Overview Features Instructions Performance Forum Downloads Products Reseller Contact

Welcome to the Apollo Forum

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



All TopicsNewsPerformanceGamesDemosApolloVampireCoffinReleases
Performance and Benchmark Results!

200 Mips (overclock) Vampire

Gunnar von Boehn
(Apollo Team Member)
Posts 3869
15 Jun 2019 15:43


Vampire V4 (overclocked)
 
Running at x16 == 113 MHz
 



Sebastian Blanco

Posts 116
15 Jun 2019 19:19


cool !


Roger Andre Lassen

Posts 55
15 Jun 2019 19:41


Phone me now !!!!


Vojin Vidanovic

Posts 1005
15 Jun 2019 20:29


Love the 113Mhz,200+ MIPS,100+ MFLOPS :)
I wonder why chipRAM speed vs A600 is zero, likely too many times faster to be shown (DDR3 fastRAM via Apollo controller mapped as Chip)


Adam A

Posts 51
15 Jun 2019 21:31


I know that FPGA implementations are already running very close to their max operating speed.

I think with a custom firmware and proper cooling it will be easier to even get a higher speed


Mike Kopack

Posts 100
15 Jun 2019 21:43


DAYUM!!!!

That's 2x+ what I'm seeing on my A500 w Vamp 2.2 running core 3 beta!


Hugo Pereira

Posts 35
15 Jun 2019 21:48


Congratulations on this excellent result.
If an Amiga with just 0,8 MIPS works wonders, imagine what one can do with 205 MIPS and 109 MFLOPS!
It makes us dream about an ASIC implementation, even in the distant future.

Thanks master Gunnar and Apollo Team.


Barana V

Posts 48
16 Jun 2019 03:25


=:-0
Wow.
How would this relate to the real world?
  If this core had a CPU only decoder ,would it be able to say - play a DVD?


Gunnar von Boehn
(Apollo Team Member)
Posts 3869
16 Jun 2019 06:00


Vojin Vidanovic wrote:

Love the 113Mhz,200+ MIPS,100+ MFLOPS :)
  I wonder why chipRAM speed vs A600 is zero, likely too many times faster to be shown (DDR3 fastRAM via Apollo controller mapped as Chip)

The chipmemSpeed==0 is caused by coding Bug in SYSINFO.
Our result overflow...



Barana V

Posts 48
16 Jun 2019 06:10


Heh. Gunnar 'broke the guages' !
Brilliant!
Does Gunnar think mpeg-2 playback is possible on the v4 core?


Gunnar von Boehn
(Apollo Team Member)
Posts 3869
16 Jun 2019 06:19


Barana V wrote:

Heh. Gunnar 'broke the guages' !
  Brilliant!
  Does Gunnar think mpeg-2 playback is possible on the v4 core?

RIVA works great as MPEG player.
Maybe you have tested RIVA?

RIVA plays back MPEG1.
I have often played with RIVA videos on AMIGA in 640x400 and even 800x480. Both MPEG2 and MPEG1 have similar CPU requirements for decoding.

Judging what size of video we can play with RIVA, playback of DVD sounds likely to me.


Michael Borrmann

Posts 66
16 Jun 2019 08:11


Nice stuff... :)

Looking forward to my V1200... :)


Hugo Pereira

Posts 35
16 Jun 2019 12:51


Gunnar, with all due respect for your work and the Apollo Team, I would like to ask a question.

Some people suggest using SysSpeed v2.6 to perform the Vampire benchmarks because they say to be more accurate in the values obtained. What do you think of this?

Thanks.


Pedro Cotter
(Apollo Team Member)
Posts 180
16 Jun 2019 17:41


Hugo Pereira wrote:

Gunnar, with all due respect for your work and the Apollo Team, I would like to ask a question.
 
  Some people suggest using SysSpeed v2.6 to perform the Vampire benchmarks because they say to be more accurate in the values obtained. What do you think of this?
 
  Thanks.

Them both are not that good! Much better to use real programs that makes use of processing power. Programs Like Lightwave and Povray really tell you how fast the processor is. MIPS can be deceptive.



Crow Mohikan

Posts 76
16 Jun 2019 19:16


Can I see someday in the future this mips on my A4000?
Maybe


Gunnar von Boehn
(Apollo Team Member)
Posts 3869
17 Jun 2019 07:56


Hugo Pereira wrote:

Gunnar, with all due respect for your work and the Apollo Team, I would like to ask a question.
 
Some people suggest using SysSpeed v2.6 to perform the Vampire benchmarks because they say to be more accurate in the values obtained. What do you think of this?
 
Thanks.

 
You have to understand that "measuring" the speed of a CPU is a complex question.
 
Compared to this measuring a car speed is pretty easy.
 
Lets explain is a little more:
Lets say you CPU support 30 basic Operations
Lets say it support on those 3 different sizes BYTE/WORD/LONG
Lets say it supports on those 15 different Address-modes.
 
This means it can do 1350 different operations.
 
 
How can you correctly measure this and express in one number?
Will you run and count each of the 1350 operations together?
Would this be fair?
 
In real live some operations are more often needed than others.
Generally ADDITION is used more often than DIVIDE.
Would a fair and realistic measurement mean to weighten ADD more than DIV?
 
SYSINFO has a certain mix of operations and measures their total time.
 
SYSPEED has another random mix of operations and measures their total time.
 
None of them both is really "correct" or "wrong".
A much more detailed report gives MINIBENCH.

Real workloads like watching a MOVIE with RIVA, or compressing/decompressing some files, or rendering a 3D scene
are also very good ways to measure and compare performance.
 


Vojin Vidanovic

Posts 1005
17 Jun 2019 10:09


Gunnar von Boehn wrote:

  A much more detailed report gives MINIBENCH.

Speaking of, when the old dead link will be fixed
(minibench download) or uploaded to Aminet?
(old dead link that needs fixing)
EXTERNAL LINK


Andy Hearn

Posts 224
17 Jun 2019 15:41


well, i'd certianly be interested in a POVRay bench, and a quake1 or quake2 demo fps :)
 
  ... just ... you know... if you're looking for things to do :D


Mike Kopack

Posts 100
17 Jun 2019 23:10


Andy Hearn wrote:

well, i'd certianly be interested in a POVRay bench, and a quake1 or quake2 demo fps :)
 
  ... just ... you know... if you're looking for things to do :D

Well there’s already one posted in another thread here for Lightwave and some previous stock clocked Doom ones on YouTube....

posts 19