@Hammer

Quote:

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

1) LISA was a 64bit update


2) correct, the Blitter did not make the 32bit/64bit upgrade in AGA - because the team did not had the development time for this. This is true. We all know this and we understand the reason.


Let us look at the bit numbers which we posted earlier:

GUNNAR:
Quote:

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.

We all agree AGA would have been stronger if the Blitter would have been 64bit too.
But we all know that the team did not had the development time for this in the AGA release.
We all know this and we understand the reason.

Saying that the Blitter is 16bit in AGA is correct.
Everybody knows this.


You repeat 100 times now "no new chips"
Could it be that you misunderstand this and think Commodore did forbid a 64bit blitter?
This was not the case.



I can fully understand Daves wonderful enthusiasm working on the Amiga 3000.
And the Amiga 3000 is and was a wonderful machine.
I can understand that Dave was motivated and keen on adding more GFX feature to the Amiga 3000.

Which engineer would not want to make his baby the world best in every aspect?

But you have to mind that Commodore planned the Amiga 3000 as very high price, serious machine, for more serious office work.

You can surely understand that adding more sprites and more games feature is not important for office work.

I see that Dave wanted to make the world number one machine.
I would have wanted the same.

I can also understand why Commodore wanted to their Office machine to market - instead delay it while waiting for more game features that an Office machine not requires.

Last edited by Gunnar on 08-Mar-2024 at 07:46 AM.

@Gunnar

Read your own post from http://www.apollo-core.com/knowledge.php?b=4¬e=29181&order=&x=0

Gunnar von Boehn:

Generally I would recommend not to use the Blitter in the future.

I can give you some reason for this:

A major part of the "beauty" of the Amiga was using the very coder friendly 68K CPU.
Because of the elegant to code 68K CPU the Amiga had such a big coder scene.

While the 68K CPU family is great to code, only the 68030 or higher CPUs are really fast in Bitshifting operations.
The CPU used on the low end AMIGA, the 68000 has a performance problem here.
The Blitter is a "fix" for this.
The Blitter allows Bitshift/Bitcombine operations useful for GFX operations at a speed like 68020/68030 CPU.

This means the AMIGA Blitter is a performance fix for the 68000 CPU.
But the Blitter also has some "drawbacks".

The AMIGA OS is a Multitasking OS.
The Blitter does NOT support saving a context. This means several task can NOT use the Blitter or share the Blitter elegantly. Each task always has to WAIT for the previous Blitter job of the previous task to fully finish. Also manipulating GFX in parallel with Blitter and CPU can create problems and in general the AMIGA OS will "busy-wait" for a Blitter to finish to prevent problems.
Also one taks using the Blitter has to carefully wait for the first job to finish before starting the next job, as starting a job while the Blitter is not fully finished will cause bad things = even crash.

This means the Amiga Blitter is a compromise.
It improves bitoperation speed to be comparable to 68020 Level but it also introduces drawbacks for Multitasking and deadlock problems.
Also it makes coding more complex and the reason for many games problems is not 100% correct coding of the Blitter.


What happens when the CPU target is 68LC040 @ 25 Mhz against the ALICE's Blitter?



Quote:

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.

ECS Denise was included with A500Plus (A500 Rev8), A600, and unreleased "A1000 Jr".

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Hammer


Quote:

What happens when the CPU target is 68LC040 @ 25 Mhz against the ALICE's Blitter?

The Motorola 68040 was at its time a high end CPU.
The 68040 purchase price alone - was over the total production cost of the Amiga 1200.

The Amiga 1200 was aimed to be a low end entry level system with low cost.
Using an Motorola 68EC020 CPU matched the market segment.


When you order a Happy Meal at McDonalds, would you expect to get Argentinian Steak and Lobster for your $5?
Your argument is not logical.

@Hammer

your post give the impression that you desperate look for Someone to blame.
That you want the someone is responsible for Amiga not being no 1.

