@kolla

Quote:

kolla wrote: How did one develop CD32 games at the time?

Did developing for CD32 require an Amiga?

Quote:

How did one develop games for any of the other consoles at the time?

Did developing for other consoles also require special dedicated systems, or was a beefed windows PC good enough to target all of them?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

There was the Net Yaroze for the PS1. Most were custom development kits back in the day! The Amiga was unique at the time for being it's own development system.

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

@BigD

Quote:

BigD wrote: @kolla There was the Net Yaroze for the PS1. Most were custom development kits back in the day! The Amiga was unique at the time for being it's own development system.

Right. Was that good or bad for development of games for the CD32?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@Turrican3

No I get what you're saying, and I totally understand the skepticism.

One of my fields of study is Systems Dynamics, so I can't help but look at how a stronger CD32 footprint would've disrupted the landscape.

It's true that up until 1993, Commodore were not particularly focused on partnerships or other relationships with entities in the Amiga hardware and software ecosystem. While I might be putting that mildly, there's was predominantly a C64 mentality of "build it and they will come".

As you point out, game console success lies in the strategic relationships with game developers.
What I am postulating is that with ~1 million units, Commodore could've been incentivised to get better at these relationships and partnerships. Because to get to ~1 million units in 12 months, C= would've had to adopt the more console like pricing model of low console profits + game sales royalties.
Once ones profits are tied to game sales, it behoves one to ensure greater game sales, so the commercially driven focus moves over to game developers.

With a weakened SEGA, we may never have seen the Dreamcast, wherein Microsoft was the software platform partner. Without MS dipping their toes in the game console market with SEGA, they may not have the key ingredient that prompted them to pivot over to what we know as the original Xbox.

Now, even if Commodore properly executed a console strategy, and the CD32 sold upwards of 3 million units by 1997, it doesn't guarantee that Commodore would be good at game publisher and game developer relationships and partnerships.
The follow-up CD64 could've gone the way of the Dreamcast if MS still decided to spend the billions to enter the game console market in the late '90s.

While we're stuck in this timeline, we can never know how things would've played out differently if the polled key contributors to CD32's failure to manifest its destiny were absent.
As a thought experiment its interesting to look back and see just how close they were, if only this or that.

_________________
All the way, with 68k

agami Quote:

One of my fields of study is Systems Dynamics, so I can't help but look at how a stronger CD32 footprint would've disrupted the landscape.

In the short term, the CD32 wasn't going to be a major disruption to the console market even if it sold over a million units which I believe was likely at a lower price. At introduction in 1993, the hardware was competitive in the high performance console market but it was toward the beginning of the next generation of consoles with significantly better hardware. The 3DO was too expensive for the console market but the Saturn and PS1 were going to take the high performance console market share back. I believe the CD32 could be cost reduced enough to remain on the market as a budget console. Increasing the installed base of CD32s would have been important to attract developers but the combined AGA computer and CD32 market meant the CD32 may have survived with fewer units sold than other consoles. I already mentioned my theory of why C= did not introduce the CD32 at $299 or quickly drop it to that considering that sales were mediocre at best. They needed to make money up front off the hardware sales instead of off game royalties later because of their financial situation. I expect the development and launch of the CD32 was likely less than $10 million USD while the major consoles players spent hundreds of millions and soon to be billions on development and promotion of a new console. The CD32 could be profitable with much fewer units sold because of this although too low of an installed base would have resulted in the end of developer support and new games. Both the Amiga 1200 and CD32 would quickly be tight roping between budget and obsolete due to C= choosing the cheapest possible hardware over basic and relatively cheap performance upgrades that improved value. If the CD32 survived, it may have ended up like the budget Sega Genesis with a large game library which could be cost reduced to a toy. The Sega Genesis has practically been on the market continuously to today despite the limited support and promotion by Sega after the Saturn came out. The Sega Genesis Mini came out in 2019 for $79.99 USD with 40 games and sold at least 1.5 million units. This was adequate for a follow up Sega Genesis Mini 2 in 2022 with 53 games and improved 6 button controller. It's not necessary to sell a lot of units and they aren't making money off game royalties anymore. Sega bet big on newer console hardware and lost which put a pause on the Sega Genesis revenue stream for a few years. The C= bankruptcy did the same for the CD32 while THEA500 Mini is the renewed revenue stream despite deficiencies. The CD32 design has an advantage over the Sega Genesis in that it can be upgraded more without breaking compatibility. Consoles usually don't allow upgrades so the Genesis and NeoGeo lose more compatibility than they gain benefit from upgrading while the Amiga, Atari ST and x68000 can be upgraded more easily as more software is aware of and supports upgrades. The CD32 could have been upgraded quickly to better compete with modern high performance consoles but that was an uphill battle due to C= falling behind in Amiga technology (Bill Sydnes), the long product development cycles at C=, the C= lack of licensing and business cooperation, the poor C= marketing and promotion skills and the financial situation at C= relative to how much it cost to remain competitive in the high performance console market. Still, the CD32 may have survived as a budget console with a relatively small installed base. How successful the CD32 would have been is highly dependent on how much the CD32 could have been cost reduced.

