@Rob

I wonder if Cloanto could have provided a license to the name? Even if the device couldn't be named an 'Amiga', some sort of 'Amiga Forever' branding could be close enough. But I don't know if they have the rights to license any form of hardware.

I agree that to sell properly, any device would need the Amiga name in there somewhere. Or Commodore, I guess... maybe.

Jens' Clone A cores in an ASIC + ARM JIT for CPU? Safer route than the Apollo all in ASIC right now.

All hypothetical and probably moot, but it's sometimes just fun to think about what could be possible if the rights holders weren't all insane.

@Nameless

Quote:

Nameless wrote: @olegil FPGA-to-ASIC was basically what I had in mind when I mentioned my 'dream amiga' being a 68K retro handheld device w/screen. Otherwise it wouldn't really fit so well, or be cheap enough. Another option which would certainly be a lot cheaper in regard to upfront costs, is to take a simple ARM device, be it something like this $9 computer, or Pi 2, paired up with JIT ARM. Slap it into a little Amiga-ish case, include all roms, and it may be close enough, at least for some people. How much would it cost to create something like this, but pair it up with a super cheap FPGA, one just big enough to hold AGA/Amiga chipsets, but not the CPU? Then use JIT ARM for the CPU stuff? Does it make sense at all to even go that route? I seem to recall Clone A thinking of doing something like that, but with MIPs instead of ARM.

The $9 computer shows you what is possible but you shouldn't settle on just this chip because you should want a faster ARM chip with more GPIO (General Purpose Input Output) pins.

This chip may be the place to learn on but you have to decide if the stated number of input / output pins are enough I/O and if it is fast enough or has enough storage:

"8 digital GPIOs, one PWM pin, SPI, TWI, UART, USB, MIPI-CSI, Parallel LCD output, touchpanel input, and a whole bunch of power rails in and out"

Last edited by Chuckt on 16-May-2015 at 01:49 AM.

@olegil

Quote:

olegil wrote: @Raffaele In short, do you want it to be made legally by a european company, or below the radar by some shady chinese outfit?

Do they look shady to you?

http://dangerousprototypes.com/2014/07/29/geek-tour-visit-shenzhen/

Or them?

http://dangerousprototypes.com/2012/12/06/shenzhen-hua-qiang-bei-market-visit-report/

Tell me where this exists in Europe. I'm waiting.

http://dangerousprototypes.com/2012/11/19/shenzhen-day-1-hua-qiang-bei-market/

China is the largest exporter of electronics and everything.

http://www.globalsources.com/ST/Made-In-China-101.html

@Nameless

Quote:

Nameless wrote: @olegil FPGA-to-ASIC was basically what I had in mind when I mentioned my 'dream amiga' being a 68K retro handheld device w/screen. Otherwise it wouldn't really fit so well, or be cheap enough. Another option which would certainly be a lot cheaper in regard to upfront costs, is to take a simple ARM device, be it something like this $9 computer, or Pi 2, paired up with JIT ARM. Slap it into a little Amiga-ish case, include all roms, and it may be close enough, at least for some people. How much would it cost to create something like this, but pair it up with a super cheap FPGA, one just big enough to hold AGA/Amiga chipsets, but not the CPU? Then use JIT ARM for the CPU stuff? Does it make sense at all to even go that route? I seem to recall Clone A thinking of doing something like that, but with MIPs instead of ARM.

Well, considering this is basically identical to a minimig, just a different CPU, you're still looking at the same price as the minimig, since the CPU is not the priciest item on there.

ASIC which includes the CPU/RAM/MCU/logic is the way to cut costs here. There just isn't a way to magically get a "super cheap FPGA big enough to hold AGA chipsets".

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@Chuckt

Ka-Ro i.MX6Q SODIMM modules work very nicely, but for some reason Ka-Ro decided to pair a GbE enabled device with a 10/100Mbps ethernet phy. Sort of a big let-down for me. However, getting up and running is not a lot of work, allthough a carrier board must be made.

