Overview Features Instructions Performance Forum Downloads Products OrderV4 Reseller Contact

Welcome to the Apollo Forum

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



All TopicsNewsPerformanceGamesDemosApolloVampireAROSWorkbenchATARIReleases
Documentation about the Vampire hardware

1080p With Vampirepage  1 2 3 4 5 6 7 8 

Olaf Schoenweiss

Posts 646
01 Jun 2020 09:34


for some amiga in fpga is "not real" enough as for some only amigaos is amiga. So I think there is a market for both products (Vampire and Warp).
 
  At least the high degree of activity around 68k shows that there is a growing market here.
 
  In comparation "NG" is dead


Sean Sk

Posts 368
01 Jun 2020 10:24


Olaf Schoenweiss wrote:

for some amiga in fpga is "not real" enough

 
Yeah I have noticed that a few on EAB in the Warp 1260 thread have been of the opinion that they will only use an accelerator card that has a "real" Motorola CPU on it. I figure that a 68060 is still a 68060 even if it's in a FPGA. Maybe the Warp team could have gone the FPGA 060 route rather than relying on old stocks of 060's being available.
 
Chucky did an excellent youtube video on the Warp 1260 card and it looks to be an excellent card and performer. The only problem is that it's a pretty bulky and ungainly card because of the heatsink required for it. You'll see what I mean when you watch the video. I get the impression that the CPU would run pretty hot being overclocked.
 
Would it have run cooler if the 68060 had been put into a FPGA? Plus you wouldn't need to rely on sourcing old chips.?
 


Gunnar von Boehn
(Apollo Team Member)
Posts 4791
01 Jun 2020 10:35


sean sk wrote:

Maybe the Warp team could have gone the FPGA 060 route

   
But how should they have done this?
There is no 68060 FPGA core available.
 
One option for the WARP team would have been to approach us and work together with us and talk about using th 68080.
The 68080 is not only advanced over the 68060 and faster - it would have also solved the problem that you can not buy any 060.
 
And would have allowed many people to use the WARP.
This number of available, good 68060 will limit the number of WARP users a lot.
   
sean sk wrote:

I get the impression that the CPU would run pretty hot being overclocked.

Getting hot is not so good.
Overclocking and getting very hot usually translates to "reduce livetime"
 
sean sk wrote:

Would it have run cooler if the 68060 had been put into a FPGA? Plus you wouldn't need to rely on sourcing old chips.?

   
Yes of couse you can buy the FPGA new direct from vendor.
This means you have band new chips - not 20 or 25 years old used parts.
 
You can today place an order of 1000 FPGA, and you will reliable get 1000 chips of which you know they are new, tested and all in perfect shape and will last for many years to come.
How can you compare this to the refurbished 68060 CPUs from china?


Gunnar von Boehn
(Apollo Team Member)
Posts 4791
01 Jun 2020 10:57


Olaf Schoenweiss wrote:

for some amiga in fpga is "not real" enough

   
Well the funny part is that the "real 68060" is full if real bugs, issues and problems.
   
Read the many pages of erratas from Motorola to the chip then you know what I mean.
 

 
If you know CPU history then you know the big problems the Motorola had with the 68040, and resulted in the decision of Motorola to drop the 68K platform altogether. And the 68060 came out AFTER this decision was made.
 
Basically you can say it came out: "after the war was over and lost already".

I've deep respect for the developers of the 68060. 
And I think the 68060 is overall a good chip.

But you will know that the 68060 is NOT the chip the development team wanted to have like this.
The team wanted to do a much better chip.

And the team would have done a better chip - if it would have not been developed in a time, when team already knew the war will be lost, and they became under time pressure, and funds started to get cut.

Its all understandable and of course the 68060 needed to be rushed and corners needed to be cut at some point as Motorola did not want to pour more millions of development cost into a project they decide to cancel.


Ray Couzens

Posts 65
01 Jun 2020 11:17


Gunnar von Boehn wrote:

Olaf Schoenweiss wrote:

  for some amiga in fpga is "not real" enough
 

 
  Well the funny part is that the "real 68060" is full if real bugs, issues and problems.
 
  Read the many pages of erratas from Motorola to the chip then you know what I mean.
 