Why do you want this?



Look my Apollo 68080 CPU is in many aspects many times better 68040 CPU and 68060 CPU.

As a friend of mine, Dave Haynie said:
Quote:

If Commodore would have had the Apollo 68080, then Amiga would have ruled the world.

Why is the Apollo 68080 so much better?
This is easy to understand.
I had knowledge of many CPU invention made after the releases of the 040/060.

Therefore I could base my work on stuff not existing at their time.
And I did work without project time pressure and could take all the time I wanted to include what I like - what the engineers at Motorola could not.

Did I ever complain or blame the Motorola guys for not making the perfect 68k?
Of course I never did this and would never do this.
At their time with their knowledge they did absolutely great jobs.


I dont understand why you frantically look for a weakness and someone to blame for.

@Gunnar

Quote:

1) LISA was a 64bit update

For context, LISA's each of eight hardware sprites are 64-bit pixels wide. LISA's sprite has a low color display capability. AGA Blitter objects can have up to 256 colors.

It's too bad the upgraded eight hardware sprites effort wasn't the Blitter.

LISA is not a true 32-bit raster display chip since it lacks a 32-bit per-pixel color.

Quote:

We all agree AGA would have been stronger if the Blitter would have been 64bit too. But we all know that the team did not had the development time for this in the AGA release. We all know this and we understand the reason.

The Blitter update should match the 32-bit memory bus at 14 Mhz.

64-bit Blitter on 32-bit bus would need two memory access cycles.

Quote:

You repeat 100 times now "no new chips" Could it be that you misunderstand this and think Commodore did forbid a 64bit bitter?

Where's Commodore's fully functional 64-bit Blitter?

Quote:

I can fully understand Daves wonderful enthusiasm working on the Amiga 3000. And the Amiga 3000 is and was a wonderful machine. I can understand that Dave was motivated and keen on adding more GFX feature to the Amiga 3000.

Commodore didn't allow new graphics IP to be made on 68030/Ramsey's 25 Mhz 32-bit bus infrastructure.

Quote:

Which engineer would not want to make his baby the world best in every aspect? But you have to mind that Commodore planned the Amiga 3000 as very high price, serious machine, for more serious office work.

Commodore repeated C128's high resolution with a low color display business resolution modes and aging game hardware mindset.

ECS's 4-color 640x400p productivity modes are not competitive. It's worse than the baseline IBM VGA let alone ET4000AX equipped PC clones.


Quote:

You can surely understand that adding more sprites and more games feature is not important for office work.

ECS's 4-color 640x400p productivity modes are not competitive. You thinking like an Apple Mac in 1985.


Quote:

I see that Dave wanted to make the world number one machine. I would have wanted the same. I can also understand why Commodore wanted to their Office machine to market - instead delay it while waiting for more game features that an Office machine not requires.

ECS's 4-color 640x400p productivity modes are not competitive. Commodore decided to bundle a TIGA card with Unix instead of fast-tracking Amiga's upgraded graphics capability with RTG.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Gunnar

Quote:

Gunnar wrote: @Hammer Quote: What happens when the CPU target is 68LC040 @ 25 Mhz against the ALICE's Blitter? The Motorola 68040 was at its time a high end CPU. The 68040 purchase price alone - was over the total production cost of the Amiga 1200. The Amiga 1200 was aimed to be a low end entry level system with low cost. Using an Motorola 68EC020 CPU matched the market segment. When you order a Happy Meal at McDonalds, would you expect to get Argentinian Steak and Lobster for your $5? Your argument is not logical.

With a full 68040 CPU @ 25Mhz, A3640's price in 1993 was about $400 retail.

Do a diff between A3000T/030 vs A3000T/040.

68LC040 is cheaper than the full 68040.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Hammer

Quote:

The Blitter update should match the 32-bit memory bus at 14 Mhz.

1) actually there is no 14MHz Memory Bus.

