@OlafS25

Quote:

At a certain point these concepts were not competitive anymore but still unique.

Original IBM PC also had unique architecture. Glad PCs today evolved massively since then.

Quote:

For me FPGA based systems would be similar to "Amiga" again.

Ideal as replacement for aging early 1990s hardware. However, if you want more performance, commodity hardware is needed. What speed reaches most powerful FPGA based 68k CPU now? 68030? 68040? Can FPGA based chipset deliver similar 2D (not even speaking about 3D!) performance as cheap PCIe GFX card?

That is why there is still market for over-priced custom build computers using mainstream components. Tolerable mix between performance, feeling and compatibility.

Quote:

A normal motherboard with PPC does not create the same interest to me (and propably many others).

That is why there is market for various FPGA solutions.

@pavlor

"Can FPGA based chipset deliver similar 2D (not even speaking about 3D!) performance as cheap PCIe GFX card?"

There are FPGA system than can have 1800 DSP units. it cost too much today, but prices are getting lower every year. However, the idea of custom co-processor via FPGA or xmos chip is really nice, imho.

An old amiga user have built custom rendering engine (Real time ray-tracing) with 4 virtex 5.

_________________
retired

@AmigaBlitter

Modern graphic card based on FPGA is just a insane, people have hard time mimic a simple AGA chip, how many man hours/years do you think it will take to make some thing that can compete whit Nvida or AMD.

Beside a custom design will be hugely expensive, and I'm also doubt full, about being able to operate the FPGA at optimal speeds because of timing and other issues.

Last edited by NutsAboutAmiga on 09-Feb-2014 at 04:17 AM.
Last edited by NutsAboutAmiga on 08-Feb-2014 at 09:51 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@AmigaBlitter

On the other hand a FPGA has many advantages, I'm not saying its not useful or interesting.

Some of the advantages, re-flashable no need to rewire your electric-curiet, a faulty production run can bankrupt a company.

No dout, its a nice glue chip.

The cost of FPGA is low compared to ASIC,

ASIC costs around $ 250.000.-, and maybe your chip does not work.
In other words unless your going to sell 10 000 or more units its not cost efficient.

http://electronics.stackexchange.com/questions/7042/how-much-does-it-cost-to-have-a-custom-asic-made

And also it compact, if you where to design some thing complicated you do not want 100's of NAND, NOR, AND, OR gates, 74LS0X chips.

Last edited by NutsAboutAmiga on 08-Feb-2014 at 09:43 PM.
Last edited by NutsAboutAmiga on 08-Feb-2014 at 09:33 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@KimmoK

Quote:

@NutsAboutAmiga AAA was in prototype/testing in y1993/1994.

In 1995, the chip be outdated, 3D chips that where coming in PC's and PS1 the year after commodores bankruptcy.

Besides a prototypes not a working chip, it might have taken 2 more years before it was ready, and another 2 years to get it onto a prototype motherboard, maybe 4 years later you have working motherboard whit your custom chip.

So maybe you get product out to 1998, by then AAA chip be really outdated no doubt.

Last edited by NutsAboutAmiga on 09-Feb-2014 at 04:18 AM.
Last edited by NutsAboutAmiga on 08-Feb-2014 at 10:00 PM.
Last edited by NutsAboutAmiga on 08-Feb-2014 at 09:59 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

Only for the fact that "Nyx" boards were allready designed when C= went under.

So realisticly we could have seen AAA based Amigas in production just intime for 1995 XMas.

Wethere that would have been to late or not is pretty much academic.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Kronos

I really doubt you have working chip on first revision a, your most likely need to design revision b and c, before the chip was problem free, and also maybe revision b of the motherboard before it was ready.

Beside if it was ready, way did not ESCOM or VISCORP bring it to market?

I can also point to another product a product that came out in 1995.
Draco designed by MarcoSystem, it also used a normal PC graphic chip the NCR 77C32BLT.

It's also interesting that they also did not go whit AAA chipset.

I think my estimate for when product be ready is not half bad, 1998 or 2000 at best.

