Apollo Team announces developing of Vampire standalone version to run as AMIGA and ATARI ST

[1st1]

Hello, the Apollo team announces the developing of a standalone version of the Vampire. They preview that it could be ready early next year. But the availability of the ATARI version also depends on if they can get support from the ATARI community because of experience with the chipsets.

Souce: http://www.apollo-core.com/knowledge.php?b=1&note=2723

Gunnar von Boehn
(Apollo Team Member)
25 Oct 2016 06:07

Roadmap Standalone for AMIGA and ATARI

As many of you will know already, we are working right now on our new card layout - which is optimized for mass production.

Based on the same features we will do both cards for all AMIGA models as also new Standalones.

We expect the first bigger batch of these Standalone to receive this year, and plan to be able to start normal sales beginning of next year.


These standalones can be made both AMIGA and ATARI compatible.

In regards of AMIGA/ATARI chipset one could either instantiate already existing and tested chipsets - or we could work together on merging our SAGA chipset with an ATARI chipset.

As already announced some time ago - we will open source SAGA when its done.

In regards of ATARI Testing and TOS hacking, and also Chipset development - we would need some help - as none of us if experienced here.

So if you are an ATARI expert and want to help to make this happen then please contact us on IRC.

We hope that our standalone from its feature set can give a new live to both retro platforms.
While we have no final price yet - our plan is to offer you the highest possible value, most memory, and best performance for the money.

Currently Vampire is a turbo card for Amiga 500/600/2000 and soon 1200. It's based on an FPGA which integrates the Apollo "68080" core at 80-100 Mhz, including 68040 compatible FPU (some instructions which are not used on Amiga missing), MCU (a kind of memory protection unit, not a full MMU and not compatible to 680x0 PMMU and 64 bit "AMMX" instructions for Multimedia. Full PMUU is planned in the far future. The FPGA also in the AMIGA version also contains a SAGA graphics (Super AGA, AGA is the graphics chipset from Amiga 1200/4000, SAGA can have more colors and higher resolutions), it contains SD-Card interface and IDE. The SAGA graphcics is also open for developers to make VDI driver, or replace by some graphics core where VDI driver already exists. The Vampire also is prepared to be an advanced soundcard, the VHDL realisation of this function still misses, this could be for example something like Falcon-Sound...? The Apollo core also contains a kind of "dual bus interface", I think this is one "leg" into the original 68000 socket and one highspeed bus for new peripherials. They already runned EmuTOS on an Amiga 2000 with it, a screenshot shares impressive performance boost compared to a TT in CPU speed. They also did already mechanical test of Vampire 500 to fit in a few ST boards.


Impressive numbers for CPU!


No running test yet, due to lack of ST chipset integration. The Apollo Team needs help here by the ATARI community.

What they really need a re competent peoples who do the integration of ST special chipsets and enhancenemts. The most should be already there, for example SuperVidel, ET4000-VHDL-core, MiST-Viking-card, and so. The board layout already seems to be ready, as I understand unfortunatelly no PC standard formfactor like MiniITX, MicroATX, etc. But not public yet.

The discussion is open now and I hope this will attract developers to join their team and make it running as a Super-ST.

[ericgus]

Please keep us posted on this .. I for one would be very interested in a standalone.. (though news is welcome on the addon card too)

[Orion_]

I thought that Rodolphe (CT60 designer) was already working with them ?

[1st1]

I just had a longer phone call with Gunnar (he called me...) and we talked more than an an hour. My impression is that he and Rodolphe know each other, but no cooperation yet. And, yes, Gunnar knows what he does, at least fpr the CPU core and the Amiga chipset. With ATARI he is not familar, help is required here.

A call from me and Gunnar: If you want Apollo as a turbo card inside an ST or as stand alone machine running TOS and you are technically able to help then, contact them. It's not necessary to be able to produce VHDL code, no need to be able to develop new VDI driver, etc. SAGA grahics is able to do all ST / STE / TT / Falcon video modes, even in higer resolutions as the original. It supports also VGA modes (ET4000 ???) and Amiga modes. The only thing is needed is to tell them how they must map the registers to be ATARI compatible. As an entry point I told them to study the Profibuch. Maybe this is basically enough, I don't know.

Besides the stand alone version also a new turbo card is comming, with a bigger/faster FPGA, so "Vampire2" board is already allmost old. The new board will be better for mass production.

