Overview Features Coding ApolloOS Performance Forum Downloads Products Order Contact

Welcome to the Apollo Forum

This forum is for people interested in the APOLLO CPU.
Please read the forum usage manual.
Please visit our Apollo-Discord Server for support.



All TopicsNewsPerformanceGamesDemosApolloVampireAROSWorkbenchATARIReleases
Performance and Benchmark Results!

World Record ATARI FreeMiNT Kronos 68080-FPUpage  1 2 3 4 5 6 

Peter Slegg

Posts 22
24 Nov 2017 16:03


As an Atarian I can't work out the model range. Is the Products page up to date ?

Currently, are they all Amiga specific devices or are there cpu replacements without the Amiga add-ons ?


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
24 Nov 2017 17:34


Peter Slegg wrote:

As an Atarian I can't work out the model range. Is the Products page up to date ?

Yes
 
Peter Slegg wrote:

Currently, are they all Amiga specific devices or are there cpu replacements without the Amiga add-ons ?

All the card offer
  * very fast memory
  * and the world the fastest 68K CPU
 
The extra features like video out can be used / but you don't have to use them.

So all the VAMPIREs can also be used (just) as the fastest CPU accelerator.


Peter Slegg

Posts 22
25 Nov 2017 22:09


Thanks, good to know.

Depending on the orientation it may fit the Milan in the square 68060 socket.

EXTERNAL LINK 
I currently have 144mb of edo simms and tried to fit 256mb but they don't work for some reason. Do you think you might go beyond 128mb ?




Gunnar von Boehn
(Apollo Team Member)
Posts 6207
25 Nov 2017 22:35


Peter Slegg wrote:

Do you think you might go beyond 128mb ?

The VAMPIRE-2 cards have 128 MB on board.
The VAMPIRE-4 cards have 512 MB on board.



Peter Slegg

Posts 22
26 Nov 2017 12:05


Does the Vampire 4  replace a 68000 or a 68060 ?

I am looking at the pcb photos and trying to work out what they will fit.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
26 Nov 2017 13:44


Peter Slegg wrote:

Does the Vampire 4  replace a 68000 or a 68060 ?
 
I am looking at the pcb photos and trying to work out what they will fit.

 
The Vampire models on the market today have as connector options:
  - DIL 68000
  - PLCC 68000
 
 


Peter Slegg

Posts 22
26 Nov 2017 14:33


I think the 68040 and 68060 are pga. Three rows of pins on each side.


Renee Cousins
(Apollo Team Member)
Posts 142
30 Nov 2017 02:48


Gunnar von Boehn wrote:
Coldfire is very restricted in instruction length with disallowed many EA modes. This makes Coldfire often need several instructions doing the equivalent of one 68K instruction.
 
  For example FPU:
  68K you can use flexible instruction like the below:
  FADD.S #1.8,fp0
  Easily adding an immediate float to a floating point register.

This only matters if you're coding in assembly and/or you're executing from somewhere with a lot of wait-states. Moreover, C/C++ will almost never use the missing addressing modes. In the MIPS/MHz category, the ColdFire v4e is a bit better (1.54) than the 68060 (1.33) and it's also clocked significantly higher (266MHz versus 100MHz with overclock). The never-released v5 core would have bettered 2.0 MIPS/MHz and hit around 400MHz, but everyone was moving to ARM at that point, so Freescale scrapped it. Too bad.


Peter Slegg

Posts 22
30 Nov 2017 10:50


Motorola/Freescale/NXP sat on the V5 Coldfire for so long.

You would think that an updated Coldfire would be cheaper to make than the "old" tech the V4 uses.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
30 Nov 2017 11:07


Renee Cousins wrote:

 
Gunnar von Boehn wrote:
Coldfire is very restricted in instruction length with disallowed many EA modes. This makes Coldfire often need several instructions doing the equivalent of one 68K instruction.
     
    For example FPU:
    68K you can use flexible instruction like the below:
      FADD.S #1.8,fp0
    Easily adding an immediate float to a floating point register.

  This only matters if you're coding in assembly and/or you're executing from somewhere with a lot of wait-states.
  Moreover, C/C++ will almost never use the missing addressing modes
 

 
No you misunderstand this.
The Coldfire misses not just some rare EA modes
but misses fundamental basis Address modes - which programmers and compilers used to use all the time.

Not only that the FPU can not uses float immediates, it can also not address memory pointed to by absolute $32bit addresses.

C/C++ compilers do of course support floating point constants and will happily use them.
Just look at GCC internals and look at code created by GCC.
 