Unfortunately by the time you've gotten the IO you want, the ARM benefit was mostly gone, as it now costs a few hundred USD

And then why would I go for that over a T1024 board which would cost the same?

Seems almost like the old expression "no free lunch" still applies. Almost.

TL;DR:
ARM is very cheap when it has only low-frequency peripherals, but as soon as you start talking PCIe and GbE the price difference compared to a PowerPC chip is negligible (woohoo, only three attempts to spell it right this time. Unfortunately I think I only used three attempts last time as well). And you'll need more cores to do the same work, so software becomes more complex.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@Chuckt

Yes, Chinese products are sometimes EXTREMELY dangerous to use, as they (in those cases) don't include ANY consumer safety features, leading to potential electrocution, fires etc. Lighting products are the worst.

Biggest != not shady.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@olegil

Guess I was hoping something like this $9 ARM computer or a Pi would be like the base, then add cheap FPGA costs (if it could be had cheaply). That was going with the assumption the FPGA cost would be less if it could be smaller, since it wouldn't have to hold the CPU/Ram, just graphics/sound.

I'm not sure of the cost differential in parts between something like this $9 PC vs the Minimig though. I just figured something in mass quantities (ie Pi), would have to be cheaper than low production runs of the Minimig. In the end, it might not be a big enough difference to matter, if the FPGA, even cheapo ones, end up costing $150+. Thought some ran in the $20ish range though, but they may be too wimpy for the chipset.

ASIC is the way to go though if wanting to go cheap. But that does require a backer with somewhat deep pockets, at least relatively speaking in Amiga-land.

Last edited by Nameless on 16-May-2015 at 05:05 PM.

@Nameless

Quote:

I wonder if Cloanto could have provided a license to the name? Even if the device couldn't be named an 'Amiga', some sort of 'Amiga Forever' branding could be close enough. But I don't know if they have the rights to license any form of hardware.

Or Hyperion could license them the AmigaONE brand also but ideally you want Amiga, although the AmigaONE X1000 got quite a good deal of press attention when first announced so AmigaONE of Amiga Forever could be enough to get the general tech media on the story.

Hyperion already have an agreement with Individual to supply kickstart 1.3 and 3.1 roms for use in the accelerators for A500, A1000, A2000 and A2500.

https://icomp.de/shop-icomp/en/33/items/hyperion-and-individual-computers-sign-license-agreement.html

Quote:

Jens' Clone A cores in an ASIC + ARM JIT for CPU? Safer route than the Apollo all in ASIC right now.

Using and ARM cpu to emulate the 68k to interface to what is essentially real hardware presents it's own challenges.

Jens' plan was to do the same with the 68k as he did with the Amiga chips and and put the whole Amiga on the ASIC anyway. With ASIC you also have the potential to have a faster CPU than any real 68k CPU too, if that's what you want.

@Rob

Is Hyperion allowed to license the name 'AmigaOne' to non-PPC hardware? The whole issues of who owns what rights confuses me.

There is also the potential issue of what the rights would even cost. If going for a cheap retro system, every dollar counts. If one of those rights holders wants something like $20+ per device sold, it's not so feasible.

"Jens' plan was to do the same with the 68k as he did with the Amiga chips and and put the whole Amiga on the ASIC anyway. With ASIC you also have the potential to have a faster CPU than any real 68k CPU too, if that's what you want."

I agree that the whole thing on ASIC is the ideal away to go. Just thinking aloud as to alternatives if there wasn't a suitable fast 68K core (Apollo) fully tested. It'd probably make more sense to just go with a slower core if need be then.


Out of curiosity, how much is the cheapest FPGA that could hold the Amiga chipset, minus CPU/Ram? At least ECS, although ideally AGA?