[ryan]

At the point that you replace both the CPU and the graphics chipset, how much of the original machine is there and worth keeping? They should just go full FPGA, make something built around a large FPGA that can be mounted in a Mini-ITX case and can run either the Apollo core with supporting VHDL cores for Atari or Amiga chipsets (or any other 68k platform for that matter.) Put it in those new Amiga 1200 cases that are being manufactured, with a new keyboard, and sell it as a stand alone product. I would donate some cash to help make it happen and I'm sure others would too.

[1st1]

For the turbo card, the original machine is the interface to the outside, to peripherials, to the user (keyboard/mouse). Interfaces except for graphics, this can be handled by the SAGA chipset. If the machine is an ST, maybe Vampire is able to mirror screen on original monitor and at the same time to HDMI if you choose a Shifter resolution. Same would be for a vampirized TT or Falcon. By the way, Apollo team is open for such projects, if one wants seriously to create a turbo card for TT (over TTRAM-Socket) or Falcon-System bus, contact them, they support you fully. The computer specific chipsets would be integrated into SAGA register mappings for graphics or attached to the inside system bus. Such components could be closed source or open source (Apollo stays closed, SAGA will be open as soon as it works 100%, Amiga and ATARI compatible (ad a few more...))

The currently planned standalone version is a very small thing only featuring the most neccessary interfaces, so for example no ROM-Port, no Midi, no ATARI specific interfaces. Just keyboard, mouse, joysticks, video, IDE, SD-Card and maybe a few more. It has much less interfaces as Firebee or maybe even MiST. But If ACP and/or Wolfgang F. wants, they can get the Apollo plus SAGA core and make a Firebee II with it instead of Coldfire or more powerfull Suska. Or Apollo Team does something in that direction, ITX/ATX, supporting all interface an Amiga and ST needs. I already explained Gunnar the importance of the ROM port for apps like Cubase 3.1 and Audio. He still does not understand the neccessaryity of DSP support as Apollo with it's 64 Bit AMMX instruction set is much faster, but we know that for some Falcon apps this would be neccessary. It's our job to explain them this, or to present him ready made VHDL-DSP-Core. In Apollo there is well defined and documented industry standard interface for attaching hardware addons. And FPGA has still a lot of space.

[alexh]

Good luck to everyone involved.

I am really surprised they didn't got for Falcon030 first. The user base is much smaller but they already have wide support for 68040 based accelerators and the SuperVidel is also FPGA based and could potentially be ported over instead of SAGA.

I guess it is because the Vampire 500 fits into an ST(fm).

I would imagine the Vampire 600 might fit some STe motherboards where the 68000 is not socketted.

[1st1]

Saga is able to do all the video modes of Videl, SuperVidel and Firebee graphics. So maybe no need to replace SAGA with SuperVidel, just create the corresponding hardware registers fpr VDI driver and map their setting into SAGA. Or, if someone likes to do, make additional SAGA-VDI driver...?

[AtariZoll]

.... He still does not understand the neccessaryity of DSP support as Apollo with it's 64 Bit AMMX instruction set is much faster, but we know that for some Falcon apps this would be neccessary. It's our job to explain them this, or to present him ready made VHDL-DSP-Core. In Apollo there is well defined and documented industry standard interface for attaching hardware addons. And FPGA has still a lot of space.

I guess that idea is that CPU is enough fast so no need for old and in compare to CPU slow DSP - unless it will be made much faster too than original.
Actually, I think that they don't understand why compatibility is important. Who will patch hundreds and thousands of programs to make them working on Vampire II bases Atari clone (actually something like Atari, but as I see low compatibility level) ? As it is now, Falcon SW will have lot of troubles because of different PMMU - some well patched TOS may improve, but someone must do it right.
Good thing is that they started to see that it will be not so simple, and that will need help.
I think that it will take pretty much time. Starting with Falcon may be not good idea. Maybe some ST for beginning.
Without pretty high compatibility level I don't think that it will be attractive for many people.

[Orion_]

unfortunately I'm not skilled enough to help, I hope some atari guys will help them, because it's an awesome project.
Making that kind of card for ST/STe is a great idea because, right now the only solution to get a fast atari is to get a Falcon which is rare and becomes more and more expensive.
Whereas ST are cheap and common !
I think, this is the future of Atari, a fully compatible fast 68k, on original hardware. (or standalone with fpga emulation for original chips)

[calimero]

I guess that idea is that CPU is enough fast so no need for old and in compare to CPU slow DSP - unless it will be made much faster too than original.
Actually, I think that they don't understand why compatibility is important.

