Overview Features Instructions Performance Forum Downloads Products OrderV4 Reseller Contact

Welcome to the Apollo Forum

This forum is for people interested in the APOLLO CPU.
Please read the forum usage manual.



All TopicsNewsPerformanceGamesDemosApolloVampireAROSWorkbenchATARIReleases
Information about the Apollo CPU and FPU.

Is the 68080 Soft Core Available On Its Own?page  1 2 3 

Nonesuch Everwas

Posts 4
28 Mar 2020 07:35


Hi all,

I chanced upon this forum while researching post-68060 implementations of the M68k ISA, intending to see if anything had ever been made which was "close enough" to potentially use in some sort of accelerator card for 68k Macintoshes. What I found here was beyond my wildest imaginings!

The '080 Apollo core is a stunning achievement, incorporating almost everything I might have thought to include if I'd proposed such a project myself -- and then _improving_ on it! I cannot overstate how impressed I am by what this project has implemented so far, and most especially by how few missing features still prevent the Apollo core from qualifying as completely interchangeable with the original Motorola CPU line.

The only slight hiccough, to my eyes, is that the '080 core appears to only be available preloaded in hardware... specifically, in the heart of an accelerator card designed for a completely different computer platform. The Vampire project is extremely impressive in its own right, but I do not believe anyone would claim it to be terribly useful in a Macintosh of any stripe, whether 68k-based or otherwise.

I am very careful to say "APPEARS to only be available" in that last paragraph, because for a newcomer such as myself, it's almost impossible to actually determine that from the very sparse data available. The semi-random assortment of websites -- which, I must add, currently sprawls across at least three different domain names -- which collectively document the heavily intertwined Vampire and Apollo-core projects is far too Byzantine, disorganized, and inadequately cross-referenced for comfort.

As a simple and highly visible example... one fellow here claims to have spent quite a bit of time consolidating various scraps of hearsay and rumour to make a first stab at an FAQ for the project(s). The reaction from the team members seems to have varied between irritation and indignation, because they had already made an FAQ... which is not linked to from anywhere on this site! I was flabbergasted to discover that there are at least two whole other websites involved here, and I only became aware of either one because the developers posted _one_ solitary link halfway through that other guy's failed proto-FAQ thread.

I've since discovered several other places here where highly relevant information, of the sort that would obviate the need for quite a few "obnoxious newbie questions," is only visible by means of a casual mention partway through some other thread. You've got a fancy menu-driven navigation bar running all the way across the top of this site; why in the heck doesn't it mention any of this stuff? An added menu for "Here's these other sites where we keep the important technical information instead of on this technical-information site" would probably make everyone's lives less stressful, in the long run.


Ronnie Beck
(Apollo Team Member)
Posts 178
28 Mar 2020 10:11


Great question!  The only thing I am going to call you out about is the usage of the term "soft core".  The "core" which is loaded on the vampire accelerators contains a mix of things including the FPGA configuration.  This configuration describes how the FPGA should wire itself internally.  So the CPU is a hardware CPU (because the FPGA is a programmable chip) and not a programmed software emulating the m68k which now runs on some pre-existing CPU.  Perhaps you understood that already but I thought I would clarify that point just in case.  And many a flame war has ensued over this very point.

It would be possible to bring the 080 to other platforms but the stumbling block is hardware.  The Vampire cards are purpose built for the Apollo Core (as the CPU and related logics are known as).  An analogous version for the MAC would need to created to bring the 080 to the Mac, as per your example.  BigGun, the principle author of the 080, once did extend an offer to hardware makers who wanted to do such boards.  But I don't know if anyone took up his offer.  Or at least there isn't any hardware around to confirm that they did.

As for the FAQ which you mention, I believe that is the work of a gentleman named Vojin.  He has put considerable effort into his FAQ.  Our official FAQ is here:

wiki.apollo-accelerators.com


D K

Posts 1
28 Mar 2020 20:42


I'm a long long time follower (ever since natami days really) but I don't believe I've posted anything so far, although I can only express full support for the team. Any Amiga related development gets a huge thumbs up from my side.
 
  This thread reminded me however of something I wanted to ask a while ago and Everwas pretty much stole my words, although my interests don't reside on Mac side of things.
 
  So, to keep it short on my end, I'm also interested in similar and my question basically also boils down to: Is it possible to license the CPU core, regardless of licensing model for now?


