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!

FPU VAMPIRE2 Beat All Available AMIGA Acceleratorspage  1 2 3 

Gunnar von Boehn
(Apollo Team Member)
Posts 5421
02 Sep 2017 23:48


FPU performance improved a lot with new FEMU and GOLD 2.7

As you all know the VAMPIRE-2 is over 10 times faster than any other produced and sold AMIGA accelerator on the market.

With the new FEMU also the FPU performance is higher than any other produced and sold AMIGA accelerator on the market.




Vojin Vidanovic

Posts 770
02 Sep 2017 23:52


Great results for softFPU. When will new Femu be downloadable?
Hope to see hardFPU results :-)


Gunnar von Boehn
(Apollo Team Member)
Posts 5421
02 Sep 2017 23:58


Vojin Vidanovic wrote:

  Great results for softFPU. When will new Femu be downloadable?
 

 
The new FEMU depends on GOLD 2.7 features.
Jari did some awesome performance improvements on FEMU.
It would not surprise me to see soon 3 MFLOPS or 4 MFLOPS scores.
 


Gregthe Canuck

Posts 274
03 Sep 2017 02:50


Wow looking better and better. Nice work Jari!
 
From another posting I just saw on Amigaworld it looks like an A3640 040@25 accelerator does 4.65 MFLOPS with Sysinfo.
 


Steve Ferrell

Posts 402
03 Sep 2017 03:39


Congratulations!  Great work Gunnar and Jari.  Sadly, your greatest critics will claim that your latest accomplishment was Photoshopped!  LOL!


Daniel Sevo

Posts 299
03 Sep 2017 08:28


Gunnar von Boehn wrote:

    8><---
    As you all know the VAMPIRE-2 is over 10 times faster than any other produced and sold AMIGA accelerator on the market.
  ----><8
   

 
  What we all know is that the Vampire 2 / Apollo core is a great product that has re-ignited the 68k Amiga world and has potential to outperform most Amiga stuff out there, even PPC..
  ... but if you say something like that ppl will take their 060 numbers and multiply with 10. It would probably give you something like ~1000 MIPS and ~350MFLOPS.
  If you mean memory performance is 10x then you should probably be a bit more specific. Especially when you back it up with good(?) old(!) Sysinfo ;-)
 
  Also, when are we going to abandon SysInfo for something more accurate? Maybe SysSpeed would be better from now on? ;-)
 
  [I]P.S. it wasn't obvious if you meant "new accelerators produced and sold *today*" and only looking at ACA 030 stuff and assuming that new Phase 5 is vaporware, then it makes more sense. ;-)[/I]
 


Vojin Vidanovic

Posts 770
03 Sep 2017 09:29


Daniel Sevo wrote:
 
    Also, when are we going to abandon SysInfo for something more accurate? Maybe SysSpeed would be better from now on? ;-)

Wasnt MiniBench exact "replacement"?
EXTERNAL LINK 


Gunnar von Boehn
(Apollo Team Member)
Posts 5421
03 Sep 2017 11:36


Daniel Sevo wrote:

  What we all know is that the Vampire 2 / Apollo core is a great product that has re-ignited the 68k Amiga world and has potential to outperform most Amiga stuff out there, even PPC..
 

 
Apollo 68080 offers very good performance, and the team works continuously on improving this.
The GOLD 2.7 release as again improving the speed for the people.

Both the Power CPUs and the VAMPIRE are fast.

In a number of benchmarks the Vampire outperforms PowerPC systems already. And with each new CPU release the number of benchmarks where the VAMPIRE wins does increase.

The Vampire did outperform 300 MHz or even 800 Mhz PowerPC in several benchmarks already.
 
Vampire2 runs ADOOM faster than 300 MHz PowerUp Systems.
And also the game 194x-Deluxe and the DRAGON-CROWN demo are showing a performance which even GigaHerz AMIGA-PPC system fail to reach.
 
 
 
 
Daniel Sevo wrote:

[I]P.S. it wasn't obvious if you meant "new accelerators produced and sold *today*" and only looking at ACA 030 stuff and assuming that new Phase 5 is vaporware, then it makes more sense. ;-)[/I] 

 
New produced and sold for AMIGA are today accelerators using
 


  68000
  68020
  TG68
  68030
  and
  68080 = Vampire

 
All of the other CPU cards are very slow compared to the Vampire.
Depending on the used benchmark, the Vampire is 10 times to 20 times faster than the next fastest available accelerator.
 
 
Regarding FPU Performance, the Vampire beats already the other cards too. The next fastest FPU cards you can buy today have an 68882 FPU.
 
According to SYSINFO the VAMPIRE with FEMU is already over 2 times faster than the fastest 68882@50MHz.



Simon Terry

Posts 1
03 Sep 2017 12:32


my shty worst a3640@66 68060    from 1992  on sysinfo

fpu:37 mflops  and says  MMU IN USE


Przemyslaw Tkaczyk

Posts 151
03 Sep 2017 14:31


simon terry wrote:

  fpu:37 mflops  and says  MMU IN USE
 

 
  I am pretty sure that the only use of the MMU in this case is the MapROM feature (Kickstart copied into FastRAM and protected from writing by MMU). And since there's no point of doing that on the Vampire (as the ROM resides in the fastest memory available already) the difference between your MMU and Vampire noMMU is null.
 