I can't put enough emphasis in how important cost is for a console, which you understand but maybe not others. A good example is the Sega Saturn vs Sony PS1. The Saturn had a better launch, it had an advantage with 2D games which were popular at launch especially fighting and platform games, it had a better controller, games often loaded faster (better code density and dedicated CPU for CD-ROM) and the sound is often better than on the PS1. It can even hold its own in 3D games against the PS1. The Saturn lost due to being significantly more expensive and not enough better to justify the large cost difference. The games at launch for both the PS1 and Saturn were underwhelming and even the CD32 likely could have held its own until Christmas of 1995 or 1996.

When Sega Saturn DESTROYED the PlayStation !
https://www.youtube.com/watch?v=1lwQszYEZZo

Sega Saturn Vs PlayStation - Tomb Raider
https://www.youtube.com/watch?v=y6ZzjgJuo54

Sega Saturn Vs PlayStation - Mortal Kombat Trilogy
https://www.youtube.com/watch?v=OXu3QiRnMPc

More 3D games started to come out and the PS1 was moderately better for 3D and development of these types of games was easier. Sony did a great job with developers and promoting the PS1, especially for their first entry into the console market. The Playstation has this over hyped hardware advantage due to them coming out the major winner and because they are the de facto console standard today (XBox and Nintendo market share has been shrinking). There is a documentary called "The Story of Playstation" which discusses the Playstation but even it is hyped (free on Tubi with commercials).

https://tubitv.com/movies/100009124/the-story-of-playstation

There is no mention of the Amiga or CD32 but there is a mention of the SNES sound setting the standard for the PS1 to copy when the Amiga and CD32 often sounded better despite the SNES sounding good on paper. The PS1 was evolutionary and a major accomplishment to bring semi-modern 3D to the console market before the desktop market but it was not untouchable hardware and the 3D quality was low but better than the competition. The PS2 3D was much higher quality and outsold the PS1 making Sony a household name and practically untouchable at that point. The PS2 had a DVD CD-ROM drive for a competitive price to DVD only players adding extra value. Vertical integration was an advantage of Sony over Sega which lowered hardware costs for the PS1 and PS2. C= was vertically integrated at one time with their chip development but squandered it. Jeff Porter did figure out a way to get a lower priced Sony CD-ROM drive for the CD32 at least.

Last edited by matthey on 20-Sep-2023 at 04:06 AM.

@matthey

Yes, the CD32 uses a CD drive originally made for just audio.

How fast can it go?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@agami

Quote:

One of my fields of study is Systems Dynamics, so I can't help but look at how a stronger CD32 footprint would've disrupted the landscape.

I can understand that, it can be a funny exercise after all.

Quote:

As you point out, game console success lies in the strategic relationships with game developers.

Or a VERY strong first/second party offering, as Nintendo has proven over and over even during this (as of now) uninterrupted Sony dominance. Commodore, unfortunately, had neither at that point in time.

But let's ignore this.
I'd argue the basic issue was that very soon the CD32 would simply be outdated to say the least.

The 2D renaissance we're still experiencing would take more than a decade (I'd date it back to the huge success of New Super Mario Bros on the Nintendo DS) so even assuming decent HW sales and a - previously unheard of - major developer support, the CD32 would have likely found itself in a limbo: at the same time it would have been more powerful than a SNES/MD but with a comparatively weaker software offering, and hugely underpowered compared to PSX/Saturn with (assumed!) decent but ultimately outdated software lineup.

That's why even the best case scenario (a healthier Commodore able to get decent software support, affordable HW pricing, good sales) seems one that would have led to a failure nevertheless to me.

@kolla

if i remember good it was run about 100/200 kb/s in reading.
for that time was a good speed. if we think how much small are iso of the cd32 games and how much small are the SNES and MD images

Last edited by tlosm on 20-Sep-2023 at 02:14 PM.

_________________
I love Amiga and new hope by AmigaNG
A 500 + ; CDTV; CD32;
PowerMac G5 Quad 8GB,SSD,SSHD,7800gtx,Radeon R5 230 2GB;
MacBook Pro Retina I7 2.3ghz;
#nomorea-eoninmyhome

kolla Quote:

Yes, the CD32 uses a CD drive originally made for just audio. How fast can it go?