I also have impression that they want to "revive" (just like thread on EAB forum says: "Will the Vampire 500 V2 revive the Atari ST scene?") Amiga and maybe Atari as mainstream (?) platforms.

Who will patch hundreds and thousands of programs to make them working on Vampire II bases Atari clone (actually something like Atari, but as I see low compatibility level) ?

they way of thinking is: 'if you want to play old games: use UAE or Hatari' on Vampirized Amiga/Atari

As it is now, Falcon SW will have lot of troubles because of different PMMU - some well patched TOS may improve, but someone must do it right.
Good thing is that they started to see that it will be not so simple, and that will need help.
I think that it will take pretty much time. Starting with Falcon may be not good idea. Maybe some ST for beginning.

There is no blueprint for Falcon custom chips (and Motorola DSP) available, right? It would be to difficult to reproduce Falcon custom chips - Wolfgang spend years on reimplementing ST custom chips in VHDL... http://experiment-s.de/en/

Without pretty high compatibility level I don't think that it will be attractive for many people.

I would certainly like to see where rabbit holes end
but compatibility would be my first concern (I did not like FireBee because of, for my taste, not enough compatibility)

[lp]

.

[Orion_]

the Firebee is open source and almost no hardware guys contributed to make the video better, why would it be different with apollo being open ?
I personally want something useable, that just works.

[calimero]

But if it achieves Hades 060 speeds, CT60 speed, or exceeds those, then you got my money.

Appolo in Vampire is already faster than CT63 100MHz!
128MB of RAM on Vampire is 300MB/s...

[1st1]

@lp : Gunnar is currently studying SuperVidel. SuperVidel is enhanced version of Videl, the graphics chipset of Falcon. He is very positive about it, that it can be mapped to SAGA graphics core. SAGA supports up to 1920x1280 /FullHD) in 30 Hz, or 1280x720 in 60 Hz (650 Mhz HDMI). SuperVidel does as well. So it fits. Apollo includes an industry standard bus protocoll, well and public documentation. So you can attach any hardware you want. If you want, you can also join their team and impleent the PMMU as a subset of their MCU which does all the logic adress mappings.

[lp]

.

[1st1]

Apollo 6808ß0 at 80-90m MHz core clock. Gunnar did some benchmarkings:

Code: Select all

      CPU      MHz   CARD                MIPS
      68020     14    A1200                 2
      TG68            Mist                  4
      68020     28    Blizzard 1220         4
      68030     26    ACA1233               4.2
      68030     50                          6
      Coldfire 266    Firebee               8
      68040     28                         19
      68040     40                         27 
      68060     50                         32
      68060     80    A1200+Apollo1260     46
      68080     78    Vampire GOLD         86
      68080     92    Vampire GOLD2       165

Gold2 is the current core version. But yesterday Gunnar added function for running two AMMX commands in parallel, but that's not part of the benchmark. With AMMX Apollo plays MPEG video smooth, better than 600 Mhz PPC. Firebee is very bad here because the test loop is using DBF instruction which has to be emulated by software lib on Firebee. mfro from atari-home created firebee conform test loop.

This is original benchmark from Gunnar:

Code: Select all

addil #16,%d0
     addil #16,%d1
     addil #16,%d2
     addil #16,%d3
     addal #16,%a0
     addal #16,%a1
     addal #16,%a2
     dbf %d6,0x486

This is from mfro to be coldfire friendly:

Code: Select all

diff bench/loop1.S bench_fb/loop1.S
5c5,6
<  movem.l d0-a6,-(sp)
---
>  lea  -15 * 4(sp),sp
>  movem.l d0-a6,(sp)
9,10c10,11
<  moveq   #1,D6
<  addq    #1,D5
---
>  moveq   #2,D6
>  addq.l    #1,D5
22c23,25
<  dbra    D6,L2
---
>  subq.l #1,d6
>  bne.s  L2
>  // dbra    D6,L2
25c28
<  bne     L1
---
>  bne.s     L1
27c30,31
<  movem.l (sp)+,d0-a6
---
>  movem.l (sp),d0-a6
>  lea  15 * 4(sp),sp

Result:
FireBee scores 194 MIPS

So Apollo is far in front of Falcon with Ct6x or Amiga with 040/060 turbo card. Firebee is still faster, but not that compatible.