 

I think the argument saying only a 'real' 680x0 CPU is any good is a pointless one.  If we were given a black box containing either a 'real' CPU or FPGA equivalent and not told which, but it behaves exactly as expected or better, then what does it matter what type of processor it is?  A little like the Turing test, if we are none the wiser, then we may think it's an original Motorola 680x0, but in either case not disappointed.

That's my perspective, but each to their own of course.

The obvious advantage to the FPGA version is that it's new and can have many more improvements over the older 'real' chips, and will be available for the future.


Gunnar von Boehn
(Apollo Team Member)
Posts 4791
01 Jun 2020 11:28


Ray Couzens wrote:

Saying only a 'real' 680x0 CPU is any good is a pointless one.

Its common practice for all companies to prototype their CPUs in FPGAs.
This means all the "real" CPUs from IBM or INTEL or AMD or ARM are first prototyped and tested in an FPGA before they "bake" the ASIC.



Ray Couzens

Posts 65
01 Jun 2020 11:52


Gunnar von Boehn wrote:

 
  Its common practice for all companies to prototype their CPUs in FPGAs.
  This means all the "real" CPUs from IBM or INTEL or AMD or ARM are first prototyped and tested in an FPGA before they "bake" the ASIC.
 

It makes perfect sense.  In the case of the 68080, the FPGA performs better than the last Motorola 68060, so an ASIC version of it would be even better.  I'm not sure just how much faster it could be, but imagine by a considerable amount.

Can you see a day when an ASIC version becomes possible?  The cost is obviously too high at the moment.




Michael AMike

Posts 130
01 Jun 2020 12:19


sean sk wrote:

I get the impression that the CPU would run pretty hot being overclocked.

Check the numbers in the video from chucky. After 4 hours clocked@95Mhz the heatsink had a temperature of 35° degrees. (Fan in silent mode) There is also a software tool available which measures ongoing the temperature of the CPU. Thats pretty impressive for a 060.



Sean Sk

Posts 368
01 Jun 2020 12:53


Michael AMike wrote:

Check the numbers in the video from chucky. After 4 hours clocked@95Mhz the heatsink had a temperature of 35° degrees. (Fan in silent mode) There is also a software tool available which measures ongoing the temperature of the CPU. Thats pretty impressive for a 060.

     
Yep I saw that, but how hot would it have gotten without the heatsink? I probably didn't word it properly in my original post, but the point I was trying to make is that it requires a heatsink in the first place. The Apollo core 68080 in an FPGA doesnt even need one. I posed the question (for the purpose of healthy discussion)  that if the Warp team developed a 68060 core in an FPGA, instead of using actual old 68060 CPUs, would it run cooler? Do they have the knowledge and experience to do this? That I dont know the answer to. 


Gunnar von Boehn
(Apollo Team Member)
Posts 4791
01 Jun 2020 14:00


sean sk wrote:

would it run cooler?

Yes of course it would run cooler.
They would not have need the heatsink.

And also the card would reach twice the performance :)

sean sk wrote:

Do they have the knowledge and experience to do this? That I dont know the answer to. 

They only need to create the card. For which they have of course the know how.
The knowledge to create the VHDL file provides the Apollo-Team.


Vladimir Repcak

Posts 327
01 Jun 2020 17:38


It's all slowly sinking in, but from my current understanding, that 105 MHz 060 in the Warp in the video, is basically a Unicorn, right?

Not much luck getting one such on eBay, right?

So, what's -realistically- the next fastest one that one could obtain?

  66 MHz? That would change the performance by a lot...


Gunnar von Boehn
(Apollo Team Member)
Posts 4791
01 Jun 2020 17:57


Vladimir Repcak wrote:

  It's all slowly sinking in, but from my current understanding, that 105 MHz 060 in the Warp in the video, is basically a Unicorn, right?
 

   
The 68060 was sold from Motorola with 50Mhz and 60Mhz.
All other clock rates are overclocking.
   
To give a rough comparison
Using latest SYSINFO 4.3
  - A 50MHz 68060 will score 39 Sysinfo Mips
  - the normal/current x12 Vampire core scores 154 Sysinfo Mips
  An overclocked x14 Vampire core scores 180 Sysinfo Mips
   
   
If your 68060 support overclocking and how much depends on the core revision and on luck or a lot of luck.
   
