@Gunnar

Quote:

The Commodore C64 on the other hand was sold over 10 million times. Who had more success? Which System was better?

C64 was released when the PC had CGA. For low-cost game machines, the C64 was KO'ed by the Nintendo Entertainment System (NES).

NES has about 61 million units sold and it was released in 1985 USA and 1986 EU.

C128 was released in 1985 and it was C64's "ECS" job i.e. reuse aging game hardware and attach "business" high resolution with low color display modes.

The pattern is similar when Commodore released the Amiga ECS against Sega Genesis and SNES.

In 1989, the Sega Genesis (Mega Drive) was released in the USA, the Amiga remained in OCS/half ECS. Sega Genesis sold 30 million units by the end of 1997.

Meanwhile, C65's 256-color display chipset was completed in December 1990 and canceled in 1991. This should have been the C128 to counter NES.

In 1991, SNES was released in the USA, Amiga remained in ECS. Commodore had C65's 256-color display chipset to counter SNES.

Meanwhile, AA3000+'s 256-color display AGA chipset was completed in Q1 1991. Commodore's IBM Jr management import was pushing "A1000 Jr" which is 68020 with ECS machine i.e. this is effectively A1200 with ECS. No Commodore sales team ordered "A1000 Jr". Commodore management wasted time with the ECS push. AGA should have been released with the rumored A1400(?) and AA3000+ in Q4 1991.

For Q4 1992, C= AGA and SNES were released in the EU. SNES had a headstart in building its install base since 1990 (JP) which is important for 3rd party software development-related investment decisions.

Mortal Kombat (1994) and Street Fighter 2 (1992) ports were concerning on the Amiga.
PC Master Race's Mortal Kombat (1994) port beats SNES version. AGA's install base was a concern for 3rd party developers. AGA target could have delivered a better Mortal Kombat 2 port.
A game (e.g. https://www.youtube.com/watch?v=LYZdho3DDX8, proper Steet Fighter 2 AGA demo) designed for A1200 will not run and the code base is unsharable with ECS/OCS Amigas. Amiga OCS/ECS large install base was the "Atari ST" against Amiga AGA.
The cited proper Steet Fighter 2 AGA demo delivered results similar to 16-bit SNES.

Commodore doesn't have 1st party game developers to assist well-known game title ports.
Microsoft's DirectX team ported Doom to Windows 95 as Doom95. Microsoft's developers ported Doom to Windows 3.1's Win32S+WinG as WinDoom. IBM''s OS/2 team is a joke.

32-bit PC during 1992 to 1994 era has 2.5D/3D texture map games that are untouchable by SNES. 32-bit PC with VGA hardware target has been building its install base since 1987. The "32-bit" gaming PC reached critical install base mass and delivered "32-bit" 2.5D/3D gaming experiences around 1992 to 1993.

From 1993, PC was building its Pentium install base against PlayStation's Q4 1995 western market release.

Software sells the hardware platform.

Last edited by Hammer on 08-Mar-2024 at 02:28 AM.
Last edited by Hammer on 07-Mar-2024 at 05:36 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Gunnar:
Quote:

Fact is that AGA did get a big memory controller upgrade. AGA Alive has 4 times the memory bandwidth compared to OCE/ECS Agnes. So what you say is not really true.

Hammer:
Quote:

Wrong. Refer to Amiga 3000's 32-bit Chip RAM with 7MB/s CPU access which is similar to AGA's 7MB/s CPU access on its 32-bit Chip RAM. The Alice is not efficient with the given 32-bit memory bandwidth.

I see where your misunderstanding is.

You are confused between CPU and Amiga chipset.
BUSTEST the benchmark that you does measure only the CPU memory transfer speed
but you spoke all the time about the DMA performance of the Amiga DMA chipset.


Let me explain you:

The Amiga chipset is based around DMA.
DMA means Direct Memory Access.
Direct Memory Access means the Amiga Chipset
can read and transfer by itself memory.
Sprites are put on Screen with DMA.
Audio / Music is played with DMA.
Screen is painted with DMA.
Colors on Screen can be changed and reloaded with DMA.
Floppy data is loaded with DMA.

This means the AMIGA can do all these jobs without using the CPU.
Some other systems need the CPU for doing this.
The Amiga not.

The Amiga concept is based around DMA.
While other systems need a stronger CPU to do more memcopy.
The Amiga concept relies on using DMA chipset for this.

The Amiga chipset is physically split into 3 main chips.