2)
Correct, the Blitter was not upgraded
We all know this.

Why was the Blitter not upgraded?
The Blitter was not upgraded because the project time frame was over.
Would it had been nice to have a 64bit blitter? Yes!
But they decided to NOT wait for this - and instead to release the machine.

Who can Hammer blame for it?
Why do you always need to blame someone?



Nothing on this planet is perfect.

Was the 68020 better than the 68000?
Yes.
Why did Motorola release the 68000 as it and not develop some more year?

Was the 68030 better than the 68030?
Yes.
Why did Motorola release the 68020 as it and not develop some more year?

Was the 68040 better than the 68030?
Yes.
Why did Motorola release the 68040 as it and not develop some more year?

Was the Apollo 68080 better than the 68060?
Yes.
Why did Motorola release the 68060 as it and not develop some more year?



Every project gets release at some point.
Every CPU, every chip on this planet could in theory been made a little better if they would have waited longer.

Your game to complain about a release which in theory could have been better if it would have been delayed - is not intelligent. Why do you do this?

@Hammer

SNES resolution is 256x224 compared to 320x256 so if Amiga developpers reduce the resolution to the SNES level they could put more stuff on screen.

@Hammer

GUNNAR:
Quote:

You repeat 100 times now "no new chips" Could it be that you misunderstand this and think Commodore did forbid a 64bit bitter?

HAMMER:
Quote:

Where's Commodore's fully functional 64-bit Blitter?

Everyone understands that the 64bit Blitter was worked on but not debugged and AGA does not have it.
What is so complicated on this situation, that you not understand it?



I think you seriously misunderstand the "no new chips".
The "no new chips" post was in context of the Amiga 3000.
Commodore wanted the Amiga 3000 for office use.
Commodore not wanted to delay the Amiga 3000 release, for adding extra game features to it.
Commodore was the opinion that an Office PC not requires more game features.

You are free to disagree with Commodore, but Commodore reasoning has some logic to it.

Hammer

You always have release dates and dead lines in development.

Maybe it helps you to think of it as of "news paper".
News paper get printed - and you need to print it latest in the evening before to get it to the people in the morning.
There is a deadline at which time the sunday release needs to be done - and has to go to print.


Engineers always likes to make his development better. This is normal.
And if you wait for them to get all 100% then nothing will ever get released in time.
The job of management is to "define" the dead line and then give the newspaper to print.

I hope this is simple to understand.

@Hammer

Quote:

64-bit Blitter on 32-bit bus would need two memory access cycles.

What exactly is the problem?
What do you think how AGA works?

The memory chips that Commodore uses in A1200 support burst mode.
ALICA the AMIGA AGA DMA controller can do a 64bit bursts access
consisting of two reads each 32bit in one memory cycle.
This is how AGA works.

But how would Zorro3 work with AC68080?
How would AMIX work with AC68080?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

Quote:

But how would Zorro3 work with AC68080?

As it does with other 68K?

@Gunnar

Quote:

Quote: But how would Zorro3 work with AC68080? As it does with other 68K?

Which one? The 68030 way, or the 68040/68060 way? Both with.. pain and MMU tricks? Have you tried yet? How's the "Kraken" coming along?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

Quote:

Which one? The 68030 way

Yes the Apollo CPU can speak native 030 and 000 bus protocol.

There are two types of Amiga mainboards.
Amiga 1000/ Amiga 500/ Amiga 2000 / Amiga 600 / CDTV
These modes "speak" the Motorola 68000 bus protocol.

The other Amigas : A3000, A4000, A1200 and CD32 speak the Motorola 68020/30 Bus protocol.

This means if you want to plug an 68000 CPU into an A500 - this is easy.
As all you need todo is connect each wire to wire - as the Amiga mainboard does speak the same language as the 68000 CPU.

The same is true for making an 68030 Accelerator for Amiga 1200.
This is super easy as all you need to do is connect each wire to wire.


