Unreleased Amiga chipset better than anything today?

[hammer]

Originally Posted by grond

Only if you add a CPU that at the same time would have been ludicrously expensive to then waste a good part of the processor speed waiting for the slow chipset. We can do that today, now that an emulated PIII-class CPU costs very little but it wouldn't have been viable back then. A competitive product would have had to be competitive in each technical category and not lack significantly in one aspect and compensate that by putting far more effort and cost into another aspect. If putting in a more powerful CPU had been a solution to overcome the slow chipset, the Playstation would have had a more powerful CPU than it had and no slow chipset and still been better.

https://archive.org/details/amiga-wo...e/n57/mode/2up
Amiga World Magazine (November 1993), page 58 of 100,
A1200 price $379
A3000 5MB, 105HD, price $899
A3000T/030, 5MB, 200MB HDD, price $1199
A3000T/040, 5MB, 200MB HDD, price $1599
Cost for 040 card = $400

A3000 is missing the AGA chipset.

The cost estimate for the 68040 cards, $1599 - $1199, hence the cost for 040 cards is about $400.

For a limited production run, Commodore was able to produce an A3640 product that includes the full 68040 @ 25 Mhz CPU and glue logic for A3000's requirements for $400.

68040 socket infrastructure mass production is needed for 68040/68LC040/68040V and 68060/68LC060 CPUs

A1200's $379 + 68LC040 card's $400 = $779 which is cheaper than Apple's $1000 Quadra 605 with 68LC040 @ 25Mhz.

Against PS1 (33 MIPS CPU + 66 MIPS co-processor) and Doom (needs 386DX-33 or 68030 @ 40 Mhz), it's integer workloads, hence the 68LC060 and 68LC040 are sufficient.

$779 in 1993 is similar to A500's $699 price in 1987.

https://techmonitor.ai/technology/mo...0_next_quarter
Date: April 19, 1994.
Motorola Inc yesterday finally launched the long-promised 68060 follow-on to the 68040, claiming that it matches the performance of the Intel Corp Pentium at less than half the price – it costs $263 at 50MHz when you order 10,000 or more and will sample next month

----------------

Without 68040 socket infrastructure mass production, the Amiga couldn't shift baseline math compute power.

Prices are in USD.

-----------------
Another example of the "Phase 5" effect in the Amiga land...

There's a large price difference between Intel Cyclone V SoC FPGA-based Chameleon96's $122.65 USD and Amiga-related / German-related tax-overheads /German profit expectation Apollo Standalone V4's €570.59 (exc. VAT).

Chameleon96's Cyclone V includes dual-core ARM Cortex A9 @ 800Mhz. Chameleon96's $122.65 asking price is from the US site Arrow.com

A low-cost CPU / SoC doesn't guarantee it will remain low-cost when Amiga tax or German profit expectation/overheads are added on top of it.

PiStorm remained low cost due to Sony's "Made in UK" mass production of Raspberry Pi 3A+ and 4B SBC. Raspberry Pi's per unit profit expectation is different from Apollo Core.

Chameleon96 was designed by Novtech located in FL, USA. Novtech is a partner for both NXP (ARM-based SKUs) and Intel.

[hammer]

Originally Posted by Promilus

Well so is GD6426 VLB... and even better at higher resolution. But it hardly was paired up with anything kind of like Pentium II/III to try it out in software quake rendering. What was the point anyway? At the time CD32 was released there was absolutely no CPU fast enough to do even half of what AC68080 does and quarter of what Pi4 emu68k does. So it was impossible to pair ANYTHING with AGA and survive the crash with PS1.

GD6426 doesn't exist, but GD5426 exists.

From https://thandor.net/object/631 with CL GD5440 chipset.
For Doom, Genoa 8540 1MB PCI (with Pentium 100) scored 53.81 fps.
For Quake, Genoa 8540 1MB PCI (with Pentium 100) scored 27.40 fps. (the bottleneck could be the Pentium 100 CPU). Genoa 8540 is not the fastest video card for Pentium 100.

Problem 1:
For 1993, CD32 and A1200 need CPU accelerators and Fast RAM as complained by Team 17, Bullfrog (associated with EA), and Psygnosis.

PC's Doom and 3DO are the 1993 targets.

EA backed 3DO. Doom was released for 3DO with frame rate issues.

This is a high-speed speed 68030 or 68LC040 level CPU to address. Doom doesn't use FPU, hence it's the integer compute power.