Last edited by NutsAboutAmiga on 08-Feb-2014 at 10:28 PM.
Last edited by NutsAboutAmiga on 08-Feb-2014 at 10:28 PM.
Last edited by NutsAboutAmiga on 08-Feb-2014 at 10:15 PM.
Last edited by NutsAboutAmiga on 08-Feb-2014 at 10:13 PM.
Last edited by NutsAboutAmiga on 08-Feb-2014 at 10:12 PM.
Last edited by NutsAboutAmiga on 08-Feb-2014 at 10:12 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

Obviously it wasn't ready, but also not more unfinished than AA was in 1990 (which could have gone on sale with A3000+ in 91 if that hadn't been axed).

Escom never had any expertise on chip design andf their plan was to move Amiga away from custom chips anyways (partly because AAA would have been 100% obsolete by 1996 or 97).

Gateway never did anything with Amiga, excet trying to put put a badge on unrelated tech (and to be honest I'm not sure they even tried to get that any further than some techno-babble).

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Kronos

Quote:

Now offcourse anybody with a basic clue about the Amiga knows very well that Zorro3 is 32Bit and that it's rather buggy implementation is the main reason why it never lived up to it's full spec.

Well I wrote Zorro not ZorroIII, the Zorro bus on Amiga500 is 16bit because it was connected directly to the 68000 CPU on Amiga500, and the CPU did have a 16bit databus, the ZorroII bus was inverted Zorro bus, and there for the ZorroII bus also was a 16bit bus.

Quote:

useing the Zorro2 slots to implement Zorro3, should have skipped the ISA slots on the A3000/4000 and put much simpler Zorro3 only slots there.

I whit a 16bit databus on the 68000, its hard to see how that will make any sense at all on a A500/A2000. Zorro3 was standard on A4000 at least, don't know about A3000.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@Kronos

Well so I'm right AGA and AAA was obsolete tenology.

And what I do want to say is that if AGA and AAA was obsolete technology, this is way its not so bad thing that AmigaONE hardware is using standard PC chipsets.

I'm happy its not a Amiga like they designed it back in 1992, because that is what killed Commodore Amiga in 1994.

As for PowerPC road map, and IBM selling there CPU division, well its no big surprise, that's what IBM do when things is not profitable, remember the IBM desktops / laptops, IBM printers, maybe IBM was to big for its own good.

IBM printers became Lexmark, at lest its not dead, if CPU division becomes a new company then that might improve focus whit in the new company.

Anyway last time I checked we where using AMCC, Freescale, PaSemi CPU's, the IBM server CPU's Power 5,6,7 are not really some thing we want, because they are deigned to get high performance and IBM do not really care about heat.

Motorola CPU division, became Freescale, so we are kind of kept whit amiga traditions and using Motorola/Freescale CPU, and so there I might argue AmigaONE hardware are more Amiga then a Toshiba for example.

Last edited by NutsAboutAmiga on 09-Feb-2014 at 05:53 AM.
Last edited by NutsAboutAmiga on 09-Feb-2014 at 05:49 AM.
Last edited by NutsAboutAmiga on 09-Feb-2014 at 05:46 AM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

i think that aeon should make an acc card with an arm,plus an x86 alone with a coule of p5040 on a boad that could work ass an acc card for the amiga one's but the very same card should work as the semise twin or the amiga inside out. an pci card or pci-e that could be puttet in an pc and then the power would be an doongle like for the os 4.1 and the card would serve as an development card but all so an power up card for amiga one's make 3 products in on wicht there could make money on,
i would buy one to put in my pc, i would buy an ekstra or two to put in my amiga x2000 cyrus board ,maby one for the rasbary pii and hope for the aron team to make an server solution, then i would buy one to give a way to an developer.. there should buy in alot of cpus and then consolidate it and then make some hardwere the holde communaty would buy multiplytimes. something that would expand the usser based and give even more power to the ove wee have, and be something we can relay on in the furture too insted of just playing cath up, i mean hyperion would get hardwere to develop arm and x86 on to and ther wount undermine the os 4.x ppc sale that you need one to run amiga os on x86 plus it will even work as an acc card for that solution too..

