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
Documentation about the Vampire hardware

Any Chance for a Vampire Core for MISTer?page  1 2 

Mark Mc Fadden

Posts 10
30 May 2020 23:24


Gunnar von Boehn wrote:

Adam Polkosnik wrote:

  I was wondering if there's a chance for a core that could be running on MISTer?
 

 
  Yes, in theory yes.
 
  Porting the core will take significant time and effort.
  If people pay us this time, then we can talk about this.
 
  As side note:
  I have myself some high end FPGA boards here (with Stratix). Those would be able to reach 400 MIPS in Sysinfo.
  But I never cared to port the core because the effort is not justified for one user.
 
  On the Vampire there is a significant user base.
  (Thousands) and the Vampire is continuously produced.
  And will also continue to be produced in the future.
  We regularly produce core updates with improvement for the Vampire.
 
  For a new FPGA system to support I would like to also see this.
  a) Continous support from the team
  b) big enough user base to justify this
  c) a roadmap from the FPGA card producer to guarantee that the cards will be produced the coming years
 
 

Would it be possible to license the 68080?
Fund Dev through preorders and then reap the benefits when released through licensing?


Gunnar von Boehn
(Apollo Team Member)
Posts 4797
31 May 2020 05:00


Mark Mc Fadden wrote:

Would it be possible to license the 68080?

Yes of course this is possible.
   
But lets be truly honest here.
Many people have bought V2, V1200 and V4 recently.
And it is very important for us to support our existing customers to the best possible.
This means for this year our team will certainly invest every single minute to make the Vampire experience the best and better for all our faithful customers.
 
And until all our work items for the existing Vampires are done we honestly have very little free time.

To sum this up: expecting a core for Mister is totally unrealistic.
The Apollo-Team is putting their effort in making the Vampire cores great. And as you know there are many Vampire models in the work.
The V1200 is out now, the V1200 can also work in CD32 and A4000 - so there will be more than enough work for the team.



Markus B

Posts 192
01 Jun 2020 22:04


I clearly get it that the team is limited and pretty packed with the current tasks.

Have you considered to open source the core?
Others out there could be able to contribute and assist in moving the platform forward.
Just like it's a good idea that AROS is open source, it would beneficial to open the Apollo Core to a larger audience.


Gerardo González-Trejo

Posts 41
02 Jun 2020 01:26


Markus B wrote:

  I clearly get it that the team is limited and pretty packed with the current tasks.
   
    Have you considered to open source the core?
    Others out there could be able to contribute and assist in moving the platform forward.
    Just like it's a good idea that AROS is open source, it would beneficial to open the Apollo Core to a larger audience.
 

 
  I don't see any reason to port the core to MISTer/DE or open sourcing it. Who will pay the bills? Do you have any profitable business plan or idea?
 
  Open Source is profitable when widely used by companies and businesses, because they need support and customisations, so they pay the bills...
 
  AROS started as a Hobby OS and actually what it needs is money and pay developers. Power2people (as a crowdfunding platform) collected money to complete many tasks to allow AROS advance and pay developers for the effort done. It was a good idea, but as many efforts on Amiga community, never gave enough money to keep a continuous/intense development. As usually: Good idea, bad implementation :(
 
  Open sourcing the Apollo core/Vampire core/68K80 when still there is not a big users base (100K or more) would only kill/freeze the project at this stage too.
 
  Is there any kind of possibility to open source the core? Well, theoretically yes, but I don't think Apollo team guys would accept give for free their several years amazing work and effort. So selling boards is the only minimally profitable way to advance at this point.
 
  Anyway, as I suggested many times, there will be a point where it will be needed to go for a subscription model on core updating to allow a continuous development on it and AROS too. As soon as that business model would be accepted, as soon as we will see faster development.
 
  I earned my bucks during years giving support for companies on this subscription model implementation, so I know it works despite at the beginning nobody trusted on it. All of them went from low/mid income to a really profitable business model in less than 1 year. But seems "subscription" word still scares lot of people, more on this Amiga community.
 
  ...So the next question is: How much money would we pay for keeping our loved Amiga related hobby alive  monthly, quarterly or yearly? ;)


Steve Ferrell

Posts 393
02 Jun 2020 03:09


@Gerardo González-Trejo

Your remarks are spot on.  Every time I hear someone say that the solution to any IT related problem is to open source it, I roll my eyes and shake my head.  The ones who say this usually have zero coding experience and seem to live under the misconception that there is an army of coders out there just waiting to take on projects and work on them for free whenever such projects pop up on github or other open source repositories.  They don't understand that the code-test-debug cycle is usually quite lengthy and that most coders who are worth their salt have day jobs where they work on difficult projects that pay their bills.  And when these guys and gals have free time on their hands, which is quite rare, they'd rather spend it doing something NOT related to coding.  For proof of what you've been saying, one only has to visit github and look at the thousands of projects that have been open sourced there that haven't had any activity in years.  I'm a professional coder in the field of LIDAR and the last thing I want to do at the end of the day is pick up a keyboard and continue coding for free.  I'd rather spend time with my family or take care of obligations that I have outside of my work place...or travel, or just sleep.


Gunnar von Boehn
(Apollo Team Member)
Posts 4797
02 Jun 2020 06:08


Markus B wrote:

it would beneficial to open the Apollo Core to a larger audience.

Are you sure about this?
What is our goal?
Our goal is to revive AMIGA and to revive 68K.

Our CPU core is constructed like a modern CPU in its internal pipeline. Our goal is at some point to produce a real ASIC.