One chip puts the colors on the Screen.
Name of this chips is Denise (OCS)/ Lisa (AGA) / Isa (Super-AGA)

One chip makes the audio
Name of this chip is Paula (ocs/ecs) / Arne (Super-Aga)

And one Chip is the DMA controller and does all the Data transfers.
On OCS this is named AGNUS. AGNUS has a max transfer speed of 7 MB/sec
On AGA this Chip is called ALICE. ALICE has a max transfer speed of 28 MB/sec
On Super-AGA this Chip is called ANNI. ANNI has a max transfer speed of 800 MB/sec


As you see ALICE max transfer speed is 4 times that of AGNUS(OCS/ECS)
So YES the memory performance of the AGA DMA Chipset / AGA was improved a lot = 400%

I hope this explanation help you understand the Amiga chipset better.

@Hammer

Quote:

32-bit 14 Mhz RAM's raw bandwidth is 56 MB/s and 68EC020 CPU (any fast 68K) only has about 7MB/s from it.

I see where your problem is.
You think about memory like modern syncronoues DDR ram.
You talk about 14 MHz Ram while there was no such things as 14 MHz memory in the 80th and 90th!
I think you have here a misunderstanding how DRAM works and how CPUs work.

Let me try to help you and explain:

During the 80th and 90th the world used asyncronous memory.
Asyncronous Ram has access time = e.g. 110 nanosecond or 80 nanosecond

The 80 nanosecond does NOT mean that you get every 80 nanosecond a data value.
80ns is one part of several timing / steps for retrieving or sending data.
Doing a DATA transfer requires has several steps.
Sending Chip select, Address: Row and Column, then waiting then getting Data.

Running the CPU at 7/14 MHz also not means that a CPU can do every cycle a memory transfer.
Also the CPU needs several steps, sending address, waiting, sending data for one transfer.

This whole concept did change later with the invention of Burst memory modes and later with the invention of syncronous memory.


Lets look at the AMIGA 500.
The Amiga 500 has a chip memory speed of 7 MB/sec

The 68000 CPU used in the AMIGA 1000/500/2000/600 runs at 7 MHz.
The 68000 has a 16bit = 2 Byte Data bus.
The 68000 needs four clock cycles for one Bus access.
This means the 68000 CPU max memory read speed is 7*2/4 = 3.5 MB/sec

This means the 68000 CPU in the Amiga 500 can at maximum use halve the memory bandwidth.
The Amiga Blitter can use 100% of the memory bandwidth.

Please mind that the CPU is very weak for doing memcopy.
This is because CPUs at the time of the Amiga creation did not have caches.
Without Cache the CPU will for every instruction that it runs, also read the instruction from memory.
This means if you use the 68000 CPU for doing a memory copy = e.g. to for painting a screen.
It will be relative slow.
This is why the Amiga uses DMA to avoid memory copies.


Over time CPUs did change and improved here.
The 68000 needed 4 cycle for one memory access.
The 68020 improved this and can at beast case do one access in 3 cycle (with zero wait states)
...
And the Apollo 68080 CPU today has a 64bit memory interface and can do one 64bit transfer every clock cycle.

The AMIGA design was based around doing the graphic and audio with DMA.
The job of the CPU in the Amiga was running game logic - its job was NOT to push/move data around.

The Amiga Chipmemory access was "divided" between chipset and the CPU.
This means 50% of the possible memory were always reserved to the Amiga DMA.
And the CPU could at peak only use the rest.

The AMIGA knew CHIP memory and FAST Memory
And when working in FAST Memory the CPU could use the FULL memory bandwidth.



Lets try to sum this up.

The concept of the Amiga was to use a very good to program CPU for game logic.
And to use DMA chipset - so that the CPU never needs to push data around.
The whole idea of the Amiga design is based on the CPU NEVER NEED TO MOVES DATA!

This concept allowed the AMIGA to make many times better looking games than a CPU of its time was able to do alone.


The DOOM game is coded / made on the idea that the CPU moves ALL the data around.
As better your CPU is as pushing DATA around as better for such games.
The APOLLO 68080 is about 300 times better than the 68000 in pushing DATA around.
This makes games like DOOM benefit a lot on the Apollo 68080.


The design goal of the Amiga 1000/500 was to minimize CPU Data copies and to instead on DMA.
This enabled the AMIGA to do more than the memory can do!

Why is DMA good?
This has 2 reasons.

1) DMA was at this time a lot faster than the CPU!

2) DMA is more efficient.

Why is DMA more efficient?

When you copy with the CPU a sprite/game object to screen
then the CPU needs to read the sprite data, and has to write the sprite to screen memory.
READ + WRITE = 2 memory access

The DMA Amiga Chipset only needs to READ the sprite data.
This means the DMA chipset give you twice the efficiency.

