Bloatware AmigaOS?

[downix]

HenryCase wrote:

uncharted wrote:
@HenryCase

Don't mind me, I'm just being a grumpy sod today.

I should be the one apologising uncharted, I was a little out of order with my tone before. Sorry.

Getting (almost) back on topic, how long do you think OS4 (not the version we have now) would have taken to come out if Commodore hadn't stopped producing Amigas?

Here's some release info for key versions of Workbench:
v1.0 - 1985
v2.0 - 1990
v3.0 - 1992
v3.1 - 1994

So I'm thinking v4 would have been around 1995/1996? Of course it wouldn't have been as good as the version we have now. I suppose it would have been launched with the AAA chipset Amigas.

No, '95-'96 would have been post-AAA, Hombre chipset.  AAA was to be 3.0, but CBM put it's development on pause, instead releasing the interim AGA.  When they restarted AAA development, they soon found themselves too far behind the curve, so they began Hombre, slated for release in '95.

[downix]

HenryCase wrote:

downix wrote:
No, '95-'96 would have been post-AAA, Hombre chipset.  AAA was to be 3.0, but CBM put it's development on pause, instead releasing the interim AGA.  When they restarted AAA development, they soon found themselves too far behind the curve, so they began Hombre, slated for release in '95.

Thanks for this info.

Just out of interest, if AAA had been released instead of AGA (i.e. at the same time) how would it have compared, tech specs wise, with IBM-PC compatible and Apple graphics h/w?

AAA compared well to IBM/Apple from roughly 1995 levels, but was due to come out in 1990/1992.  The main advantage to AAA was the ability to grow the system, that is the chipset was no longer bound to the CPU, but was modular, using an interconnect bus (originally the AMI bus, but later replaced with PCI) enabling upgradeability to Hombre when that was due to ship in 1995.  

AAA was based on ECS, not AGA, don't forget, so the talk of releasing it today is still a step backwards, as it would break AGA apps.

[downix]

A6000 wrote:
I hope that the next minimig version does not copy AGA but goes straight to an HDTV ready graphics mode, IE: 1920 x 1080 x 24 with an external DSP for 24 bit audio. The custom chipset meant that the amiga could do more than a pc yet cost less, if the amiga is to survive another 25 years we must return to that concept.

Fully agreed.  While, sure, modern day GPU's are more than adequate, truth is, the rest of a PC or Mac's chipset is downright anemic for performance.  I build these things every day, and deal with these limitations.  Example, the common AC97 sound system that's universal nowadays.  It is bogged to the CPU, which means you loose performance with it even existing in the memory map, and not all BIOS allow for disabling of them.  Same with the disk controllers, USB controller, etc.  

A from-scratch design would definately stand out, and with the reduced costs for prototyping and custom fabbing, could really be done.  I've been working on my chipset for how long now?  10 years as of last friday.  I know these costs, and I've witnessed the cost to produce go through the floor.  My first-gen design would have cost me $480,000 for each of the 8 chips it used.  Today, $1,850 for the same chipset.  My current design of 2 chips would cost me roughly $4500 for a full speed prototype, vs the millions back in 1997.

Look at the rise of nVidia and XGI, two fabless semiconductor companies that have risen over the past few years.  See where they're going.  It's more than possible for us.

[downix]

AmiGR wrote:

I must disagree with you here, if the last 12 years has taught us anything it is that it is IMPOSSIBLE to build an amiga at a competetive price using off the shelf parts.

Huh? All the problems associated with building an Amiga with off the shelf parts are due to no serious company trying. It's even more impossible to build an Amiga with custom chips than it is with off the shelf parts when you're producing anything less than 100,000. Remember Phase5? They needed to sell 20,000 PPC cards to get even and they sold half that and their boards did not even have custom ASIC's. Imagine how many more you'd have to sell to get even with a fully custom chipset.

Exactly.  Phase5 by not using custom logic was forced to pay far more per-board than if they had ASIC'd the parts together, to reduce the overall cost of production.  That is why the VIC-20 could price-undercut the TI-99A so much, Commodore custom-made the chips, resulting in lower cost to produce.  Yes, the R&D and initial cost is higher, but the end-price is far lower.

The MiniMig, don't forget, uses a "custom made" single chip to replace 4 chips, which themselves were custom made to reduce the cost to produce the original multi-thousand-chip Lorraine unit.  Your arguement about cost is a paper tiger, the cost of producing is nothing when compared to the cost savings by having reduced the overall number of parts in the product.

