Was anyone else disappointed with the A1200?

[Bruce Abbott]

Originally Posted by hammer

Instructions need data. SysInfo shows stock A1200's MIPS degradation.

SysInfo doesn't tell the full story. AIBB gives you a better idea of how real code will perform. With the instruction cache on:- Memtest is 30% faster, Dhrystone is 35% faster, Matrix is 75% faster, Sort is 98% faster, Sieve is 138% faster.

[hammer]

Originally Posted by Bruce Abbott

Let's say you didn't want to plot pixels one at a time, what then?

Great! Now show us the circuit made out of standard logic gate ICs. Then show us the extra circuitry you need for 1, 2, 3, 4 and 6 bitplanes too, and how you switch between them.

Jay Miner chose bitplanes for one simple reason - he could make a circuit that did 1 bitplane and apply it to as many as he wanted without a lot of tricky logic. And he didn't have any fancy HDL tools to create and debug it. To test the design he had to build the circuit on a wire-wrap board. These photos of the Lorraine prototype give you an idea of what a massive undertaking that was. If you've ever built logic circuits this way you will appreciate the effort that went into it. I've done a few, but I would never attempt something this large and complex.

The Blitter had to work in bitplane mode for doing 1 bitplane anyway, but guess how much extra circuitry you need to do multiple bitplanes? That's right, nothing. So you design your single bitplane blitter and that's it! Then to get more colors you just stack up multiple identical 1 bitplane display boards, and feed their output into the LUT. Perhaps even separate them into 2 groups, each generating its own display (dual playfield). And the blitter doesn't have to know anything about this.

But hey, perhaps you are right and it wouldn't be any more complex. Unfortunately - unlike today's armchair engineers - Jay Miner didn't think to do all it in packed pixel mode. Strangely nobody else did either. I guess they just weren't as smart as Amiga fans of today - probably due to all that tetraethyl lead they were breathing in back then.

I can think of more copper optimizations too, but this doesn't mean what we got was inadequate. It was far better than anything else out there. And it actually worked, unlike so many bright ideas that turned out to be impossible to implement. The AGA chipset was complex enough already, without loading it up with feature creep that would delay its introduction even more.

With EGA, you can write pixels to video RAM in a chunky method via Write Mode 2, BUT it's slower than writing plane by plane.

Amiga OCS's 6-bitplanes are suitable for 1985-era memory bandwidth. It's too bad the color register wasn't expanded to the proper 64-color display mode.

A proper 64-color display with a 4096 color palette is superior to Mega Drive's 64 colors with a 512 color palette.

AGA includes 6-bit to 8-bit plane modes with AAA's 16 million color palette.

VGA was designed beyond the 1987 memory bandwidth and it scaled with faster VGA clones.

Amiga has Copper Chunky but this path is not a proper chunky pixel screen mode.

The Graffiti retrofits chunky pixel support after Lisa/Denise's raster, but Graffiti has its overheads.

If AGA had chunky pixel support, there's the problem with a weak CPU e.g. refer to SNES's SuperFX2 Doom port despite Mode 7's chunky pixel support. For Doom-type games, SNES is facing a single problem while AGA is facing two problems.

Nintendo lost its game platform leadership to Sony's PS1.

Nintendo had less experience with 3D, hence hiring SGI for N64. SGI's key engineers for N64 project founded ArtX which is later purchased by ATI.

SNES's Super FX chip was designed by Argonaut Games. For the hardware, Argonaut Games contacted Flare Technology, the same people who designed the Atari Jaguar. https://www.gamedeveloper.com/progra...super-nintendo


3DFX is also stacked with ex-SGI's engineers. SGI was bleeding engineers who wanted the "power without the price" 3D direction.

PS1's Ken Kutaragi designed SNES's S-SMP audio solution that includes a 24.576 MHz 8-bit CPU, 24.576 Mhz 16-bit DSP, and 64 KB of SRAM. The approach is repeated for PS1, PS2's Emotion Engine, and PS3's CELL SPU. STI's CELL has limits and contracted NVIDIA for GeForce 7-based 3D accelerated GPU.