If you copy the Sprite with the CPU into the background then you trash the background and you need to repair it. This repair requires again memory read and write from the CPU.

This means that in real live hardware Sprites are 2-4 times more efficient than CPU rendering.


We need to mind that affordable CPUs at the early years of the AMIGA could NOT reach 100% memory speed.
And a CPU can NEVER do the job as efficient as using e.g. DMA Sprites.


The DMA design of the Amiga allowed the Amiga to do much nicer games than it would have been possible with CPU.

The NEOGEO relies to 100% on sprites.
The CPU does never render anything a NEOGEO game. The CPU even can NOT render anything in NeoGEO. All the NEOGEO game display is 100% done with DMA.

The NEOGEO could never do a DOOM game.

The AMIGA design allows to make both 2D games and also DOOM game.
But the low price Amigas were sold with relative slow CPUs.
And the DOOM game benefits from a higher clocked CPU more.

Over the years CPUs did significantly increase in speed.
This means CPU like 68030/040/060 and 080 give a lot more performance for CPU based games.


As you know Commodore aimed mainly for making very affordable Computers this means their main systems were based around very good to code CPUs but that had not the highest punch.

@Hammer

Quote:

6LC8040 @ 25 Mhz with nearly 20 MIPS with local 32-bit Fast RAM can do a better composing job. Lisa can display the completed frame buffer after it's copied from Fast RAM.

Hammer you try to make arguments with posting snippets from Wikipedia -
but the point makes no sense if you consider the goal.


Your talk about 68040 CPU is like talking to FORD why they not use a 5 liter engine in the Ford Fiesta. The point of the Fiesta is being low cost and affordable.
Asking them to such a very pricy 5 liter engine into the Fiesta would work against their goal.
I think if you consider the goals of Commodore then you will agree.



Let me try to help and explain some background.

Commodores main goal was to sell in expensive entry systems

The Amiga chipset has DMA power.
On the AMIGA you have 3 options:
- you can do games with CPU
- you can do games with DMA
- and you can do games using both CPU and DMA together


PC games were based on the CPU to do all.
NeoGeo did always do all with DMA.
The Amiga design is more flexible and has the above mentioned 3 options.


Having a faster CPU in the AMIGA (e.g. the Apollo 68080) means of course that you have more CPU power.
Having a faster Amiga Chipset (Super-AGA) means you have a lot more DMA power.
All this gives you more options.


The 68040 was a good and very expensive CPU.
As mentioned Commodores goal was to sell many low price entry level systems.
This totally ruled out to use expensive 040 CPUs in their main systems.

If you understand the goal of Commodore than this all makes a lot of sense.

@Hammer

Quote:

AGA's Alice is mostly 16-bit hence there's something common with Jack Tramiel's mindset.

You are misunderstand this.


Mind, the AMIGA concept was around doing DMA.
All the AMIGA DMA is done by
OCS=AGNUS (16bit)
AGA=ALICE (16/32/64bit)
Super-AGA= ANNI (256bit)

While the Amiga has very man DMA channels.
Some are more hungry than others in terms of bandwidth.

The top 3 hungry users are:
- GFX display DMA
- Blitter DMA
- Sprites DMA


Lets compare and look the improvements:

	(OCS AGNUS)	(AGA ALICE)	(Super-AGA/ANNI)
GFX display	16 bit	64 bit	256 bit
Blitter DMA	16 bit	16 bit	256 bit
Sprites DMA	16 bit	64 bit	256 bit

* AGA did significantly improve Display and Sprites x4 times
* It is correct that AGA did not improve Blitter.
It goes without saying that if AGA would have improved Blitter too then this would have made AGA stronger.


You know that Super-AGA does improve Blitter.
Super-AGA Blitter DMA can do 64bit operations and can do 256bit burst DMA.
Developing the Super-AGA DMA was complicated and took a lot of time.
I've worked several years on the Super-AGA design.


You need to mind that the Commodore team was under time pressure to bring our AGA.
They started late ... and needed to limit the scope.
I can fully understand that they simply not had enough time to bring out a 64bit Blitter.

Would a 64bit blitter made AGA stronger? Sure of course.
But they not had the time for this.

This is unfortunate.
But this is how live.

Last edited by Gunnar on 07-Mar-2024 at 08:09 AM.

@Hammer

Quote:

Sega Genesis sold 30 million units by the end of 1997.

This was good for them.

Hammer you know when I was younger I've played Football myself and later I coached kids in sport.

My experience is that you can do a good job but you can still loose.
Maybe because the others did a even better job,
or sometimes because of plain luck, or sometimes because of misfortune like an accident/injury.


