Was anyone else disappointed with the A1200?

[pandy71]

Originally Posted by Bruce Abbott

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.

Most VGA cards doesn't have capability to pool and distinguish between V and H sync so usually you simply perform series write operations into RAMDAC adress.

Originally Posted by Bruce Abbott

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.

Simple comparison - please provide dot product time for 56k clocked with 32MHz and 68030 @ 50MHz - this is ultimate answer which one is faster in 3D arithmetic.

[Thomas Richter]

Originally Posted by pandy71

Most VGA cards doesn't have capability to pool and distinguish between C and H sync so usually you simply perform series write operations into RAMDAC adress.

All VGA cards have the possibility to detect a VSync, and have a interrupt request output. Unfortunately, that interrupt request was sometimes not connected such that cards could not trigger an interrupt. You could read the status bit, though. All VGA cards also have the possibility to detect whether the display is in in the active area. That, of course, does not mean that it was convenient to do so - just the reverse. It wastes a lot of CPU time. What graphics cards offered instead later on was a much more flexible mechanism, namely true color graphics (so you are not tight to the raster timing in the choice of colors) and overlays, which allows you to replace one rectangular region of the display with data from another region, often also in another graphics format such as YUV or high-color.

Originally Posted by pandy71

Simple comparison - please provide dot product time for 56k clocked with 32MHz and 68030 @ 50MHz - this is ultimate answer which one is faster in 3D arithmetic.

Not necessarily. A DSP may be faster to compute it, but you also need to get the data into it and the results out of it, which may cost even more time than the actual computation.

[pandy71]

Originally Posted by Thomas Richter

All VGA cards have the possibility to detect a VSync, and have a interrupt request output. Unfortunately, that interrupt request was sometimes not connected such that cards could not trigger an interrupt. You could read the status bit, though. All VGA cards also have the possibility to detect whether the display is in in the active area. That, of course, does not mean that it was convenient to do so - just the reverse. It wastes a lot of CPU time. What graphics cards offered instead later on was a much more flexible mechanism, namely true color graphics (so you are not tight to the raster timing in the choice of colors) and overlays, which allows you to replace one rectangular region of the display with data from another region, often also in another graphics format such as YUV or high-color.

Once again - not many cards offers such things as V and H sync pulses detection - most of them not. As such it was uncommon thing to use this - instead developers rely on internal HW design to provide support for access to graphics buffer content.
You are right about theoretical feature and you are right also on this that it was not used commonly.

Originally Posted by Thomas Richter

Not necessarily. A DSP may be faster to compute it, but you also need to get the data into it and the results out of it, which may cost even more time than the actual computation.

Usually you not calculating single dot product but set of them and this is same story as behind HW graphic acceleration - setting up cost time but when it is set then in overall you have significant speedup.
So using DSP for 3D calculations was not uncommon thing in past, in fact it was one first primary task for DSP's, later specialized 3D solutions was introduced and DSP's was usually no longer used to accelerate 3D operations but as 3D is special case of DSP then DSP's usually outperform general CPU's on this task.
Btw to address my question - dot product on 56K take 12 cycles so 32MHz DSP like 56k was capable to perform over 2.5 million dot products per second (over 50000 per frame - assumption 50fps).
This is way frequently DSP's was present on many arcade machines boards - they act as programmable math processor.
If you design your architecture wisely (Falcon is example of poor system architecture) then overall benefits from having DSP on board can be substantial (but even in severely crippled Falcon design DSP provide substantial processing power).

[sandruzzo]

What do you need to do to go faster on 2d/3d games? Parallel operations. It doesn't matter Mips, peak rate of this or that, if you can't use them at 100% Look at the Theoretical TFLOP of RDNA3 61.4, near 3 time RDNA2 peak rate! Did they get 3 times the performance of the latter? No! Because in the real world application, peak rate is not the holy grail of performance!

[oscar_ates]

Originally Posted by AestheticDebris

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.

Then the Amiga 1200 got a very weak CPU, because commodore could get them for 2-3 dollars. Users needed to install an accelerator card that costed many hundred dollars, few did that and professional companies left the Amiga market because of user base numbers.

[Amigajay]

Majority of publishers don’t give a monkeys how powerful something is. Publishers didn’t jump ship because the Amiga 1200 was underpowered, they were slowly leaving because of piracy, the issues Commodore were having, and then Commodore went bankrupt, they saw no future in that market then.

[Cyprian]

Originally Posted by pandy71

Btw to address my question - dot product on 56K take 12 cycles so 32MHz DSP like 56k was capable to perform over 2.5 million dot products per second (over 50000 per frame - assumption 50fps).

how did you calculate these 12 cycles per dot?

Originally Posted by pandy71

If you design your architecture wisely (Falcon is example of poor system architecture) then overall benefits from having DSP on board can be substantial (but even in severely crippled Falcon design DSP provide substantial processing power).