  If I use "CPU FASTROM" command on my full 030, SysInfo also says "MMU IN USE" because it does exactly what I described above. And nothing more.


Crow Mohikan

Posts 78
03 Sep 2017 15:04


Przemyslaw Tkaczyk wrote:

simon terry wrote:

    fpu:37 mflops  and says  MMU IN USE
 

 
  I am pretty sure that the only use of the MMU in this case is the MapROM feature (Kickstart copied into FastRAM and protected from writing by MMU). And since there's no point of doing that on the Vampire (as the ROM resides in the fastest memory available already) the difference between your MMU and Vampire noMMU is null.
 
  If I use "CPU FASTROM" command on my full 030, SysInfo also says "MMU IN USE" because it does exactly what I described above. And nothing more.

Think again.It seems you dont have zorro3.

all 68040/68060s built into an Amiga are full CPUs containing a working MMU. This is because Zorro III requires MMU mapping of IO space.



Gunnar von Boehn
(Apollo Team Member)
Posts 5421
03 Sep 2017 15:20


crow mohikan wrote:

  all 68040/68060s built into an Amiga are full CPUs containing a working MMU. This is because Zorro III requires MMU mapping of IO space.
 

 
Thank you for the good example.
Its another example of something, which is solved on 68080 automatic.

The 68080 has also in MMU internally.
And during power-on the MMU setting of the 68080 are automatically correctly setup.




Lorenzo Pistone

Posts 21
03 Sep 2017 21:04


EXTERNAL LINK


Rod March

Posts 119
04 Sep 2017 02:46


Awesome!
 
  Great work Apollo team, can't wait for his to get released.

I think my 68882 @ 50Mhz gave me 0.85 Mflops on sysinfo, so to get 2.2 with software is remarkable!
 
  As a registered user should we expect an email when it's ready?


M Rickan

Posts 177
04 Sep 2017 04:51


This has likely been covered elsewhere but with FPU support included as part of the core for V4 will FEMU be used to provide compatibility for V2?


Chris Dennett

Posts 67
04 Sep 2017 05:05


This was posted in a Vampire group on Facebook, so I assume it is applicable here.
   
   
   
    Vampire V2 X13 Gold 2.7 core, MIPS is integer, MFLOPS is soft-FPU. I believe the difference here to the other screenshot is some important soft-FPU instructions have been put to the core (not all of them).


Chris Dennett

Posts 67
04 Sep 2017 05:33


m rickan wrote:

This has likely been covered elsewhere but with FPU support included as part of the core for V4 will FEMU be used to provide compatibility for V2?

Yes, on the v2 femu is patched into the kickstart and loaded before any FPU detection to trap FPU instructions not implemented in the core and show the Amiga as having an FPU without having to do any manipulations later on. Totally transparent to the user.


Amiga Noob

Posts 1
04 Sep 2017 09:13


Have any fixes been made to v4.00 of Sysinfo? Even the Swedish guy who took it over said he heard of problems back in April. I'm just wondering if the numbers shown are flawed on the Sysinfo v4.00 pic in this thread after all?

I don't have a Vampire myself. I'm still patiently waiting for the 1200 version.


Gunnar von Boehn
(Apollo Team Member)
Posts 5421
04 Sep 2017 14:57


Amiga Noob wrote:

Have any fixes been made to v4.00 of Sysinfo? Even the Swedish guy who took it over said he heard of problems back in April. I'm just wondering if the numbers shown are flawed on the Sysinfo v4.00 pic in this thread after all?

 
All Sysinfo scores worldwide are fake.
 
Sysinfo uses "some" fantasy code to measure Mips and Flops.
All versions of SYSINFO use the same "selfmade" code since forever.
All versions of SYSINFO show numbers which are compatible to previous to previous SYSINFO numbers.
 
SYSINFO never used the correct Dhrystone code, also the MIPS and MFLOPS are fantasy values and never matched the official results of MOTOROLA for the chips.
 
The Sysinfo code wrongly not uses the 2nd Pipe on Super Scalar CPUs like 68060 and 68080.
There for a 68060@50 scores basically only halve the Dhrystone and MIPS numbers in SYSINFO - compared to the real official results from MOTOROLA.
 
Basically for 68060 and 68080 you can assume to x2 the score shown in SYSINFO to get a result close to a real Dhrystone score.

So if SYSINFO "guesses" 160 MIPS for 68080 a correct value would be 320 MIPS


Samuel Devulder

Posts 247
04 Sep 2017 17:24


Gunnar von Boehn wrote:

            The Sysinfo code wrongly not uses the 2nd Pipe on Super Scalar CPUs like 68060 and 68080.
           

Correctly using the 2nd pipe indicates the skills of the ASM coder or, more likely, of the C compiler. So by mis-using/non-using a feature which it is not aware of, sysinfo gives a speed that reflect the bunch of amiga softwares that are also unaware of 060+ features.
           
So I would'nt say that sysinfo results are fake per se. They just give an indication of performance for non highly hardware-optimized code, i.e. average programs.
         
It's only the test-names that sysinfo displays that are fake, not the results. The results do say something interresting for average-joe programs, and can help comparing performances between various amigas with typical old & under-optimized programs (i.e. most of aminet indeed).
 

posts 54page  1 2 3