If you watch American Football e.g. Super Bowl, then you will see always two excellent teams competing. Both sides have excellent athletes and excellent coaches - and still one will always loose. Loosing does absolutely not mean that one side was all idiots and all lazy.

If we look back at Commodore and Amiga.
Then of course mistakes were made. This is normal.
And its unfair with 30 year hindsight to look back and compare what you see now with what they could not foresee 30 years ago.

Everybody can have here different opinion
But I think that both the Amiga engineers did a good job and the Commodore Management tried to follow their plan and goals.

If my kids team would today play against the Kansas Chiefs.
They would for sure loose.
But this does not mean that the Kids are bad or their coach is incompetent.
It simply means the other are super strong.


I think whole narrative of trying to blame someone is not right.

Developers did not pushed AGA as they pushed OCS but looking at what Daniel can do give us some ideas about its possibilities.
https://youtu.be/ACQU3_17JtA?feature=shared
https://youtu.be/k-6Q4ZBijlM?feature=shared

Last edited by kamelito on 07-Mar-2024 at 09:35 AM.

@kamelito

Quote:

Developers did not pushed AGA as they pushed ECS but looking at what Daniel can do give us some ideas about its possibilities. https://youtu.be/ACQU3_17JtA?feature=shared https://youtu.be/k-6Q4ZBijlM?feature=shared

Good post. Thank you Kamelito!

Daniel is a good Amiga coder and his Game WIP shows what type of games the AGA / A1200 was designed for.

I think that the "goal" of Commodore to make a low cost machine every kid could afford,
able to play games like this was not all wrong.

@Gunnar
Quote:

You are confused between CPU and Amiga chipset.

I'm aware of Amiga's DMA and CPU cache coherency issues.

3DO has 36 channels of DMA and its custom chips are limited by the familiar 25 Mhz clock speed.

Quote:

BUSTEST the benchmark that you does measure only the CPU memory transfer speed but you spoke all the time about the DMA performance of the Amiga DMA chipset.

My point with the CPU's Chip RAM bandwidth is the magnitude range.

Fact: Commodore's AGA and memory configuration do not have 3DO's memory bandwidth.

Quote:

Let me explain you: The Amiga chipset is based around DMA. DMA means Direct Memory Access. Direct Memory Access means the Amiga Chipset can read and transfer by itself memory. Sprites are put on Screen with DMA. Audio / Music is played with DMA. Screen is painted with DMA. Colors on Screen can be changed and reloaded with DMA. Floppy data is loaded with DMA.

That's a red herring. DMA activity from non-CPU devices consumes memory bandwidth allocation and DMA cycle slots.

There are gotchas when using tiled 64-bit wide sprites for the background layer e.g. low color registers.

Sega Saturn's and 3DO (ex-Amiga key engineers)'s four-sided sprite-distortion engine supports texture handling for the quadrilateral 3D accelerator. These are examples of when the hardware sprite engine evolved without Commodore's "read my lips, no new chips" directive.

Furthermore, the entire Chip RAM is cacheless since Amiga custom chips are cache coherence incompetent.

68000 7.1 Mhz with barely 1 MIPS which opens opportunities for value-added fixed-function ASIC providers.

Sega maintained higher-cost arcade machines with higher hardware specs when compared to the Mega Drive. Sega's higher-cost arcade machines' use cases have influenced the low-cost Saturn's R&D.

Sega's higher-cost arcade machines are fully working and shipped products which is important for developing the in-house engineer team's skills and knowledge base.

My argument centers around corporate skill development for their engineers.

Quote:

The Amiga chipset is physically split into 3 main chips. One chip puts the colors on the Screen. Name of this chips is Denise (OCS)/ Lisa (AGA) / Isa (Super-AGA) One chip makes the audio Name of this chip is Paula (ocs/ecs) / Arne (Super-Aga) And one Chip is the DMA controller and does all the Data transfers. On OCS this is named AGNUS. AGNUS has a max transfer speed of 7 MB/sec

I'm already aware of this information.

Again, 68000 7.1 Mhz with barely 1 MIPS opens opportunities for value-added fixed-function ASIC providers.


Quote:

On AGA this Chip is called ALICE. ALICE has a max transfer speed of 28 MB/sec

Meaningless.

AGA delivers faster bitplane DMA to provide enough bandwidth for 8 bitplanes in 35ns modes, which under ECS could only do a maximum of 2 bitplanes, and 256 color registers. Everything else is pretty much unchanged from ECS e.g. BLITTER and COPPER.

Lisa Chip has the bulk of the AGA updates.

Alice BLITTER is still 16-bit with ECS timings.

