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
Information about the Apollo CPU and FPU.

Could a Vampire On Latest Core Emulate PPCpage  1 2 3 

Anton Gale

Posts 14
11 Apr 2018 15:12


Gunnar von Boehn wrote:

But of course in the FPGA there is a natural clockrate limitation.

You can not expect to run swift applications in an FPGA which normally need a Gigaherz CPU. An FPGA simply can not reach high enough clockrates.
 

I assume that the upper frequency limits of FPGAs will rise with time as new technologies are developed, but is there ever a time that you could turn the 68080 core into an ASIC or is it just too cost prohibitive?

I've so many questions about this topic that they will become annoying as my rudimentary understanding isn't at a high enough level.

Could multi-core be a possibility in the future (and how would the OS support that)
Can more instructions per cycle be squeezed onto a bigger FPGA?

OK... I am going to stop rambling.


Gunnar von Boehn
(Apollo Team Member)
Posts 3505
11 Apr 2018 15:49


Anton Gale wrote:

I assume that the upper frequency limits of FPGAs will rise with time as new technologies are developed,

FPGA speed is a directly connected with FPGA price.
The VAMP-2 reaches 80-100 MHz

If you use FPGA which cost $500 more, then you can reach 150-180 MHz.

In one project that APOLLO was used, over 200 MHz were reached in a high end FPGA. But those FPGA cost over $1000.
I think this is way to expensive.




Vojin Vidanovic

Posts 485
11 Apr 2018 17:14


Gunnar von Boehn wrote:

  If you use FPGA which cost $500 more, then you can reach 150-180 MHz.
 
  In one project that APOLLO was used, over 200 MHz were reached in a high end FPGA. But those FPGA cost over $1000.
  I think this is way to expensive.
   

  Real life begins at 200Mhz+ ;-)
 
  Good thing is that FPGA prices go down each year.


Anton Gale

Posts 14
11 Apr 2018 20:13


Vojin Vidanovic wrote:

Gunnar von Boehn wrote:

    If you use FPGA which cost $500 more, then you can reach 150-180 MHz.
   
    In one project that APOLLO was used, over 200 MHz were reached in a high end FPGA. But those FPGA cost over $1000.
    I think this is way to expensive.
   

    Real life begins at 200Mhz+ ;-)
 
  Good thing is that FPGA prices go down each year.

I believe some people would pay.  Are they pin for pin compatible or a different family?  I also assume they would need some form of active cooling.


Vojin Vidanovic

Posts 485
11 Apr 2018 21:02


Anton Gale wrote:

  I believe some people would pay.  Are they pin for pin compatible or a different family?  I also assume they would need some form of active cooling.

The same family grows. Current v2 uses Altera Cyclone 3 FPGA.
Upcoming v4 is Altera Cyclone V A5 (77k LE, 28nm technology).

Even within Cyclone V family, A5 is "mid range" or optimal choice in performance/price ration. But Cyclone Vs go up to 300K Les - or up to 4,5 more space for logic and higher clock.

Cyclone V family datasheet
EXTERNAL LINK 
At present time Altera offers Stratix 10 family up to up to 5.5 million logic elements (LEs). It even includes quad core ARM which could bring Android and Linux :-) Mid ranged Aria family, Apollo core was once mentioned to be tested with, is made now in 20 nm technology and offer DSP, integrated LAN, DDR4 ... World of FPGA seems to be interesting day by day, but would be far less if there wasnt things like Vampire.

Within same family, core transfer and further development should be realitevely easy. From teams perspective, there are a lot of new features to add, so core development first.

So there is a space for Vamp to grow, but each generations needs to be developed and used first. v2 now, v4 soon, and who knows what after ...?


Wawa T

Posts 478
11 Apr 2018 22:14



  Good thing is that FPGA prices go down each year.

isnt that they increased last year?


Gunnar von Boehn
(Apollo Team Member)
Posts 3505
11 Apr 2018 22:21


wawa t wrote:


  Good thing is that FPGA prices go down each year.

isnt that they increased last year?

The answer is very simple.

An FPGA model has typically a very stable price, and get usually NOT cheaper over the years.

Every other year a New models generation comes out, and always gets bigger - at the same price.
This means every few years the new models offer more LE for less $.

Ranges which have a certain age, get un-announced.
At this point of time these models become more expensive.

All clear now?


Anton Gale

Posts 14
11 Apr 2018 22:32


Very clear.  Thank you for the explanation.


Mallagan Bellator

Posts 375
11 Apr 2018 23:41


tango one wrote:

Gunnar von Boehn wrote:

 
Adam Whittaker wrote:

  is it powerful enough now to actually emulate PPC processors?
 

 
  Yes 68080 is very powerful, clock by clock more powerful than a PowerPC.
 
 
 

 
 
  but apollo would need to pay a licence to A-eon. ?

