A600 + Vampire + Phoenix-CPU = running texture demo

[Gunnar]

The demo uses 320x256 screen with 64 colors (EHB)

Link to demo (Amiga executable)
http://www.apollo-core.com/phoenix_demo4

Link to video showing the result
http://www.apollo-core.com/bringup/onlyamiga.avi

[lukassid]

Nice 1. Progress is a good thing!

[Lord Aga]

Great work
Nice to see those little Amigans back on their quality control jobs

[mfilos]

Haha nice one
Only Doomy missing now (rofl).

Keep us posted \o/

[NovaCoder]

I think this is the most interesting development in the Amiga scene over the past few years

We have our first functioning FPGA accelerator board (thanks to Majsta) and an advanced CPU core under development (Phoenix).

Personally I find this direction much more exciting than a completely new motherboard design like the over ambitious Natami. The Vampire board itself is obviously a bit limited but it's only the first attempt at creating a Classic FPGA accelerator. If the Phoenix core can be properly developed by the community hopefully somebody will be given enough of an incentive to develop more advanced FPGA boards for A1200's and big box Amigas.

The Classic community has a real opportunity here to develop a very fast and compatible CPU that can be based on a combination of instructions from 020/040/060. Because a modern board can have much faster memory access (and a larger cache) than the old Classic accelerators it should be possible to create a very special board.

As far as I can tell, all we need now to move things forward are for some talented FPGA coders in the Classic community to invest in Vampire boards and help develop this CPU core.

Would it be possible to set up some kind of bounty for this important project?

[alexh]

Going fully open source is what's needed. Documentation, PCB schematics & layout , HDL and synthesis scripts, plus firmware.

None of this closed source... "released when its ready" rubbish

[pandy71]

Perfect, interested in A500 version.

[matthey]

Originally Posted by NovaCoder

As far as I can tell, all we need now to move things forward are for some talented FPGA coders in the Classic community to invest in Vampire boards and help develop this CPU core.

Programming an fpga is specialized and more difficult than most software programming. Designing and implementing a modern pipelined processor in fpga is even more specialized. I doubt there would be very many qualified individuals in the Amiga community that could contribute to the fpga CPU design and programming. The community could be instrumental in testing and debugging the CPU. The community would also be needed to fund fpga hardware with purchases.

Originally Posted by NovaCoder

Would it be possible to set up some kind of bounty for this important project?

Money could encourage and allow the core developers to work on the project more. I have my doubts that enough money to make a difference could be raised without being connected to the production of some hardware. Even if a bounty for a finished sythesizable core that anyone could use was a goal, there would still be obstacles. The core would have to be adapted for each implementation which is not "easy". Also, what would be the definition of working and debugged? I'm not opposed to the idea and it's not my decision to make. I'm just pointing out some of the possible difficulties.

Originally Posted by alexh

Going fully open source is what's needed. Documentation, PCB schematics & layout , HDL and synthesis scripts, plus firmware.

None of this closed source... "released when its ready" rubbish

It will be "released when it's ready" is better than the perpetual "2 more weeks", IMO. Open source is a great concept from the beneficiary point of view. It's almost like putting in a request for government handouts and an inheritance. Maybe enough charity could bring back the Amiga though .

[wawa]

@gunnar

run some genuine amiga benchmarks please. a synthetic one cannot give the public an idea about the actual performance as you may imagine.

[vitux]

Oh! another project more

[NovaCoder]

Originally Posted by matthey

Programming an fpga is specialized and more difficult than most software programming. Designing and implementing a modern pipelined processor in fpga is even more specialized. I doubt there would be very many qualified individuals in the Amiga community that could contribute to the fpga CPU design and programming. The community could be instrumental in testing and debugging the CPU. The community would also be needed to fund fpga hardware with purchases.

I never said it would be easy, just that it was a great opportunity for the retro community. Let's face it, real Amiga 060 accelerators aren't getting any younger and are also on the expensive side even if you do manage to find one.

You'd be surprised at the FPGA skills that exist either in the community or at least are lurking around the edges. It looks like this Phoenix core is already basically 'working' so it might not take too much work to complete it.

My dream would be for AmigaKit to sell FPGA accelerators with a warranty for A1200/A2000/A3000/A4000 at less than $200. These accelerators would all share the same basic design to reduce costs and would have a decent amount of FASTRAM (say 512 MB). The FPGA core would be constantly improved (speed/compatibly/enhancements) either open source or not.

It would be really cool to have a new boot menu and choose if you wanted to start up with a 14Mhz 68020, Blizzard 1230 Mrk4 or 100Mhz Apollo 060....how freaky would that be

I was thinking, how practical would it be to have an IDE/SATA controller on an FPGA based accelerator card (using a patched scsi.device) and then disabling the on-board IDE. I was thinking about the performance benefit of offloading the HD controller to a separate small FPGA CPU instance, not sure how practical or complicated that would be to achieve.

[turrican3]

great job !
Could we expect an improved a1200 version or 4000 ?
But you did an amazing job already.

[matthey]

Originally Posted by NovaCoder

I never said it would be easy, just that it was a great opportunity for the retro community. Let's face it, real Amiga 060 accelerators aren't getting any younger and are also on the expensive side even if you do manage to find one.

The Amiga computers aren't getting any younger or cheaper either. At some point we will need whole new replacement boards. It's kool to show off a '90s computer with a fast accelerator but it would be good to get rid of the bottlenecks at some point. Also, modern I/O like USB, ethernet, DVI and/or SATA would be convenient.

Originally Posted by NovaCoder

You'd be surprised at the FPGA skills that exist either in the community or at least are lurking around the edges. It looks like this Phoenix core is already basically 'working' so it might not take too much work to complete it.

