Overview Features Coding Performance Forum Downloads Products OrderV4 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!

V1200 - the Closest Performance In WinUAE

Mateusz S.

Posts 53
07 Oct 2020 20:58


Hi,
As I am waiting for the V1200, I would like to do some
performance tests on emulation. Can you tell me if this WinUAE
setup is comparable with A1200? I mean: 060 + P96 + RTG (winuae drivers).. I found it quite fast, but I don't know if winuae RTG driver can be comparable with A1200. I don't know if it some kind accelerated by PC or what. Can you say its faster or slower that V1200.




Roy Gillotti

Posts 478
07 Oct 2020 21:11


WinUAE will be considerable faster, that said it's also not accelerating your A1200 and it's just running on a PC.


Mateusz S.

Posts 53
07 Oct 2020 21:20


Roy Gillotti wrote:

WinUAE will be considerable faster, that said it's also not accelerating your A1200 and it's just running on a PC.

Thats not what I am asking :)
WinUAE makes good job for testing apps when emulating 040, 060 with different clock rates when wmulating standard video outputs. But I don't know what about RTG. You can emulate CGX or PICASSO or other cards, but I would like to know if using winaue 'card' is coparable to Vampire RTG?


Mateusz S.

Posts 53
07 Oct 2020 21:25


In WinUAE I have selected the UAE Zorro III as graphic board RTG,
can it be comparable to V1200 RTG?


Roy Gillotti

Posts 478
07 Oct 2020 21:28


It acts like a Picasso 96 card, speed wise, not sure it's plenty fast for me.


Vojin Vidanovic
(Needs Verification)
Posts 1896/ 1
07 Oct 2020 21:29


Mateusz S. wrote:

I found it quite fast, but I don't know if winuae RTG driver can be comparable with A1200. I don't know if it some kind accelerated by PC or what. Can you say its faster or slower that V1200.

Emulation is not a good testbed for such a comparison. Overall fastest Vampire is as fast as emulating A1200 on Celeron PC, so anything far modern beats it.

Also, whatever emulation options you pick, its can slow down emulation but its overall dependent on your PC CPU, GPU and given task. In some task emulation speed can vary much more then real world hardware and its mostly in high end Amiga games and heavy render tasks. Vary - as not giving constantly same result and not being slower then real 060.

Picasso RTG is what Vamp has, but on your emulation sys speed is dependent on DirectX - given RTG task - your GPU interaction more then on version of RTG driver used in emulation.


Olaf Schoenweiss

Posts 684
07 Oct 2020 21:34


it is very propably faster than V1200, even without JIT
 
  it is not possible to make precise emulation regarding speed

you can only emulate A500 precise


Gunnar von Boehn
(Apollo Team Member)
Posts 5483
08 Oct 2020 07:44


Olaf Schoenweiss wrote:

it is very propably faster than V1200, even without JIT

Olaf how can you make such posts?
You don't have a Vampire and never used it

In my experience Vampire is faster than UAE without JIT.


Mateusz S.

Posts 53
08 Oct 2020 09:49


Sorry, I made a couple mistakes in first post :) (it was late)

In other words:
Can UAE Zorro III as graphic board in winUAE running as 060/50mhz be comparable to V1200 RTG? If I make some graphic tests I would have similar fps?

The best option would be running some rtg games or graphic tests and compare the fps. Maybe someone has done it? :)


Olaf Schoenweiss

Posts 684
08 Oct 2020 11:04


it is not possible to adjust speed of UAE exact like you want. The only exception is A500 with 68000. You can make it slower by turning off JIT but the speed then still depends on the hardware you run UAE on. Having something to adjust a certain speed would be nice but is not available. Also SAGA is of course much faster than the old AGA but neither SAGA nor 68080 is supported by UAE and I do not expect that to happen.


Gunnar von Boehn
(Apollo Team Member)
Posts 5483
08 Oct 2020 11:20


Mateusz S. wrote:

The best option would be running some rtg games or graphic tests and compare the fps. Maybe someone has done it? :)

To emulate an exact CPU speed is very complicated.
UAE can relative good emulate the exact speed of the A500 (0.5 Mips total speed)
UAE can also run faster then this (e.g. with JIT)

But UAE can not emulate any other speed properly.
This means if you want to code a game that runs well on an 68030@25 MHz - then UAE can not be used to code this well.
As you can not "select" any speed in UAE.




Mateusz S.

Posts 53
08 Oct 2020 12:00


Gunnar von Boehn wrote:

 
Mateusz S. wrote:

  The best option would be running some rtg games or graphic tests and compare the fps. Maybe someone has done it? :)
 

 
  To emulate an exact CPU speed is very complicated.
  UAE can relative good emulate the exact speed of the A500 (0.5 Mips total speed)
  UAE can also run faster then this (e.g. with JIT)
 
  But UAE can not emulate any other speed properly.
  This means if you want to code a game that runs well on an 68030@25 MHz - then UAE can not be used to code this well.
  As you can not "select" any speed in UAE.
 
 
 

 
  Thanks for reply

edit:
OK, but what about benchmark applications? like AIBB or SysSpeed. We can compare V1200 results with some winUAE configurations.. I don't know if AIBB will work under RTG resolutions, but maybe other benchmarks that shows different OS graphics operations per second?
 


Gunnar von Boehn
(Apollo Team Member)
Posts 5483
08 Oct 2020 12:31


Mateusz S. wrote:

OK, but what about benchmark applications?

   
Sorry I have no clue what your question is?
What do you want to do?
Do you want to code for Amiga but use UAE as development tool?
 
Experienced coder told us that this will not always work 100% - you will need to re-test your stuff on real Amiga.
   
Do you want to "tune" code for speed with benchmarking on UAE?
This will also not work.
   
some Code like this:

add.l #1,D0
add.l #2,D0
add.l #3,D0
add.l #4,D0

 
UAE will internally change this to

add.l #10,D0

This means 4 instructions will be exchange to 1 instructions.
This will change the measurement by factor 4.

This means benchmarks trying to measure the speed are fooled by UAE this way - and will very often report unrealistic scores.
And the same way you can not use UAE for profiling your code or for comparing which code version is faster for the same problem.
You will need to benchmark your code always on real HW.



Mateusz S.

Posts 53
08 Oct 2020 13:58


Thanks for answer
I wanted to check performance of code using RTG somehow, before I get real V1200 hardware


Nick Fellows

Posts 111
08 Oct 2020 17:31


We will only find out how fast emulation is , when emulation supports the full apollo 68080 instruction set. Why would you want that ? Well - it makes developing software for it much more convenient.

For example - when i develop C64 demo's i use VSCode and and Emulator (C64Debugger or Vice). Then when im nearing completion test on real hardware.

As far as i know its currenly not possible to develop for the vampire except on physical hardware.


posts 15