7.1 MB/s memory copy can deliver +50 hz 320x200 256 color results e.g. SW Dark Forces with Pistorm32-Emu68 https://youtu.be/hj-QSr1P-jQ?t=99

If Alice BLITTER has 28MB/s capability, it should have a reasonable 50 hz 640x480p 256 colors. Despite having a hyper-fast CPU, AGA is relatively slow with 640x480p 256 colors.

AGA's primary pixel block manipulator hardware a.k.a. BLITTER is slow when compared to
1992-1993 contemporaries. Alice's Copper still operates in 16-bit.

The Amiga BLITTER is no faster in AGA than it was in OCS.

There's a reason why my YouTube "proper Street Fighter 2 port" tech demo example was running with 128 colors instead of 256 colors.

Dave Haynie's intention with AGA is to bundle with low-cost AT&T DSP3210 (13.3 MIPS, 25 MFLOPS FP32) at 50 MHz.

Quote:

On Super-AGA this Chip is called ANNI. ANNI has a max transfer speed of 800 MB/sec

I'm aware of this. V4 is backed by a 16-bit DDR3L chip which is 21st-century memory tech that is out of the scope of this topic.

For V4, DDR3L chip example is ISSI IS43TR10256A-15HBL

https://au.mouser.com/ProductDetail/ISSI/IS46TR16256AL-15HBLA1
Data width: 16-bit
Clock speed: 666 Mhz

The competition, RPi 4B's LPDDR4-3200 SDRAM reported about 5 GB/sec memory bandwidth.

Last edited by Hammer on 08-Mar-2024 at 04:19 AM.
Last edited by Hammer on 08-Mar-2024 at 03:49 AM.
Last edited by Hammer on 08-Mar-2024 at 03:46 AM.
Last edited by Hammer on 08-Mar-2024 at 03:43 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@kamelito

Quote:

kamelito wrote: Developers did not pushed AGA as they pushed OCS but looking at what Daniel can do give us some ideas about its possibilities. https://youtu.be/ACQU3_17JtA?feature=shared

That's mostly a "late 16-bit" era 2D gaming experience.

AGA install base didn't reach critical mass to take over from OCS/ECS.

https://www.youtube.com/watch?v=EqFjQmfO6F4
Reshoot Proxima 3 has superior animated backgrounds.

The late 16-bit SNES competition.
https://youtu.be/QAhBwwg0EEw?t=2
https://youtu.be/QAhBwwg0EEw?t=483
https://youtu.be/QAhBwwg0EEw?t=742

SNES started to build its install base in 1990 (Japan) when it transitioned from 8-bit NES.

When SNES arrived in Amiga's core European gaming market in 1992, SNES had a large enough install base to take over from NES while Amiga AGA started from nearly ground zero.

Nintendo has superior hardware generation transition management. Commodore has no ready answer against SNES's 1990 release.


Install base is important for 3rd party developers. For this topic, John Carmark's statement against the Amiga has an install base with 040 and chunky graphics context.


Meanwhile, PC world focused on making VGA standards and CPU faster and cost reduction.

For the Wintel platform, Microsoft intervenes with two Doom ports for Windows 3.1 i386/Win32S/WinG (i.e. leaked WinDoom build) and Windows 95 (i.e. Doom95). For Microsoft, Doom95 is the standard banner for Windows 95's DirectDraw/DirectX. The platform holder needs to intervene as a 1st party games developer.

Microsoft's public Win32 transition started with the 1993 released Win32S for Windows 3.1 386 Enhanced Mode and Windows NT 3.1 which set the groundwork for Windows 95 and NT 4.0.

Business management has a major influence on engineering results.

Quote:

https://youtu.be/k-6Q4ZBijlM?feature=shared

That's pretty good.

Last edited by Hammer on 08-Mar-2024 at 04:53 AM.
Last edited by Hammer on 08-Mar-2024 at 04:51 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

1) Hammer:
Quote:

Wrong. The Alice is not efficient with the given 32-bit memory bandwidth.

2) Gunnar:
Quote:

ALICE is the DMA controller of AGA. ALICE has a max transfer speed of 28 MB/sec. This is a 4 times improvement over OCS/ECS 7 MB/sec. The BUSTEST benchmark you showed is a CPU test not DMA test. You are confused between CPU and Amiga chipset.

3) Hammer:
Quote:

I'm aware of Amiga's DMA and CPU cache coherency issues.

1) You say incorrectly that ALICE, the Amiga DMA Controller would not be improved.
Fact is that AGA DMA bandwidth was improved by 400%
ALICE can do 28MB/sec

2) You then post as "proof" a CPU benchmark result of 7 MB/sec.
I tell you that this Benchmark is a CPU test and not an AGA DMA / ALICE test and has nothing todo with the AMIGA DMA controller.