Matthew Langtry

Posts 176
28 Mar 2020 22:19


I think gunnar explained before has only business to business model (for example individual computers make accelerator card with cyclone fpga buy core from team) there not interested in selling core to so say MiSTer users nor de 10 nano user with there own hats nor own sdrams. I could be wrong people change minds.
 
  original thread EXTERNAL LINK 
I think alternative firmware/ custom kickstart is available for atari st users maybe they could do something similar for mac users?


Ian Parsons

Posts 217
28 Mar 2020 23:34


I've not seen anybody have an issue with the term "soft core" (only emulation). Soft core is the long established term for a processor implemented by re-configuring the FPGA logic and now that many SoC FPGAs are available with hardened or "hard core" processors included, it's important to differentiate between soft and hard cores in an FPGA.



Ian Parsons

Posts 217
28 Mar 2020 23:42


I don't know much about Apple Macs but I always had the impression Apple didn't usually design their Macs with processor upgrades in mind unlike PCs and most Amigas. Apple products tend to be closed systems so if you want an "upgrade" you have to buy the newer models.


Marlon Beijer

Posts 177
29 Mar 2020 19:02


Ian Parsons wrote:

I don't know much about Apple Macs but I always had the impression Apple didn't usually design their Macs with processor upgrades in mind unlike PCs and most Amigas. Apple products tend to be closed systems so if you want an "upgrade" you have to buy the newer models.

The 68k macs DID have processor upgrades. The sonnet PCI PPC cards that Hedeon have gotten to run on Amiga via Mediator expansion were originally made as PPC accelerator cards for Macs.

There were 68k upgrades for Macs as well.



Nonesuch Everwas

Posts 4
29 Mar 2020 21:48


It's great that so many folks have already answered my questions! Since there doesn't seem to be a method to post to the thread without specifically replying to one of the posts within it, I have chosen Ronnie Beck's post as having a slightly higher number of points I wanted to touch on.

In chronological order as they came up in the thread:

- Ian Parsons is exactly correct: the Vampire's use of the 68080 design is implemented as a soft core (a specific but alterable configuration of an FPGA), not as a hard core (a fixed and unalterable hardware block within an FPGA, analogous to an embedded ASIC).

- I am aware that, should someone want to put a 68080 in their Mac, the stumbling block is a lack of supporting hardware. That is the case for ANY use someone might have for an '080! However, no such support could ever be devised without a picture of what it actually needs to _do_. The very first step from the perspective of someone wanting to create such hardware -- in other words, wanting to use a 68080 -- is to discover what form it takes. Does the Apollo core's design implicitly require instantiation on the same family of FPGAs as used in the Vampire? Either way, what characteristics does a specific FPGA need to have if you want it to act as a full-featured 68080?

The sole and entire purpose of my having visited this forum in the first place was to find answers to these questions. There are a lot more places than the Amiga (or the Atari, or the Macintosh) where 68k processors were used in the past, and even today, I very much doubt I'm the only person who would enjoy getting hold of a physically instantiated 68080 derivative purely for hobbyist use.

- I do know that the official FAQ is at (wiki.apollo-accelerators.com) -- but I only discovered it by accident, linked to midway through a different discussion thread. I reiterate: Why on Earth is it not better advertised that vital information about the Apollo core is not in fact located on apollo-core.com? It's not like it would be difficult or unsightly for the navigation bars here to link to any of the other major websites involved in the Apollo core or Vampire projects.

- Matthew Langtry expands on Ronnie Beck's statement that someone called "BigGun" or "Gunnar", who presumably is in a position to negotiate (or at least veto) licensing of the 68080 core, prefers to do so on a business-to-business basis. Between the two of you, that gives me a few more things to investigate; thanks especially to Matthew for providing an actual link.

Again, this sort of thing is the very first step for anyone who wants to _use_ the Apollo core; why on Earth is finding out how to do it so poorly documented? If this Gunnar person is hoping to discourage frivolous, time-wasting inquiries by making it very difficult to figure out where to send them, this is about the least effective way to do it; all it means is that a successful querent is likely to have been irritated into a collossally foul mood by the time they finally do make contact.

- Ian Parsons mentioned having an impression that Apple didn't make much provision for CPU accelerators. That was true of the very earliest Macs and it is even more true of their current offerings, but as Marlon Beijer points out, accelerators for Macs have been widely available regardless of Apple's opinions. In any case, most Macs from the very late 1980s through the mid-2000s were at least somewhat easier to work with. The majority had _some_ form of internal expansion slot, sometimes directly on the system bus. (Later on, especially in the Power Mac era, several models had their CPUs on daughtercards that could be swapped out wholesale -- but I digress.) While there are a lot of models which were either impractical or simply not worth the effort to upgrade, there were a great many more that could and often did benefit from CPU accelerators of one sort or another.

