New 68060 accelerator idea

[dirkies]

I would not mind someone buying the rights to the cyberstorm accelerator for a3/4000, rework it a bit (more memory through a single so-dimm slot instead accepting 1gb module and also onboard CF slot replacing scsi module) and then produce it.

[rarefaction_]

Originally Posted by dirkies

I would not mind someone buying the rights to the cyberstorm accelerator for a3/4000, rework it a bit (more memory through a single so-dimm slot instead accepting 1gb module and also onboard CF slot replacing scsi module) and then produce it.

I have been researching schematics of older accelerators to see what could be improved upon or brought into this millennium

They generally aren't overly complex which is why i wanted a bit more out of this project than just an accelerator.

I think i will start with just the basics first however and then build on that.

the Apollo core is interesting but nothing I really want the beating heart to be a real '060. As for the PPC, well it's not really that important to me.

DSP and high quality audio is however, as well as a modern NIC and RTG

[ptyerman]

68060, memory slots and RTG, the perfect things to work for. Anything more is a bonus!

[rarefaction_]

Originally Posted by ptyerman

68060, memory slots and RTG, the perfect things to work for. Anything more is a bonus!

I think this is a good starting point thanks for the support.

I have read around the place that there may not be such a shortage of 68060 chips as some might think. They are very expensive though

[mech]

Originally Posted by rarefaction_

OK an ambitious but sincerely serious project. Something we've all seen as the 'Unicorn' of Amiga upgrades that never seem to make it past vapour-ware. I have yet to see anything from the UltimatePPC project and mine is similar but initially for the A1200.

Anyways.

Here is my pipe dream so far.

A1200 Accelerator Card:

- full MC68060 (mmu+fpu)
- PowerPC processor (yet to be decided - possibly PWRficient, possibly something beefy like POWER7+)
- 4 x DDR/2/3 SO-DIMM slots
- XMOS DSP for hi-fi onboard audio plus programmable DSP.
- Gigabit Ethernet
- PowerVR or AMD GPU with HDMI output
- 802.11.ac Wifi
- Bluetooth
- MicroSD
- SATA III

Due to the similarities in Specifications and design it may be somewhat compatible with the Amiga X-1000 systems.

With current embedded design techniques this card could easily be made to fit and run nice and cool in a space much smaller than the A1200 expansion bay - That's not where we will run into problems. It's more about compatibility with existing extensions such as Picasso96, AHI and PowerUP/WarpUP.

My goal is to have this thing as compatible as possible with Classic Amiga software. Especially access to the CPUs, RAM, GPU through Picasso96 and OpenGL, DSP and Audio through AHI, ethernet etc.

What are your thoughts guys?

I will probably be flamed for this but here is my opinion.

First you are starting with the wrong base. The 1200 is the worst base to build on because you have so many variables. some revisions have bus problems,some have weak power supplies,crazy amounts of dongle added in to contend with etc,these are variables you will have to contend with. these things will turn into untold support emails and questions.

It makes more sense to build on a proper 32bit design like the A3000 or A4000.They have known,decent power supplies,room for the expansions etc.
Looks what jens has run into with building 1200 addons- out of spec 1200 psu's etc that cause havoc with todays low voltage chips and level shifters etc.

I think you may be underestimating the software compatibility issue. getting the hardware done is no problem, its usually the software and drivers etc that are the hard part.

The xmos is a oddball thing that brings nothing to the design imho. why worry about x-1000 compatibility ever.

68060 chips are still available in quantity unlike what others claim. You have to do some serious digging to find the good masks and watch out for counterfeits but they are out there.

Like other have said, start small, build a 060 accelerator for a machine with a single Dimm or Sodimm slot(we dont need 4 slots since anything from 256MB to 4GB can be on 1 stick.)
I would build it with 1GB ram so there is room for things like mediator in config space etc. Sata would be a great addition. Forget the MicroSD and use a cf or regular sdhc card,the micro ones are a pain and you cant exactly label them.
put the Ethernet and Usb on this card so it is local bus(read direct to the cpu) and does not suffer from zorro3 speeds.
Not sure if Bluetooth is needed but if its easy to add,then go for it.
When you have completed this and made a successful production run,then you will have cart blanch to start some kind of Kickstarter for the next one and are likely to get backing.
Pointing out flaws in a plan is not hating or trolling, building a accelerator is not a trivial task especially one that is feature rich.
If i had to guess, the UltimatePPC guys realised this task and how hard to get compatible software done now,but that is only speculation. the 040 was never a good idea,but i see why they did it,and in some ways it does make sense.