3) Then you answer totally off-topic about Cache-Coherence

Cache Coherence has nothing to do with the topic.
Cache Coherence is the feature that 2 chips having caches see each other changes.
The A500 CPU 68000 has no cache.
The A1200 CPU 68020 has no Dcache.
So why should they by coherent?
Your post makes no sense.
Btw Super-AGA and 68080 (which has caches) are coherent.


Hammer, you seem very confused.

Why do you talk about cache-coherence of CPUs that have no cache?
Do you not understand much about Computers?
Or do mix just randomly topics?

Last edited by Gunnar on 08-Mar-2024 at 05:14 AM.
Last edited by Gunnar on 08-Mar-2024 at 05:12 AM.
Last edited by Gunnar on 08-Mar-2024 at 05:05 AM.

@Gunnar

Quote:

Gunnar wrote: @Hammer Quote: Sega Genesis sold 30 million units by the end of 1997. This was good for them. Hammer you know when I was younger I've played Football myself and later I coached kids in sport. My experience is that you can do a good job but you can still loose. Maybe because the others did a even better job, or sometimes because of plain luck, or sometimes because of misfortune like an accident/injury. If you watch American Football e.g. Super Bowl, then you will see always two excellent teams competing. Both sides have excellent athletes and excellent coaches - and still one will always loose. Loosing does absolutely not mean that one side was all idiots and all lazy. If we look back at Commodore and Amiga. Then of course mistakes were made. This is normal. And its unfair with 30 year hindsight to look back and compare what you see now with what they could not foresee 30 years ago. Everybody can have here different opinion But I think that both the Amiga engineers did a good job and the Commodore Management tried to follow their plan and goals. If my kids team would today play against the Kansas Chiefs. They would for sure loose. But this does not mean that the Kids are bad or their coach is incompetent. It simply means the other are super strong. I think whole narrative of trying to blame someone is not right.

You contradicted David Pleasance's and David Haynie's argument positions.

The difference is Sega was committed to higher-cost arcade machine R&D and their shipments for developing the use case and skillsets for its engineering team. Sega had the engineering skill depth to develop Saturn hardware. RISC CPU core provided by Hitachi via SuperH.

For leading-edge tech companies, corporate employee skills development is important.

Sega's downfall is the confusing Amercian-driven X32 add-on and Japanese-driven Saturn projects and quadrilateral 3D programming difficulty which are similar problems for 3DO.

Sony's Playstation management is focused when compared to Sega's confusion.

Commodore's management wasn't excellent and its failure pattern started with C128. The original Amiga technology injection helped Commodore which didn't exist in the early 1990s.

Commodore's external technology injection for the early 1990s was paper spec HP PA-RISC and Dr Ed Helper. Time, money, and luck run out for Commodore.

Hitachi is capable of implementing a PA-RISC clone. Hitachi's PA-RISC implementation is Commodore's selection for desktop Amiga Hombre.

Sony's Ken Kutaragi designed SNES's S-SMP audio chip which consists of a custom 8-bit CPU, custom 16-bit DSP, and 64 KB of SRAM. It has about 24.5 MHz clock speed. MIPS RISC was the technology injection for PlayStation. Rendition's Verite has MIPS-like with 3D extensions.

Commodore lost 65xx key 8-bit CPU designers e.g. Chuck Peddle and Bill Mensch.
Commodore lost key original Amiga engineers after 1987.

For Commodore's AAA, its Copper was an incomplete lite-RISC-like processor.

VS

NVIDIA has proven ex-AMD microprocessors and ex-SUN GX 3D accelerator engineers.

3DFX has proven ex-SGI engineers and was purchased by NVIDIA.

ArtX has proven ex-SGI engineers and was purchased by ATI. Tseng Labs was purchased by ATI. ATI was purchased by AMD.

SGI was bleeding engineers.

SGI's remaining 3D business units were purchased by NVIDIA.

Intel purchased Lockheed's Real3D. Intel purchased 3DLabs. Intel Ontario was established to attract ATI engineers near ATI's R&D center in Ontario.

AMD under Lisa Su purge the old Bulldozer and Radeon HD VLIW engineers and management. Intel, Qualcomm, and Apple are happy to grab good Radeon engineers from Radeon.

Intel, Qualcomm, and Apple have critical engineering talent and depth to design their respective GpGPUs.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Quote:

Commodore's "read my lips, no new chips" directive.

You posted this many times now.
But is your quote correct?