If I remember anything else I wanted to mention, I'll post a follow-up.


Gunnar von Boehn
(Apollo Team Member)
Posts 4790
30 Mar 2020 09:59


Hello Gordon,
 
Thanks for your compliments to the 68080 CPU.
As you correctly said the 68080 is code compatible to previous members of the 68K family and in addition provides many technological advancements, resulting in much improved performance.
 
Building a PCB with modern components as DDR memory and utilizing an Soft-Core in such a platform requires a lot of specialized experience and know-how.

In our experience such projects often requires and greatly benefits from consulting, and support from our logic and 68080 expert team.
If you are interested in building a system using the 68080 and acquiring professional support in this, then please get into direct contact with us.
The contact details can be found on the top button labeled "contact" 
 
 


Tim D

Posts 63
30 Mar 2020 10:52


Of course I'm not an expert and not part of the Apollo team. So Gunnar can correct everything that is wrong about following statements :)

Realise that the Apollo core contains more than "just" the 68080. It also contains the rest of the Amiga chipset.

I don't know whether all the improvements and additions (e.g. MMX, initial 3D engine,...) is all part of the 68080 logic or whether it's actually enhancing the other Amiga "chips" logic. I assume e.g. audio enhancements are not done in 68080 but in the audio chip representation (Pamela).

So before getting your hopes up, it may be that the 68080 logic only contains a subset of all the improvements you read about.


Nixus Minimax

Posts 416
30 Mar 2020 12:59


You can always get an Amiga and a Vampire and then just run Shapeshifter to emulate the Macintosh. The Amiga has always been the better and faster 68k-Macintosh... ;)

EXTERNAL LINK 



Peter Slegg

Posts 21
03 Apr 2020 13:33


This is a brilliant project and I can fully understand that your prime objective is to target Amiga owners. As an Atarian I am a bit jealous (never thought I would say that ;-)
 
  One day it would be nice to see a board that could help other m68k communities.
 
  Perhaps a minimalist cpu board that could be used in a wide variety of machines.
 
  This might appeal most to those who have all ready upgraded our machines. Anyone still using an unchanged 68000 probably isn't interested. Speaking selfishly, an upgrade for 68040/68060 machines would be fantastic.


Gunnar von Boehn
(Apollo Team Member)
Posts 4790
03 Apr 2020 13:54


Peter Slegg wrote:

This is a brilliant project and I can fully understand that your prime objective is to target Amiga owners. As an Atarian I am a bit jealous (never thought I would say that ;-)

Thank you.

BTW Do you know that ATARI EMUTOS runs very well on VAMPIRE?


Gsteemso Del Canuckistan

Posts 7
07 Apr 2020 03:53


Hello again Gunnar,

Thanks for taking the time to correspond with me by email. (I see I also get to thank you for straightening out my account here! "None such" person "ever was" signed up here, so it's a relief to have access under my own name.)

Gunnar von Boehn wrote:

(...) Building a PCB with modern components as DDR memory and utilizing an Soft-Core in such a platform requires a lot of specialized experience and know-how. In our experience such projects often requires and greatly benefits from consulting, and support from our logic and 68080 expert team. (...)

Speaking in absolute terms, you're quite correct. Getting truly competetive levels of performance requires a very exacting and rigorous development process. It's not something anyone sane should attempt without the degree of integration and collaboration you're talking about.

That said, not every use case for "a better 680x0" requires the kind of jaw-dropping, orders-of-magnitude improvement that you are achieving with the Vampire.

From reading through a few of the "performance" threads in this forum, it seems clear that the Amiga community considers 68k-related topics from a very different frame of reference than most other people ever will. I'm going to digress a little to explain why I think that is; its relevance becomes obvious later in this message.

Commodore imploded in what might be the only way that could have provoked the Amiga user base to invest so much energy in their chosen platform's continued viability. Real Amiga hardware was purely 68k-based, and Commodore ceased to exist before it became clear that Motorola wasn't going to keep investing in that processor line beyond the '060.

Once the writing was on the wall, the other contemporaneous 68k platforms either came up with migration strategies  and moved on, or failed to remain relevant and were abandoned... but the Amiga situation was unique.

The Amiga was a viable platform with an aggressively enthusiastic user base when it was orphaned, despite Commodore's vigorous incompetence, and that same highly motivated user base has ended up more or less fending for itself ever since... with astounding success, really. To this day, it's a significant population of knowledgable users who have never completely abandoned the dream.