Being a long time amiga person from 1987,i have seen it all,there have been a million let downs and only a few finished projects that came to light in all these years. The well is filled with good intentions,but when people really dig into stuff they see just how complex and hard it is to build stuff like this for amigas. You questions have been asked a million times

Mech

[ptyerman]

Yes but, the A1200 is more common, especially here in Europe. Limiting oneself to a 3000/4000 is suicide really.

[rarefaction_]

Thanks a lot for the input Mech

These are all things I have been taking into consideration. The real reason I chose the A1200 as a base is because A3000s are really hard to find these days. I would love to have one - but I will be paying $500-1000 USD for one.

Thanks also for the links from Amiga30 i look forward to watching them all.. I love videos of Dave Haynie - he's a real inspiration of mine. He has a scruffy mind like me, and by the way he talks about his projects I love the way he attacks problems and finds funny and funky solutions.

One interesting schematic is the A3640 card which I believe he designed.. or at least the A2630.. These cards would be quite easy to evolve into something new-school

I realise drivers and community support are going to be the hardest aspects of this project as you said, getting it to work electrically is easy but drivers and support will be difficult unless it's something that people simply cannot live without in their setup. I see so many a1200 builds with 20 add-on cards and dongles and bits and pieces all hanging together to produce what could technically done on the one expansion card.

I still have a long way to go even in just the planning aspect before i get to the prototype stage.

Keep the ideas coming guys it's great feedback. Thank you.

[mech]

Originally Posted by ptyerman

Yes but, the A1200 is more common, especially here in Europe. Limiting oneself to a 3000/4000 is suicide really.

Seems phase 5 and DCE did well with it as well as Elbox. The only way to fix the 1200 is a whole new motherboard without all the problems and bottlenecks.
Putting this in a 1200 is akin to strapping a jet engine to a vw bug, it may work,but the results wont be good. The complaints back in the day with 1200 users i remember was the slow ide and no way to add a gfx card(no, indivision is not a solution to a gfx card). mediator somewhat fixed this,but there again you have the terrible bottlenecks.

[ptyerman]

The Blizzard 060 cards seemed to do fine, and still do. Apollo cards? Well, best forget them, they either worked or didn't! The PSU is a no brainer, most people have probably already upgraded their PSU if they have upgrades installed, it's not much to upgrade it if they don't.
IDE won't be a problem as SATA has already been mentioned as a possible built in expansion, ditto for the RTG.
As for strapping a jet engine to a VW Beetle, you can say the same for A3000/4000. At the end of the day, it's the designers/developers decision which machine he/she goes for, but I still say it's suicide to limit yourself to the small number of big box Amigas compared to the A500/A1200. That's a possibility where the Ultimate PPC got it wrong and hence went quiet. Right from the beginning folk were dumbfounded why the A1200 was been left out, just because of the sheer numbers of them around!

[Mr.Flibble]

Yes, targeting the A1200 seems very sensible from a numbers point of view, as much as I'd love more options for the A4000, it's even more of a limited market than the A1200...

060 or fast FPGA 68K core, onboard (modern) ram & RTG would be ideal - if you could squeeze ethernet/wifi/sata in, then great.

[mech]

Originally Posted by ptyerman

The Blizzard 060 cards seemed to do fine, and still do. Apollo cards? Well, best forget them, they either worked or didn't! The PSU is a no brainer, most people have probably already upgraded their PSU if they have upgrades installed, it's not much to upgrade it if they don't.
IDE won't be a problem as SATA has already been mentioned as a possible built in expansion, ditto for the RTG.
As for strapping a jet engine to a VW Beetle, you can say the same for A3000/4000. At the end of the day, it's the designers/developers decision which machine he/she goes for, but I still say it's suicide to limit yourself to the small number of big box Amigas compared to the A500/A1200. That's a possibility where the Ultimate PPC got it wrong and hence went quiet. Right from the beginning folk were dumbfounded why the A1200 was been left out, just because of the sheer numbers of them around!

Just the opposite. No one was dumbfounded, if you can read the specs one would ask WHY you expand a 500/1200. they have bottlenecks everywhere and motherboard bugs,no proper expansion bus,terrible psu and keyboard,etc...dont get me started on the crappy clockport.
The window of addressable space/speed on the 1200 mediator is pretty bad too.
Lets make no mistake i own 1 of most common amigas,2 of some,and i do have expanded 1200 with blizzard060 etc, it's why i talk from experience. The clockport is a horrible piece of crap port that should never be used for expansion-even though Jens has done miracles with it. its slow as dirt. At the end of the day a 4000 with csppc,deneb etc runs rings arund the best ppc 1200, and i do mean runs rings around it. the best expanded 1200 runs maybe a third as well as a expanded 3000/4000. If you dont know this you must of not owned one.
the Goal is to move forward,not backward with expansions if at all possible
You get what you pay for and building on the worst foundation never makes sense. Would you build your house on mud?