Problem 2:
PS1 is the late 1994 and 1995 target.

This is 68LC060 CPU level to address since PS1 is mostly an integer-based games console.

When A1200 was returned by Escom in 1995, SkidMark/Gloom's developer complained about baseline A1200 specs in major Amiga magazines.

Problem 3:
Quake is a 1996 release and Intel released Pentium 120/133 and, Pentium Pro 150/180/200 in 1995.

I purchased Pentium 150 (easily overclocked to 166Mhz with a 60 Mhz to 66Mhz FSB jumper) with S3 Trio 64UV PC clone for Quake. Hollywood standard's 24 fps to 33.90 fps (Pentium 150) Quake was acceptable.

https://thandor.net/benchmark/33
1994-era Pentium 90 delivers 24 fps Quake. At this point, 68060 remained at the 50 Mhz to 75 Mhz clock speed range, hence it wasn't sufficient. This is an AC68080 V2 "What If" situation. Pentium 90 can be overclocked to 100 Mhz via the 60 Mhz to 66 Mhz FSB jumper. Intel Pentium multiplier is locked at this point.

PS1's Quake was refactored for integer compute.

Escom's Amiga Walker's 68030 @ 40 Mhz specs in 1996 were LOL.

[hammer]

Originally Posted by Bruce Abbott

Yes, in 1993 it was enough.

It wasn't enough.

Originally Posted by Bruce Abbott

Being a 'cut down A1200 with CDROM' was a good thing. It meant that Amiga developers (like me) could hit the ground running, leveraging their existing experience and infrastructure to quickly produce good titles for it. The diversity of software which was the Amiga's strength would be the CD32's strength too.

The Sony PlayStation was released in Europe in October 1995. So Commodore had a good 2 years to establish the CD32. Furthermore the CD32 wasn't an unknown quantity, it was part of the Amiga family. This meant that, unlike consoles, it would be supported by the Amiga community. By 1994 CDROM drives were becoming standardized, so the next Amiga computers would probably come with one too. The CD32 could also be turned into an Amiga computer with little effort. That meant each machine helped keep the others alive.

During late 1993 and into 1994, Sony was building 1st party studios, buying 3rd party game companies, making time-exclusive title deals, and creating PR hype to gimp 32-bit game console competitors.

Commodore UK, Team 17, Bull Frog, and Psygnosis were concerned.

PlayStation hardware was operational sometime in 1993.
In 1994, PlayStation hardware was in loyal 3rd party developers' hands.
Sony learns from the failed Betamax.

Originally Posted by Bruce Abbott

Imagine if Commodore had developed a console like the PlayStation that was completely new and not in any way compatible with the Amiga or anything else. Hell, let's imagine Commodore developed the actual PlayStation (and Sony didn't). Do you really think they would have shipped over 100 million units in less than a decade? Or would it have done poorly due to lack of software, low manufacturing volume and weak marketing? Just look at some other 32 bit consoles like the Jaguar to see where it was likely headed.

Commodore International's last CEO Mehdi Ali told Team 17, Bull Frog, and Psygnosis to fu_koff in the meeting.

After Commodore International's disastrous meeting, Psygnosis entered into negotiation to be purchased by Sony. Bull Frog entered into a negotiation to be purchased by EA.

Mehdi Ali's game developer relations weren't his strong point.

Originally Posted by Bruce Abbott

Commodore simply didn't have the clout that Sony had, and never would. They had a small window of opportunity during the 16 to 32 bit console transition. If they made a success of that then they could look at a successor with more advanced chipset. This could possibly have been made availible as an upgrade to the CD32 similar to Sega's 32X (only a lot neater when hidden inside the CD32).

Sega has its own corporate mess.

Sony focused on developers, developers, developers, and learn the lesson from the failed BetaMax, but PlayStation hardware is good and low cost.

[Bruce Abbott]

Originally Posted by Promilus

Commodore had already well over decade of experience doing stuff like that, Sony did not.

You're kidding, right? Sony had vast experience in the consumer electronics industry. Sony partnered with Philips to introduce the Compact Disc format (Sony made the players) so they were far better positioned to make a CD based console. Sony produced their own chips for most of their devices. They also had previously produced their own computers (Sony Hitbit HB-75B MSX in 1984, HB-G900P MSX-2 in 1985, PalmTop PTC-500 in 1990). Sony had enormous resources to plough into a games console if they chose to. Once they entered the market they were virtually guaranteed to crush the competition provided they had a competent product.