That, right there, that's a frame of reference that _no_ other population of 68k users can embrace in numbers large enough to be relevant. Even today, I strongly suspect that many Amiga folks would leap at a realistic opportunity to ditch their commodity Intel machine in favour of returning to an uncompromisedly Amiga alternative.

For that market, only the most performant 68k system is a viable option; designed with every scrap of optimization and performance-tuning that can be brought to bear, exactly as you are doing with the Vampire series.

On the other hand, most of the other use cases I can envision don't need the fastest possible implementation. They just need one that's noticeably faster (and, preferably, more available) than whatever else they might be able to scrounge up.

Speaking of available processor upgrades, it's important to note that '060 CPUs in regular use are almost unheard of outside the Amiga world. '060 accelerators targeted for other platforms have been made, but they never attained the kind of relevance they have to Amiga users. For large portions of the world, particularly those with vintage Macs, the most performant thing that someone can readily acquire has hitherto been an overclocked 68040.

I repeat: A _68040_.

As far as I can tell, no one ever sold one of those rated for clocks over 50 MHz. Considering the sheer capability currently shown by the Apollo core, that's not much of a height to surpass.

It's clear from commentary in the "performance" discussions that a large part of the Vampire's astounding efficacy is due to its fast, wide connection between the Apollo core's notional 68080, its cache, and the very fast modern DRAM the Vampire also has on board.

Obviously, peeling all of that off of the design to isolate a bare "68080" processor for individual sale would drastically cut down its performance... but, and this is the part that makes no sense at all from the Amiga community's perspective, to everyone else it _would still be vastly superior to the alternatives._

The best part is that, once the innovators of the world have access to such a thing, they will first be inspired to create modest wonders by directly employing the bare '080, and then to reach out to the Apollo core team in collaboration to make _large_ wonders, through integrating it more naturally into whatever they are building.

This is a dream _I_ can get behind.


The contact details can be found on the top button labeled "contact"

Oh, that was the first place I looked, but it only says who wrote the website. Until someone coincidentally named you as one of the most central developers of the Apollo core, I had no idea there was any connection.

It's in no way certain that a person who operates a website dedicated to a specific development project will have any deep involvement on the hardware end, and the "impressum" (I'm guessing that's a traditional book-publishing thing? I've never seen them outside of European sites, and rarely outside of German ones) mentioned exactly zero topics besides the website. What other conclusion should anyone draw from that, in the absence of prior knowledge?


Gunnar von Boehn
(Apollo Team Member)
Posts 4790
11 Apr 2020 11:41


OK I understand you think about building a CPU accelerator for an 68k-Mac. And for this you look for a 68080 Chip, or for a 68080 mini module.

Such a CPU card you want to build would need a couple thinks to be functions :
- Voltage converted
- Bus levelshifter
- local flash
- ideally local memory (SDRam/DDR)
- JTAG port

Actually such a module will in the end be very similar to the today already existing Vampire accelerators.




Vojin Vidanovic

Posts 1697
11 Apr 2020 12:24


nonesuch everwas wrote:

    As a simple and highly visible example... one fellow here claims to have spent quite a bit of time consolidating various scraps of hearsay and rumour to make a first stab at an FAQ for the project(s). The reaction from the team members seems to have varied between irritation and indignation, because they had already made an FAQ... which is not linked to from anywhere on this site! I was flabbergasted to discover that there are at least two whole other websites involved here, and I only became aware of either one because the developers posted _one_ solitary link halfway through that other guy's failed proto-FAQ thread.
   

   
    Fellow asked team how to help.
   
    With Gunnars blessing he decided to improve FAQ as non coding job.
   
    Did spend quite a lot of time gathering info that is not in official FAQ posting progress both on forum and regularly sending progress to Gunnar with no answer.
   
  Once v1.0 was reached and time to go public,  did not get any official aid, acknowledgement or even proofreading - team edit.
Moved FAQ as purely community one.
   
    I knew they have official FAQ and ideas was to build upon it and expand to troubleshooting (which is separate on official website) I consider FAQs should be extensive, not only technical (difference in approach).
   
    Major difference is also language (official one is more done as tech manual, community as friend2friend help)
    and surely more careful recommendations (user experiments vs fail safe procedures) - company spirit vs community spirit.
   
    Motivation was same as yours: centralize vital info on Vamp as progresses, update and add a bit of overall Amiga info and magic.
   
-------------------------

    On soft core:
   
    Seems it was one time teams answer to existing criticism of close IP ownership of core.
   