The Apollo-Core today is about twice to three times as fast as an 68060 ASIC with the same clockspeed.
Putting our project into an ASIC would allow us to reach 1-2 GHz clockrate. This means in an ASIC the Apollo-Core would reach speed of an ~4 GHz 68060 CPU.

We believe that this would be the ideal basis for fully reviving the Amiga.

Building an ASIC is a very long term goal which will a huge amount of money and ideally will need support of big companies.
While we like Open source a lot - every expert in the field tells us that our chances to make this happen and to get support of a big company behind us are much higher when we not open source.


Markus B

Posts 192
02 Jun 2020 08:37


@Gunnar

That's a valid point. If you look at funding the ASIC, commercial perspective is for sure a thing. I won't try to judge on that.

If producing an ASIC is in the range of 1-2 million Euros, that's a long way to go. I'd imagine you're saving some bucks from every card you sell, but still, that's a long way to go.

Developing the core under open source doesn't mean you won't sell cards anymore. But it gives you access to some real skilled developers in this community. Providing it under suitable license can give you legal freedom and commercial protection.

You mentioned the Warp 1260 yourself. Nice card, talented developer. But as you said, very shitty situation with the CPU which can't be resolved.
Interested people could create a branch of the core which is not as advanced but more like the 68060 (most of it FPU and MMU). You made that offer yourself some time ago, to strip down the AC to 060 level. The core could be dual licensed so that commercial products like the Warp 1260 could contribute to the funding of your ASIC plans.

I can totally understand that you don't have the resources to do that and do adoptions for the DE10 or the FPGA Arcade etc.
Let others contribute and make the AC more available and accepted through the community.


Gunnar von Boehn
(Apollo Team Member)
Posts 4797
02 Jun 2020 09:23


Markus B wrote:

Let others contribute and make the AC more available and accepted through the community.

There are 2 problems.

A) To make the ASIC happen the best will not only be saving a penny each day - but to get support from a real big company like SONY or HUAWEI at some point. The likely hood to get their support to make the Amiga happen again is significantly higher if the project is not open source - everybody in the industry tells us this.

B) "contribute" sounds much easier than reality.
To make this clear:
TG68 is a very simple core.
Its easy to understand, simple in design,
and will always be very limited in its performance.

APOLLO-CORE is a very modern CPU - like some beast from INTEL.
Its very complex in its design, its very complex to understand.
To be able to significantly produce contribute you will need to get used to its core design for several month and you ideally should have a history and background as experienced CPU architect.




Markus B

Posts 192
02 Jun 2020 22:35


You can always decide to change the type of license once you get into details about such a deal. Or setup a license which covers commercial aspects of it in advance.

I'm really stunned by the level of expertise you guys have by doing such an interesting development.
But don't underestimte other people. And the possible dynamics in such an approach. I guess you'd be surprised as well.


J Bilander

Posts 6
03 Jun 2020 13:46


How easy or hard would it be to derive an 060 "exact replica" from your 080-work? I think if you could offer an 060-core (stripped down 080) that can be put on a cheap FPGA as a replacement CPU for existing or newly made accelerator cards such as the Warp 1260 this would be of great value to the Amiga community as the 060s starting to get expensive and hard to source, not to mention all the fake relabeled ones. With a reasonably priced 060-core without the 080-finess it could bring some revenue stream to further pay for the 080-improvements and ASICs down the line, and would perhaps make people more appreciating of the hard work the apollo team have put into all this. Rather than competing with all the coming 060-cards, join them with a CPU-option that can be clocked high. Food for thoughts...


Gunnar von Boehn
(Apollo Team Member)
Posts 4797
03 Jun 2020 14:23


J Bilander wrote:

  How easy or hard would it be to derive an 060 "exact replica" from your 080-work? I think if you could offer an 060-core (stripped down 080) that can be put on a cheap FPGA as a replacement CPU for existing or newly made accelerator cards such as the Warp 1260 this would be of great value
 

You will need a small PCB with some extra chips and also a reasonable FPGA will cost some money. So $100 in parts will be reached very quickly. Maybe $200 sales price.
 
As you might know the APOLLO Core has a memory controller build in ..
The APOLLO-CPU can with this reach over 900 MB/sec speed.
 
As you will know the 68060 has no memory controller build in but depends on an external controller. The WARP offer this.
I think inside the WARP, the 68060 will reach about 74 MB/sec speed.
 
This is much less than the Apollo Core would reach directly.
In fact the WARP reaches 10% memory speed of the APOLLO-CORE.
 
If you look at this then its clear that
if you "force" the 68080 inside the 68060 you loose 90% of the memory speed.

You need to understand that APOLLO-CORE can directly speak DDR3 memory protocol - while 68060 can not do this.
Therefore the WARP will have to translate 060 "language" to "DDR".

The solution that you propose will be highly inefficient.
I'm not sure this will make you happy.
From an engineering point of view this looks like a very expensive  solution which is much slower than the Vampire...

Do you think that this really in the best interest of customers?




Markus B

Posts 192
03 Jun 2020 15:32


Absolut valid point, the memory controller.

I don't think the people would be looking for an exact implementation of the 060. Give them fp80 and a 040/060 compatible MMU and nobody will complain.

Except those who always complain about the AC.


J Bilander

Posts 6
03 Jun 2020 16:04


Ok $200 in sales price would probably be far too much, I agree. I thought removing bits and pieces from the 080 would make it possible to fit in a much smaller cheaper FPGA-chip. Performance wise I think an 060-core replica that can be clocked just north of 100 MHz or so running real cool at that speed would be a good CPU-option for such a card and it wouldn't cannibalize on the Vampire sales due to the huge difference in performance.


Adam A

Posts 99
03 Jun 2020 16:53


I think Aros for Mister will be a better idea

posts 34page  1 2