It nearly didn't happen though. Sony management wasn't interested in producing their own games console. They partnered with Nintendo to make a CD drive addon (called the 'Play Station') for the NES, but Nintendo backed out leaving Sony with a worthless product design. The developer of the PlayStation's graphic chipset worked on it in secret without approval from management. If it wasn't for his efforts and getting the ear of the CEO, the Sony PlayStation as we know it would never have seen the light of day.

In early 1993 when the CD32 was designed it would be reasonable to conclude that the machine had at least a couple of years useful life, which turned out to be true. Unfortunately due to Commodore's weakened financial state it wasn't able to reach full potential. However a more advanced machine would have required much more development effort and probably would not have been launched at all.

I think basing the CD32 on the A1200 without going overboard on new technology was the right call given Commodore's situation at the time. Just a pity they didn't have the same policy previously, intead of trying to produce an all singing all dancing AAA chipset that proved too much for them. But that was because the engineers were designing stuff they themselves wanted, not looking at the actual market.

[Promilus]

Originally Posted by Bruce Abbott

You're kidding, right? Sony had vast experience in the consumer electronics industry.

You are joking right? So what? Same can be said about Philips (and Philips had a IC fabs) or Panasonic. And neither of those would be able to just enter game console market without solid plan. And that solid plan from Sony was basically 1 pissed off person forcing his way.

Sony partnered with Philips to introduce the Compact Disc format (Sony made the players) so they were far better positioned to make a CD based console.

Console ain't about game medium... and it's a very long way from CD player to gaming machine.

Sony produced their own chips for most of their devices. They also had previously produced their own computers (Sony Hitbit HB-75B MSX in 1984, HB-G900P MSX-2 in 1985, PalmTop PTC-500 in 1990).

Yeah, and HB-75B is made out of the Yamaha, Sharp, TI and NEC chips as well... so majority of the chipset.

Sony had enormous resources to plough into a games console if they chose to. Once they entered the market they were virtually guaranteed to crush the competition provided they had a competent product.

LoL, no! Hardware is just a tip of the iceberg...

It nearly didn't happen though (...)

No way! So Ken who previously had absolutely no experience with gaming consoles or games worked his ass off to produce something better than whole Amiga (or Atari for that matter) R&D team? DUH! I think you involuntarily just made my point...

I think basing the CD32 on the A1200 without going overboard on new technology was the right call given Commodore's situation at the time. Just a pity they didn't have the same policy previously, intead of trying to produce an all singing all dancing AAA chipset that proved too much for them. But that was because the engineers were designing stuff they themselves wanted, not looking at the actual market.

Yeah... not.

[AestheticDebris]

I think people overlook how much working alongside Nintendo probably have Sony insight into how to go about designing a games console, including a laser focus on features that are useful for games specifically and not just generic computing.

Meanwhile, console-ifying an existing computer literally never worked and the CD32 was no exception to that. And it shouldn't really be a surprise, planar graphics are great for low colour depth generic bitmap drawing but absolutely awful compared with a hardware tiled display for 2D games of the era. And for 3D games a bitmap without hardware polygon drawing was never going to hit high frame rates either. And that's before you even consider the the CD32 was supposed to pull off 256 colour graphics, which is literally the point at which a planar display becomes nothing but a hindrance. Heck, every CD32 has actual silicon dedicated to providing a floppy disk controller - sure that saved them design costs but it meant that were literally wasting materials in every unit.

[Promilus]

My dear friend... Ken was working on SOUND CHIP for a console with architecture fundamentally DIFFERENT than PS1. So I'd say this didn't give much of an insight...

console-ifying an existing computer literally never worked and the CD32 was no exception to that.

Correction. It kind of did work with first Xbox and obviously everything past PS3 and Xbox360 is based on PC derived APUs.

every CD32 has actual silicon dedicated to providing a floppy disk controller

Oh not at all... AFAIK floppy functions are spread between paula, gary/gayle and cia and CD32 has only Paula.

[CCCP alert]

The NES, with it's 1983 Famicom guts, is the only time pimping has-been old tech really worked. It didn't even work that well outside the USA.

Success mostly comes down to the games, if the system doesn't have enough great/system selling games that make full use of the unique advantes of your hardware then you're in trouble. With the C64 GS it's too late (10 years!) for the tech and on top of that they didn't even pick the best games of all time on the system to put on the carts (which had blanked off peripheral ports still on the motherboard).