[downix]

itix wrote:

Exactly. Phase5 by not using custom logic was forced to pay far more per-board than if they had ASIC'd the parts together, to reduce the overall cost of production. That is why the VIC-20 could price-undercut the TI-99A so much, Commodore custom-made the chips, resulting in lower cost to produce. Yes, the R&D and initial cost is higher, but the end-price is far lower.

Too bad this scheme did not continue on Amiga series. I wonder why Commodore was selling Amiga 1000 with such insane price tag.

Hrm?  $1000 for a machine that beat $4000 workstations, graphically?

Incidentally, the cost to produce the VIC20 was $165, vs $485 for the similar-spec TI-99A4

[downix]

AmiGR wrote:

Exactly. Phase5 by not using custom logic was forced to pay far more per-board than if they had ASIC'd the parts together, to reduce the overall cost of production. That is why the VIC-20 could price-undercut the TI-99A so much, Commodore custom-made the chips, resulting in lower cost to produce. Yes, the R&D and initial cost is higher, but the end-price is far lower.

According to Laire, the licences to use the VHDL synthesis software were half a million. Plus R&D, they really would have no chance to break even, even if they used ASIC's. At the numbers they sold, the cost savings of the production would not outweight the R&D and setup costs.

Funny, at the same time I paid a lot less for my software.  For Eddas development, I shelled out approx $1200 for my software package

The MiniMig, don't forget, uses a "custom made" single chip to replace 4 chips, which themselves were custom made to reduce the cost to produce the original multi-thousand-chip Lorraine unit. Your arguement about cost is a paper tiger, the cost of producing is nothing when compared to the cost savings by having reduced the overall number of parts in the product.

That is true when we're talking about numbers but look at the post I replied to and tell me, what would be the chances of the AmigaOne, for instance, being cheaper had it not been based on off-the-self hardware? This market does not really have the numbers to allow companies to produce and sell enough to cover the cost of custom hardware and make profit.

Actually, very good chances.  I did a cost breakdown for a similar move at about the same time, by migrating to a fixed ASIC and integrating as much as possible, saved almost $45 on production costs.  The toolup would have cost approx $37000, mind you, so you'd have to sell 825 boards to break even.  But this would have eliminated the whole Mai-supply issue, and given you a faster chipset to boot.

[downix]

persia wrote:
Custom graphics chips were a bad idea.  First of all you are reinventing the wheel, companies like NVIDIA and ATI spend millions of dollars on designing video cards, don't tell me that a company that can't pony up an additional $7K for it's OS could possibly do what NVIDIA and ATI do.

Also, having the video on a card means that you are in control, maybe you by your system with a low end video card and then expand later or you replace an old card with new.  Either way you are in control.  

Of course if you make the computer then a custom video chip set can be a big dongle I suppose, but the bigger and better dongle is intel's trusted platforn technology, that's what Apple uses.

I admit speculating on what Amiga would have done had they survived is difficult and there was virtually zero chance of survival, but the point is that had Amiga survived, Amiga OS today would look abosolutely nothing like Amiga oS 4.

Who said graphics?  Above I said that the current crop of graphics chips do a fine job.  It's the cost of the rest of em that can drive you up.

My suggestion, if you were to take it, would be to license one of the GPU's out there (nVidia or XGI would be my suggestion) then design a chipset around that and the CPU, giving you a trifecta of performance.  GPU's are underutilized even under heavy loads in several cases due to bad system management thanks, in part, to the system chipsets under them.  A solid supporting chipset could cut production costs, and make a more powerful solution.

[downix]

AmiGR wrote:

Actually, very good chances. I did a cost breakdown for a similar move at about the same time, by migrating to a fixed ASIC and integrating as much as possible, saved almost $45 on production costs. The toolup would have cost approx $37000, mind you, so you'd have to sell 825 boards to break even. But this would have eliminated the whole Mai-supply issue, and given you a faster chipset to boot.

And if you add the costs of hiring someone to design and test such a chipset? It's not as if Eyetech had anyone with the skills of doing that, they tried to hire Escena to design a custom chipset (on FPGA, iirc, but that's a different story) but failed.

Quite correct.  Don't forget, I'm talking as a guy that does know VHDL and Verilog, and finds hardware a fun thing to play with.