The Motorola 68040 CPU has a different protocol, which makes connecting it to Amiga 3000/4000 more complicated as there is a "language" translation needed.
As you know the 3640 has many small programable chips on it and their job is protocol translation.
The different protocol is also the reason the 3640 is very slow in memory performance.


The Apollo 68080 CPU provides 2 memory controllers.
One controller support natively the 16bit and 32bit Bus protocol of the 68000 and 68030 CPU.
This means the Apollo 68080 CPU can be directly connected to ANY/ALL Amiga main-boards, simple wire to wire.

The other memory controller of the Apollo 68080 is for connection of modern Fastmemory.
This fastmemory controller can directly speak SDRAM and DDR Ram protocol.
This means you can directly connect memory chips to the CPU.


With its two controllers the 68080 can in parallel do memory operation in its local memory
and also at the same time do chipmemory transfers on the Amiga bus.
This is a special feature of the 68080 - all previous 68K only had 1 bus controller.


To answer your question - as the Apollo 68080 speaks both 68000 and 68030 bus protocol natively.
There is never a translation needed and it can also directly talk with Zorro2 and Zorro3 -
as Zorrobus is "copy" of the Motorola CPU bus.

@Gunnar

So 68030 way, with pain and MMU tricks? Have you tried yet? How's the "Kraken" coming along?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

Quote:

So 68030 way, with pain and MMU tricks

There are no MMU tricks needed.

I know why you think this
But what you think about is not the actual technique of the bus protocol
but behaviors of the CPU around it. In other words doing something that it should not do.

Let us explain this and dive into how the Motorola bus protocols works:

The 68020 Bus protocol is relative easy to explain.

Several participants can have the bus.
This means either the CPU or e.g. a DMA controller can "ask" for the bus and then use it.
When you have acquired the bus you can use it to read or write. This works fine.

The 68030 bus protocol works the same as the 020
plus the 030 has the option to do not only single reads but it can it it wants do also line bursts.
A line burst will do always 4 reads in a row the idea is to reduce the time to read cache lines.

It goes without saying that doing a burst access is generally bad idea
when talking to Amiga chipset or when talking to Zorro cards.
Therefore you should not do this.

So all you need to do is to talk like the 020.
Or "tell" your 030 to only burst the fastmemory and never to Amiga chips.
Its that simple.


As I explained before the older 68K have 1 bus only
This means they run their fast memory access and the access to the Amiga bus over the same interface.

Its obvious that having 2 fully independent controllers is a big advantage.
As for the Apollo 68080 the situation very simple.
One controller is talking to the Amiga IO and Zorro cards.
And this does not burst.

The other controller 68080 talks to its own local fast memory and will happily do high speed burst all the time. Therefore there is never a confusion.

Kraken is moving along I received an A3000 and A4000 for test.


In general a CPU can see "external" items on the bus as either Memory or IO.
I hope what Memory is is clear.
IO can be items like chips which you program = Blitter,
or chips that you read that change their data by itself = E.g. SCSI or Network card buffer.

If a CPU has caches then its obvious that you should never cache IO ranges.
Its clear that if the CPU would e.g. cache the serial port content - it would always read the same byte form it.
So every normal computer design would correctly setup cache/nocache reason.
The CPU have extra signals for this and you DO NOT need to use an MMU for this.

If you read the errata (known bugs) of 040 and 030 and 060 CPU then you can find that there are known bugs in the old Motorola chips. You setup a Motorola 030 system that run into bugs on the bus.
This does NOT mean the 020 /030 Bus protocol is inherently wrong.
Its good.

It just means the chip you got from Motorola had a bug you can trigger.
The Apollo 68080 does no have these bugs.

To cut a long story short,
The Apollo 68080 has many improvements over the old Motorola chips

Last edited by Gunnar on 09-Mar-2024 at 07:45 AM.