Well there were thee different Falcons - released home edition Falcon 030; ready to release Proedition Falcon Microbox and under developed Proedition Falcon Painter (with Jaguar chipsets)

I guess you do mean home edition Falcon 030, then yes it was quite poor compared to PC. Because PC were modular with higher bus bandwidth than any Amiga/Atari.
But when you compare it to A1200 or A4000 then Falcon looks very good:

Architecture:
- Amiga: Address range: 24bit; Data bus: 32bit
- Atari: Address range: 24bit; Data bus: 32bit

32bit Chipsets:
- Amiga: the only Lisa
- Atari: the only Videl

CPU:
- Amiga: 14MHz 2.4 MiPS
- Atari: 16MHz 3.5 MIPS; DSP 32MHz 16~48 MIPS

Max Video resolution:
- Amiga: 1280x512 interlace 256 color
- Atari: any resolution within 32Mhz pixel clock (fully programmable video chip), e.g. 1280x512 interlace 256 color or 640x480 16bit color or 800x600 256 colors

Video colors:
- Amiga: bitplane 256 color register
- Atari: chunky 16bit direct color

Audio:
- Amiga: 8bit 4channels
- Atari: 16bit 8channels plus unlimited with DSP

Max Video RAM:
- Amiga:2MB
- Atari: 14MB

RAM Bus Bandwidth for Video:
- Amiga: 28MB/s
- Atari: 32MB/s

RAM Bus Bandwidth for CPU 32bit access:
- Amiga: Read: 4.5MB/s Write: 6.9MB/s (2 chipset cycles 3.5MHz per 32bit) (BusSpeedTest 0.19)
- Atari: Read: 5.4MB/s Write: 6.5MB/s (2 chipset cycles 8MHz per 16bit)

RAM Bus Bandwidth for CPU 16bit access:
- Amiga: Read: 2.2MB/s Write: 3.5MB/s (2 chipset cycles 3.5MHz per 32bit)
- Atari: Read: 5.4MB/s Write: 6.5MB/s (2 chipset cycles 8MHz per 16bit)

[sokolovic]

Originally Posted by Amigajay

Majority of publishers don’t give a monkeys how powerful something is. Publishers didn’t jump ship because the Amiga 1200 was underpowered, they were slowly leaving because of piracy, the issues Commodore were having, and then Commodore went bankrupt, they saw no future in that market then.

That sum it all.
Publishers had no problem to develop for the GameBoy Color released in 1998.

[TCD]

Originally Posted by Bruce Abbott

So?

Isn't your whole 'shtick' that the Amiga was so much cheaper than a PC? I'm sure every Amiga user that was still using the platform in late 1997 was very happy to get a Doom port. That isn't exactly the same thing as getting it to work on a pregnancy test though [ Show youtube player ]

Edit: PS Doom got a cool new soundtrack, colored lighting and sound effects though. Really not too shabby especially if you consider the price

Edit 2: I should mention that I owned Doom on the Jaguar

[d4rk3lf]

Originally Posted by TCD

I'm sure every Amiga user that was still using the platform in late 1997 was very happy to get a Doom port.

I was using my A1200 in the late 1997, and believe it or not, I couldn't care less about Doom.

[TCD]

Originally Posted by d4rk3lf

I was using my A1200 in the late 1997, and believe it or not, I couldn't care less about Doom.

Now look what you have done! Now Bruce has to find something else to... yeah, I don't think that will be a problem for him

[pandy71]

Originally Posted by Cyprian

how did you calculate these 12 cycles per dot?

This figure is provided by Motorola "DSP56000/DSP56001 USER’S MANUAL, APPENDIX B, BENCHMARK PROGRAMS"

Originally Posted by Cyprian

Well there were thee different Falcons - released home edition Falcon 030; ready to release Proedition Falcon Microbox and under developed Proedition Falcon Painter (with Jaguar chipsets)

I guess you do mean home edition Falcon 030, then yes it was quite poor compared to PC. Because PC were modular with higher bus bandwidth than any Amiga/Atari.
But when you compare it to A1200 or A4000 then Falcon looks very good:

You get me totally wrong - i've criticized design limitations done for Falcon not compared Falcon to Amiga - also you was unfair doing your comparison - Amiga can and frequently use HAM modes to increase number of colors (HAM is crude delta RGB compression offering 3:1 compression) not as you tried suggest having only 256 colors - HAM8 is largely fine due bigger CLUT (64 vs 16 colors in HAM6).

Contrary to your comment - PC bus has many problems so it is less efficient than Amiga or Atari design (illusory 8.33MBps shared between slow and fast devices and also video - real life ISA bus performance rarely exceed 6MBps in total) - but even compared this to Zorro II (slower, 5.3MBps) it is used in Amiga better than ISA. Zorro III offered way higher bandwidth (so any Zorro III Amiga easily outperformed in theory not only theoretical performance ISA but also EISA and even VLB/PCI - theoretical Zorro III performance is 150MBps).
TBH - i always have difficulty to understand why either CBM or Atari didn't introduced something like LPC version of inter/between IC communication to reduce IC pin count (reduce cost), simplify PCB layout (reduce cost) and improve bandwidth - but now is easy to say what and where was wrong and/or too late.