Note, I can make as good a case for using commodity chipsets in such a solution as well.  I just don't like seeing both sides of any arguement dismissed so casually, as most likely the best solution would be a mixture of both.

[downix]

AmiGR wrote:

Note, I can make as good a case for using commodity chipsets in such a solution as well. I just don't like seeing both sides of any arguement dismissed so casually, as most likely the best solution would be a mixture of both.

Agreed. I won't pretend to have any real-world experience with chipset design or that I've produced any at any point myself anyway. ;-)

Well, how about I present the proposal I made to one of the guys awhile back.

Xilinx offers free cores for a few functions, including a HT module and PPC bus.  Take those, add a DDR2 controller, and viola, you now have a fully functioning module that can substitute for an AM2 Athlon on a motherboard.  Using the CPU fan mount for support, you can even fit it into an existing socket.  Now, the performance wouldn't be worth the work, but if you swapped out the BIOS with an OpenFirmware, you'd have a fully functional PPC based machine w/o needing to develop a new motherboard beyond custom making the firmware, which would limit you to a specific motherboard or a limited selection of motherboards, which you retail for a slight markup.

[downix]

AmiGR wrote:

Xilinx offers free cores for a few functions, including a HT module and PPC bus. Take those, add a DDR2 controller, and viola, you now have a fully functioning module that can substitute for an AM2 Athlon on a motherboard. Using the CPU fan mount for support, you can even fit it into an existing socket. Now, the performance wouldn't be worth the work, but if you swapped out the BIOS with an OpenFirmware, you'd have a fully functional PPC based machine w/o needing to develop a new motherboard beyond custom making the firmware, which would limit you to a specific motherboard or a limited selection of motherboards, which you retail for a slight markup.

I like the idea and I'd be willing to bet that the performance, worth the effort or not, would be better than any ArticiaS machine. ;-)

Um... yeah, you could say that again.  

The beauty about the design was that it was made to support up to 4 CPU's.  A single CPU would have been a waste of memory bandwidth.  But, still, the cost for the unit would have more than offset the cost savings on the motherboard.  But performance would have been better I'd have bet.

[downix]

itix wrote:

Too bad this scheme did not continue on Amiga series. I wonder why Commodore was selling Amiga 1000 with such insane price tag.

Hrm? $1000 for a machine that beat $4000 workstations, graphically?

Amiga 1000 sales were struggling while cheaper (yet still expensive but technically inferior) Atari ST was selling like hot cakes here in Europe. Consumers simply could not afford an Amiga 1000.
 
But of course they got it finally right and inexpensive Amiga 500 made Amiga successful. I just wonder why it did not happen with Amiga 1000 when they are based on the same chipset.

but they're not.  Actually your arguement is a perfect example of Commodores willingness to cut costs by the use of Custom Chipsets.  The A1000's chipset was just the three main chips + CIA's.  While the A500, they migrated logic that previously was on the board into Agnus, making the Fat Agnus, *AND* consolidated yet other logic to make the Gary chip.  In short, they cut the cost to produce by almost a third through the use of custom chip logic, consolidating chips from the previous design into newer, cheaper units.

[downix]

A6000 wrote:
Every day our community shrinks, for the amiga to survive we need new users or when we die the amiga will die.

We need to pursuade new users to buy a 3rd gen amiga ( when they become available) instead of :-
1. pc
2. games console
3. mac
4. linux box
5. anything else

To do this the new amiga must be as far ahead of the pc as it was in 1984, and be reasonably priced.
This can only be done by a manufacturer with a business plan to make, market and sell a minimum of 500,000 units a year.
The cost of designing a custom chipset will be amortised over 500,000 units and production costs will allow the amiga to compete on price.
The major costs come in persuading people to buy the machine (marketing).
The amiga does not have to replace the pc, simply! carve out a niche for itself.

Quite right, and it is more than doable.  I would even strongly suggest partnering with nVidia, as they have a GPU and chipset design with some merit, but need to utilize it on a non-PC.  Some smart development, custom ASIC for the CPU, maybe a better memory controller, we'd have something.  Imagine a true legacy-free system paired with a super-thin OS embedded into the mobo.  

[downix]

koaftder wrote:

Quite right, and it is more than doable.  I would even strongly suggest partnering with nVidia, as they have a GPU and chipset design with some merit, but need to utilize it on a non-PC.  Some smart development, custom ASIC for the CPU, maybe a better memory controller, we'd have something.  Imagine a true legacy-free system paired with a super-thin OS embedded into the mobo.  

