Apollo Team announces the Vampire V4

[jj]

Gunnar may not be making much money now, but he has always been clear that he wants to market his softcore to the embedded market and that he makes decisions based on that rather than every single piece of 30 year old Amiga software. It's in his interest to appeal to the type of people who will evangelise his core.

The friction comes in because there are enough people who will be happy to see vampire as an expensive whdload card, with the added promise of the Amiga having completely new development that will make it viable against PC's. They fight off anyone who tries to get FPU/MMU on the table because they are scared that it will upset Gunnar. Who sees this support and has no reason to expand his base, because it's enough.

Which cuts out software that would actually quite cool to run. Because it's closed source, because of the embedded licensing, then Gunnar is able to exert whatever control he wants.

Hopefully someone else will produce a core for the vampire that does appeal to me.

 is the speed all about the core or is some of the underlying fgpa.  Could minimig or one of the others be converted to run on this hardware ?
 
 Do any of them have MMU/FPU ?

[Gulliver]

No embedded commercial endeavour and no hope for any asic will even surface if the Apollo core doesnt get rid of Intel and Motorola (now NXP) instructions (AMMX and ColdFire).

If you have a company will you use a design that will cause you an assured lawsuit?

Would you invest hundreds of thousands dollars for an asic on a design that infringes the IP of two of the biggest industry leaders, so they can sue you till death?

They havent sued Apollo guys yet, because their design is sold in a tiny niche on a very small market, that is why they have gone under the radar. And also there is still not much money to make out of them yet if they sue them right now.

[guest11527]

If you have a company will you use a design that will cause you an assured lawsuit?

Strangely, I made the very same comment years ago when Gunnar first told me about the Natami. He was quite confident that no such problem would arise. Whether that's true I cannot say, but at least any patent on the instruction set itself run out by now.

On the other hand, it's now Qualcomm, and yes, they are known for a very strict patent policy. )-:

Thus, if there are any IPs in it, they are probably not back from Motorola times, though IBM might claim them in worst case. Gunnar is AFAIK an IBM empolyee, and the typical statement in contracts is that IPs you generate in your spare time are with the employer. Probably not enforcable by German law, though.

So, while the instruction set itself is probably not quite the source of the problem, there might be other know-how in the design that may be "interesting" for the usual suspects.

[Iggy]

NXP actually allows Coldfire V1 to be incorporated into designs free of charge.
And I've never heard of anyone facing legal challenges for recreating 68K functionality in FPGAS.
In fact, many companies selling proprietary cores for the 6809 and the 68000, so I don't see this as being a serious problem.

The use and promotion of operating system software based on pirated code (ApolloOS)? That is a problem.

[Djole]

is the speed all about the core or is some of the underlying fgpa.  Could minimig or one of the others be converted to run on this hardware ?
 
 Do any of them have MMU/FPU ?

I am sure they could, but the soft cores used in minimig and any other (Amiga) FPGA 68k board are not even in the same ballpark as Apollo core....

[Nickman]

The lack of MMU is clearly a big problem for developers who make use of the tooling built around the MMU for debugging. Given the new custom instructions, this seems lacking to me - even if they have a developer core that is slower but has an MMU, and a fast core without. To me this feels more of a priority, but yeah, a lot of people are waiting on the FPU that has been promised.

I can understand people being a bit peeved that new custom instructions like their integer AMMX stuff have been done before a real FPU or MMU. On the other hand, these people are mostly doing it for not a lot, if they want to have some fun (likely hardware ALU blocks on the FPGA made wiring up AMMX a lot easier than doing a full FPU implementation) who am I to complain...

None of that takes away from the achievement of everyone involved in Vampire and the concept of replacing the Amiga CPU with an FPGA implementation initially. It feels like there is a lot more to come.

Think Gunnar said that a FPU takes like 15000LE and AMMX is less then 1000LE. Plus the foundation used in AMMX could be reused for a simpler FPU then the full 80bit apollo FPU that they have now.
And Gunnar and the rest of the team has on several occations said that buy the Vampire and only look at existing features not things that can or will come in the future. Dont think they have forced anyone to buy anything?

[kolla]

So what are the existing features?