A-eon donít own the rights to emulate PPC, and I donít think they, or anyone else, even could own the right to do that. You can make an emulator, and no one has any right to stop you, what so ever


Gregthe Canuck

Posts 264
12 Apr 2018 06:29


Vojin -

Next logical step past Cyclone V is Cyclone 10. Slightly more expense but with more LE and other functionality (as Gunnar mentioned). However this series has only really been out a few months. It needs to mature for a while. Could be a good candidate for a theoretical V5 series in 2019-2020?

More info here: EXTERNAL LINK 



Vojin Vidanovic

Posts 485
12 Apr 2018 16:44


gregthe canuck wrote:

Vojin -
 
  Next logical step past Cyclone V is Cyclone 10

Surely, its easier to transfer the work within same Cyclone family, as V4 Vampires should demonstrate. We first have to live their potential used to a maximum.


Samuel Crow

Posts 301
04 Jul 2018 01:40


http://zero.eik.bme.hu/~balaton/qemu/amiga/ has instructions on how to emulate PPC with QEmu.  It doesn't have a 68080 back-end but at least it has a SAM460 front-end.


M Rickan

Posts 150
04 Jul 2018 03:18


Interesting... although internalizing the emulation would pretty much corner the market and in time outperform the other options.


Calaminici Davide

Posts 17
06 Jul 2018 14:46


Well,

maybe the hardware guru can start a PPC core design, but we are in the risc realm here, different from 68000 lineup.

I wonder what kind of design the hw dudes could realize


Vojin Vidanovic

Posts 485
06 Jul 2018 15:00


Calaminici Davide wrote:

Well,
 
  maybe the hardware guru can start a PPC core design, but we are in the risc realm here, different from 68000 lineup.
 

Why? IBM does that quite well. Plus 200+Mhz Vamp could emulate PPC at half that speed with good JIT, which is about Blizz speed.

Only real world use would be emulation of 603/604 at decent speed and enabling WarpUp/PowerUp software where there were no 68k versions (few games namely). However performance there is often tied with 3d, where Vamp does not excel currently.


Steve Ferrell

Posts 217
06 Jul 2018 15:51


@thread

What would be the point of emulating PPC on a Vampire? Simply to run OS4 and OS4 apps? That wouldn't open up very much software in terms of what's already available via 68K and the performance hit would be terrible. OS4 has been dead for years and after having been a user of that OS from 2007 to 2011, in my opinion it should be allowed to remain in the grave.  I spent thousands on the hardware and operating systems (I had my PegII dual-booting OS4 and Morphos) and speaking from experience OS4 has nothing to offer that I couldn't get back in 1998 from my copy of Windows 98. The performance of OS4 on my Peg was pretty terrible (MorphOS would just fly though) and with the lack of useful software I had a very expensive doorstop.  It became a curiosity that sat unused in my office.  68K Amigas certainly have a more robust selection of apps to choose from too.  It might be an interesting academic exercise to emulate PPC on a Vampire but I can't see anything useful coming of it. And if all you want to do is run OS4, there are already better options for doing so, both real and emulated.


Adam Whittaker

Posts 133
06 Jul 2018 16:10


because - just simply becasue


Gunnar von Boehn
(Apollo Team Member)
Posts 3505
06 Jul 2018 16:57


Steve Ferrell wrote:

@thread
 
What would be the point of emulating PPC on a Vampire?

Hey Steve,

I agree with you that emulating x86 will give you much more Software.



Vojin Vidanovic

Posts 485
06 Jul 2018 17:28


Steve Ferrell wrote:

68K Amigas certainly have a more robust selection of apps to choose from too.  It might be an interesting academic exercise to emulate PPC on a Vampire but I can't see anything useful coming of it. And if all you want to do is run OS4, there are already better options for doing so, both real and emulated.

Until Libre Office is seen, only example of productivity OS4 software well done seems to be Emotion video player.Since we have Riva there really isnt a point there.

Emulating x86  at 66-100Mhz 486/P5 does not help that much either.

Bringing  back 68k Linux and having a Vamp kernel could be a fastest way to bring a bit more updated apps faster and more native.


Samuel Crow

Posts 301
06 Jul 2018 17:51


Vojin Vidanovic wrote:

Steve Ferrell wrote:

  68K Amigas certainly have a more robust selection of apps to choose from too.  It might be an interesting academic exercise to emulate PPC on a Vampire but I can't see anything useful coming of it. And if all you want to do is run OS4, there are already better options for doing so, both real and emulated.
 

 
  Until Libre Office is seen, only example of productivity OS4 software well done seems to be Emotion video player.Since we have Riva there really isnt a point there.
 
  Emulating x86  at 66-100Mhz 486/P5 does not help that much either.
 
  Bringing  back 68k Linux and having a Vamp kernel could be a fastest way to bring a bit more updated apps faster and more native.

Running Linux on less than 256MB sucks.  Running OS4 on less than 256MB sucks too.  MorphOS would be more interesting but still not on Efika-level RAM.  AROS already has 68k support and just needs bugfixes to ATA.device .  What's left?  MacOSX Leopard sucks on anything less than 2 GB!

posts 60page  1 2 3