3D - F030 vs Amiga 1200

[Flash951]

That is pity, the A3000 is the best Amiga ever made. It has SCSI, built in VGA scandoubler and stylish case. The A4000 is more like A1200 with a accelerator in ugly 286 like case. Still both machines are based around the same chipset as the A500+/A1200 as Commodore newer developed a new chipset for its highend machines.

That exactly Dave Haynie said!

please check youtube video that I put in my previous post!

A4000t also has SCSI built in, using the fastest SCSI chip available at that time, I think I got about 10-20Mb/sec on my A4000t with 2GB SCSI2 f/w.. Another thing that seems to "not count" is that the Amiga's are probably the most expandable computers that ever existed.

I needed RTG graphics nand real chunky mode for productivity speed and resolution, so I fitted a EGS 23 Spectrum graphic card in one of the seven Zorro III ports, 1600x1280 resolution. I wanted to play games on my VGA monitor, so I fitted a flicker fixer in one of the video slots (flicker fixer using the same chip as on the mainboard flicker fixer in the A3000. I wanted video and 16 bit sound, so I fitted a board for that in the other video slot, the output was true the normal Amiga outputs for sound and video. I used to play music on the desktop while working, using .mp2 music files with 16 bit output true the Amiga sound output, as the card decoded layer II without CPU time.

I will say, for me, the expand-ability of the Amiga made the big difference.

[ctirad]

He measured the CHIP RAM speed of the Economy model 030 version of the A4000, this RAM is not used by the CPU, only fast RAM is used by CPU.

The ChipRAM is of course used by the CPU, becuase all the graphics data, samples, I/O registes and ports are there.

If you want compared FastRAM speed, comapre FastRAM on a both machines.

Another example: "VGA output from the Amiga is useless"
Why is that useless? I used VGA output all the time for productivity.

Because it is very limited and uncompatible. Just the fact it has to be switched by software is problematic, because you have to have properly configured system fort that and still the whole boot time you are stucked in PAL/NTSC and you don't have access to the early start up menu. Also forget to start anything from the floppy in VGA. On the F030 the machine detects the type of the monitor after power on and immediately ouputs only the right signal.

The compatibility with the games/demos is virtually zero. On F030 (and TT) VGA you can play ST games or watch most of the demos without any problem.

The Amiga VGA resolutions and refres rates ends where the F030 starts due to the limited 28MHz pixel clock.

On my early A500 I upgraded the chip set to ECS (A500+) to be able to output VGA,

Yes, teh ECS is able to output VGA, but on the A500+ the redraw speed is soooo slow it cannot be used for virtaully anything. I did many experiments with my A500+ and later A1200 back then.

[calimero]

4 years, because of a key engineer left Atari.

so Roy Good come as replacement? Who was TT engineer that left Atari?
I have impression that Shiraz Shivji was not involved in TT design.

[ctirad]

I needed RTG graphics nand real chunky mode for productivity speed and resolution, so I fitted a EGS 23 Spectrum graphic card in one of the seven Zorro III ports, 1600x1280 resolution. I wanted to play games on my VGA monitor, so I fitted a flicker fixer in one of the video slots (flicker fixer using the same chip as on the mainboard flicker fixer in the A3000. I wanted video and 16 bit sound, so I fitted a board for that in the other video slot, the output was true the normal Amiga outputs for sound and video. I used to play music on the desktop while working, using .mp2 music files with 16 bit output true the Amiga sound output, as the card decoded layer II without CPU time.

It looks like you spent a lot of money to rebuild your highend Amiga to mid end PC

[ctirad]

than Amiga coders really kick ass!
This is 14MHz 020 with FastRAM: https://www.youtube.com/watch?v=8GmzmOQ6VWA
(there is even motion blur! )

Yes, the Amiga demos are nice mainly due to the design that Atari demos usualy lacks.
This particular demo shows what one would expect from the 020 + FastRAM. The 3D is 2x2 or maybe even worse. The FastRAM is very helpful not only because it is fast, but also all the C2P stuff can be made here much faster and then just copied to the ChipRAM.

