• Updated 2023-07-12: Hello, Guest! Welcome back, and be sure to check out this follow-up post about our outage a week or so ago.

Fastest 68k CPU available?..

sh4

New member
Hello folks,

I've tried searching the forum but cannot quite believe it hasn't been discussed here yet: there's an ongoing work of creating a custom 68k "CPU": https://www.buffee.ca/beta-buffees-ordered/

While it has originally targeted Amiga market, I believe it can be used for SE/30 (via some adapter board) as well... There's not much info regarding availability yet (they're in their closed beta), but here's what's known so far:

- it behaves completely the same as the original 030;
- it has ARM CPU inside running custom m68k JIT in bare metal mode (no OS);
- they promise GHz speeds...

GitHub seems pretty abandoned, but work is certainly being done.

Thoughts?
 

chelseayr

Well-known member
@krishnadraws well if we want pure 68k cpus like the one physically in the macintoshes then yes the last one was 68060 which was clocked to 75mhz although its been reported to run up to 133mhz overclocked, almost twice its factory speed!

but on the other hand the nxp coldfire is basically 'newer' abit due to not being linearly compatible I won't be surprised if the system/mac os didn't want to behave on it in theory
 

cheesestraws

Well-known member
pure 68k cpus

None of the 68k processors are strict supersets of their preceding CPUs; this is why the MacOS won't run an on 060. In theory it could be made to work, but in practice it's very unlikely to happen, partly because if you want faster classic MacOS machines, there are perfectly good PowerPC ones out there.

The Coldfire is even further off: it is related to the 680x0, but it's meant to be assembler-level compatible, rather than object code level compatible: and even then, if you look at an OS like EmuTOS that runs on both 68k and Coldfire, there are huge swathes of the lower level code that are #ifdef-ed depending on whether it's running on a real 68k or not. Without source, the chance of MacOS running on a coldfire is essentially zero.

The Buffee project is certainly very Amiga, and has been discussed here before. I am reminded a little of the third of the Twelve Networking Truths from RFC1925, though:

With sufficient thrust, pigs fly just fine. However, this is not necessarily a good idea...

In that community, with its rather strange approach to the passing of time and its rather dodgy view of history, this makes a lot of emotional sense. On a Mac, though, and for this community, what does the buffee actually achieve, other then "well, that's cool, it works"? Nobody in this community has the slightest excuse for not realising that our machines are relicts; and a multi-gigahertz ersatz CPU on an SE/30's bus is not only making the pig fly by main force, but putting lipstick on it before takeoff.

Looking at threads here, the recurring theme is honestly more breaking memory ceilings than CPU ceilings; there are hot-rodders who upgrade all the things past recognition, but the general mood seems to be to explore what would have been possible at the time....
 

Daniël

Well-known member
I do believe the Buffee was talked about, but it got lost in the Great Crash of '21.

None of the 68k processors are strict supersets of their preceding CPUs; this is why the MacOS won't run an on 060. In theory it could be made to work, but in practice it's very unlikely to happen, partly because if you want faster classic MacOS machines, there are perfectly good PowerPC ones out there.

This was actually also the reason Daystar ditched their 060 accelerator, supposedly. They were working on both the 060 and 601 cards, then just came to the conclusion that there wouldn't be any point to releasing both, so the 060 was simply a casualty of the PowerPC switchover.

I am amused by the fact the name of this product is meant to be a (literal?) jab at the Vampire accelerator. The Amiga community is just classy.
 

chelseayr

Well-known member
@daniel yeah I can see why apple didn't take a good look at the 68060 because there isn't a lot you can differ inbetween a 68040@40 and 601@60
(and I guess the question is, if the 601 had not happened was there ever going to be any good beyond-68040 roadmaps at all? I suspect maybe not as there was likely some bound limits in the 'old' design by there)
 

bdurbrow

Well-known member
Mmm… well, I suppose the 68k architecture could have gone the way the x86 did: a pre-execution translation stage to turn the CISC opcodes into RISC-like micro-operations; issued to multiple execution units in parallel, and a virtual core architecture to extract parallelism from multiple threads. For that matter, there’s no reason why that can’t be done now on a FPGA with enough resources.
 

cheesestraws

Well-known member
I can see why apple didn't take a good look at the 68060

I mean, they probably did look at it internally, given that we know that DayStar did. It wouldn't have been particularly hard to make it work, certainly compared to the PowerPC transition, it's just that they didn't.
 

sh4

New member
I do believe the Buffee was talked about, but it got lost in the Great Crash of '21.
Ahhh that's why I cannot find the topic about recreating SE/30 motherboard... it's on archive.org though, but latest snapshot is from March I think.

In that community, with its rather strange approach to the passing of time and its rather dodgy view of history, this makes a lot of emotional sense. On a Mac, though, and for this community, what does the buffee actually achieve, other then "well, that's cool, it works"? Nobody in this community has the slightest excuse for not realising that our machines are relicts; and a multi-gigahertz ersatz CPU on an SE/30's bus is not only making the pig fly by main force, but putting lipstick on it before takeoff.
I agree but in part; wasn't bus overclocking also possible? I mean obviously noone is going to run sysinfo on Mac, but imagine if we remove SE/30 from equation and take some late Quadra?..
 

LaPorta

Well-known member
The real reason you can’t find the SE/30 board topic is that Kai got pissed and ran of with it.
 

demik

Well-known member
The real reason you can’t find the SE/30 board topic is that Kai got pissed and ran of with it.

No Kai didn't have anything to do with it. Bolle did the SE/30 topic and that was lost in the great crash. What I know is that the board are going through some testing right now, but it's somewhat stuck due to real-life constraints.

The topic will be restored when testing is done.
 
Last edited:

mg.man

Well-known member
Kai got pissed and ran of with it.
Yes, but he didn't run off with it... just asked that it be locked...

 
Top