@Nameless

Quote:

Nameless wrote: @Rob I wonder if Cloanto could have provided a license to the name? Even if the device couldn't be named an 'Amiga', some sort of 'Amiga Forever' branding could be close enough. But I don't know if they have the rights to license any form of hardware.

I'd be very surprised if the trade mark side of things would have been left untouched when they made the deal to purchase the operating system.

Quote:

Jens' Clone A cores in an ASIC + ARM JIT for CPU? Safer route than the Apollo all in ASIC right now.

One big (pun intended) problem here is endianness. The 68k is true big endian, and so is the PPC. This is what makes PPC such a great target if you want to run old Amiga S/W at a native binary compatible level (albeit the instructions being translated of course, like in Trance/Petunia). Because on PPC this is possible thanks to the endianness. On ARM and x86 it is not.

You may say "but hey, isn't ARM bi-endian"? Well not really, not in a true sense (and the same goes for PPC which is also "bi-endian"). At least not the interesting CPU's for a purpose like this (CPU's in the ARMv7 family). They are "bi-endian" in the sense they can load in big endian data. But instruction format is still little endian, addressing modes are still little endian. Emulation would be necessary. Which contradicts the purpose of the Clone A in the first place (having a physical H/W Amiga). So if not use an Apollo, then a PPC chip would be the best choice. Both these options would however result in a computer closer to $900 rather than $9, and with "Clone A" bolted to it, it would be a $900 A500 albeit with CPU on steroids compared to the original A500. Which also would break compatibility of much S/W, and the point of Clone A was to be as true to the original as possible.

But another point is that in 2015 things like running ancient code natively has kind of lost its importance. And frankly, so has stuff like "Clone A".

Go here instead: http://www.amigaforever.com/



And for non-retro fans, hasn't things like legacy Amiga compatibility run its course as well? Isn't it time we look at true NG features now? Stuff that aren't Amiga compatible per design? Like SMP, true MP, 64-bit? And new CPU's? Little endian ones? Like ARM, and more importantly x86?

@TRIPOS

Go here instead: http://www.amigaforever.com/

Thanks. Of course I'm aware of Amiga Forever (mentioned it above in this thread). And yep, I agree stuffing everything, amiga chipset + apollo + ram into ASIC would be the best way to go. It's also a pricey way to go with upfront costs, unfortunately. And nobody would put Apollo into an ASIC until all the bugs were ironed out.

As for emulation being necessary, yep, I assume it would be, if using ARM. I was ust wondering if JIT ARM for CPU + FPGA (or ASIC) for chipsets would offer any advantage over full emulation... or be possible at all. By the sounds of it, it'd still be pricey and might not even be possible anyway.

@Nameless

CPU + FPGA-big-enough-for-chipset-emulation IS what the minimig IS. So no, you're not going to make it significantly cheaper by replacing the 68000 with an ARM. Sorry.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@TRIPOS

Quote:

TRIPOS wrote: One big (pun intended) problem here is endianness. The 68k is true big endian, and so is the PPC. This is what makes PPC such a great target if you want to run old Amiga S/W at a native binary compatible level (albeit the instructions being translated of course, like in Trance/Petunia). Because on PPC this is possible thanks to the endianness. On ARM and x86 it is not. You may say "but hey, isn't ARM bi-endian"? Well not really, not in a true sense (and the same goes for PPC which is also "bi-endian"). At least not the interesting CPU's for a purpose like this (CPU's in the ARMv7 family). They are "bi-endian" in the sense they can load in big endian data. But instruction format is still little endian, addressing modes are still little endian. Emulation would be necessary. Which contradicts the purpose of the Clone A in the first place (having a physical H/W Amiga).

The big-endian support provided by ARMs is enough for emulators/JITer like Petunia and Trance.

@olegil

My mistake then, I thought only the early Minimigs relied on a separate CPU. I thought it included a 68K core in the FPGA nowadays.