    Idea was to offer "unlockable basic free core" down in speed and or features for small licensing, but I havent seen any interest, leave alone end product using it.
 
  More over, I believe team has left the idea, as it isnt mentioned anymore or included in official FAQ. So I decided to skip it from community one too. IF This is changed, it is VITAL INFORMATION for possible partners (ability to license core).
   
    Its quite possible, since Vamps are ordinary FPGA Cyclone cards and similar designs could use it. As well as it would be possible to adopt Mister cores to Vampire if that was the focus (and it isnt).
   
    Community FAQ is developed and I will try to incorporate (and link) "best of" official info too. For the sake of wondering users and seeing same and same questions again and again - a thing that every good FAQ tries to alleviate :)
   
    I highly doubt team policy will change, but thanks for the effort.
 
 
Ronnie Beck wrote:

    As for the FAQ which you mention, I believe that is the work of a gentleman named Vojin.  He has put considerable effort into his FAQ.  Our official FAQ is here:
   
    wiki.apollo-accelerators.com
 

 
  Dear gentle-man,
 
  Can you please provide what should be removed from community FAQ on teams opinion (as you mentioned?). Thanks.


Markus B

Posts 187
11 Apr 2020 15:01


Wasn't there an offer by the team to license the core for free for interested parties?
A somehow limited version?


Vojin Vidanovic

Posts 1697
11 Apr 2020 15:54


Markus B wrote:

Wasn't there an offer by the team to license the core for free for interested parties?
  A somehow limited version?

It was, it was short lived, and no one applied. Never appeared anywhere again. Thats what I know, if anyone from team sais differently, it is so.

2017
EXTERNAL LINK 
EAB flames that probably drove BigGun off
EXTERNAL LINK 
From Böhn’s offer provides that a hardware manufacturer can license a free version of the 68060, which is throttled after booting to the speed of a real 68060 processor, but otherwise has the full range of features of the 68080. This “shareware output” of the CPU could then end users by acquiring a Keycode from Böhn from throttling.


Gunnar von Boehn
(Apollo Team Member)
Posts 4790
11 Apr 2020 17:34


Yes there was a "free offer" and to this offer some requirements where tied.

1) it is required that the project will benefit many users
    e.g. a project plan to produce several hundreds or thousands of cards
2) As our experience shows that 68080 is a complex CPU and needs support to be implemented, it is required that the project company is ready and willing to acquire pay for implementation support if needed. 


Gsteemso Del Canuckistan

Posts 7
12 Apr 2020 00:54


Upon rereading this thread, I think I owe the forum an apology! I have managed to bring up so many topics, in each of my posts thus far, as to make it hard to single out which one I considered to be _the_ topic. I'm sorry for that.

For clarity, my question is:

What steps are needed to make a bare "68080" processor available to the public?

Important background points:

- I do not need (or want) the Apollo Core's actual source code, or whatever the term is. My only interest is in useable physical specimens of it.

- I know that a "bare '080" is unusable without a great deal of supporting components. That's pretty much the whole point.

- The Vampire gets something like 2 orders of magnitude higher performance than a bare 080 would. That's irrelevant.

- If (for example) a compatible 68k processor had been available from Motorola/Freescale/etc., we all know that by now, various hobbyists would have produced accelerators based upon them. We watched it happen, when the '060 came out — although the market forces at the time meant that very few people were both willing and able to pay retail prices for '060 accelerator cards.

- At this point in time, almost no one would go to the trouble of designing a totally new accelerator around the '060. They're hard to get. I also believe that, in most cases, they couldn't possibly bring enough of an improvement over an '040 to justify the effort. Since nothing better is available to purchase, that's the end of the matter. That's the whole reason you had to create the Apollo core in the first place!

- If a bare '080 were readily available for purchase, people would purchase it. If they could get their hands on one, they would begin developing things based upon it. Once such an invention was brought to a functioning state, they would be in a position to collaborate with you on a properly-integrated revision akin to the Vampire.

- However much we might wish otherwise, it's hilariously unlikely that anyone is going to walk up to the Vampire folks with a well-funded plan to use the Apollo core in something. Dragging a new design all the way from "concept" to "cash flow" is _hard_. The Vampire project's folks don't have the time or energy to be in on the beginning stages of every Tom, Rick & Gerhardt's vaporware fantasy; an effort serious enough to be worth their while would almost need to have reached a pretty advanced stage. The trouble is, very few — I would argue zero — people are going to drag their own in-house efforts up to that advanced stage of development if they can't even get a working '080 to experiment with! It's a chicken-and-egg problem.


posts 46page  1 2 3