exxos's Atari, Amiga & retro forum

terriblefire wrote: 03 Jul 2020 13:17

Cyprian wrote: 03 Jul 2020 13:02

is there any 060 FPGA core around?

Not open source.

Sorry for off topic. How many hours of work do you think fpga 060 implementation would take?

arkadiusz.makarenko wrote: 03 Jul 2020 14:01

terriblefire wrote: 03 Jul 2020 13:17

Not open source.

Sorry for off topic. How many hours of work do you think fpga 060 implementation would take?

It would be years of effort unless we got the coldfire verilog sources.. which is where i think some other projects started from.

terriblefire wrote: 03 Jul 2020 14:21 It would be years of effort unless we got the coldfire verilog sources.. which is where i think some other projects started from.

V....re one? :)
anyway sounds really cool

terriblefire wrote: 03 Jul 2020 11:56

nibiru wrote: 03 Jul 2020 11:40 I think this would be the greatest project ever. Full Rev6 060s are close to depletion, prices have skyrocketed already, and all the new boards are going to be sold without an 060, possibly reducing a lot the interest in them (or making them way more expensive than they could have been).

Whereas a FPGA core would provide way more reliable hardware, built in 2020, more features, speed, and probably cost the same if not less.
I'd make a Kickstarter out of it. Perks: once the core is done, a socket + CPU to all backers :mrgreen:

No i'll never do a kickstarter.. It puts me under too much pressure and its not like i need money.

Yeah I agree with that. It's not for the money but at least to cover the effort and expenses in sourcing components, testing, assembling,
etc.
My idea was more: once it's done and working, collect orders from KS, so the only pressure would be making the final boards, not to deliver a working core. But it still involves pressure, so... back up!

I've no problem getting boards once i have a working core. Thats the easy bit.

terriblefire wrote: 05 Jul 2020 17:06 I've no problem getting boards once i have a working core. Thats the easy bit.

I agree, but I said that since I believe extra money is a way to say "thank you for your hard work and dedication". And I think people would gladly spend quite more for a FPGA 68060 than it actually cost as they are already doing now to find decent CPUs.

But I guess this also goes against the open-source spirit, and that would be quite necessary in developing a new core from scratch.

Then to add something to the topic: DO REALLY Amigas/STs need FPGA 68060s?
Wouldn't a pumped-up 68030 be somewhat easier to accomplish?

Hi,

The 060 and even the 030 designs are complex, that's because due to manufacturing process limitations you had to go complex to get performance, caches, pipelining, etc

With modern FPGAs the 30 years of process improvements means we get clock frequency for free. So you don't need complexity to get performance, we just brute force it with clocks.

Designs like the Vampire 1 and Mike Sterling's designs are pretty much the tg68 open core, clocked fast, and they deliver performance past 060s.

Tg68 does 020 at 40mhz with very fast ram and a single cycle cache access. Remember that CPUs have to fetch instructions from memory unless they are cached, and older devices had very small caches compared to today.

Basically no point doing a complex CPU architecture on a FPGA, the Vampire4 has pushed it pretty far as they want to stay on a, relatively cheap FPGA.

nibiru wrote: 06 Jul 2020 16:38 I agree, but I said that since I believe extra money is a way to say "thank you for your hard work and dedication". And I think people would gladly spend quite more for a FPGA 68060 than it actually cost as they are already doing now to find decent CPUs.

But I guess this also goes against the open-source spirit, and that would be quite necessary in developing a new core from scratch.

Then to add something to the topic: DO REALLY Amigas/STs need FPGA 68060s?
Wouldn't a pumped-up 68030 be somewhat easier to accomplish?

Kickstarter is more pain to me than a help. So i'll never do one.

amimjf wrote: 06 Jul 2020 17:18 Hi,

The 060 and even the 030 designs are complex, that's because due to manufacturing process limitations you had to go complex to get performance, caches, pipelining, etc

Incorrect. 030 is linear. 060 is superscalar.

With modern FPGAs the 30 years of process improvements means we get clock frequency for free. So you don't need complexity to get performance, we just brute force it with clocks.

Designs like the Vampire 1 and Mike Sterling's designs are pretty much the tg68 open core, clocked fast, and they deliver performance past 060s.

Tg68 does 020 at 40mhz with very fast ram and a single cycle cache access. Remember that CPUs have to fetch instructions from memory unless they are cached, and older devices had very small caches compared to today.

Mike's core doesnt get to 060 levels. Nowhere near the last time i checked. It clocks to 66Mhz but has no super-scalar pipeline. Its good but its not good enough for me. Also no FPU and no MMU.

The Apollo core is a complex design and given its closed source we want a new FPGA design to break the monopoly. Its superscalar and pipelined.

Basically no point doing a complex CPU architecture on a FPGA, the Vampire4 has pushed it pretty far as they want to stay on a, relatively cheap FPGA.

Yes there. Because Vampire is not open.

My information based on https://www.mike-stirling.com/2020/03/a ... m-sd-card/

Results against 040 show it's pretty close to what an 060 would do, and that's on a pretty standard/cheap FPGA.

amimjf wrote: 06 Jul 2020 20:13 My information based on https://www.mike-stirling.com/2020/03/a ... m-sd-card/

Results against 040 show it's pretty close to what an 060 would do, and that's on a pretty standard/cheap FPGA.

But no 060 instruction set... no FPU.. no MMU.

So you cant play the 060 demos' which is the whole point.

EDIT: also please dont tell me not to do a project. I've been working and planning on doing this 060 FPGA since 2012.