[Promilus]

@pandy - ZIII never reached quarter of that so ... talking about theoretical ISA throughput then showing how it was actually shitty solution due to the fact real bandwidth was quarter of that... 3rd at best. Then comes Zorro III which not only destroys EISA (being actually THE SAME KIND OF DESIGN) but reaches VLB/PCI with... theoretical bandwidth. Shame actual bandwidth is much, much, much smaller and most likely even slowest PCI in PC world (so shady 486 chipsets) can outperform it. ISA (and by extension EISA) were x86 processor buses adapted to work as expansion slots. Zorro II (and by extension Zorro III) were 68k processor busses adapted to work as expansion slots. EISA was 32bit extension of ISA which was kind of backward compatible? Zorro III was 32b extension of Zorro II which was kind of backward compatible. So... yeah. Also how many native ZII/ZIII chips were produced... ever? Most of ZII/ZIII add-ons uses ISA/PCI chips with glue logic. Also... how long typical zorro card was? :P

[sokolovic]

Originally Posted by Cyprian

But when you compare it to A1200 or A4000 then Falcon looks very good:

The Falcon was much more expensive than the A1200. There is a reason why nobody bought one back then.

That's interesting btw how Commodore and Atari switched their policy between the ST/A1000 and the A1200/Falcon030.

[TEG]

Originally Posted by sokolovic

The Falcon was much more expensive than the A1200. There is a reason why nobody bought one back then.

And the distribution was very confidential. In Paris there was only 2 or 3 very specialized shops were you can get one. I think it never reach more well established distribution points in France.

[sokolovic]

Originally Posted by TEG

And the distribution was very confidential. In Paris there was only 2 or 3 very specialized shops were you can get one. I think it never reach more well established distribution points in France.

Yep. And France was probably one of the biggest Atari friendly market.

[hitm4n]

196 pages guys!? I see this thread every damn day and i don't even understand why the question was asked. I'm sure everyone here hated their 1200, on this Amiga forum...

I wasn't disappointed with my 1200.

Is there a way to block threads?

[sandruzzo]

repetita iuvant

[Thomas Richter]

Originally Posted by pandy71

Once again - not many cards offers such things as V and H sync pulses detection - most of them not. As such it was uncommon thing to use this - instead developers rely on internal HW design to provide support for access to graphics buffer content.

Input status 1 is a regular VGA register, and it includes the information you seek. This is a standard feature. What was unfortunately less standard is to wire the interrupt output of the VGA chip to an interrupt of the ISA slot.

[Cyprian]

Originally Posted by pandy71

Btw to address my question - dot product on 56K take 12 cycles so 32MHz DSP like 56k was capable to perform over 2.5 million dot products per second (over 50000 per frame - assumption 50fps).

Originally Posted by Cyprian

how did you calculate these 12 cycles per dot?

Originally Posted by pandy71

This figure is provided by Motorola "DSP56000/DSP56001 USER’S MANUAL, APPENDIX B, BENCHMARK PROGRAMS"
to say what and where was wrong and/or too late.

ach nice finding.
I know that PDF very well. It is very valuable document. It shows how to reach ~48MIPS (instead of 16MIPS) from 32MHz DSP (the same is with AT&T32010 it can boost its performance by 3 in the same manner)

But you know, those "dots" are not points/pixels on the screen but something more complex:



Here is algorithm:

Regarding dot/pixels DSP has the same as 68k dedicated instructions for that - BSET/BCLR (single cycle or two cycles each, don't remember now).

Originally Posted by pandy71

You get me totally wrong - i've criticized design limitations done for Falcon not compared Falcon to Amiga.

It wasn't "design limitations" but outdated design in both cases Falcon and A1200. I think both computers were late and should be released two years earlier

Originally Posted by pandy71

also you was unfair doing your comparison - Amiga can and frequently use HAM modes to increase number of colors (HAM is crude delta RGB compression offering 3:1 compression) not as you tried suggest having only 256 colors - HAM8 is largely fine due bigger CLUT (64 vs 16 colors in HAM6).

You can't use HAM in Workbench or inside your own GUI window.
Regarding games - I'm not really interested in this area, but I know only one game preview (shown here this year - really nice screen) with HAM but sprites there are static.

Originally Posted by pandy71

TBH - i always have difficulty to understand why either CBM or Atari didn't introduced something like LPC version of inter/between IC communication to reduce IC pin count (reduce cost), simplify PCB layout (reduce cost) and improve bandwidth - but now is easy to say what and where was wrong and/or too late.

good point, I'm wondering the same thing.

Atari had a new fast architecture - Jaguar. It had 108Mb/s bus bandwidth (compared to Falcon's 32MB/s Falcon or A1200's 28MB/s), and Falcon Painter (with 68040) had integrated Jaguar chipsets.
Amiga had Ranger chips if I'm not wrong.
Both companies had an interesting technology, but both were very slow to implement it.