[Flash951]

The ChipRAM is of course used by the CPU, becuase all the graphics data, samples, I/O registes and ports are there.

You are completely wrong. All I/O, sound and graphic data has direct access to CHIP RAM on a completely separate memory bus in the Amiga, even in the first Amiga model, there is 24 DMA channels. That's why the Amiga was powerful compared to other platforms based on 68k cpu, because many things doesn't affect CPU in the Amiga. In the Amiga, the CPU has completely separate bus to Fast RAM, and the co-processors has completely separate bus to chip ram.

[Cyprian]

even in the first Amiga model, there is 24 DMA channels.

to be honest, number of channels means nothing.
there is no magic here 4 channels for blitter, 4 for sound, 8 for (16pixels 4 color)sprites, 6 for video, one for copper, one for floppy disk.
They don't improve a computing power in any way.

That's why the Amiga was powerful compared to other platforms based on 68k cpu, because many things doesn't affect CPU in the Amiga.

not exactly true,
In terms of blitting (moving data from one memory area to the other) you right, amiga is more powerful than ST which has no any blitter chip.
But in terms of real power - in a computing terms, amiga is less power than Atari because of lower CPU clock. And also can be easily more slowed down by usage of DMA channels.

In the Amiga, the CPU has completely separate bus to Fast RAM, and the co-processors has completely separate bus to chip ram.

nothing special, exactly the same story is in case of ST: CPU has separate bus to ALT-RAM, and TT/Falcon - CPU has has separate bus to Fast-RAM/TT-Ram

[alexh]

You are completely wrong. All I/O, sound and graphic data has direct access to CHIP RAM on a completely separate memory bus in the Amiga

Rubbish. CHIP RAM is CHIP RAM. It is a single resource which is shared with everything including the CPU. If the CPU needs to read/write custom chip registers it uses the CHIP RAM bus. If the CPU needs to set-up any data the other chips are accessing (i.e. render a 3D screen) it uses CHIP RAM!

[ctirad]

You are completely wrong. All I/O, sound and graphic data has direct access to CHIP RAM on a completely separate memory bus in the Amiga, even in the first Amiga model, there is 24 DMA channels.

I see. You probably have a very fuzzy idea what the DMA really is and why it has nothing to do with a separate BUSes at all. There is really only one data BUS between all chips and ChipRAM (see schematics if you don't believe me). When the CPU or DMA access the BUS, all other chips have to wait.

The only exception is FastRAM, that can be accessed by CPU only (!) in the same time in another chip access the ChipRAM. This significantly speeds the CPU operations and that is why it is called FastRAM, after all. The drawback is no other chip besides CPU can access the FastRAM area. No videoram, no samples, no DMA in FastRAM are possible.

That's why the Amiga was powerful compared to other platforms based on 68k cpu, because many things doesn't affect CPU in the Amiga.

In the terms of raw CPU power the A500 is actually the slowest from the Amiga/ST/MAC just because it has less MHz. It shines in the tasks where the blitter and sprites can be utilised. Shortly, in 2D games and demos.

[shoggoth]

Flash951: Many of your arguments here are in my opinion based on myths and lack of knowledge about the Amiga.

[Frank B]

The ChipRAM is of course used by the CPU, becuase all the graphics data, samples, I/O registes and ports are there.

You are completely wrong. All I/O, sound and graphic data has direct access to CHIP RAM on a completely separate memory bus in the Amiga, even in the first Amiga model, there is 24 DMA channels. That's why the Amiga was powerful compared to other platforms based on 68k cpu, because many things doesn't affect CPU in the Amiga. In the Amiga, the CPU has completely separate bus to Fast RAM, and the co-processors has completely separate bus to chip ram.

Did you know the Amiga ins't fully compatible with the 68k? If you execute the TAS instruction from chip memory the chipset dies and needs power cycled.

[AnthonyJ]

Did you know the Amiga ins't fully compatible with the 68k? If you execute the TAS instruction from chip memory the chipset dies and needs power cycled.

I came into this thread hoping to find lots of links to the "best" 3D engines on both platforms, and some interesting technical discussions of the merits of both machines. I expected Doug's work on Bad Mood + Quake 2 to be the declared "the winner", but was hoping to find links to other similar projects on the Amiga which would be an interesting read.