The CD32 uses a dual speed CD-ROM drive which most of the console market started with including the 3DO, Jaguar CD, Saturn and PS1 but excluding the Sega CD and Neo Geo CD. The audio CD-ROM drive lacks controlling hardware which is convenient for computers so the console hardware has to provide it and it needs to be robust (CD32 had logic in Akiko ASIC for CD-ROM). The Sega CD also used an audio CD-ROM drive and had more issues than the CD32.

https://en.wikipedia.org/wiki/Sega_CD#Development Quote:

Latham and Sega of America vice president of licensing Shinobu Toyoda assembled a functioning Sega CD by acquiring a ROM for the system and installing it in a dummy unit. The American staff were frustrated by the Sega CD's construction. Former senior producer Scot Bayless said: "[It] was designed with a cheap, consumer-grade audio CD drive, not a CD-ROM. Quite late in the run-up to launch, the quality assurance teams started running into severe problems with many of the units—and when I say severe, I mean units literally bursting into flames. We worked around the clock, trying to catch the failure in-progress, and after about a week we finally realized what was happening." He said the problems were caused by the need for games to use more time-seeking data than the CD drive was designed to provide.

An audio CD-ROM drive may be lower quality but problems may be simply due to lack of staying within specified parameters and lack of error tolerance in a custom roll your own drive controller. The data read rates of double speed CD-ROM drives should have been similar but there was a difference how fast consoles could decode and buffer the data streams. The CD32 seemed to have fast loading times to me but I haven't measured against other consoles. Programs were smaller for the 68k (CD32, Sega CD and Neo Geo) and SuperH (Saturn) due to good code density while ARM (3DO) and MIPS (PS1 and N64) were roughly 50% larger and slower to load. This reduced the time to the startup screen but not for loading data which included other factors like data compression.

Turrican3 Quote:

That's why even the best case scenario (a healthier Commodore able to get decent software support, affordable HW pricing, good sales) seems one that would have led to a failure nevertheless to me.

Failure for a business probably should be defined here. The purpose of a business is to make money for its shareholders (otherwise it is a non-profit organization). C= suffered the ultimate business failure due to bankruptcy. However, the CD32 should be judged on whether the CD32 project made money or, more fairly, would have made money. I expect the CD32 would have been a success due to the low cost of entry into the console market for C= and the low cost of the CD32 hardware. They likely only needed to sell hundreds of thousands of units for the project to have an overall profit. This is comparable to THEA500 Mini which was relatively cheap and quick to bring to market, used cheap underwhelming but adequate hardware, leveraged the advantage of an existing large game library to add value and appealed to an existing customer base with an enhanced product. I expect THEA500 Mini project has already made money with unit sales of a few hundred thousand as they are planning for an Amiga Maxi, likely with cheap and disappointing emulation hardware for an Amiga compared to what is possible if trying to create a sustainable Amiga market again. Likewise, C= may not have been concerned if they had to exit the console market as long as the CD32 project made money. They could reenter later if they saw another opportunity. This is a much different strategy compared to Sega which bet so big with spending to gain console market share that over 9 million Saturn units sold was a failure. This is also much different than Microsoft which lost $4 billion dollars on the original XBox project to gain a temporary distant 2nd place position in the console market for later generations of XBox consoles that would hopefully be profitable against Sony with a vertical integration advantage. C= was financially conservative but too conservative for technology related markets. They needed to be more aggressive as it takes investment to make money but it was likely already too late for C= by the time the CD32 was released. Maybe they could have obtained a bridge loan if there was enough demand for the CD32 at $399 USD but that was unlikely in the console market where console hardware is commonly heavily subsidized.

@agami
I found a great resource that gives sales and marketing info on C= through 1993 and perhaps into 1994 near the bankruptcy.