It may sound like i am way down on the 1200,but its not the case,it has its uses,but it has crippled us with the mentality of building on the least common denominator. Pc clone people dont try to put I7's in 386 boards

[ptyerman]

That's fine, you own a A4000 and want a new expansion for it, most people don't own one and never will. I don't think you and maybe just a few others would fund this project through purchases, that's why the A1200 is the target machine.
Who mentioned clockport? I didn't, I don't use it and possibly never will. Same as mediator, I never mentioned it and I don't use it.
I don't have or want a A4000, you are obviously happy with yours and kudos to you, but don't try and ram it down my or others throat, I don't care if it runs rings round the A1200, the discussion was about the number of machines around which will maximize the sales potential of an upgrade board. A3000/4000 sales just won't cut it.

[Thorham]

Originally Posted by rarefaction_

- full MC68060 (mmu+fpu)
- PowerPC processor (yet to be decided - possibly PWRficient, possibly something beefy like POWER7+)
- 4 x DDR/2/3 SO-DIMM slots
- XMOS DSP for hi-fi onboard audio plus programmable DSP.
- Gigabit Ethernet
- PowerVR or AMD GPU with HDMI output
- 802.11.ac Wifi
- Bluetooth
- MicroSD
- SATA III

If I want all that then I'll just use my peecee

[Michael]

A good ppc can do a good emulation of 58K cpu. AmigaOS4 has is already.
And the card can have boot options menu (like uae, as to what core to emulate in 68k mode, making it even more compatible to classic software)
Plus when running on single cpu we will have much faster system (no cpu switching slowdowns, and ultra fast 68K jit emu for OS progs)

But the project should be the opposite way round.
Basically an OS4 PPC compatible mother board in the form of a trapdoor expansion that uses the A1200 board as a legacy chipset expansion.

[Amiga1992]

Originally Posted by Jope

Go for it, but do it in quiet, only emerging into the limelight once your product is ready.

So much this. As you can see, you already have seen what it creates.

[strim]

Originally Posted by rarefaction_

@strim - I do have a degree in electronics engineering and I was thinking of doing this as a project for my Masters next semester... but i might just work on it in secret if this is the kind of feedback i am bound to get..

Please don't misunderstand me here. I'm not trying to troll. I'm an open source enthusiast and I'd be delighted to see a community-driven project like that. As someone mentioned above, Amiga marked was plagued by projects that were announced, but nothing at all came out of them (well, weren't you annoyed by UltimatePPC failure? seems like you're just making a similarly big announcement now!).

So far I have hard time beliving your project has any chance of success. Why? Because you presented nothing besides what you, yourself called a pipe-dream. At least do your howework and give us a damn block diagram of how you imagine such board should work. There's nothing to work with now, even if I wanted to join the project.

Also:

getting it to work electrically is easy

No, it's not damn easy. Try to design a PPC-m68k bus translator and you'll see.

Before anyone jumps to conclusions, I at least am trying to do my homework: http://eab.abime.net/showthread.php?t=75417&page=3

[Locutus]

Originally Posted by Michael

Basically an OS4 PPC compatible mother board in the form of a trapdoor expansion that uses the A1200 board as a legacy chipset expansion.

Fuck, wouldn't that be irony. exactly like the original Eyetech AmigaOne 1200 'concept' board

[ptyerman]

Originally Posted by Locutus

Fuck, wouldn't that be irony. exactly like the original Eyetech AmigaOne 1200 'concept' board

It would be nice and the way to go if it could be produced cheaply, AND if it was feasible! Definitely getting complicated now though.

[Locutus]

still remains a deluded pipedream (-:

[rarefaction_]

I didn't mean to make it a big announcement that's the thing I am at pre-alpha-concept mid-research phase I haven't even started to make up a bill of materials or anything I haven't even decided on hardware 68k or FPGA / PPC emulation. I think a decent PPC could emulate a 68060 with jit etc.. but I feel like it goes against what makes a Classic Amiga what it is. I didn't really want to cross over into X1000 territory either, I just want some form of useful DSP similar to the Delfina board for the a1200, which uses the motorola 56k DSP... plenty of support and code already exist for this chip, it was also used in the Falcon as many of you would know.