Using a PC as an accelerator? Had an idea last night....

[bloodline]

@BillHarrison,

If we abstract your idea, you basiclly want to connect a modern x86 to the CPU bus of an Amiga. Then run a 68k emulator on the x86 so that the it looks like a 68k to the Amiga. Perfectly resonable idea and one I toyed with back in 2002... I expect my old posts are still in the archive here somewhere...

But now there are 3 problems...

1) As Piru has pointed out the market for this is so small it would be impossible to make it a resonable price, the cost of developing the hardware and software alone could NEVER be recouped.

2) Once you've built this new CPU board, you would want to add USB, a sound chip, a GFX chip, perhaps S-ATA (you pretty much get all these things for free with the x86 support chips)... then you have to ask yourself what is the Amiga actually doing? Why have I bothered to build an interface to the Amiga when everything is now offloaded to the new board...

3) On my machine at least (a MacBook Pro), it can, with UAE/WinUAE emulate any real Amiga I have... better than my real Amigas...since all output is by default on a nice monitor and via better audio... I don't have to use one of my aged Amiga Mice with the dodgy left click...

In short, it would have been nice 8 years ago... but not now :-)

[Tron2k2]

Here's my two cents here

It seems like modern USB2 or firewire would be more than fast enough for anything that ever took place in a real Amiga's memory bus-think around 30 megs/sec which is well within firewire's scope.  So you could theoretically have the Amiga as a device in the firewire chain ;-) and the Amiga would just have a nice, fast CPU.  I know this whole idea isn't practical or, really, even sane :-)  But we're all insane here, so who cares?!

We've all long since gone past the point of being practical.  I mean really, how many of us here bought a GBA1000 and are fully intent on soldering it together, all 476 components or however many it has, so we can have the fastest Amiga 1000 in town?

I'd buy a microITX PeeCee and stick it next to a 1200 in a heartbeat if I could run those old Haujobb demos twice as fast as I could with a Cyberstorm in a 4000.

I've tried UAE but just couldn't ever get it to run the old hardware banging AGA demos right, if at all.  But that was a few years ago, so is the luck better now with them?

[BillHarrison]

Thats basically the idea.  Anytime you want to use your amiga accelerated, plug it in the pc, fire up the emu software, then boot the amiga and enjoy.  Should be no aga demo etc problems because you will be using the real deal.  Atleast no more than with any other accelerator.  

I think the software part should be simple, there are a vast number of open source 68k emulators.  Put together a hardware interface would be the hardest part.  What bus, is firewire actually fast enough etc.

[Nlandas]

   What if we twist your idea around and look at having a Clone-A or NatAmi on a PCI card for UAE to access for custom chip access. (That is assuming that Clone-A or NatAmi FPGA code would be more compatible with the custom chips that WinUAE software emulation.) It sounds like the Clone-A might be.

I think Clone-A was going to be ECS in it's first version but still. I've thought of using modern PC hardware to run Amiga with a PCI card with custom chips for backwards compatibility for years.

-Nyle

[BillHarrison]

Lots of possibilities, but I figured the simplest and one of the most needed was simply a faster processor.  I see no reason you could not also expand it to do rtg, usb, and some other things in the future, just like many accelerator cards did. But for now, simplest would be plugging it in, and just having it do the "Brain work".  

Having it do more down the line is a definite possibility, and probably one of the better things about the idea.  Once the interface is designed, you can do about anything you want on the pc side with the data.

[BillHarrison]

So on this though, does anyone have some information on the cpu slots of different amigas?  Some information on how an accelerator actually works on an amiga?  Etc? I would like to do some reading on it.

[HenryCase]

@BillHarrison
If you're looking for a 'classic Amiga on steroids' you should check out the Natami:
http://www.natami.net/

[bloodline]

But all this revolves around the flawed assumption that UAE is less compatible than real hardware, this really isn't true anymore.

[Dr_Righteous]

I've often wondered why no one bothered making an x86 version of the PPC Amiga accelerators. Same concept, different CPU. 680x0 + x86. Call the new multi-tasking kernel "Warp86". :idea:

I'd be all for a PCI card with FPGA custom chips as well, designed for use with WinUAE in a similar fashion to Catweasel (perhaps even incorporating it). This way UAE controls the environment and CPU emulation, but lets the hardware functions be passed to, well, hardware.