https://archive.org/details/commodore-post-bankruptcy

There is some good CD32 info. There were 75,000 CD32 units sold in the 2nd half of 1993 at introduction generating $25 million in revenue which I calculated as $333.33 per/unit average sale price. The production cost of the CD32 is given as $235 which is $98.33 of profit per/unit and $7,375,000 of total profit. Other included literature mentions they had sold 100,000 CD32 units. It wasn't quite as cheap to produce as I thought but costs were dropping all the time. Surprising to me was that the CD32 was $9 more to produce than the Amiga 1200 according to the literature. The biggest difference comes from what is marked "Floppy Drive" which is $33 for the Amiga 1200 and $57 for the CD32 where I assume it is the CD-ROM drive but it doesn't fit with Jeff Porter's claim of a $15 CD-ROM drive but maybe C= included custom parts with the CD-ROM drive. It's still not a bad price for the drive which was likely to be one of the fastest falling costs for the CD32. I wouldn't be surprised if the production cost would have dropped below $200 in a couple of years. Other than the in-house CSG/MOS custom chips, custom chips had one of the longest lead times of 12-14 weeks which was not helpful at the end. I believe the AGA chipset cost is claimed to be less than $13 and the 68EC020 less than $7 while "Single Chip with Integrated CPU Feasible" is mentioned as double performance and dropping the cost to less than $20 if I am reading some blurred text correctly. A diagram shows what I expect to be a SoC with 68030 and AA+ chipset running at 57MHz but only shows twice the performance of AGA even though the chipset is clocked twice as fast and the CPU would be clocked about 4 times faster than a 68EC020@14MHz and due in early 1995. A 64 bit PA-RISC Hombre chip@135MHz was due out in late 1995 and was shown elsewhere as being producible for $35. I wonder if they would have combined the $35 Hombre CPU/3D chip with the $20 68k Amiga SoC to retain backward Amiga compatibility, give performance a boost and make development easier for their console that would compete with the Saturn, PS1 and N64. Retaining very good CD32 backward compatibility would have been very beneficial, the CPU performance would have been better than other consoles, the price shouldn't have been too bad based on this data and it should have been one of the easiest to program consoles among the competition. I don't know what the 3D performance would be like and the timeline is unbelievable. The C= financial results from 1993 are bad enough that it may have shut down development just as they were planning to ramp it up. Amiga sales and revenue were trending down even with the addition of the CD32 while PC clone sales fell off a cliff down to about 14% of what they had been. C= certainly could have lowered the CD32 price which would have generated more revenue but maybe not as much up front profit. The document mentions they added a 4 game bundle like I suggested to add value and they may have lowered the price as even no profit is shown in one place. There are some market share percentages in some places which are interesting as well. Enjoy.

Last edited by matthey on 21-Sep-2023 at 04:00 AM.

@matthey

Something doesn't smell right. I find the $67 for "Other Components" for CD32, being higher than the $50 for A1200, very questionable.

While the base CD drive mechanism may have been $15, the fully functional drive ended up being $57. I wonder if they're amortizing any of the Akiko dev costs in there, or perhaps the entire chip.

If Manufacturing is only $10 per unit, and Sea Freight is only $1 per unit, which sound good, then they're doing something else very wrong to have the CD32 cost more than $200 per unit to produce.

Anyway, I do agree that many of the components would see massive price drops within 12 months. $235 in mid 1993 would quite possibly be $135 in mid 1994.

I still think they should've gone with an EC030/25 and some SRAM (1MB) if they're going to price for performance at $399. As it was, it should've had an MSP of $249.
They'd clearly identified that the home entertainment and video-game market worldwide was worth billions. They should've acted accordingly. They were right f#@&ing there! In the right moment in time with all of the key ingredients.

The best thing the CD32 had going for it was the A1200 architecture. While existing Amigans at the time saw an upgrade path to a home/personal computer equal to the A1200, for the broader console market it could've meant life-extending upgrades to stay competitive with the Saturn and PS1 as they rolled out into the western markets.

By Xmas 1995, even if RISC/3D (Hombre) was not ready, they could've worked with an emerging 3D chip shop to produce an add-on card. It didn't have to beat the PS1, it just needed to place. Making sure devs continue to cross-publish titles.

Last edited by agami on 22-Sep-2023 at 04:13 AM.

_________________
All the way, with 68k

@agami

Like many have pointed out both A1200 & CD32 system should of been a bit more powerful, either 030EC25mhz or adding 2mb of Fast Ram, I mean both would of been amazing, but we all know margins are important.

When I look at the games that did take advantage of higher spec Amigas,

Genetic Species
Shadow of the Third Moon
OnEscape
Breathless
Alien Breed 3D 2
Flyin High
Payback

Just to name a few off my top of my head you can see how games / 3d engine where much closer to what PC and Console could do in 94/95.

But even with these specs bump, I think really we needed a better graphic chip.

_________________
AmigaNG, YouTube, LeaveReality Studio

@amigang

That be said Games can be a system seller, look at Sonic / Mario, Amiga could of had Zool, but my gut feeling, Mascot wise and character wise, Zool was very good, but I feel like the Game just wasnt quite up to Sonic level (which it was clearly more aimed at).

Everyone in 93/94 was falling over them self having Doom or a Doom type game on there system. Doom would not run on CD32 unless it was a whole new engine / version of the game like the nutter who make a kinda doom engine for Zx Spectrum https://www.youtube.com/watch?v=3v7cFGneuaw . I mean Commodore could of tried to get the name, and rename / skin Gloom to Doom. I dont know if ID software would of been up for that deal.

But one of the best selling games in 93 was Myst, now i know a port came over to Amiga in 97, which required 8mb Fast Ram, but I do wonder with a bit of scale down and a better port it might of just about ran on a CD32. Would that of sold more systems?

Is there any game that could of be made / ported within the spec of AmigaCd32 that could of been a system seller?

Last edited by amigang on 21-Sep-2023 at 12:53 PM.