Curious still as to the cost of the FPGA... figured prices had come down by now. What does a 400K gate FPGA generally cost?

Last edited by Nameless on 17-May-2015 at 09:48 AM.

@Nameless

Quote:

My mistake then, I thought only the early Minimigs relied on a separate CPU. I thought it included a 68K core in the FPGA nowadays.

The Minimig core has been used on the Commodore One, Turbo Chameleon, FPGA Arcade replay and a few other FPGA based systems that don't have a 68k CPU. That's probably where the confusion came from.

@Nameless

Quote:

Thanks. Of course I'm aware of Amiga Forever (mentioned it above in this thread). And yep, I agree stuffing everything, amiga chipset + apollo + ram into ASIC would be the best way to go. It's also a pricey way to go with upfront costs, unfortunately. And nobody would put Apollo into an ASIC until all the bugs were ironed out.

The 68k core used in the Vampire 600 is fast because of the manufacturing process it is built on rather than because of the design itself. The 68020 was built on a 2um/2000nm process. The Cyclone II FPGA used by the Vapmire 600 is built on a 90nm process. Smaller circuits use less power and generate less heat meaning you can clock them higher.

If you make an exact clone og the 68020 and build it on a much smaller process and you can clock it much higher without having to do anything fancy with the design.

The reason Gunnar and others are adding extra features to the Apollo core is because the FPGA they are targetting has more space than is needed for the core currently used in the V600. They are utilising the extra space to gain more performance by making it a superscalar design like the 68060 so that it can execute more than one instruction per clock cycle.

@Nameless

Around 20 USD if buying more than a handful, best guesstimate without consulting anyone. FPGA prices have not been reducing much over the years, unfortunately.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

remotely related FPGA stuff:
http://www.excamera.com/sphinx/fpga-j1.html#fpga-j1
http://www.excamera.com/sphinx/fpga-cpu.html

JAVA co-processor... http://www.jopdesign.com/

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@cdimauro

Quote:

cdimauro wrote: @TRIPOS Quote: TRIPOS wrote: One big (pun intended) problem here is endianness. The 68k is true big endian, and so is the PPC. This is what makes PPC such a great target if you want to run old Amiga S/W at a native binary compatible level (albeit the instructions being translated of course, like in Trance/Petunia). Because on PPC this is possible thanks to the endianness. On ARM and x86 it is not. You may say "but hey, isn't ARM bi-endian"? Well not really, not in a true sense (and the same goes for PPC which is also "bi-endian"). At least not the interesting CPU's for a purpose like this (CPU's in the ARMv7 family). They are "bi-endian" in the sense they can load in big endian data. But instruction format is still little endian, addressing modes are still little endian. Emulation would be necessary. Which contradicts the purpose of the Clone A in the first place (having a physical H/W Amiga). The big-endian support provided by ARMs is enough for emulators/JITer like Petunia and Trance.

No it obviously isn't. On legacy 68k Amiga everything is big endian. On OS4/MorphOS on PPC the legacy code, datablocks, pointers, registers, whatever exists and runs completely down-to-the-metal native (after being translated by Trance/Petunia). This is only possible thanks to PPC being TRUE big endian. But on little endian architectures (including "bi-endian" that still has little endian addressing modes) this will crash and burn, simply not possible, and an emulator (a program running that mimics a 68k CPU and thus takes care of all the S/W execution, like in UAE) will be necessary. If endianness hadn't been a problem then OS4/MorphOS would have been ported to ARM or x86 a long time ago already. But the key objective of both these operating systems has been to provide a native environment and true binary compatibility. This can only happen on a TRUE big endian CPU like the PPC. This is the only environment where legacy Amiga stuff can exist and function, natively, not inside emulators.

@Nameless

Quote:

Nameless wrote: And nobody would put Apollo into an ASIC until all the bugs were ironed out.

I think there might be some boring stuff like Intellectual Property issues of the 680x0 as well...