1) ECS in AMIGA 3000 is a new chipset.
2) AMBER, BUSTER are new chips
3) AAA was a planned new chipset
4) AGA was a new chipset.

We very clearly see that many new chips were produced.

Posting something 10 times does not make it correct.

@Hammer

HAMMER:
Quote:

The difference is Sega was committed to higher-cost arcade machine R&D and their shipments for developing the use case and skillsets for its engineering team. Sega had the engineering skill depth to develop Saturn hardware. RISC CPU core provided by Hitachi via SuperH.

What exactly is your point?

In my opinion that CPU of the Amiga 1200, the Motorola 68020 CPU is in many ways the much better CPU compared to the SuperH.


Hammer, it is not clear to me what your "position" and "argumentation" is.
I'm sorry to say this but your posts look to me like a wild mix of wikipedia snippets,
without order and without logical connection.
Can you be so kind and try to explain your point in English?


Let me start with explaining you what my opinion on the Amiga history is.

In my opinion the Amiga is a very good combination
of a very clever DMA based chipset, with a very nice to program CPU.

What was the job of the CPU?
The job of the CPU was NOT to move Mega Bytes of Data around.
For this the Amiga chipset was there - and the chipset could do this more efficient than any CPU.
The job of the CPU was do execute the control logic - and for this selecting a very nice to program CPU made the live of every Amiga coder much better.
This means selecting the 68K was a very good decision!


I think that the Amiga chipset was in many ways brilliant.
Let me explain you why I think this:
(If you understand Chipdesign then the Amiga brilliance is very visible.)

The Amiga chipset added very exciting features to the Amiga -
and the engineers were able to do this with using minimal amount of transistors.


The Amiga NEVER had the world best and most expensive chipset.
And this was also NEVER the goal of the Amiga.
The Amiga chipset was efficient and for the "budget" what it did cost - the Amiga could do very cool stuff.

Its this simple.


You can list as many other system with nice features.
But this is still a NO ARGUMENT - as no one ever said the Amiga has the world best chipset.
This was also never the goal.

The AMIGA was an good entry level system.
With a good balance of features and price.

@Hammer

Quote:

For leading-edge tech companies, corporate employee skills development is important.

For sure but where is the context?
Commodore never was a leading-edge tech company.


Could it be that you misunderstand what the Amiga is that you bought?


The Amiga was an very good entry level system.
Its key features were
- the excellent programmability of the CPU
- the "bonus features" that the efficient DMA chipset provided on top of the nice to program CPU.


The Amiga was a great budget system that your parents could afford to buy for you,
it never was a Formula One car. And it never tried to be a Formula One car.

The Amiga did compete in the "budget" home computer range were C64, and Atari ST were.

The combination of great to program CPU, a Multitasking OS with nice GUI, and efficient chipset - was a very good combination and for sure the Amiga did had a lot potential.


I dont really understand why you post 100 times off-topic other systems.
What do you try to "proof" with this?
Would you go into an "Vespa" Roller forum and post there 100 times pictures of some Suzuki Hayabusa Motorcycle or some Harley Bike and state that have more PS than the Vespa?

Would you do this?

Last edited by Gunnar on 08-Mar-2024 at 06:58 AM.

@Gunnar
Quote:

You say incorrectly that ALICE, the Amiga DMA Controller would not be improved. Fact is that AGA DMA bandwidth was improved by 400% ALICE can do 28MB/sec

2 cycles fast page mode, 16-bit, and 14 Mhz, nope. A1200's Chip RAM is not made of hyper-fast SRAM.

A1200's memory clock speed is not high enough to mitigate 2-cycle latency.

Quote:

2) You then post as "proof" a CPU benchmark result of 7 MB/sec.

https://youtu.be/OqMIioWB9YI?t=74
Emu68's RPi 4B's CPU to Chip RAM 7.1 MB/s writes as of June 2023.

What can the AGA "display adaptor" do with 7.1 MB/s writes from a fast CPU?

https://www.youtube.com/watch?v=5scuw8lgEDc&t=144s
Amiga 1200 AGA Beats Of Rage with PiStorm32 Lite Emu68 RPi 4B

https://www.youtube.com/watch?v=hj-QSr1P-jQ
Amiga 1200 AGA SW Dark Forces with PiStorm32 Lite Emu68 RPi 4B, Results: +50 fps with 320x2x0 256 colors.

Quote:

I tell you that this Benchmark is a CPU test and not an AGA DMA / ALICE test and has nothing todo with the AMIGA DMA controller.

7.1 MB/s is based on Amiga's memory bandwidth division e.g. multiply 7.1 MB/s by 2 is 14.2 MB/s.

Quote:

Cache Coherence has nothing to do with the topic.