_________________
AmigaNG, YouTube, LeaveReality Studio

@amigang

Thinking abit more on this I was thinking for the A500, it was the Batman Pack that sold the most for Amiga, and Amiga systems having the best version of that game, in 1993 Jurassic Park movie was the big hit, but the game of that just wasn't that good, plus it had a very poor mans doom level, maybe if Commodore had the money they could of struck a similar deal with Ocean, maybe help develop the game a bit better to suit Amiga graphics, Im thinking like maybe a driving level like Batman game had, but Jurassic Park themed. That might of been pretty cool. Then of course make a Jurassic Park Pack for Cd32. would that of sold more?

Other big ips could of been Star Wars, Robocop, Terminator. I doubt Commodore had the money to get the ip to make a new game based on them franchise, but it might of helped.

I guess we will never know, but it fun to speculate.

_________________
AmigaNG, YouTube, LeaveReality Studio

@amigang

Jurrasic Park AGA is just annoying. I sacrificed an evening to it but no incentive to go back!

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

agami Quote:

Something doesn't smell right. I find the $67 for "Other Components" for CD32, being higher than the $50 for A1200, very questionable.

There are more than a few things in the document that are suspicious but most of the numbers are reasonable even if it is all a very elaborate fake. The CD32 doesn't have many "Other Components" on the PCB assembly. The Akiko chip and 1MiB ROMs (Amiga 1200 has 512kiB ROMs) should be listed under "Semiconductor Chips". There is also the CD-ROM audio mixing support which the Amiga 1200 does not have but that shouldn't amount to much. The CD32 was relatively new so maybe the cheapest suppliers hadn't been found yet or the cost included large volume purchases of new parts necessary to get good pricing.

agami Quote:

While the base CD drive mechanism may have been $15, the fully functional drive ended up being $57. I wonder if they're amortizing any of the Akiko dev costs in there, or perhaps the entire chip.

It's possible but my expectation would be for Akiko to be listed under "Semiconductor Chips". I would expect the chips price of $6 to include the AGA custom chips yet elsewhere the AGA chipset is mentioned as costing $15 and $13. The cost was falling all the time so $15 down to $13 sounds reasonable but down to $6 is hard to believe especially if including more chips than the AGA chipset. The CD32 doesn't have a lot of other chips but it has the 1MiB ROMs which if so cheap then why not provide more in ROM to save on $25/MiB memory (ROM can be accessed at full speed like fast memory too)?

agami Quote:

If Manufacturing is only $10 per unit, and Sea Freight is only $1 per unit, which sound good, then they're doing something else very wrong to have the CD32 cost more than $200 per unit to produce. Anyway, I do agree that many of the components would see massive price drops within 12 months. $235 in mid 1993 would quite possibly be $135 in mid 1994.

That fast of a cost reduction sounds optimistic for 1 year due to many of the cheap older parts nearing a bottom close to production cost. For example, the cost of a newer 68040 could drop by more than 25% in a year while the 68EC020 may drop $0-$2, especially if they were getting a great deal at $8. My guess would be more like 10%-30% cost reduction in 1 year for the CD32 without major cost reduction measures like improving integration.

agami Quote:

I still think they should've gone with an EC030/25 and some SRAM (1MB) if they're going to price for performance at $399. As it was, it should've had an MSP of $249.

From the document, I can better predict a 68EC030 at ~25MHz to cost $5-$10 more than $8 for the 68EC020 and memory was $25/MiB. Other circuitry, chips, possibly oscillator and a larger PCB would likely be required but $40-$50 for about a 3 times performance boost is my rough guess. A 68EC030 omits the MMU but has a 32 bit address bus unlike the 68EC020 24 bit address bus which only supported addressing 8MiB of memory. A single fast memory SIMM socket with 1MiB SIMM probably would have cost a few dollars more but this may be worthwhile for the versatility and modularity for full computer upgrades, audio/video use, embedded use and allowing more software for an upgraded CD32+ to run on the older CD32.

From the document, Perhaps the reason why they didn't upgrade the CD32 to a 68EC030 with AGA is that the chipset couldn't operate at ~28MHz due to the ancient chip process used except for Lisa. Jay Miner called for more modern CMOS upgrades long ago but C= didn't think it was necessary and since they hadn't upgraded CSG/MOS fabs, the new chips couldn't be produced by them and had longer development cycles and lead times. A fully upgraded AGA chipset in CMOS would have offered the following advantages.

o the chipset could have been clocked at double the frequency (28MHz vs 14MHz) up to doubling the chipset performance
o a 68k CPU could have been clocked at 28MHz without requiring another relatively expensive oscillator
o CMOS chips would have reduced power requirements allowing for cheaper power supplies or higher performance CPUs without using a fan
o the chipset would have been easier to enhance and integrate further. Integrating chips together reduces chip and PCB costs and allows for smaller and cheaper devices