Overclocking from 50 to 55Mhz all Revisions will do.
But extrem overclocking like from 50 MHz  to 75 Mhz - will more likely work with Rev6.
My experience is that the 68060 FPU tends to calculate wrong before the CPU fails when you do overclocking.
Overclocking to 105 Mhz needs a lucky chip.

 
   


Michael AMike

Posts 130
01 Jun 2020 19:19


Based on the data from chucky I made a comparison with the V1200. As expected is the V1200 in PAL mode faster than the Warp1260 - on the other hand is the RTG speed of the Warp1260 a bit faster. BUT that is only based on one testcase. (Quake)
 

 
  EXTERNAL LINK


Vladimir Repcak

Posts 327
01 Jun 2020 19:38


39 Mips for 50 MHz 060 ? Ouch...

OK, but that changes picture drastically, then.

Disregarding the old Vampire core in that video, it was a comparison between supremely overclocked 060 vs standard clock vampire.

To be fully fair, either it should be a standard clock comparison (e.g. 50 MHz 060) or then the Vampire should have also been overclocked.

I'm still excited about the potential of the 400 MHz ARM in it for gfx purposes (it has some serious potential, if a 3D library was written in hand-optimized ASM), but we need a realistic comparison target (e.g. 50 MHz 060 baseline).


Olaf Schoenweiss

Posts 646
01 Jun 2020 19:55


to have a 400 Mhz ARM processor for graphics is nice but has to be really used by the OS and you need the drivers for it


Michael AMike

Posts 130
01 Jun 2020 20:23


Vladimir Repcak wrote:

39 Mips for 50 MHz 060 ? Ouch...
 
  To be fully fair, either it should be a standard clock comparison (e.g. 50 MHz 060) or then the Vampire should have also been overclocked.
 

That makes no sense - the Warp can deliver the 95Mhz permanently and not for a limited time - this is just the potential of this card. On the other hand - what is the base clock of an Apollo Core? There are 10x cores just like X12 or X14 cores. A comparison of the available stable performance makes sense. If a stable 2.12 X14 core is available, this is the new basis for comparison - same for the Warp if they can reach stable 120Mhz.



Gunnar von Boehn
(Apollo Team Member)
Posts 4791
01 Jun 2020 21:03


Michael AMike wrote:

  That makes no sense
 

 
I think Vlads point is that every Vampire1200 comes with a x12 core.
But the Warp comes without no CPU.
The official clockrating of the 68060 is max 60MHz.

I think the score that were published were for a heave overclocked CPU.
I think Vlads point is that very many 68060 can not be overclocked this extreme.

Mike, you must agree this point is valid?

 
 
 
 
 


Claudio Guglielmotti
(Apollo Team Member)
Posts 182
01 Jun 2020 21:11


@Michael AMike
really, ClickboomQuake gives 17,90 fps in pal mode on my V1200@x12 and 9,98fps in RTG 640x480...



Gunnar von Boehn
(Apollo Team Member)
Posts 4791
01 Jun 2020 21:11


Michael AMike wrote:

    Based on the data from chucky I made a comparison with the V1200. As expected is the V1200 in PAL mode faster than the Warp1260 - on the other hand is the RTG speed of the Warp1260 a bit faster. BUT that is only based on one testcase. (Quake)
         
    EXTERNAL LINK     

Mike, now this is funny what you post.
We all have complete opposite result than you.
Why are your Quake Scores for V1200 so very low?
 
       
For default clock (x12) core of V1200 I get
  ~18 FPS AGA
  ~27 FPS RTG 320
  ~10 FPS RTG 640
     
To me it looks that the default clock V1200 wins in each row against the overclocked 68060.
 
Mike what clock was the 68060?
Was at a regular 68060 or again overclocked to 95Mhz?
 
 

 
 
If I understood your numbers correct,
then the Warp in RTG gets 18 FPS while the Vamp 27 FPS.
 
So it looks like default clocked Vamp is nearly 10 FPS faster than overclocked 68060 - is that what you wanted to say?

Mike can you please explain again?
And help us understand your post?


Claudio Guglielmotti
(Apollo Team Member)
Posts 182
01 Jun 2020 21:20


Here is my clickboom test on a Pal screen
  EXTERNAL LINK

posts 142page  1 2 3 4 5 6 7 8