Instead I seem to have gone back into the 80s/90s where people are fighting the platform wars, and just looking to bash the other platform while defending their own platform to the death...

[calimero]

^
Flash951 is in wrong thread and in wrong centuries

take a look at eab thread: http://eab.abime.net/showthread.php?t=69697 there is also lot of noise but at least you will find links to very impressive Amiga 3D demos!

[DarkLord]

Instead I seem to have gone back into the 80s/90s where people are fighting the platform wars, and just looking to bash the other platform while defending their own platform to the death...

It was a good discussion (and will be again, I'm hoping) until basically -1- Amiga user came
in and well...you know. (to be fair, he claimed that someone from this thread went to an
Amiga area somewhere and did the same thing).

So lets hope that's over with and the subject topic can be resumed.

[calimero]

^

I open both thread (here and on eab). It is really hard to talk openminded about Amiga and Atari, only few people manage this (I am not among them :/) but I also want to learn technical things.

On eab I ask them to point me to best A1200+FastRAM 3D demos since I have trouble to find these kind of demos since almost all demos on A1200 require accelerator.

[DarkLord]

I understand perfectly.

It's like when I, as a Pittsburgh Steelers fan, try to talk to a Dallas Cowboys fan. Or Coke vs Pepsi.
Or D&D 4e vs Pathfinder. Or the conversations between UK Wildcat fans and Duke U. fans. Or
how the British didn't really sink the Bismark. They destroyed her fighting capability, but the
Germans scuttled her themselves.

Wow, have I raised enough "hot button" topics yet or what?

[ctirad]

On eab I ask them to point me to best A1200+FastRAM 3D demos since I have trouble to find these kind of demos since almost all demos on A1200 require accelerator.

That is very true and it is just another proof that stock A1200 is not taken seriously even by the Amiga users. Most Amiga user says "A1200" but in fact they mean a 68060 machine with a plenty of FastRAM.

[ctirad]

It's like when I, as a Pittsburgh Steelers fan, try to talk to a Dallas Cowboys fan.

The discussion was not about personal preferences but about technical facts.

[DarkLord]

The discussion was not about personal preferences but about technical facts.

Exactly. So when I mention the stats of the 1975 Pittsburgh Steelers vs the stats of the
1975 Dallas Cowboys with a Cowboys fan, regardless of the "hard" technical facts, it
almost always devolves into a semi-spiritual, quasi-religious discussion about the merits
of each.

Kinda like Atari vs Amiga "discussions"...

PS but enough of this - lets let the thread get back on topic!

[mc6809e]

I see. You probably have a very fuzzy idea what the DMA really is and why it has nothing to do with a separate BUSes at all. There is really only one data BUS between all chips and ChipRAM (see schematics if you don't believe me). When the CPU or DMA access the BUS, all other chips have to wait.

Actually THERE IS a separate bus called the register bus (I'm surprised you didn't see this while scrutinizing the schematics). It's used to move data into chip registers after a DMA cycle so that the CPU or another DMA channel can access chipram immediately as quickly as once every two CPU cycles. Without the register bus, DMA would proceed as it does it does using the blitter on the STe where four cpu cycles are required for just one access. The register bus effectively doubles the speed at which DMA can occur. A copy with the Amiga blitter can run as quickly as four CPU cycles per word.

In the terms of raw CPU power the A500 is actually the slowest from the Amiga/ST/Mac just because it has less MHz. It shines in the tasks where the blitter and sprites can be utilised. Shortly, in 2D games and demos.

The MAC is actually the slowest. It doesn't do interleaved memory access. When video is displayed the CPU and video take turns every four CPU cycles accessing memory. This causes most CPU memory accesses to take eight cycles. The only time the CPU in the MAC runs at full speed out of RAM is during the blanking intervals. The effective CPU speed is about 6MHz on typical code.

I say typical because all these CPU speed comparisons between the machines depend critically on the code that's actually run and when that code executes relative to other DMA.

