Apollo Accelerators

I read somewhere on an Atari forum that a possible stand alone version is planned.

From what I understand about modern CPU's, it is possible to use microcode to translate instructions,
I wonder, would it be possible to write new microcode for the newest intel processors, thus enabling them to
execute 68k instructions natively instead of x86-64 ?

(and NO intel management engine)

Then a custom motherboard could be made where the remaining needed stuff is handled for example by multiple FPGA's (one for GPU and one for everything else?)

maybe even with pcie 3 x16 for gtx1080 vulcan ?

This would potentially enable an Amiga compatible computer that is as fast, if not faster than when the CPU's were used for intel instructions ?

As well as enable for insane amounts of memory, like on intel based computers... ie. Something like 64 GB fast ram and xxx GB chip?

for RTOS, I think several open source RTOS exists for 68k if I am not mistaken... theese would need additional drivers... but besides that...

then it could be an ultra fast platform, with guarantee that there are no debug or backdoors etc to worry about ... as well as several very light OS options...

maybe even let all x86-64 needed virtualization stuff remain as part of the microcode and voila...
1:1 virtual pc that could run for example windows 10 without any effort at all...

it would revolutionize the IT world...

So what are peoples thoughts on this? possible? or?

---

edit..

just did a speed test of how fast I could emulate an amiga using a i7-6700K at 4 GHz...
I got 226 MIPS emulated amiga goodness in 32 bit winuae...

If the i7-6700K could run 68k directly with its microcode, then I suspect that number would increase dramatically

I still think the vampire project is fantastic though and it is a project that has taken amiga in to the new millennia...
I suspect that both atari and amiga will be revitalized by the awesome vampire project... MY deepest respect!

A standalone system is very much planned but this will use the Apollo core in an FPGA just as the current accelerators do.

I don't know to what degree microcode on Intel CPUs can be modified, it's an interesting idea but could end up being a distraction if the Apollo team themselves were to start looking at it when there's still so much to do with the FPGA (and maybe in the future moving to creating their own ASIC). If an expert in the field was to look at it and if doable take it on then great

The problem with graphics cards like the GTX1080 Vulcan is that the GPUs are usually closed so it's very difficult if not impossible to write good drivers for them on another architecture. Drivers are one of the biggest issues with the new hardware even when it's fully documented due to the small number of developers able and willing to devote the time required for little or no financial reward.

IanP wrote:

A standalone system is very much planned but this will use the Apollo core in an FPGA just as the current accelerators do.

I don't know to what degree microcode on Intel CPUs can be modified, it's an interesting idea but could end up being a distraction if the Apollo team themselves were to start looking at it when there's still so much to do with the FPGA (and maybe in the future moving to creating their own ASIC). If an expert in the field was to look at it and if doable take it on then great

The problem with graphics cards like the GTX1080 Vulcan is that the GPUs are usually closed so it's very difficult if not impossible to write good drivers for them on another architecture. Drivers are one of the biggest issues with the new hardware even when it's fully documented due to the small number of developers able and willing to devote the time required for little or no financial reward.

I guess you are Right...

But wouldnt the biggest problems be to find someone with expert knowledge of intel microcode and
To find someone with the time to write a generic linux driver wrapper for various other operating systems like amigaOS etc ?

Linux has drivers for almost anything incl gtx1080
the vulkan api would still need to be added to the target os aswell though...

I just cant help but dream of how nice an 68k intel CPU could be

I am looking very much forward to the FPGA based stand alone solutions though .. Since dispensing with the ancient amiga motherboards All together most likely Will translate to increased performance compared to the current vampires

cbmfreak wrote:

To find someone with the time to write a generic linux driver wrapper for various other operating systems like amigaOS etc ?

This has been done a number of times using a thin Linux host, most notably for Amithlon.

Gunnar is the person to ask about microcode - you can catch him on IRC.

ne_one wrote:

cbmfreak wrote:

To find someone with the time to write a generic linux driver wrapper for various other operating systems like amigaOS etc ?

This has been done a number of times using a thin Linux host, most notably for Amithlon.

Gunnar is the person to ask about microcode - you can catch him on IRC.

thanks for the info but as it has been mentioned in this thread before... I suspect that now is not the time.... I think the vampire team needs time to complete all the current projects and all planned projects before using any time or energy on something like this.

It would make for an interesting experiment and could potentially be a future project for them at one point perhaps? It would be quite the challenge though.

I just wish that AMD would think outside the box and make some 68k versions of their hardware includimg drivers... but I guess its to small a market for them and the AMD/ATI guys currently are way to busy with updating drivers and catching intel and nvidia I suspect.

cbmfreak wrote:

ne_one wrote:

I just wish that AMD would think outside the box and make some 68k versions of their hardware includimg drivers... but I guess its to small a market for them and the AMD/ATI guys currently are way to busy with updating drivers and catching intel and nvidia I suspect.

It's a safe bet that they have a bit more at stake than the Vampire crew.

I think many people would like to see 68K Macintosh Apollo accelerators. I know for example that there is a big community around the Mac SE/30. But targeting other Macintosh models might be a better idea. This is because the main reason the people are interested in the Mac SE/30 is that it supports 128MiB of RAM which a Vampire II equivalent board could easily work around on other Macs.

Here are some Macs that might be interesting targets in my opinion:

• Macintosh 128K
• Macintosh 512K
• Macintosh Plus (longest-produced Macintosh ever)
• Macintosh SE
• Macintosh Classic
• Macintosh SE/30
• Macintosh Classic II
• Macintosh Color Classic

I highlighted the Macintosh Plus and Macintosh SE because they support a wide range of Mac OS versions (up to 7.5.5) and they are widely available.

The only real issue with these old Mac's is that there wouldn't be much point in getting HDMI on them (although 24/32 bit gray scale depth would be nice). Maybe 3D acceleration at low resolution to get OpenGL games would be nice too (think Quake or Quake II).

Well, the problem is that the ressources of the team are not infinite. We can't go on newer design when the already planed are not finished Currently, MacOS run very well thanks to Fusion and Shapeshifter on Amiga, but maybe in the future there is a possibility to make a new board for these Macs, how know.

guibrush wrote:

Well, the problem is that the ressources of the team are not infinite. We can't go on newer design when the already planed are not finished Currently, MacOS run very well thanks to Fusion and Shapeshifter on Amiga, but maybe in the future there is a possibility to make a new board for these Macs, how know.

It's clear that the team should finish what they started.

That said what is planned can always change. The real question is what is more profitable currently? Also if we are to ever realistically see an ASIC version of Apollo 68080 you will need much more than the Amiga community alone to make it happen (the costs are just tremendous). We would need a board that is a drop in replacement for most 68000/68020 sockets and ideally be mostly agnostic to the hardware (with addional enhancements for popular platforms like HDMI and SDcard on the Amiga).

I think the vintage Mac market is possibly much larger than the Amiga market that is why I suggest they look into it. More cash flow means better product for everyone.

ericdc30 wrote:

I think many people would like to see 68K Macintosh Apollo accelerators.

I think that there would be almost no interest in 68K Mac accelerators.

68K Mac production stopped around the same time that Commodore went bankrupt.

Apple still produces Macs.

Most people interested in 20 to 30 year old Macs want to use them as they were. (Only 2 of the models that you listed had color displays!)

Fans of the Amiga, Atari ST and X68000 (strong gaming computers) would likely have the most interest in the Apollo core.