[lp]

.

[1st1]

Everything except the core itself will be opensource. Layout, SAGA, atari chipset support, ...

But not Apollo core. Motorola 680x0 CPU and Coldfire is also not open source and you use it. Gunnar has some patents on the Apollo core to protect some technologies he developed for it. Like running 64 bit instruction in 32 bit mode. (ARM, Intel and AMD can't) The Apollo core is fully pipelined with 1 or even two (AMMX) instructions per clock. Apollo core is up to time, like Intel/AMD(ARM CPU technology, but it runs 680x0 code. It's not a 20-30 years ago design. The reason for not going to open source is that he has an investor who wants to do something with Apollo. That means the investor finally wants to have an ASIC. ASIC costs millions of bucks. Speed of the product will be like original thing. But ASIC can run at 1 GHz from what Amiga and Atari would profit. So what is better, open source and allways FPGAs with 100...200 MHz, or finally ASIC with 1 GHz? ASIC and opensource is not possible as the investor don't want that anybody else can use the core to make money. But anyhow, if someone wants to make turbo card for TT or Falcon, or replace Coldfire in Firebee, Gunnar is open for it.

[lp]

.

[rpineau]

There is no blueprint for Falcon custom chips (and Motorola DSP) available, right?

Code: Select all

------------------------------------------------------------------------
----                                                                ----
---- ATARI Falcon COMBEL compatible IP Core                         ----
----                                                                ----
---- This file is part of the SUSKA ATARI clone project.            ----
---- http://www.experiment-s.de                                     ----
----                                                                ----
---- Description:                                                   ----
---- Atari's COMBEL with all features to reach                      ----
---- ATARI Falcon compatibility.                                    ----
----                                                                ----
---- Top level file for use in systems on programmable chips.       ----
----                                                                ----
---- To Do:                                                         ----
---- -                                                              ----
----                                                                ----
---- Author(s):                                                     ----
---- - Wolfgang Foerster, wf@experiment-s.de; wf@inventronik.de     ----
----                                                                ----

straight from Suska vhdl

[lp]

.

[AtariZoll]

Maybe I'm really stupid and don't know enough about CPU's - but is it fair to do benchmarks involving some AMMX instructions (operating with multiple data at once), which are of special use in some MIPS test ?
I will believe this speed ratios when someone neutral will perform some tests.

Good thing that it will be open source. But somehow I don't like all this bombastic CPU talk. How can say that some FPGA is Intel/AMD/ARM technology ? It has 1000x less gates, 40x lower clock . Sure it is not 30 years old design. There is too much if yet. Maybe at 1GHz it will be interesting, of course if price will be not higher than current one - huh ! What I talking - nothing special - when some quad core at 4 GHz costs under 80 Euros then maybe better idea is to do what Apple did already - emulating old 68000 SW on much faster CPU. Doing complete new TOS with new CPU code ...
I don't know is 68000 architecture at all suitable for some concurrent CPU now. That would be great for me - as I like it's ASM. But that seems really unlikely.
I understand why it will be not open source. After all, it's not usual that some CPU is open source. Will small team achieve what Motorola and IBM were not able, and went on Power PC ?

[lp]

.

[1st1]

Small team can achive this with today technology. It's possible to simulate this before baking silicon. CPU design knowledge today is much better than in old times. Gunnar worked on some known CPUs before. The other team members as well. They know what they do. They know things like microcode, branch prediction, pipelining, amd so on. And the are usung some of these technologies, and probably some very new thing.

If you compare Apollo with ARM/Intel/AMD/PPC of today...

ARM is so called "RISC", but has many many instructions and goes more and more to CISC. Can run one instruction per clock per core.
INTEL/AMD is originally CISC, blews up with more and more instructions. Can run one instruction per clock per core. Inside Microcode is RISC.
Motorola 68K is CISC, but has much less instructions than today's ARM, and Intel/AMD. Can not always run one instruction per clock.
Apollo is M68K, but can run one instruction per clock, even 2 AMMX instruction per clock.

My conclusion: Compared to today's ARM, Intel, AMD Apollo is more RISC follwoing the definition of RISC from about 20 years ago (Acorn Archimedes Time) as ARM today. (what a difficut sentence, I hope you understand.)