The CD32, and CDTV's, problem is simply there wasn't enough system selling games to justify 300 notes for a console who's true rivals were only about 125 quid. The Amiga hardware is also much much more complex to extract the maximum out of than the MegaDrive.

Even the 1985 Amiga 1000 is really powerful, it still runs things like Lionheart and Lotus II 'console quality' games. The problem is it is really complex to juggle the chipset's power vs a simple kick-ass leading edge hardware sprite/hardware parallax focused technology. I can imagine a really awesome port of things like Mortal Kombat and SF2 etc is possible with AGA sprites but the problem is no such thing existed.

Actually, same problem as Jaguar, pretty powerful but hugely complex and didn't really sell any better than the CD32.

Unless you make awesome FMV based or massive games then on the CD32 the 175 quid extra for the CD32 over a SNES is wasted. Putting Zool 2 or Super Stardust etc on a CD does nothing for the game.

Now if Novastorm had actually been converted instead of the pretty abysmal Microcosm for CD32 as a launch title then we're talking. If you sell an expensive CD based system you have to actually make games that use CD's advantages as much as the custom chips on the motherboard I reckon.

[CCCP alert]

Originally Posted by AestheticDebris

I think people overlook how much working alongside Nintendo probably have Sony insight into how to go about designing a games console, including a laser focus on features that are useful for games specifically and not just generic computing.

If you haven't watched Bedroom's to Billions:The Playstation Years then you are in for a real treat I wish there was a similar commercial quality documentary about the Saturn.

[pandy71]

Originally Posted by grond

Not quite right. The 020 and 030 are officially the same cores with just some minor enhancements (burst read and data cache for the 030). The same was true with the 68000 and 68010. This was the official numbering system where the odd numbers designated bug-fixed/enhanced versions of the preceding even numbered processor.

Originally Posted by hammer

68030's 256-byte data cache is tiny. CPU data caches can't be used for Chip RAM only Amigas. Remember, Amiga's Chip RAM is an uncacheable zone for CPU data cache.

68020 and 68030 are different cores - 68020 has most complete instruction set from all 68K family members - each newer 68K CPU has only reduced instruction list when compared to the 68020.

Originally Posted by hammer

386DX has pin-compatible upgrades with larger +KB range on-chip caches and on-motherboard 64KB/128KB caches are available.

At first second level (L2) cache was not available for 386 motherboard, later rather small (16..32KiB) L2 cache was introduced mostly in 386SX boards (to compensate reduced performance when compared to the 386DX) and in expensive 386DX boards (those mostly workstation/server oriented).

[pandy71]

Originally Posted by Promilus

Oh not at all... AFAIK floppy functions are spread between paula, gary/gayle and cia and CD32 has only Paula.

Paula and Akiko (Akiko cover some aspects of CIA and Gayle/Gary - literally single flipflop to latch MTR state).

Mixed things - there is only Paula so floppy CIA I/O and flipflop to latch MTR state need to be implemented externally.

Btw having CSG on board - Commodore could design some simple RISC, similar to ARM (or even simpler), with 16..20MHz clock - it was doable for CSG and Commodore - adding some supporting HW they could provide 3D for Amiga earlier on market than anyone else.

[sokolovic]

Originally Posted by CCCP alert

Now if Novastorm had actually been converted instead of the pretty abysmal Microcosm for CD32 as a launch title then we're talking. If you sell an expensive CD based system you have to actually make games that use CD's advantages as much as the custom chips on the motherboard I reckon.

It was converted, at least almost finished if you read previews from midi 1994. I still don't understand why Psygnosis canned it (apparently because of Sony) but still published Flink as a CD32 only game later.

[Thorham]

Originally Posted by Bruce Abbott

I think basing the CD32 on the A1200 without going overboard on new technology was the right call given Commodore's situation at the time.

It was an extremely bad idea. Consoles of the time where better suited for many of the games of the time while the A1200 was much better at being a PC (yes, I know, home computer ). Nowadays sticking a PC in a console actually works.

[sokolovic]

Originally Posted by Thorham

It was an extremely bad idea. Consoles of the time where better suited for many of the games of the time while the A1200 was much better at being a PC (yes, I know, home computer ). Nowadays sticking a PC in a console actually works.

Commodore were ahead of their time

[Thorham]

Originally Posted by sokolovic

Commodore were ahead of their time

[MigaTech]

Originally Posted by Bruce Abbott

Commodore simply didn't have the clout that Sony had, and never would. They had a small window of opportunity during the 16 to 32 bit console transition. If they made a success of that then they could look at a successor with more advanced chipset. This could possibly have been made availible as an upgrade to the CD32 similar to Sega's 32X (only a lot neater when hidden inside the CD32).

Firstly you could be right about somethings but... Sony cared about their software developer's. That's why they tried to reclaim all the MK1 PlayStation units, because they were vulnerable to chipping.

Commodore didn't give enough power to the CD32, yet they over engineered the CD64 for a non existent launch.

PlayStation in 1994 was not that powerful compared with an A1200 with an 68030, yet it did wonders. How did they push 33MHz to perform like a DX4?

[Cris1997XX]

Originally Posted by MigaTech

Firstly you could be right about somethings but... Sony cared about their software developer's. That's why they tried to reclaim all the MK1 PlayStation units, because they were vulnerable to chipping.

Commodore didn't give enough power to the CD32, yet they over engineered the CD64 for a non existent launch.

PlayStation in 1994 was not that powerful compared with an A1200 with an 68030, yet it did wonders. How did they push 33MHz to perform like a DX4?

The CPU was not that important, instead they relied on its 3D hardware. Sure, it wasn't a graphics card like those found in a PS4 or Xbox One, but it was still extremely fast at calculating 3D graphics...at the expensive of precision, hence the wobbly polygons and textures

[MigaTech]

Originally Posted by Cris1997XX

The CPU was not that important, instead they relied on its 3D hardware. Sure, it wasn't a graphics card like those found in a PS4 or Xbox One, but it was still extremely fast at calculating 3D graphics...at the expensive of precision, hence the wobbly polygons and textures

Never noticed the imperfections because it was mind blowing compared with anything else, I had seen back then.

Resident evil was just breath taking, it made me realise in 1998 that this was the future of video games.

That FMV intro, just epic. Does it not make sense to push the boundaries of what they had, instead of what they didn't?

[Cris1997XX]

Originally Posted by MigaTech

Never noticed the imperfections because it was mind blowing compared with anything else, I had seen back then.

Resident evil was just breath taking, it made me realise in 1998 that this was the future of video games.

That FMV intro, just epic. Does it not make sense to push the boundaries of what they had, instead of what they didn't?

Most PS1, Saturn and Nintendo 64 games aimed at realism, which sadly means their graphics have aged poorly, with low-poly and weird models. I honestly think pixel art is way better in this regard, including the good old Amiga

[Bruce Abbott]

Originally Posted by Thorham

It was an extremely bad idea. Consoles of the time where better suited for many of the games of the time while the A1200 was much better at being a PC (yes, I know, home computer ). Nowadays sticking a PC in a console actually works.

It worked for Sega in 1983. Several other home computer manufacturers produced a console version of their machines too, with varying success. So it's not like it was unprecedented.

Furthermore the majority of home computers were treated like consoles most of the time. You say the A1200 was 'much better at being a PC'. I disagree. The base machine wasn't well suited to business applications, so why was it the one most sold? Because most people were buying it primarily to play games. I also dispute that the A1200 wasn't as suited to games. For many games it was much better than a typical console.

Originally Posted by Promilus

You are joking right? So what? Same can be said about Philips (and Philips had a IC fabs) or Panasonic. And neither of those would be able to just enter game console market without solid plan. And that solid plan from Sony was basically 1 pissed off person forcing his way.

Compared to Commodore, Sony was far more capable of coming up with a 'solid plan', and had far more resources to put into it. The one thing they didn't have that eg. Sega and Nintendo did was software development expertise. But it wasn't hard to attract developers when you were Sony.

What did Commodore have? Nothing in house, just independent developers who were familiar with the Amiga. That would be the CD32's strength. The main reason developers would be interested in the CD32 is the extra storage space of CD, potentially increased sales due to its low price, and - perhaps most important - lack of piracy. The number one concern of Amiga developers at the time was erosion of their market due to piracy. Amiga fans shrug it off to justify their own role in it, but the fact is that it was a serious problem which was killing the Amiga.

In other respects the CD32 was practically identical to a stock A1200/HD, which made development very easy. I was a CDTV developer so I already knew the format, but for those who didn't it was pretty much just make your title work on a hard drive and add a bit more to support the joypad, NVRAM and CD audio if you wanted. That's a very different thing from learning a new architecture that you can't even play with until you get a (very expensive and difficult to acquire) development system. Few Amiga developers would be willing to make the effort for a machine whose success was far from guaranteed.