Compelling? Not to C= who didn't get it done by AGA release in 1992 even though it looks like it was planned to run at 28MHz according to one of the diagrams. It was planned for AA+ due in 1994 according to the document and finally including a reduction of one in the custom chip count. Oddly, one of the diagrams shows a CD32+ due in the fall of 1994 still using AGA but with a 68EC030@28MHz. That should probably include AA+ but more memory than 2MiB is needed for the CD32+ for 16 bit chunky and a lot more for the CD64-3D using 16 bit and full color chunky and because the PA-RISC CPU has horrible code density. This was probably an early diagram that was never updated. Another diagram shows AA+ operating at 57MHz but only a 68030@50MHz was available and that would be quite an overclock. The "Single Chip" SoC option with 68030 and AA+ likely would have allowed a 68030@57MHz as shorter distances with better connections would likely allow this and a smaller chip die size could always be used.

One other thing. You said 1MiB SRAM above but you probably meant VRAM? C= in their infinite wisdom made sure the base Amiga chipsets did not support VRAM after Jay Miner designed the Ranger chipset to support VRAM. The high end AAA chipset was designed to support either VRAM or DRAM which likely could have been done on the base chipsets as well. Even AA+ didn't support VRAM. If VRAM was supported for chip memory, it likely would have eliminated CPU and chipset contention and allowed higher refresh rates. Would it be better to have 2MiB of VRAM chip memory or 2MiB of DRAM chip memory and 1MiB of DRAM fast memory though?

1MiB of SRAM would be cool as it could be dual ported like VRAM, it doesn't have wait states and it doesn't need refreshing. I calculate 1MiB of SRAM to use 50,331,648 transistors though. The console custom chips were typically around 1 million transistors back then and a "Single Chip" Amiga SoC with 68030 and AA+ would likely have been around that too. Caches are usually made of SRAM and many of the console CPUs didn't have any caches, not even L1.

Console | CPU@MHz | Caches in Bytes
SNES 65816@3.58MHz -
SegaCD 68000@12.5MHz -
CD32 68EC020@14MHz 256I
3DO ARM60@12.5MHz -
Atari Jaguar Custom RISC@26.59MHz -
CD32 68030@28MHz 256I+256D
Sega Saturn Hitachi SH-2@28.6MHz 4096U
PS1 MIPS R3000A@33.87MHz 4096I+1024D
N64 MIPS R4200@93.75 MHz 16384I+8192D

Some of the consoles had different memory for CPU, video and audio so accesses didn't interfere with each other but partitioning reduces the maximum available memory allocation and wastes memory which is why unified memory was becoming more popular. CPU caches reduce memory accesses making unified memory more efficient. CPU clock rates and performance were growing quickly with newer consoles because of deeper pipelines, larger caches and better chip fab processes to reduce power and heat. The larger caches used lots of SRAM transistors as the MIPS instruction caches above exploded is size due to the poor code density. In my opinion, SuperH was a better choice early on for consoles as the simple RISC CPU with pipelining freed transistors for a larger cache and the code density was good so it didn't waste as much of the cache. The only problem is that it doesn't scale up in performance because all the instructions are only 16 bits in length so weak and many instructions are required to perform the same workload. The Saturn had 2 of these SH-2 CPUs which shared the same memory but they performed well when the workloads were mostly in cache. It is easier to use a single more powerful CPU though. The 68020 and even the 68030 were pretty old by this time with less pipelining and small caches but they were cache misers and easy to program. RISC hype was in full swing and RISC CPUs did have a transistor savings until the transistor savings was spent on expensive SRAM instruction caches and more memory. Memory cost was still a relatively high percentage of a console cost.

agami Quote:

They'd clearly identified that the home entertainment and video-game market worldwide was worth billions. They should've acted accordingly. They were right f#@&ing there! In the right moment in time with the all of the key ingredients.

C= finally saw the light but too late. The Mehdi Ali and Bill Sydnes dubious duo that tried to kill the Amiga and switch to PC clones did them in. They should have unloaded their commodity PC clone business selling into a saturated market before the 86% drop in PC clone revenues at the end of 1993 as the document shows. That is revenue too where Dave Haynie said they were forced to sell clones below cost which is believable. At least selling the CD32 at cost generated royalties later.

agami Quote:

The best thing the CD32 had going for it was the A1200 architecture. While existing Amigans at the time saw an upgrade path to a home/personal computer equal to the A1200, for the broader console market it could've meant life-extending upgrades to stay competitive with the Saturn and PS1 as they rolled out into the western markets. By Xmas 1995, even if RISC/3D (Hobre) was not ready, they could've worked with an emerging 3D chip shop to produce an add-on card. It didn't have to beat the PS1, it just needed to place. Making sure devs continue to cross-publish titles.

Despite C= managerial lack of vision and negligence in the past, they had more options than most businesses and some good engineers remaining until they ran out of financing.

Last edited by matthey on 22-Sep-2023 at 12:25 AM.