and the posabilaty to stuff on some ram. i will hope pepole will discuised this option so i can capture trevors attention !!!

have a nice weekend

@retro

aeon should really make such a card so there will consolidate the sale of the p5040 cpu and bring the price down and give more power to the communaty in the same deal,

bring pci card amigans to x86 and give that solution a power up too, make old x86 hardwere be cheap amiga servers, and bost the sale to plus bring arm and x86 on the table for the long run use the card as a doongle soo it wount damage the sales of x2000 and an os 4.2
i would buy one for my amiga cyrus x2000.
maby there could make an preorder for both the cyrus and the amiga ng acc pci/pci-e / dev card

@NutsAboutAmiga

Quote:

Well I wrote Zorro not ZorroIII, the Zorro bus on Amiga500 is 16bit because

You started by writing about adding PCI to A3000/4000 and how the 16Bit Zorro bus was such a letdown for those computers, and I pointed out that they actually did have 32Bit slots....

On 68000 based Amigas anything above Zorro2 make 0 sense as it's a 1:1 connection to the CPU, read no slot system could get any faster (without bypassing the CPU).

Quote:

Zorro3 was standard on A4000 at least, don't know about A3000.

If you don't know and don't even care to look up such basic facts, it might be a good idea to stop participating in technical discussions !

Quote:

I'm happy its not a Amiga like they designed it back in 1992, because that is what killed Commodore Amiga in 1994.

Completly wrong again, Commodore killed itself by investing money brainfart products (C128, C65,A600 and to some extent the CDTV) and managment payout instead of pushing their money-cow forward.

Had they done that we would have seen AGA by 1990 (or earlier) and it's aboslutly possible that costum chip Amiga would have been competetive throughout the 90s, heck maybe even till today if the made their chips available to other vendors (read C= would have been what AMD/ATI or NVidia are today).

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@olegil

Here we don't call them either. We call them mobiles.

@pavlor

I am a little involved in 2 FPGA projects (at least I read it and hope that AROS Vision becomes standard there ) and they talk about a new generation of FPGA with 800 Mhz (I think this year) and they work on a cpu reimplementation that can be embedded in the FPGA. If this succeeds it would mean a performance far above what is possible now (68060 with up to 100 Mhz). But we will see it. In best case it will certainly not outperform standard hardware but it will have some "geek" factor and a concept that is different. And that what was Amiga to me, a system different to PC at the time. Standard hardware with a different processor is not the same to me. I understand why a custom PPC solution is expensive (the same with any custom solution) but it is not offering enough difference or additional value to justify the price. Besides I would never spend so much money for a hobby like other people that buy X1000 (or in future X5000). And I already own several computer (both X86 and X64) why spending thousands of Euro for it? And the naming alone does not justify it.

@cdimauro

Quote:

That was just for marketing, to attract people with "modern features".

If that was so they should gone for a plain 32-bit single core PPC which is cheaper.

If it's not cheaper then markettng aside 2x64-bit is way to go.

If it was cheaper then a faster plain PPC would have been better for OS4. But without real hardware would have left OS4 crippled to being single core 32-bit only.

Also, rather like 68k AmigaOS running on a 32-bit CPU or OS4 running with AltiVec, the OS can be optimised to make use of 64-bit features on CPU. And not just limited to some utility.library functions.

I know these fully 64-bit software features don't come overnight but without hardware to support it there would be no hardware platform to develop and test these modern ideas. The X1000 at least provides that base. Since AFAIK the dual core AmigaOne module fell through years ago.

Quote:

They have promised SMP, but after the scheduled 2 years nothing was shown, and then reverted to a generic "multi-core" support.

Was this on the offical roadmap? Was it planned for an update OS4.1 release? Or an OS4.2 feature?

Quote:

They also announced 64-bit "support", which is generic enough to let people think and dream that a 64-bit o.s. and applications will come.

This support can come as I've mentioned above., But expecting a fully 64-bit AmigaOS is unrealistic. Right now at least.

