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!

Comparing CINEMA4D On Apollo and 68060

Gunnar von Boehn
(Apollo Team Member)
Posts 6207
12 Dec 2015 07:28


Rendering with Cinema 4D with Vampire on A600 and A1200 with 68060 side by side.

EXTERNAL LINK


Enrique Martos

Posts 27
18 Jul 2016 08:35


both tests was made with INT version for Vampire2 and 68060 ?


Mr-Z EdgeOfPanic

Posts 189
18 Jul 2016 11:41


Guess so because the FPU on Vampire is not active yet.
In the mean time the wait for the 060 should be even longer since silver core 7 made giant speed leap :)


Wawa T

Posts 695
18 Jul 2016 20:27


whats wrong with this comparison, where a2000/060 "totally destroys" vampire2 600 in c4d and lightwave benchmarks? fpu vs int version? that wouldnt be enough, would that?
  EXTERNAL LINK


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
18 Jul 2016 22:24


The screenshots show different screen modes.
You see clearly that the A2000 used 800x600 and the Vampire rendered on 1360x768.
   
Running Cinema 4D on screens and Workbench of different screenmodes will have influenced the measurement.

The Vampire shared Video DMA and CPU Bandwidth.
If you set Screen modes like 1300x760 you will slow down the CPU.
You see this immediately in the SysSpeed score.
The Vampire should with a lower Screenres score 30 Mips More.
So here Speed was traded in for hires Screen mode

The comparison is therefore not correct.
I would even call it measurement error.

I would propose to repeat the Cinema 4D with having both times the same e.g 640x480 screenmode.
   
   


Enrique Martos

Posts 27
18 Jul 2016 22:34


thx Gunnar


Wawa T

Posts 695
18 Jul 2016 23:21


thats about c4d, but what about the lightwave results, where vampire scores 4 times longer if im not mistaken? enormous rendering times all that.

btw, something needs to be done about this rtg issue imho. its bad enoughg that higher res steals memory cycles from the cpu, is this still pure frame buffer? probably at least basic 2d hw acceleration p96 or cgx can take advantage from would be a step forward. as is this certainly doesnt help for instance when testing aros on higher res.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
19 Jul 2016 05:35


wawa t wrote:

  something needs to be done about this rtg issue imho.
 

 
There no RTG issue at all.
The Aros RTG issue that you mean is a Software problem.
 
 
 
wawa t wrote:

  probably at least basic 2d hw acceleration p96 or cgx can take advantage from would be a step forward.
 

 
You have this already!
Adding another 2D Hardware makes no sense, as the Apollo 68080 core has 2D HW instruction included already!
 
The Apollo CPU is very fast it can easily do 2D Blitting operations with 800 MB/sec.
 

Regarding Higher Res:

The Vampire Design is price optimized.
You get over ~ 130 Mips, 128 MB Fastmem
and support for up to Truecolor FULLHD Resolution.

The VAMPIRE today is designed as a compromise offering what we thought the people wanted.
It offer better Screen resolution that most Zorro AMIGA card.
It offer more CPU Speed than 68060 and it offer faster Video rendering than AMIGA Zorro cards - all this for a budget price.

Of course the performance could be increased with a more expensive card. We could easily design a $1000 card which offer you 300 Mips, and 4 times as fast memory.
If this is what you want then make a public poll and if enough people agree we will make this card.

Some hint:
If you want to get really fast Workbench then use 16bit screen instead 32bit and if your TV/Monitor supports it set your Display Refresh rate not to 60 Hz but to 25 or 30Hz.
This will save a lot DMA bandwidth and even 1360x768 will be really fast.


Thierry Atheist

Posts 644
19 Jul 2016 09:04


Gunnar von Boehn wrote:

Of course the performance could be increased with a more expensive card. We could easily design a $1000 card which offer you 300 Mips, and 4 times as fast memory.
If this is what you want then make a public poll and if enough people agree we will make this card.

Where, Gunnar, where??

(Could you make it have 2 Gigabytes?)


Gregthe Canuck

Posts 274
19 Jul 2016 10:56


Kipper already ran a poll on v1200 here back in April: EXTERNAL LINK 
 
As you can see there was more demand for the higher-performance option (priced ~500 Euro) than the current option. Of course this is a small sample size, YMMV, etc... but it shows what is likely an even split between demand for the basic and advanced versions.
 
 


Wawa T

Posts 695
19 Jul 2016 12:01


there was a poll also on a1k, almost even with slight preference for the budget option. afair about fifty people voted for each.

edit: found it:
EXTERNAL LINK 
  gunnar: yes, aros problem is aros software related, but the dma bottleneck doesnt make it better. perhaps the video memory bus should be separated from rather than shared with the general memory bus in future designs, in order not to impact each other. however current design might have some advantage as i understand, as soon as p96 could allocate video memory dynamically according to needs. i think thor mentioned on a1k that its already technically possible.

btw. bad results on lightwave are still unexplained;)


Gregthe Canuck

Posts 274
19 Jul 2016 13:24


Thanks for the pointer to that thread - hadn't seen that one. Don't speak German either but Google translate helps with that.
 
  Looking like roughly 50/50 demand for the regular/turbo models again. Just checked Mouser - the first Arria 10 are now listed with projected delivery January 2017. I am assuming that would be the target chip for the turbo version? Or was it the Cyclone V? Entry-level GX 160 priced at $382 USD in single unit quantities. Ouch. EXTERNAL LINK 


Salteadorneo Salteador

Posts 20
19 Jul 2016 16:29


Cyclone V is more economical and appears to have the same maximum operating speed as the Arria10 800Mzh. Although Arria 10 cheaper has 160 logic gates against the 77,000 of this cyclone V: EXTERNAL LINK (costs about 80 euros). The Arria 10 cheaper, I see that is valued at 291.76 euros.


Thierry Atheist

Posts 644
19 Jul 2016 16:56


Thanks Wawa T,

Well, I didn't participate in that poll, so the number would go to
77 budget, and 69+1= 70 of the super duper unit!

52.38% for the slower one to 47.62% for the superior one.

(I would buy 3, if I could afford them.)


Daniel Sevo

Posts 299
20 Jul 2016 02:12


wawa t wrote:

----><8
  btw. bad results on lightwave are still unexplained;)

It would be interesting to see a follow up test once the Apollo FPU is enabled. Core 7 apparently got some nice INT speed gains (except mine of course, which got bricked) so hopefuly, the FPU will get some attention next.


Simo Koivukoski
(Apollo Team Member)
Posts 601
24 Jul 2016 14:03


Cinema 4D ColourText scene comparison between APOLLO CORE 68080 and Apollo 1260 / 80MHz:
                       
                       
   
    Cinema 4D FPU exe will makes the calculation more accurate not faster.

                V3_3375_x13: Integer.exe  0 h 38 min 20 sec
                    Silver7: Integer.exe  0 h 43 min 30 sec
              Apollo1260@80: integer.exe  1 h 24 min 56 sec
                       
              Apollo1260@80: Phase5-FPU.exe 1 h 25 min 18 sec
              Apollo1260@80: Apollo-FPU.exe 1 h 26 min 07 sec
                   
                        Vimeo: EXTERNAL LINK


Szyk Cech

Posts 191
24 Jul 2016 14:59


Simo Koivukoski wrote:

      Cinema 4D FPU exe will makes the calculation more accurate not faster.

Any idea why is this happen? Normally FPU should speed up few times such calculations...



Gregthe Canuck

Posts 274
26 Jul 2016 02:39



Simo - thanks for running an up to date comparison.

Roughly 2x performance of 060@80. Wow!

posts 18