[BozzerBigD]

@Hattig

Most of the software that would benefit from a speed up probably needs a FPU; Quake, Lightwave, etc. AlienBreed 3D 2 seems to be the exception but Lightwave would be more fun as would Quake and Quake 2. BOR is fun but limited in longevity. MP3 playing is fun but a Prisma would do the same thing. FPU support would silence the nay-sayers to be honest. It was a good idea to get some AGA functionality and 060 speeds on A600 and A500 machines but for A1200 and Big Box Amigans I'm guessing the Vampire will be a harder sell as most of them have an accelerator already with MMU and FPU support!

[Crom00]

So is anyone making accelerators that can match an 060 in performance that we can purchase today?

Vampire is, but sans FPGA FPU. I'd be happy with a 100mhz 060ec card with 128 megs, SD card, Network capability, RTG gfx with HDMI out and flicker fixing (via ecs and AGA through HDMI). But that type pf  060 card doesn't exist at a reasonable price....it has never existed... sadly it may never exist so I Got a Vampire.

Sure it's got some kinks but 060's don't exist anymore, they stopped making them.
The late Dave Needle said in the special edition VIVA Amiga interview that if he was doing an Amiga product today he would use an FPGA. That's good enough for me.

[Iggy]

@Hattig

Most of the software that would benefit from a speed up probably needs a FPU; Quake, Lightwave, etc. AlienBreed 3D 2 seems to be the exception but Lightwave would be more fun as would Quake and Quake 2. BOR is fun but limited in longevity. MP3 playing is fun but a Prisma would do the same thing. FPU support would silence the nay-sayers to be honest. It was a good idea to get some AGA functionality and 060 speeds on A600 and A500 machines but for A1200 and Big Box Amigans I'm guessing the Vampire will be a harder sell as most of them have an accelerator already with MMU and FPU support!

Except maybe for the A2000 owners as they do not have access to really fast accelerators. These Amigans are quite likely to find this attractive, especially with AGA functionality.

[Crom00]

Except maybe for the A2000 owners as they do not have access to really fast accelerators. These Amigans are quite likely to find this attractive, especially with AGA functionality.

Damn straight. I route all the cables out of the back and AGA on an A2000 is pretty neat.

[Iggy]

Damn straight. I route all the cables out of the back and AGA on an A2000 is pretty neat.

Yep, I may come off as a bit 'dicky' because of my push for an MMU and an FPU, but outside of that, I think this will do quite nicely on my old tank A2000.
And you simply can't kill these systems, they generally don't even need recapping.
Big, solid, looks like a serious desktop computer.
Once I get the hard drives sorted out, I want to move to IDE, I'll be quite set.

Gunnar and I have had our differences (we're both an bit excentric and outspoken), but with the Vampire V4, he's definitely forgiven.

This should operate at warp speed.

[QuikSanz]

As you will see I am not knowledgeable in this area but, if a full FPU was put in the FPGA could some MMU functions be put in software as it looks as though Gunnar is opposed to a full MMU in the FPGA?

[gregthecanuck]

@QuikSanz -

The Apollo core does have an MMU implementation. However it is not backwards-compatible with previous Motorola designs.

It supports new features such as non-executable memory. This is pretty much an expectation of any modern MMU. The MMU has a "modern, forward-looking" design instead of being compatible with 20+ year old designs in the 68060 and earlier generations.

There are rumours the full FPU will make its debut in the V4 product but that is all they are at the moment. My understanding is the FPU took a back seat to getting the chipset reimplementation logic completed for the V4. An interim FPU emulator "femu" has released that works for all non-FPU Amigas. It does not run at the speed one would expect of a native FPU on a 100+ MHz 68060 but it allows programs to run that would otherwise fail.

Cheers!

[kolla]

it allows programs to run that would otherwise fail.

Or allows programs to fail that would otherwise not run

[kolla]

@QuikSanz -

The Apollo core does have an MMU implementation. However it is not backwards-compatible with previous Motorola designs.

It supports new features such as non-executable memory. This is pretty much an expectation of any modern MMU. The MMU has a "modern, forward-looking" design instead of being compatible with 20+ year old designs in the 68060 and earlier generations.

I am speculating that the MMU in Apollo Core is mostly used to move memory addresses around so that AmigaOS and its apps find what they expects at the addresses they expects. With multithreaded I/O-operatings, legacy I/O, DMA, SAGA and kickstart protection, it should be quite busy already.