Quote:

A very expensive and limited platform to make Linux running on it...

As is typical with al NG Amiga machines, it's a terrible machine to make Linux run on. I now see the point everyone in my Amiga club seems to be making: "Why would you want to run Linux on an Amiga nmachine? When you can just use it on a PC? Why would you bother?" Can't autoboot a Linux CD, kernel can be slow to load and you must fiddle to add a boot menu.

Quote:

It'll not happen, for the exposed reasons. What you'll see with OS4 is the good old Amiga o.s. 3.1 which'll be patched again to introduce SOME support for multi-core and 64 bits. That's all.

They are working on these things right now. Since the beginning they have declared old functions obsolete and depreciated others while introducing new functions. They also are "privatising" the OS. Forbidding previously common functions and hiding internal data.

And look at how RunInUAE is an offical add on for the OS. What does this tell you? To me it looks like a warning for things to come. Breaking 68K compatibility.

As it stands right now both 68K and native apps are inside a 32-bit memory space, with their own spaces allocated. With full 64-bit not only a new API but new ABI will be nedded. So both 68K (if it still exists) and 32-bit apps will need to be sandboxed one way or the other for OS4 to move on into the future. Right now the 32-bit nature of OS4 has become the "new" AmigaOS legacy.

So IMHO I am "being realistic." I also said "in theory" but perhaps "I hypothesize" would have made my point more clear. Since I didn't say it was going to happen.

Quote:

The application receive a pointer, and it only has to (de)reference it as usual. In this case the CPU will mask & cut off the tagget bits, so that a correct virtual address can be generated.

...

Quote:

It's a great advantage.

It's a great advantage until you go from a 24-bit address bus to a full 32-bit one and find the the neat trick you have been enjoying all these years just totally scwewed you over!

Does it not matter because it's 64-bit? I suppose it will take longer for CPUs to go from 48-bits to address the full 64-bits, than it did for 68K 24-bit to 32-bit. But it still looks like a hack. Your meant to learn from your mistakes. This would almost be like AmigaOS using signed 640bit integers for DOS again and then eventually finding...

Quote:

24-bit addressing was a problem only for badly written applications, which used the upper 8-bit for their stuff, totally ignoring the existence of processors which used some or all of the upper 8-bits of the pointers.

Yes, and one of thsse applications was Mac OS!

Which was a point I had already made above.

Last edited by Hypex on 09-Feb-2014 at 01:55 PM.

@OlafS25

In order to get a 100MHz 060 out of an 800MHz FPGA you would need to write a core that needs no more than 8 logik cells serialized to do what a 060 can do each clock.

Won't work....

Such an 800MHz FPGA might give better performance than an 060 if it's running an artifical CPU-core specially designed for FPGA (and there 100% incompatible with any existing CPU).

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Kronos

I only read it, I am not a low-level chip designer like some in the community (working for big companies in this area) so I cannot judge it technical. They think it is possible.

@itix

Quote:

It can be done with less instructions if some of 16-bit loads are zero.

Kind of like a MOVEQ? Yes it would be shorter that way. But I was thinking of a full 64-bit integer.

Quote:

It only works if you have base address loaded for easy relative addressing but reduced code size would compensate greater data size.

Yes, a a base would be used off, since PPC must work that way. That's how I was thinking.

@NutsAboutAmiga

Quote:

Yes the G4 CPU is 32bit, but OS is restricted to 31bit atm, only 2Gbytes of RAM is allowed, not 4Gbytes aka 32bit.

I think that goes for all 32-bit OS and CPU architectures, AFAIK for desktop. For example, like in the 80's, hardware is memory mapped, so all PCI devices and other hardware in the machine must sit somewhere in this 32-bit space. Therefore, not all the 32-bit can be used, as it must be shared between memary and hardware. So the OS4 2GB limit is normal in this case.

I recall before PCs went 64-bit that we saw 3GB being the max RAM. The other 1GB space needed for PCI and other hardware.

So, a 4GB PC, would actually need a 64-bit CPU to use 4GB of RAM. Strange as it may seem.