Code like this is normal:
FMUL #0.5,Fp0
FMUL #Pi,Fp0
 
Of course the Coldfire can not support this, and has to create the same result using several instructions.
And this is just one example where the Coldfire needs to execute more instruction to do the same work.
The Coldfire often needs more instructions to do the same work.
There are many 68K instruction like MOVEM xxx,--(sp) or DBRA
which the Coldfire can not support and needs several instructions to get to the same result.
 
In average, the Coldfire V4 is significantly slower than 68060 at the same clockrate. The V4 chips can make up for this by having a higher clockrate.
You see this results also on the ATARI benchmarks.
In many KRONOS tests, the V4 Coldfire is lot slower than a 68060 running at the same clock as the Coldfire would be.
Coldfire has the potential to be faster for some algorithms with optimized MAC code - similar as 68080 can use AMMX to get a speed gain.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
30 Nov 2017 11:28


Renee Cousins wrote:

  The never-released v5 core would have bettered 2.0 MIPS/MHz and hit around 400MHz,
 

 
To be precise, the Coldfire V5 was released and also sold to some companies - but it was not sold to everyone.
If you want a Coldfire V5 then you can refurbish them from some  printer models.
 
BTW: You can not directly compare the Dhrystone results of Coldfire and 68060 as they were using different versions of the tests.


Peter Slegg

Posts 22
30 Nov 2017 15:37


I never knew that:

EXTERNAL LINK 
Processor  400 MHz, Motorola ColdFire® V5



Michal Warzecha

Posts 209
30 Nov 2017 19:26


It was known from the ColdFire V5 beginning. Polish company Elbox try to order ColdFire V5 but they can't get it, probably quantity was too low. Elbox is no HP :)
Have You, Gunnar, any idea CF v5 was more compatible to 68k than V4, or it was just faster?


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
30 Nov 2017 20:50


Michal Warzecha wrote:

CF v5 was more compatible to 68k than V4, or it was just faster?

The V5 was super-scalar like the 68060.
While the V4 was not fully super scalar.

The V5 was clocked a lot higher and could do more work per cycle.

But it had all the limitations of the Colfire family.
- No BYTE /WORD operation.
- 6 Byte max instruction length
- many instructions missing
...




Gregthe Canuck

Posts 274
01 Dec 2017 00:47



I can think of a core that would make a great upgrade from the Coldfire V5...  ;)



Thierry Atheist

Posts 644
01 Dec 2017 02:51


Gunnar von Boehn wrote:
The V5 was super-scalar like the 68060.
While the V4 was not fully super scalar.

The V5 was clocked a lot higher and could do more work per cycle.

But it had all the limitations of the Colfire family.
- No BYTE /WORD operation.
- 6 Byte max instruction length
- many instructions missing
...


So, basically, useless to the continuity of AOS and new AMIGA motherboards.

As much as Commodore's demise led to AMIGA vanishing from the landscape, motorola's abandonment of the 680x0 family of chips, would have killed AMIGA's future AS WELL.

AND DID, because if there WERE 200 MHz 68060s developed and produced, someone would have made an accelerator with it. :-( :-( :-( :-(


OneSTone O2o

Posts 159
01 Dec 2017 09:09


Forget, Coldfire is like dead. So much legacy software which does not run on it, no original TOS, no original Amiga-OS, not many games, not many productive applications.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
01 Dec 2017 12:08


oneSTone o2o wrote:

Forget, Coldfire is like dead.

Yes, this is true.

Coldfire can make a good CPU in the embedded market.
To replace 68K Cores running legacy 68k software - the Coldfire is unfortunately not well suited.

The Coldfire is de-facto dead.
The Coldfire V5 was only sold to HP.
The Coldfire V6 which was aiming for much higher performance, was canceled.

The 68060-B would have made a much better CPU for ATARI / AMIGA market then Coldfire.

But Motorola did also cancel the 68060-B which shared a number of improvements over the 68060 - with the 68080 now.
So the 68080 is basically an 68060B++



OneSTone O2o

Posts 159
03 Dec 2017 15:19


Gunnar von Boehn wrote:

So the 68080 is basically an 68060B++ ...

... currently still without 68k compatible PMMU ...


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
03 Dec 2017 15:22


oneSTone o2o wrote:

 
Gunnar von Boehn wrote:

    So the 68080 is basically an 68060B++ ...
   

   
  ... currently still without 68k compatible PMMU ...
 

 
The 68080 MPU offers everything that a develop needs for debugging.
And has for this purpose a number of advantages over the 68060 MMU.

posts 109page  1 2 3 4 5 6