Or you could just focus on the software, which is what really matters.

And wind up on a dead end as you sink 18 months into developing "only software" just to have the one hardware piece you relied on dry up, with no alternative?

Have you not learned the lessons of the A1 and Pegasos?

[downix]

koaftder wrote:

downix wrote:

koaftder wrote:

Quite right, and it is more than doable.  I would even strongly suggest partnering with nVidia, as they have a GPU and chipset design with some merit, but need to utilize it on a non-PC.  Some smart development, custom ASIC for the CPU, maybe a better memory controller, we'd have something.  Imagine a true legacy-free system paired with a super-thin OS embedded into the mobo.  

Or you could just focus on the software, which is what really matters.

And wind up on a dead end as you sink 18 months into developing "only software" just to have the one hardware piece you relied on dry up, with no alternative?

Have you not learned the lessons of the A1 and Pegasos?

Where they went wrong was tying down to a custom board. Stuffing some CPU core on an asic and having NVidia roll out a chip doesn't do anything but make it cost more. Heres an idea, do what apple did, float your platform on standard pc hardware. Nobody cares what chips are in the box. Advances in hardware don't impress people anymore. This is the late 1980s.

And gain vendor lockin like Apple is suffering from now?  With AMD, Intel and nVidia all forcing Apple to cancel products ahead of schedule, delaying the rollout of products, and generally hampering the platform development?  Sure, sign me up, and watch as we go *poof*.  Apple can get away with it due to their user base, we can't.  We're the other guys, the guys nobody bets on!  We want a future, we can't be the other guys, we have to be the best guys.

So, I'm willing to discuss this option, tell me, how do you propose gaining the documentation to enable us to even port our OS to the next-gen Intel or AMD CPU's?  The next-gen chipsets?  Next gen GPU's?  Now now, no bringing up todays products, we're talking a new platform, we need to hit the ground with the new OS, new apps all running on the new CPU Intel ships in 18 months before anyone else has had a chance to code for it.  Propose to me how we manage that, with our market how it is today.

[downix]

koaftder wrote:

And gain vendor lockin like Apple is suffering from now?  With AMD, Intel and nVidia all forcing Apple to cancel products ahead of schedule, delaying the rollout of products, and generally hampering the platform development?  Sure, sign me up, and watch as we go *poof*.  Apple can get away with it due to their user base, we can't.  We're the other guys, the guys nobody bets on!  We want a future, we can't be the other guys, we have to be the best guys.

So, I'm willing to discuss this option, tell me, how do you propose gaining the documentation to enable us to even port our OS to the next-gen Intel or AMD CPU's?  The next-gen chipsets?  Next gen GPU's?  Now now, no bringing up todays products, we're talking a new platform, we need to hit the ground with the new OS, new apps all running on the new CPU Intel ships in 18 months before anyone else has had a chance to code for it.  Propose to me how we manage that, with our market how it is today.

Your argument doesn't make much sense on the processor side. Sure, new procs come out every year or 6 months or something, but architecturally they are the same.

SSE4 or SSE4a anyone?  Yes, the general core design is constant, but the add-ons touted by both companies is incresingly proprietory.  AMD will soon be shipping CPU's with ATI GPU's embedded in them.  Would it not make sence to be targeting something like that?

The graphics processors and other types of devices, yup, moving targets. It's just something you have to deal with. Creating your own just means that you get stuck in an expensive and never ending cycle of obsolescence while the bigger players eat your lunch.

A small company like Amiga doesn't have the funds to go an talk to nvidia or IBM and pay for customized versions of different types of hardware just for the sake of having all the specs let alone actually having them fabbed.

Processor specs are easily avaiable. The linux folks are putting together software that runs on all kinds of stuff, the bsd guys are doing it, hell even Aros is running on cheap common hardware.

Aros will have a web browser before you could even work out a spec for a custom processor or get all the paperwork done to get access to the latest specs on an NVidia or ATI up and coming device.

Amiga has plenty of funds to approach the "other white meat" GPU vendors, such as ST Micro, VIA and XGI.  Each of them is hungry for a vendor to include them by default, and are willing to be wine and dined.  But you still do not know the cost of fabbing, I do.  It is not as expensive as it once was.