I know there are some fpga experts lurking around this forum even. There are community members capable of creating whole emulated computer hardware in an fpga which is very impressive. Very few have an in depth understanding of modern processor design. There are a few people like Tobi Flex (TG68) and maybe Yaqube (or was it Mike who added the cache to the TG68) who could contribute but they would probably be learning more than they were contributing at first.

Originally Posted by NovaCoder

My dream would be for AmigaKit to sell FPGA accelerators with a warranty for A1200/A2000/A3000/A4000 at less than $200. These accelerators would all share the same basic design to reduce costs and would have a decent amount of FASTRAM (say 512 MB). The FPGA core would be constantly improved (speed/compatibly/enhancements) either open source or not.

You probably wouldn't get a full featured accelerator for $200 (unless that is AU dollars which are worth more than U.S. monopoly money). I pushed for more memory on the next accelerator but each chip over 64MB needs another memory chip which increases complexity, board size and cost a little per chip. My starting request was for 256MB but I just hope I get my comprise of 128MB. The next accelerator won't be for big box Amigas or ever the 1200 where it probably makes sense to add more memory.

Originally Posted by NovaCoder

It would be really cool to have a new boot menu and choose if you wanted to start up with a 7Mhz 68000, Blizzard 1230 Mrk4 or 100Mhz Apollo 060....how freaky would that be

Selectable slots for cores and 1MB MAPROM kickstarts would be awesome but we will have to see what is practical.

Originally Posted by NovaCoder

I was thinking, how practical would it be to have an IDE/SATA controller on an FPGA based accelerator card (using a patched scsi.device) and then disabling the on-board IDE. I was thinking about the performance benefit of offloading the HD controller to a separate small FPGA CPU instance, not sure how practical or complicated that would be to achieve.

The next target Amigas for the accelerator will probably not have any IDE to disable. SATA requires high speed transceivers in the fpga which adds significantly to the cost. More likely is a single CF slot in place of IDE but supporting better (lower CPU use) transfer modes than the Amiga 1200 or 600 IDE. The fpga may be able to handle some of the work in parallel to the fpga CPU without creating another CPU instance. There will likely not be a helper CPU (controller) for I/O like the fpga Arcade ARM processor and the original CPU probably needs to be disabled when the fpga CPU is used. I am not a hardware guru and I don't know what Majsta will come up with so I'm only letting you know what I expect.

[NovaCoder]

Originally Posted by matthey

The Amiga computers aren't getting any younger or cheaper either. At some point we will need whole new replacement boards. It's kool to show off a '90s computer with a fast accelerator but it would be good to get rid of the bottlenecks at some point. Also, modern I/O like USB, ethernet, DVI and/or SATA would be convenient.

Yep that's true but you'll end up with a completely new computer which IMHO loses the retro feel. You might as well stick a PC motherboard into a A1200 case and boot up straight into WinUAE.

The old AGA chipset can do some pretty cool things if you give it enough MIPS and memory

[wawa]

Originally Posted by matthey

The Amiga computers aren't getting any younger or cheaper either. At some point we will need whole new replacement boards. It's kool to show off a '90s computer with a fast accelerator but it would be good to get rid of the bottlenecks at some point. Also, modern I/O like USB, ethernet, DVI and/or SATA would be convenient.



I know there are some fpga experts lurking around this forum even. There are community members capable of creating whole emulated computer hardware in an fpga which is very impressive. Very few have an in depth understanding of modern processor design. There are a few people like Tobi Flex (TG68) and maybe Yaqube (or was it Mike who added the cache to the TG68) who could contribute but they would probably be learning more than they were contributing at first.



You probably wouldn't get a full featured accelerator for $200 (unless that is AU dollars which are worth more than U.S. monopoly money). I pushed for more memory on the next accelerator but each chip over 64MB needs another memory chip which increases complexity, board size and cost a little per chip. My starting request was for 256MB but I just hope I get my comprise of 128MB. The next accelerator won't be for big box Amigas or ever the 1200 where it probably makes sense to add more memory.



Selectable slots for cores and 1MB MAPROM kickstarts would be awesome but we will have to see what is practical.



The next target Amigas for the accelerator will probably not have any IDE to disable. SATA requires high speed transceivers in the fpga which adds significantly to the cost. More likely is a single CF slot in place of IDE but supporting better (lower CPU use) transfer modes than the Amiga 1200 or 600 IDE. The fpga may be able to handle some of the work in parallel to the fpga CPU without creating another CPU instance. There will likely not be a helper CPU (controller) for I/O like the fpga Arcade ARM processor and the original CPU probably needs to be disabled when the fpga CPU is used. I am not a hardware guru and I don't know what Majsta will come up with so I'm only letting you know what I expect.

matt, one step at a time please, we dont wanna natami syndrom again. as soon aa the accells sell off one might aim at the whole systems again. but let that happen first.

[Leffmann]

Very cool and you've even got your kids to sing along to the cheesy Amiga song

I was discussing this graphics effect with PMC last week, and it's one of these cool demo effect that never gets old or out of style. It can always be made to look impressive.

I can tell from how fast it runs on your A600 that there's a lot of CPU power in that accelerator board. Great job!

[majsta]

I saved some video few days ago:

[ Show youtube player ]

[tolkien]

Impressive!!!

[Lord Aga]

Well, you sure got my like
Is it possible to run this demo on a stock A600 ? Or 2CHIP+ 4FAST one ?

[majsta]

It is possible but code needs to be rewritten because it uses FastRam space that you probably don't have with your memory expansion board.

To correct myself:
On my site I uploaded demo who works on any Amiga, so you can use it. I was assuming that you were asking about the way I run it showed on my video.