[hardlink]

Dr_Righteous wrote:
I've often wondered why no one bothered making an x86 version of the PPC Amiga accelerators. Same concept, different CPU. 680x0 + x86. Call the new multi-tasking kernel ...

This seems to be in line with the original post, but we don't need no stinkin' kernel  :-) (My WarpEngine or GVP A530 has no kernal) I have plenty of M68K accelerators that plug directly into the CPU socket and mainly have a faster chip - 68020, etc.

I wonder how much hardware would it take to emulate a really fast 68K? A cpu (not necessarily intel) and an FPGA maybe? Here comes my 1 GHz CDTV!

[Karlos]

Dr_Righteous wrote:
I've often wondered why no one bothered making an x86 version of the PPC Amiga accelerators. Same concept, different CPU. 680x0 + x86. Call the new multi-tasking kernel "Warp86". :idea:

I'd be all for a PCI card with FPGA custom chips as well, designed for use with WinUAE in a similar fashion to Catweasel (perhaps even incorporating it). This way UAE controls the environment and CPU emulation, but lets the hardware functions be passed to, well, hardware.

Such a design would be totally blighted by endian conversion issues. You'd probably need to use the big-endian memory access tricks they did for amithlon.

Another problem with sticking increasingly fast cpus into a classic is the extreme latency of talking to the original hardware and handling interrupts etc raised by it. Eventually, you will have offloaded so much of the original hardware to your x86 stock kit that as bloodline says you'd wonder what it was your original system was even doing.

[Lemmink]

I hate to shoot down your daydreams in midair, but unfortunately this is all pointless.

Even if the development would cost nothing and the software would just fall from a tree it would still be not worth it as it won't accelerate the Amiga as much as you might think.

You should not forget that the rest of the Amiga can not really cope with a "limitless" faster CPU. In fact the existing 060/PPC accelerators are already pushing the limits of the custom chip architecture.

The only Software that would realy take advantage of a mega fast CPU would be software that mainly makes calculation but does not have much graphics our sound output. So you are stuck with numbercrunching software like raytracers.

3D Software will not benefit much from it, as the (fast) calculated data can only be displayed quite slow.
Same goes for Video editing software....

[amigaksi]

>Such a design would be totally blighted by endian conversion issues. You'd probably need to use the big-endian memory access tricks they did for amithlon.

But he is suggesting some interface hardware in between to do the CPU signal translation in real-time.

>Another problem with sticking increasingly fast cpus into a classic is the extreme latency of talking to the original hardware and handling interrupts etc raised by it. Eventually, you will have offloaded so much of the original hardware to your x86 stock kit that as bloodline says you'd wonder what it was your original system was even doing.

I think he's suggesting a pin-compatible replacement of the CPU card with a PC-interface card so as long as the timing is done in hardware (since 0x86 software won't be able handle it), it will be like adding an accelerator card.  The custom chips would have to by synced at the same 7.16Mhz using hardware so it's backward compatible (and NTSC/Pal compatible) with existing software.  Software that uses fastmem and CPU accelerators for computations currently would run faster while the chipset keeps running at the same speed.  Why would you offload sound/graphics to 0x86 hardware?  You are only making a "black-box" type replacement of the CPU.

[hardlink]

amigaksi wrote:
... You are only making a "black-box" type replacement of the CPU.

EXACTLY! (which is what an ICE box is)

It may not make the overall system faster than with an 060, but you can't even buy 060's new anymore, let alone for a CDTV.

[bloodline]

hardlink wrote:

amigaksi wrote:
... You are only making a "black-box" type replacement of the CPU.

EXACTLY! (which is what an ICE box is)

It may not make the overall system faster than with an 060, but you can't even buy 060's new anymore, let alone for a CDTV.

I am typing in English, am I not? Are you unable to understand the points that people are trying to tell you?

[joekster]

Came across this on ebay:
http://cgi.ebay.com/Applied-Microsystems-Code-ICE-68040_W0QQitemZ3868461965QQihZ015QQcategoryZ1504QQrdZ1QQssPageNameZWD1VQQcmdZViewItemQQ_trksidZp1638Q2em118Q2el1247

doesn't seem too expensive... I'm not sure what the onboard cpu is for - maybe there is a fallback mode???