Consider the code:

Code: Select all

     move.w 1000, d1
next:
     move.b (a0)+, d0 
     dbeq d1, next  

On the ST, this code takes always takes 8 + 12 = 20 cycles per iteration since every memory access must begin on a four CPU cycle boundary.

On the Amiga, however, the dbeq instruction can take 10 cycles, if no other DMA is occurring, such as during horizontal or vertical blanking, given a total of 18 cycles per iteration. The CPU on the Amiga may begin a chipram access on any two cycle boundary (though the access itself takes four cycles total).

This isn't enough to completely close the MHz gap, but it does reduce the spread.

On the Mac, the worst situation occurs during video display. The code will take 32 cycles per iteration as the video block CPU access.

[CiH]

Back on topic with my extremely limited and unscientific scale of measurement.

Standard Falcon including cunning DSP code = A1200 with a 50 MHz 68030 booster.

Centurbo 2 with a 50 MHz 030 and 50 MHz clocked DSP, could make 040 Miggy boards look nervously over their shoulders? (I wouldn't mind some other thoughts and views on this second one actually.)

68060 on both machines is closer, but the Falcon edges the A1200 versions as I think someone has already said.

I apologise if this is going over oft-trodden ground, but I was getting mildly bored with the "It was used for the animations in Babylon 5" type comments.

[calimero]

Actually THERE IS a separate bus called the register bus (I'm surprised you didn't see this while scrutinizing the schematics). It's used to move data into chip registers after a DMA cycle so that the CPU or another DMA channel can access chipram immediately as quickly as once every two CPU cycles. ...

heh... I just found my old high school paper that I wrote for Computer Class. I wrote about Amiga (unfortunately it is in Serbian but there is nice Amiga 500 diagram )

EDIT: I also attached AMIGA.PDF diagram.

[Frank B]

Did you know the Amiga ins't fully compatible with the 68k? If you execute the TAS instruction from chip memory the chipset dies and needs power cycled.

I came into this thread hoping to find lots of links to the "best" 3D engines on both platforms, and some interesting technical discussions of the merits of both machines. I expected Doug's work on Bad Mood + Quake 2 to be the declared "the winner", but was hoping to find links to other similar projects on the Amiga which would be an interesting read.

Instead I seem to have gone back into the 80s/90s where people are fighting the platform wars, and just looking to bash the other platform while defending their own platform to the death...

Not really. I'm a huge Amiga fan too! I'm just amused by rabid amigans who don't know much about their machine

[calimero]

I see. You probably have a very fuzzy idea what the DMA really is and why it has nothing to do with a separate BUSes at all. There is really only one data BUS between all chips and ChipRAM (see schematics if you don't believe me). When the CPU or DMA access the BUS, all other chips have to wait.

Actually THERE IS a separate bus called the register bus (I'm surprised you didn't see this while scrutinizing the schematics). It's used to move data into chip registers

maybe I am wrong but RGA is address bus?
and there is only one data bus to chip RAM (as ctirad stated)?

[galahad]

The ChipRAM is of course used by the CPU, becuase all the graphics data, samples, I/O registes and ports are there.

You are completely wrong. All I/O, sound and graphic data has direct access to CHIP RAM on a completely separate memory bus in the Amiga, even in the first Amiga model, there is 24 DMA channels. That's why the Amiga was powerful compared to other platforms based on 68k cpu, because many things doesn't affect CPU in the Amiga. In the Amiga, the CPU has completely separate bus to Fast RAM, and the co-processors has completely separate bus to chip ram.

Did you know the Amiga ins't fully compatible with the 68k? If you execute the TAS instruction from chip memory the chipset dies and needs power cycled.

Thats not entirely accurate. TAS is a special condition, it has three phases, read/modify/write, and most instructions are expected to only do read/write. Because of this extra condition, its 'possible' that the Agnus chip might also be trying to access chip memory during the same DMA cycle as the TAS instruction is trying to write, and invariably what happens is it causes TAS to simply not write its results properly.

Of course using TAS in registers is fine, and use of BSET instead of TAS overcomes the problem of TAS being unreliable.

bus lockouts are rare, but then most people simply didn't use it, so it was never really an issue.