@matthey

Quote:

matthey wrote: That fast of a cost reduction sounds optimistic for 1 year due to many of the cheap older parts nearing a bottom close to production cost... My guess would be more like 10%-30% cost reduction in 1 year for the CD32 without major cost reduction measures like improving integration.

Fair enough, my 42% in 12 months is highly optimistic. But I think 30% ($70) is doable, via multiple cost optimisation strategies, one of which is getting better pricing for higher volumes.

If 1M+ unit sell-through per year is the target, then a production cost of $150 or less should be the goal.

Quote:

Compelling? Not to C= who didn't get it done by AGA release in 1992 even though it looks like it was planned to run at 28MHz according to one of the diagrams. It was planned for AA+ due in 1994 according to the document and finally including a reduction of one in the custom chip count. Oddly, one of the diagrams shows a CD32+ due in the fall of 1994 still using AGA but with a 68EC030@28MHz. That should probably include AA+ but more memory than 2MiB is needed for the CD32+ for 16 bit chunky and a lot more for the CD64-3D using 16 bit and full color chunky and because the PA-RISC CPU has horrible code density.

I saw that too. They went from big gaps between updates, to a plan for annual updates.

Usually the market would punish a company for such frequent updates, but given the platform's excellent backward compatibility, I think it would've worked, especially if they could've worked in some forward compatibility as well.

Though based on the document, the spec of the CD32+ should've been in the original CD32.
The 12-month refresh in the CD32+ I agree should've been AA+ instead of AA (AGA)
And despite their best efforts, a big shift to RISC/3D CD64-3D would likely slip into 1996, which would actually be OK, especially if there was a basic 3D add-on for the CD32(+).

Last edited by agami on 23-Sep-2023 at 04:25 AM.
Last edited by agami on 23-Sep-2023 at 04:24 AM.

_________________
All the way, with 68k

agami Quote:

I saw that too. They went from big gaps between updates, to a plan for annual updates. Usually the market would punish a company for such frequent updates, but given the platforms excellent backward compatibility, it would've worked.

There could be repercussions to such frequent hardware updates and starting with such a low original CD32 spec. Customers may see their hardware investment quickly devalue with yearly CD32 updates. While current customer software investments would be retained, large jumps in hardware performance would mean that many games would soon stop being produced for the lowly original Amiga CD32 or they would run poorly. Hardware CPU accelerators are possible but they would have likely been more expensive than the next generation of CD32 and lacked the chipset enhancements.

agami Quote:

Though based on the document, the spec of the CD32+ should've been in the original CD32. The 12-month refresh in the CD32+ I agree should've been AA+ instead of AA (AGA) And despite their best efforts, a big shift to RISC/3D CD64-3D would likely slip into 1996, which would actually be OK, especially if there was a basic 3D add-on for the CD32(+).

A 28MHz clocked 68030 and AGA chipset would have made a big difference for the CD32 "base" standard as it looks like C= was originally planning. The CD32+ would still have a 68030 and AA+ chipset clocked at 28MHz so not as much of a performance upgrade unless reduced to a single chip SoC where the 68030 and AA+ could likely be clocked to 57MHz. AA+ supports 4MiB of chip memory which is useful for double buffered 16 bit chunky screens although AA+ still has minimal texture mapping and 3D capabilities. While I have concerns about the performance of a PA-RISC CPU using a SIMD unit for 3D, it is flexible and upgradeable with new 3D libraries in ROM back then (today flash memory for bug fixes any time). The licensed PA-RISC CPU core was likely mature so as long as the C= added SIMD instructions didn't have bugs, the PA-RISC 3D chips could be produced and the ROM based 3D libraries could be changed at the last minute. The PA-RISC CPU and SIMD unit would have been available for other more general purposes while some specialized fixed pipeline 3D hardware is more efficient at 3D but less flexible. Using CPU cores with SIMD units for 3D workloads didn't scale well as most CPU cores were too large but smaller more flexible massively parallel universal shader cores often using SIMD instructions is where we ended up for 3D GPUs today. The PA-RISC SIMD unit solution likely would have made the CD64-3D competitive enough but it may have been replaced later as the SIMD unit was integer only so low quality 3D like the PS1 and while more parallel PA-RISC CPU cores could be added, the 3D scaling likely would not have remained competitive. Hopefully, they would have been smart enough to retain the 68k Amiga SoC for compatibility but AA+ may have only supported 16 bit chunky while the CD64-3D could also render into 24/32 bit chunky bitmaps.