@Gunnar

Quote:

Let us explain

https://en.wikipedia.org/wiki/Royal_we

Quote:

As I explained before

Yes, many, many years ago. Many, many TerribleFire cards ago. Many, many Re:Amigas ago... long before PiStorm

Quote:

Kraken is moving along I received an A3000 and A4000 for test.

So soon? Well, good luck.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@Gunnar

Quote:

Gunnar wrote: @Hammer Quote: 64-bit Blitter on 32-bit bus would need two memory access cycles. What exactly is the problem? What do you think how AGA works? The memory chips that Commodore uses in A1200 support burst mode. ALICA the AMIGA AGA DMA controller can do a 64bit bursts access consisting of two reads each 32bit in one memory cycle. This is how AGA works.

Your "64-bit" is meaningless without context. 64-bit payload over several clock cycles is different from 64-bit per clock cycle.

64-bit payload with 16-bit wide hardware logic block is four clock cycles.

Your "64-bit" argument is inferior to the classic Pentium's 64-bit wide front-side bus or S3 Trio 64V's 64-bit wide video memory bus.

Your "64 bit" marketing doesn't work on me.

Last edited by Hammer on 11-Mar-2024 at 05:01 AM.
Last edited by Hammer on 11-Mar-2024 at 04:57 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Gunnar

Quote:

1) actually there is no 14MHz Memory Bus.

There's memory bandwidth magnitude.

For 50 hz Turrican 2 AGA with 256 colors from PC VGA 256 color port, it requires Fast RAM equipped A1200.

Reference
1. https://eab.abime.net/showthread.php?t=106735

Unlike the Lisa chip, Alice's 16-bit Blitter design is unable to exploit 4X memory bandwidth improvements on stock A1200.


Quote:

Why was the Blitter not upgraded? The Blitter was not upgraded because the project time frame was over. Would it had been nice to have a 64bit blitter? Yes! But they decided to NOT wait for this - and instead to release the machine.

For this topic, the blame is on Commodore's upper management.

Dave Haynie has the timeline for R&D stall i.e. during the A3000's R&D phase.

Quote:

Who can Hammer blame for it? Why do you always need to blame someone?

Refer to Dave Haynie's argument position.

David Pleasance (Commodore UK MD) has other information on Commodore's Hong Kong to the Philippines factory transfer debacle, Commodore International CEO(1993)'s rejecting Commodore UK/Bull Frog (associated with EA)/Team 17/Psygnosis's CPU accelerated A1200 bundles.

Bull Frog was working on Magic Carpet in 1993 and it was cancelled. Many British Amiga magazines were hyping Magic Carpet for the AGA platform.

Quote:

Nothing on this planet is perfect. Was the 68020 better than the 68000? Yes. Why did Motorola release the 68000 as it and not develop some more year? Was the 68030 better than the 68030? Yes. Why did Motorola release the 68020 as it and not develop some more year? Was the 68040 better than the 68030? Yes. Why did Motorola release the 68040 as it and not develop some more year?

The record speaks for themselves.


Quote:

Was the Apollo 68080 better than the 68060? Yes. Why did Motorola release the 68060 as it and not develop some more year?

Are you Hector Ruiz?

PC's 1993-1994 games with textured map 3D and 2.5D games delivered proper "32-bit" gaming experiences beyond the late 16-bit SNES on PC's higher-priced hardware.

During 1993, PC clone's 486SX-25/486SX-33 based PC prices were about A1200 with non-economics of scale 3rd party 030 @ 50Mhz accelerator. 386DX-33, 386DX-40, and fake 486DLCs (for 386 socket upgrades) are cheaper than i486SX based PC clones.

A1200 and CD32 were squeezed out from the mainstream market and Commodore Germany's A4000's Video Toaster professional market and Commodore PC clones weren't large enough to sustain Commodore.

Last edited by Hammer on 11-Mar-2024 at 05:15 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB