Was anyone else disappointed with the A1200?

[Bruce Abbott]

Originally Posted by pixie

Interestingly I've tried on similar settings that of a 50Mhz 68030 on adoom in winuae, but on rtg, and it got me about the same fps.

On real hardware with 50MHz 030 on AGA (A1200 with Blizzard 1230IV) I get 12.5 fps in ADoom, and 14 fps in DoomAttack.

[Bruce Abbott]

Originally Posted by hammer

For 1992
68000-12 = $5.5
68EC020-16 PQFP = $16.06,
68EC020-25 PQFP = $19.99,

68EC030-25 PQFP = $35.94
68030-25 CQFP = $108.75

68040-25 = $418.52
68EC040-25 = $112.50
---
Competition

AM386-40 = $102.50
386DX-25 PQFP = $103.00

486SX-20 PQFP = $157.75
486DX-33 = $376.75
486DX2-50 = $502.75

Interesting price comparison. We can see why Commodore chose the 16MHz EC020 for the A1200. We also see why the A4000 was so expensive.

The EC040 was a bargain! 4 times the compute power of a 25MHz 030 for almost the same price. But... no MMU or FPU. Good for games, not good for raytracing or program development. I wonder why no Amiga accelerator card used it?

On the 'competition' side we see why AMD was killing it with the 386DX-40, which was a popular 'mid-range' PC at that time.

The X86 PC clone market competition kept the distribution channel's profit margin relatively low i.e. it minimized the "Phase 5" tax.

Low profit margins, and lots of bankruptcies.

BTW yesterday I attached a hard drive to my 386SX-16 with 16MB RAM and booted Windows 95. Very slow starting up, but not too bad once it got going. Unfortunately I don't have a serial port mouse so I had to use the keyboard to operate it, which wasn't fun. Next problem is how to get Doom onto it. Might be easier to stick the hard drive into a modern PC and copy it across, rather then trying to zip it onto multiple floppy disks.

[AestheticDebris]

Originally Posted by Bruce Abbott

Interesting price comparison. We can see why Commodore chose the 16MHz EC020 for the A1200. We also see why the A4000 was so expensive.

Rumour always had it that Motorola had a job lot of 020s they basically couldn't give away, since everyone wanted 030s. They basically offered them to Commodore for less than cost price just to clear their backlogs and CBM snapped them up because they desperately wanted to keep costs down.

[pixie]

Originally Posted by Bruce Abbott

On real hardware with 50MHz 030 on AGA (A1200 with Blizzard 1230IV) I get 12.5 fps in ADoom, and 14 fps in DoomAttack.

Which are on pair with what was being shown on Falcon

[pixie]

Originally Posted by Bruce Abbott

Got a link to that?

The best video of DML's work I could find was this...

[ Show youtube player ]

...which isn't Quake.

I may have got the wrong idea since they seem to refer to Falcon always as Falcon 030, even when a accelerator card is present...

[Cyprian]

Originally Posted by Bruce Abbott

So unless you need an exact 24 bit color the CPU (or Copper) only needs to do one write per color register.

true, but why would anyone limit the A1200's color range to 12bit?

Originally Posted by hammer

DSP56 @ 32 Mhz delivers a similar MIPS magnitude as 68030 @ 50 Mhz. PC Doom requires 4 MB RAM.

DSP56k (and actually DSP3210) is a regular processor with some fancy features that easily outperform 68030 (and 68040) like:
- three instructions done in 2 cycles vs 68030 1 instruction per 3 cycles;
- 2 cycle multiplication (48bit math with 8bit saturation) vs 44 cycles for 68030 (no saturation) vs 20 cycles for 68040;
- one DSP instructions has its equivalent in three or more 68030/040 instructions.

In case DSP (56k or 3210) 32MHz has 48mips vs 68030 50Mhz with 18MIPS.

[Cyprian]

Originally Posted by hammer

https://ftp.fau.de/aminet/docs/misc/dsp3210info.txt
DSP3210 has 8KB on-chip memory

DSP3210's 8KB is closer to 68040's 4KB+4KB L1 cache.

I would say no, because DSP3210 can assign a whole 8kB RAM for instructions vs 4kB instruction cache in 68040

Originally Posted by hammer

Good luck building 68LC060-based Amiga 1200 for 3DO's $699 target.

true,

Now I think that installing ARM instead 68k would be better option. ARM with its lower price and much much better performance would be killer.

[Promilus]

@Cyprian - Amiga chipset itself was centered around 68k buses and signals. You couldn't possibly expect to put ARM instead for AGA and make it work. And what about binaries already made for 68000? In 92 ARM sure had decent performance but not enough to compensate for lack of compatibility. And the first chipset which was actually trying to move on and separate chipset from (68k) CPU was AAA (but still relied on 68k buses)
ARM based CD32 with AGA capabilities and half a decent chipram speed? That's basically pre-3DO...

[van_dammesque]

RE: CPU prices, would this have been nice?

- Hypothetical 'A200' with an 020 and ECS in an A600 form factor, a A500 successor.

- A1400 (i.e. A500 form factor) 040ec with AGA and HD floppy, offer 3.5 HD bundles as well.




Very nice lineup until 'hombre' arrived, allows the like of wolf3d and doom to run and keeps software companies invested too.

[Bruce Abbott]

Originally Posted by Cyprian

true, but why would anyone limit the A1200's color range to 12bit?

Because it's faster and uses less memory.

But hey, perhaps Commodore should have copied the way that VGA chips do it. First you write to the (8 bit) DAC mask register to select which color registers to work with, then you write the color register number to the (8 bit) DAC write register, and finally write 3 more consecutive bytes to send the 18 bit color value. These are all I/O writes so they run at ISA bus speed with one wait state.

And VGA doesn't have a Copper to do this automatically at precise points in the display, so you have to do it with the CPU by polling the vsync and hsync pulses and counting cycles.

DSP56k (and actually DSP3210) is a regular processor with some fancy features that easily outperform 68030
In case DSP (56k or 3210) 32MHz has 48mips vs 68030 50Mhz with 18MIPS.

That's like saying a tractor is a regular motor vehicle with some fancy features that easily outperform a saloon car.

DSP56k only had 2 ALU input registers and 2 accumulators, a 16 bit program counter and a 16 word hardware stack, not much good for general computing purposes. Which is why Motorola created the 56800 series with embedded CPU, and the 68456 which was a 68000 CPU combined with a DSP56k.

It sure would have been nice if Commodore had added one to the A1200 at no extra cost. But that wasn't going to happen. Dave Haynie put a DSP chip in the A3000+, which was a high-end machine that would cost several thousand dollars. Without also being put in low-end models it wouldn't get much support.

But even if they did put one in the A1200 it would take years for developers to get the best out of it, like it did on the Falcon. Meanwhile faster 68k CPUs were coming out that didn't require any fancy programming to get equivalent (or better) results, as well as speeding up other applications.

When the Doom source code was released on December 23, 1997 it only took 3 days to port it to the Amiga. By this time many A1200 owners already had a 50MHz 030 or better. In 1997 a Blizzard 1230-IV with 4MB of 60ns FastRAM only cost £149.95 (chicken feed compared to the price of a new PC).

[hammer]

Originally Posted by Bruce Abbott

It sure would have been nice if Commodore had added one to the A1200 at no extra cost. But that wasn't going to happen. Dave Haynie put a DSP chip in the A3000+, which was a high-end machine that would cost several thousand dollars. Without also being put in low-end models it wouldn't get much support.

DSP3210's price range is $20 to $30.

AA3000's major ASIC components are nearly 1 to 1 maps with A1200's.

Lisa = Lisa. Lisa replaced ECS Denise.

Alice = Alice. Alice replaced ECS Agnus.

Paula = Paula.

Fat Gray = Gayle. Gayle has extra PCMCIA and IDE. Fat Gray replaces A500's Gray chip.

Buster = Budgie. Budgie has buffered 16-bit connections with Gayle's PCMCIA.

Ramsey = Budgie. 32-bit memory controller for the CPU's 32-bit Chip RAM or 32-bit Fast RAM.

About 7.1 MB/s max access for CPU's 32-bit Chip RAM is common for A3000's Super Buster/Ramsey, A1200's Budgie, and CD32's Akiko.

Amber with VRAM frame buffer is less needed for AGA. Commodore released a dual 31/15 kHz frequency 1942 monitor for the job.

68882@ 25 Mhz under 1 MFLOPS is too slow for games. Motorola charges 68030 MMU a premium price. 68030-25's wholesale price tracks 386DX-25's price until AMD's releases 386DX-40.

A1200 could have been released without the A300/A600's 1991 Gayle and Budgie delay factors.

Originally Posted by Bruce Abbott

When the Doom source code was released on December 23, 1997 it only took 3 days to port it to the Amiga. By this time many A1200 owners already had a 50MHz 030 or better. In 1997 a Blizzard 1230-IV with 4MB of 60ns FastRAM only cost £149.95 (chicken feed compared to the price of a new PC

Your "many A1200" is below 650,000 units. Blizzard 1230-IV is a 386DX-40 class solution that is obsolete in the 1997 time frame.

Like many others, my Pentium 166 (overclocked 150 Mhz with 60 to 66 Mhz jumper) PC had its NVIDIA RIVA 128 video card upgrade and ditched the S3 Trio 64UV+ 2MB. In my university, we have fleets of workstations Pentium 133/166/180/200 PCs with Premedia 2 and RIVA 128 OpenGL video cards, and the University's FTP server has WinQuake (Q1 1997), GLQuake(Q1 1997), Quake II (1997), Quake III (1999), and Unreal (1999) files and they run on dual booted Windows NT 4.0/Linux. My university's mass RIVA 128 deployment influenced my selection for RIVA 128. My university has Apple's PowerPC-based Performa and DEC's Alpha (Windows NT 4.0/Linux dual boot) workstation fleets. Netscape 4.x was running on them. The university had multiple Broadband ISDN microwave towers and Novell NetWare unified them all.

[dreadnought]

Originally Posted by Bruce Abbott

When the Doom source code was released on December 23, 1997 it only took 3 days to port it to the Amiga. By this time many A1200 owners already had a 50MHz 030 or better. In 1997 a Blizzard 1230-IV with 4MB of 60ns FastRAM only cost £149.95 (chicken feed compared to the price of a new PC).

It really is something special that this thread has managed to drag on for so long with this kind of supporting arguments.

Hammer says the that the number your many beefed up A1200 is below 650k, but I think even that is pretty generous. I suppose people who invested in higher end accelerators would be in tens of thousands.

But even if it was half a mil or so it doesn't change two key facts, namely that that "chicken feed" priced card is only an upgrade - you'd have to invest heavily first to beef your A1200 up, so you can't really compare a single card to a whole new PC.

Secondly, and more importantly, in 1997 we were waiting for Quake II to drop. Doom was still great of course, but it's kinda like telling people in 2020 they can forget about Cyberpunk 2077 and play 2016 Doom remake instead.

[hammer]

Originally Posted by Cyprian

true, but why would anyone limit the A1200's color range to 12bit?

24.7 million color palettes look good on marketing which is useless for sustained 320x200p/256p gaming resolutions.

Originally Posted by Cyprian

DSP56k (and actually DSP3210) is a regular processor with some fancy features that easily outperform 68030 (and 68040) like:
- three instructions done in 2 cycles vs 68030 1 instruction per 3 cycles;
- 2 cycle multiplication (48bit math with 8bit saturation) vs 44 cycles for 68030 (no saturation) vs 20 cycles for 68040;
- one DSP instructions has its equivalent in three or more 68030/040 instructions.

In case DSP (56k or 3210) 32MHz has 48mips vs 68030 50Mhz with 18MIPS.

Officially, DSP56000 @ 32 Mhz has 16.5 MIPS, and its Doom results show near 68030 @ 50 Mhz results.

Commodore's DSP3210 runs at 50Mhz which is linked to 68030's 25Mhz front side bus with Ramsey 32-bit memory controller. DSP3210 is an additional client on the Ramsey bus like SDAC (DMA interface for SCSI controller chip).

Ramsey's functions were integrated with A1200's Budgie. Commodore's engineers and Lew Eggebrecht's intent was for all Amigas models to receive DSP3210. Under Lew Eggebrecht's administration, Commodore was moving towards SuperFX2 tactics. Commodore runs out of time and runs out of money.

A1200's expansion bus with Budgie (Buster+Ramsey+directive from IDE/PCMCIA changes) had AA3000+'s original intention baked in already i.e. one step away from Commodore's SuperFX2 tactics.

Commodore's 1992 year was financially damaging. Keeping the A500 for 1992 would be a better tactic than the sales flop A600 toy.

[TCD]

Originally Posted by dreadnought

Secondly, and more importantly, in 1997 we were waiting for Quake II to drop.

Plus Quake II was released before the Doom source code. It's fair to say that by the end of 1997 the world had moved on from being 'Doom crazy'.

Edit: By the end of 1997 you could get a PlayStation with Doom for about the price of that Blizzard card.

[hammer]

Originally Posted by Cyprian

I would say no, because DSP3210 can assign a whole 8kB RAM for instructions vs 4kB instruction cache in 68040

Data needs instructions. DSP3210's 8KB has extra flexibility.

Originally Posted by Cyprian

Now I think that installing ARM instead 68k would be better option. ARM with its lower price and much much better performance would be killer.

3DO had 12.5 Mhz ARM60 with later models has 20 Mhz ARM60. 3DO's custom 32-bit chipset runs at 25 Mhz. 3DO's custom chipset is designed to saturate 3DO's memory bus. 3DO has discrete memory pools i.e. 2MB system RAM and 1 MB VRAM.

ARM60 (ARMv3) is a pure integer CPU. DSP3210's FP32 power is for geometry processing which offers an alternative to strong fixed point INT32 compititors.

INT32, INT16, and INT8 can be used for texture and pixel processing.

DSP3210 is designed to link with 68020/68030 and i386 bus standards. ARM60 CPU would need Budgie redesign and glue logic interface with 68K based Amiga custom chips.

[Thorham]

Originally Posted by hammer

24.7 million color palettes look good on marketing which is useless for sustained 320x200p/256p gaming resolutions.

Not useless, because you have a better, more accurate selection of colors. 4096 is relatively bad.

[hammer]

Originally Posted by Bruce Abbott

Interesting price comparison. We can see why Commodore chose the 16MHz EC020 for the A1200. We also see why the A4000 was so expensive.

The EC040 was a bargain! 4 times the compute power of a 25MHz 030 for almost the same price. But... no MMU or FPU. Good for games, not good for raytracing or program development. I wonder why no Amiga accelerator card used it?

From a materials standpoint, the A3000 and A4000 are expensive beyond the CPU.

My 386DX clone PC's mini-tower case is cheap soft metal compared to A3000's hard external case and hard internal metal platform. The mini-tower case is not designed to mount a heavy CRT monitor on it.

My university has extra metal cages on Pentium PC clones to give it extra strength. A3000 or IBM PS/2 wouldn't need the extra metal cage. University's Apple Macintosh Performa PowerPC desktop fleet didn't need the extra metal cages.

My A3000 can take some abuse while I transport it from home and high school. A3000 is my large cardboard presentation.

PS/2 Model 55SX and A3000's metal material quality are good, but they are not needed for a "gaming PC" use case.

For a LAN party gaming PC... I wouldn't bring my large "fish tank" glass PCs with heavy RTX 4080/4090s i.e. the cracked PCB problem is real. I have one gaming PC with a Lian Li O11 Dynamic EVO XL "fish tank" case. My Gigabyte RTX 4080 Gaming OC's heatsink is the same as the RTX 4090 version, hence making it heavy i.e. as heavy as my ASUS TUF 4090 OC.

I already destroyed the Fractal Define R6's glass side panel by accident and it wasn't a "fish tank" glass PC case e.g. Antec C8 Constellation or Lian Li O11 Dynamic EVO.

It depends on the use case and the intended environment.

--------------
030 bus has a pin (Cache Inhibit In Pin, CIIN) to tell the CPU for cacheless memory access mode and the data cache has a write-through scheme, hence MMU-less EC030 can still be used.

68040 MMU is used to mark the Chip RAM DMA address range as cacheless. 68LC040 was used for the Amiga platform.

68EC040 with a disabled data cache could be used.

Alice would need to be modified for cache coherence with data cache enabled 68K CPU e.g. Alice+ tells the CPU there are memory content changes and the CPU reloads relevant cache lines.

68LC040 + DSP3210 would be a potent combo for 3D games. For 68EC040, a method is needed to tell the CPU that the Chip RAM DMA memory range is not data-cacheable and then it would be "power without the price" potential if this issue is addressed.

[hammer]

Originally Posted by Thorham

Not useless, because you have a better, more accurate selection of colors. 4096 is relatively bad.

VGA has an 18-bit color palette compromise.

SNES has a 15-bit color palette compromise.

AGA's 16.7 million color palette is related to the AAA chipset.

[Bruce Abbott]

Originally Posted by TCD

Plus Quake II was released before the Doom source code. It's fair to say that by the end of 1997 the world had moved on from being 'Doom crazy'.

The 'world' moved on from the Amiga in 1994, if not before. But Amiga fans didn't care because they lived in their own world - the world of Amiga. In that world Doom was a new thing at the end of 1997. And on other platforms too. Releasing the source code spiked interest on many platforms. Today people are running Doom on everything from dishwashers to Lego bricks.

Edit: By the end of 1997 you could get a PlayStation with Doom for about the price of that Blizzard card.

So? The Playstation was good for only one thing - running commercial games that were made for it. An A1200 with 50MHz 030 could do far more.

BTW the PlayStation version of Doom is missing a lot of the interesting graphics in the original PC version, possibly because it was based on the Atari Jaguar port. Unlike on the Amiga there wasn't anything you could do about that.

[hammer]

Originally Posted by Bruce Abbott

The 'world' moved on from the Amiga in 1994, if not before. But Amiga fans didn't care because they lived in their own world - the world of Amiga. In that world Doom was a new thing at the end of 1997. And on other platforms too. Releasing the source code spiked interest on many platforms. Today people are running Doom on everything from dishwashers to Lego bricks.

Your isolated Amiga fans argument wasn't able to keep 5 to 6 million Amiga users within the Amiga's ecosystem, let alone attract new Amiga users.