Despite the CD32 only being out for about 6 months and only about 100,000 known hardware units, the CD32 standard remains popular. I did a recent Indie game search for "Amiga", as there is no CD32 tag, which found 278 games.

https://itch.io/games/tag-amiga

The following are Amiga games that look like they have full CD32 compatibility and support.

https://sonicslothgames.itch.io/turrican2aga
https://electricblacksheep.itch.io/project-quest-chapter-i
https://z-team.itch.io/rick-dangerous-amiga
https://z-team.itch.io/maria-renards-revenge
https://mcgeezer.itch.io/kung-fu-remaster
https://nivrig.itch.io/turbo-tomato
https://mcgeezer.itch.io/turbo-sprint
https://retream.itch.io/skillgrid
https://retream.itch.io/memo
https://acidbottle.itch.io/amiga-wonderboy
https://retream.itch.io/follix
https://mcgeezer.itch.io/santa-run-2020
https://captainnow2.itch.io/amiga-renegades-deluxe-2014
https://amiga-cobe.itch.io/pong4kb-cdtvcd32-commodore-amiga
https://amiga-cobe.itch.io/lumberjack-revision-cdtvcd32-commodore-amiga
https://lockeddoorpuzzle.itch.io/wrong-way-home
https://amiga-cobe.itch.io/aztec-challenge-episode-one-commodore-amiga
https://youenchene.itch.io/blower
https://earok.itch.io/amishion-impossible-amiga
https://kobold-eduard.itch.io/goblinking-eduard
https://jotd666.itch.io/supercars-2-aga

The following games have some CD32 compatibility or ISOs but it's not clear if they have full CD32 compatibility and support.

https://rgcddev.itch.io/tiger-claw
https://rgcddev.itch.io/powerglove-reloaded
https://h0ffman.itch.io/metalgear-amiga
https://mcgeezer.itch.io/rygar-aga
https://danteretrodev.itch.io/akumajou-dracula-aga
https://retream.itch.io/ring-around-the-world
https://retream.itch.io/artpazz
https://kobold-eduard.itch.io/fynn-and-thorin

There are lots of comments asking for CD32 game and controller support too. Impressive CD32 support for a long dead low spec console? Compare to native PPC AmigaOS 4 games numbering 4 that I found from https://itch.io. Talk about a zombie AmigaNOne platform. Maybe we should analyze the success, or rather lack of it, of the AmigaNOne. Over 20 years of dismal sales, PPC is dead, only selling a few hundred units of hardware a year at best, reducing their minimum hardware standard, horrible hardware value, barely any software. At least we can look at C= and say they could have done this or that to survive and then maybe been right back in the game but the situation looks hopeless no matter how much some fans like AmigaOS 4.

Last edited by matthey on 23-Sep-2023 at 07:27 AM.
Last edited by matthey on 23-Sep-2023 at 07:20 AM.

I'll summarize again:

The CD32 could have been a success if C= had done everything 180° from what made C= C=....

In reality it was not only late to ship, the whole project was way to late start and even a "loaded" 1990 C= would have had problems getting the funding needed for a proper cold console launch. An already failing 1992/3/4 C= never had a chance.

Remember in that general timeframe both Sega and Atari died while shipping o.k. consoles while it took behemoths like Sony and MS to enter the market successfully (and with MS it took 2-3 generations).

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

@matthey

Quote:

There are lots of comments asking for CD32 game and controller support too. Impressive CD32 support for a long dead low spec console? Compare to native PPC AmigaOS 4 games numbering 4 that I found from https://itch.io. Talk about a zombie AmigaNOne platform.

Well that's hardly surprising, isn't it?

I mean even ignoring the potential common code baseline with other AGA hardware which would only make the comparison even more embarassing (huge installs are not an issue anymore in 2023 as it can be safely assumed most AGA owners have a decently sized mass storage, the only incompatibility I can see is wanting to have CD audio at all costs), we're talking about a potential difference of *two* orders of magnitude in the userbase.

Of course nobody knows how many CD32 units have been actually sitting, long forgotten, in a closet (or worse, got rid of) but that could be argued for PPC hardware too... of course, as I briefly mentioned before, I expect most reasonable developers to target AGA hardware and not specifically the CD32 (unless he *really* wants to, which makes little sense if you ask me) so I guess any OS4/PPC developer needs to have a very strong passion/motivation or perhaps some insurmountable technical requirement to target this tiny market

Please understand though I'd NEVER EVER blame them, I mean who am I to criticize? I'm just stating some very simple statistics (we have two platforms with a significantly different install base) but if anybody wants to spend his time for an extremely niche audience, so be it.

Quote:

Maybe we should analyze the success, or rather lack of it, of the AmigaNOne. Over 20 years of dismal sales, PPC is dead, only selling a few hundred units of hardware a year at best, reducing their minimum hardware standard, horrible hardware value, barely any software. At least we can look at C= and say they could have done this or that to survive and then maybe been right back in the game but the situation looks hopeless no matter how much some fans like AmigaOS 4.

That's why some people have been advocating for a Raspberry Pi port for a long time.

But I don't think that's the right thread to open this can of worms (not to mention we have already discussed in the past )