CPU performance which is needed for 3D.

Quote:

Cache Coherence is the feature that 2 chips having caches see each other changes. The A500 CPU 68000 has no cache. The A1200 CPU 68020 has no Dcache.

This is problematic since the mentioned CPUs sucked for 2.5D and 3D.

Quote:

So why should they by coherent? Your post makes no sense.

It makes sense when AGA needs to be redesigned for CPUs with 68030 or higher instead of a cacheless workaround.

Quote:

Btw Super-AGA and 68080 (which has caches) are coherent.

Thanks for making my argument.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

GUNNAR:
Quote:

You say incorrectly that ALICE, the Amiga DMA Controller would not be improved. Fact is that AGA DMA bandwidth was improved by 400% ALICE can do 28MB/sec

HAMMER:
Quote:

2 cycles fast page mode, 16-bit, and 14 Mhz, nope. A1200's Chip RAM is not made of hyper-fast SRAM. A1200's memory clock speed is not high enough to mitigate 2-cycle latency.

HAMMER
Quote:

Lisa has up to 28 MB/s.

Dont you see - how your contradict yourself?
You say at one time AmIga chimemory could impossible reach 28 MB/sec
while at the same time you say the AMIGA video chip LISA gives out 28MB/sec


That LISA can give out 28 MB/sec is correct.
But for LISA to give out 28 MB/sec, ALICA has provided this amount with DMA..
LISA is the AMIGA display chip.
ALICE is the postman which delivers all the bytes to LISA.



True Fact: AGA DMA Controller ALICE does reach 28 MB/sec.
OCS AGNUS did reach 7 MB/sec
AGA ALICE reaches 28 MB/sec.
This means AGA ALICE DMA speed reaches in parts 400% performance of OCS/ECS AGNUS.
PLANE DMA and SPRITE DMA are boosted to this level.
Other DMA were kept the same for project time reasons.
Getting AGA chipset to market was the key priority.


Hammer its very clear that your wrong argumentation is based
on you simply not understanding how the Amiga chipset works.

If you want then can explain you how the chipset works and why ALICE reaches 28 MB/sec.
Do you want that I explain it?

Last edited by Gunnar on 08-Mar-2024 at 07:19 AM.

@Gunnar

Quote:

Gunnar wrote: You posted this many times now. But is your quote correct?

Quote:

1) ECS in AMIGA 3000 is a new chipset.

You're making the argument at retail release. Dave Haynie made his argument being inside Commodore.

1989 era Amiga 500 Rev 6A already has ECS Agnus 1 MB and the PCB has jumpers for 2 MB Chip RAM configuration.

Commodore designed Amiga 500 Rev 6A with ECS Agnus 2MB as reserve capability.

Amiga 500 Rev 6A has ALL the ECS Agnus improvements except for the 2MB memory address.

ECS Denise's extra modes are useless for games and it works on existing 16-bit Amigas.

Quote:

2) AMBER,

Common with Amiga 2000's A2320 “Amber� card. This doesn't advance Amiga's inherent graphics capability.

Quote:

BUSTER are new chips

This doesn't advance Amiga's inherent graphics capability.

Quote:

3) AAA was a planned new chipset

Too late and run out of time. AAA wasn't proven to be functional as a shippable product.

Quote:

4) AGA was a new chipset.

AGA was released too late despite being completed in AA3000+'s Q1 1991 time period. Release timings matter when building the install base.


Quote:

We very clearly see that many new chips were produced. Posting something 10 times does not make it correct.

Who are you, Mr IBM? You're not Commodore.

Last edited by Hammer on 08-Mar-2024 at 07:18 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Gunnar

Quote:

That AGA DMA Controller ALICE does reach 28 MB/sec is a true FACT.

Fact: ALICE's Blitter is 16-bit.

ALICE's 16-bit Blitter is mismatched with the 32-bit LISA update.

Last edited by Hammer on 08-Mar-2024 at 07:39 AM.
Last edited by Hammer on 08-Mar-2024 at 07:25 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Quote:

ECS Denise's extra modes are useless for games and it works on existing 16-bit Amigas.

Some feature, but like the BIG BLIT ECS, were infact very useful.
And many Amiga games actually code a switch to use this ECS feature.

But in the big picture: Yes ECS was not there to improve games.
ECS allowed to have 1 chipset for PAL and NTSC.


Yes ECS was not there to improve games.
A3000 also was not meant to be a game machine.
Commodore did design/want the A3000 as "office" machine.

You have to judge the element in their context.
Why would you want to make a new chipset (to add more game features) for an office machine?