some fast ram.
3D games?
Well, we weren't much into those, and I am sure 386 could handle much better games like Wing Commander, Frontier (don't mention Doom, as it was crawling on 386 :) )... etc.
A500 have it's base of 3D games that are very playable like Gunship 2000, FA/18 Interceptor, ZeeWolf 1 and 2, etc.

That's just my opinion about 68000=286.

The rest of the list makes more sense, but then again, I think there's hardly any title that utilized 020 (A1200) to it's full potential, as many titles did with A500 (Commodore went bankrupt pretty soon), let alone, 030, 040. 060... and 080. :)

Oh... you're asking only for processors, not configurations? :D
Sorry for that , but then again, this was rather point of view, from average user, anyway. :)

Cheers! :)


Vojin Vidanovic

Posts 1731
16 Oct 2020 08:20


Boban Krsmanovic wrote:

      Oh... you're asking only for processors, not configurations? :D
      Sorry for that , but then again, this was rather point of view, from average user, anyway. :)
      Cheers! :)
 
  A500 have it's base of 3D games that are very playable like Gunship 2000, FA/18 Interceptor, ZeeWolf 1 and 2, etc.
   

 
  HUNTER and Trexx Warrior! I really loved this plain isometric 3D, no textures, no fancy effects, but full 3D even on A500!
   
    Oh boy that Win 3.11 was ugly and unstable, limited to 8 plus 3 chars in filenames, these days WB 2.0/3.0 looked like its future!
   
    Fully agree, remembering how I laughed at PC Speakers, mono monitor 286-386 being able poorly to play only Prince of Persia. But then again, decent 386DX with SB and SVGA could later on play Frontier First Encouters, Wing Commander 2, Descent, Doom, Day of Tentacle,. XWing etc. so in the end it really outperformed even A1200HD esp. in productivity apps.
   
    Its less known, but DPaint was actually avail for PC too, it just didnt catch up too. Probably people stick to Microsoft Paint and later Photoshop.
   
    While Amiga was cheap as basis and what you get, expanding it to 030, monitor, 4MB fast and few more things could outcost the multimedia PC with SVGA and SB, at least in our Balkan corner, where prices were inflated by non official dealers.
   
    Yes, its CPU comparison.
   
    @Gunnar
   
    I am no coder, but I believe as single CPU (not instruction set) 68 000 was worlds most popular CPU, thanks to widespread in Atari, Amiga, Mac and few consoles :) Was it because its easy to code or widespread or well documented and with many coding tools? Probably combo of all.
   


Gunnar von Boehn
(Apollo Team Member)
Posts 4964
16 Oct 2020 11:54


Vojin Vidanovic wrote:

@Gunnar, ...I believe as single CPU (not instruction set) 68 000 was worlds most popular CPU, thanks to widespread in Atari, Amiga, Mac and few consoles :) Was it because its easy to code or widespread or well documented and with many coding tools? Probably combo of all.

Yes you are fully right.
The 68K had once a total market domination.

It was used in many Arcade machines, many Consoles, in computers like Amiga, Atari, Apple, in many UNIX workstations, it was used by the industry for consumer electronics, airplanes, cars, trains, rockets, military systems.

In my opinion the reason for its success was that its good ISA,
Which is powerful and coder friendly and allowed straight forward software development.




Örjan Olofsson

Posts 9
16 Oct 2020 12:30


Vojin Vidanovic wrote:

 
Örjan Olofsson wrote:

  How hard/expensive would it be to make a 68080 in real silicon?
 

  So much we need a real world sponsor.
 

  Would it be possible to get Motorola on board (fun intented) somehow?
 
Vojin Vidanovic wrote:

  68000 = its faster and way more mature then XT
 

  With XT do you mean 8086?
 
 
  Oh by the way, I apologize writing too fast and misspelling silicon ;)


Örjan Olofsson

Posts 9
16 Oct 2020 13:39


Let's say history was different, Commodore being smarter and not going bankrupt in 94, and able to persuade Motorola to develop something similar to 68080 some time after 68060.
 
Would that have saved Amiga from following Apple along the PPC route? Or rather just postponed it?
 
My second, sad question, is if Commodore eventually had ended up using Intel the way Apple did? :-(
 
 

posts 40page  1 2