Kutaragi's motivation was when he watched his daughter play a Famicom and realized the potential that existed within video games.

Without Sony CEO's support, Ken Kutaragi wouldn't be in a happy place since Sony's executives were furious with Kutaragi's skunkworks.

Commodore's original Los Gatos Amiga team shutdown wasn't a good idea for Amiga's core logic R&D development. Commodore Germany won the corporate politics.

My point, a tech company's core assets are the proven engineers.

[Promilus]

MIPS doesn't tell anything... I can have AVR 20MHz giving out nearly 20 MIPS. It could barely get 1DMIPS probably. Because MIPS only shows how many instructions per second processor can do, DMIPS does follow standardized set of actual calculations (so indicates how much of usable work processor can do, not just how many flimsy instructions which might or might not relate to actual computational power). It's stupid to assume 16bit processor with 8bit bus running at 3.58MHz can run nearly as fast as 32b processor with 32bit data bus and small cache running at 14MHz. And no, there is no discrete main and graphic memory access on 5A22. Check schematics - 5A22 talks to S-PPU (1&2) and those have own RAM which 5A22 cannot address directly. So that design is like castrated Amiga with fast RAM. CPU has access to private work memory but does access everything else through chipset registers. In case of Amiga CPU doesn't have private memory at stock configuration but indeed does have access to chipset memory. SNES processor don't, not for PPU memory and not for DSP memory. So you can't do things you can do with Amiga with really fast CPU.

https://en.wikipedia.org/wiki/List_o...ancement_chips

Pilotwings uses DSP ... Just as I said - there were great number of games using enhancement chips on SNES. On Amiga games were designed to run on stock models, later on with RAM expanded models. Only near commodore's demise actual turbo cards etc. were required.

[hammer]

Originally Posted by Bruce Abbott

SysInfo doesn't tell the full story. AIBB gives you a better idea of how real code will perform. With the instruction cache on:- Memtest is 30% faster, Dhrystone is 35% faster, Matrix is 75% faster, Sort is 98% faster, Sieve is 138% faster.

AIBB does this by using data already stored in SRAM registers. The real-world game workload has data being fetched from memory.

[hammer]

Originally Posted by Promilus

MIPS doesn't tell anything... I can have AVR 20MHz giving out nearly 20 MIPS. It could barely get 1DMIPS probably.

Is this applicable to SNES's CPU?

Originally Posted by Promilus

Because MIPS only shows how many instructions per second processor can do, DMIPS does follow standardized set of actual calculations. It's stupid to assume 16bit processor with 8bit bus running at 3.58MHz can run nearly as fast as 32b processor with 32bit data bus and small cache running at 14MHz.

That's meaningless when Turrcian 2 AGA (with VGA 256 color artwork) needs Fast RAM to reduce A1200's shared bus problem. MIPS shows how many instructions are being completed per second.

68020 has a tiny instruction cache and real-world game workload has data being fetched from memory.

SNES has a Wolfenstein 3D port.
Mega Drive (68000@ 7.6 MHz) has a Wolfenstein 3D port with line skips like on A500's Dread/Grind.

A1200's CPU is almost 2X faster with 32-bit Fast RAM.

68000 requires 4 clock cycles to access RAM comparing only 2 cycles for 65C816.

Originally Posted by Promilus

And no, there is no discrete main and graphic memory access on 5A22.

Again, there is a discrete memory bus for the CPU's system RAM and S-PPU's video RAM.

For proper acceleration, the CPU and GPU should remain in their lanes.

Originally Posted by Promilus

Check schematics - 5A22 talks to S-PPU (1&2) and those have own RAM which 5A22 cannot address directly.

So what. The host CPU shouldn't be involved with S-PPU's graphics processing.

CPU's IO interference with the GPU's operations can disportationally degrade raster performance.
https://cdn.wccftech.com/wp-content/...40-not-176.png
PS4 example

PS5 has 512 MB DDR4 and 16 GDDR6 memory pools.
PS4 Pro has 1 GB DDR3 and 8 GDDR5 memory pools.
AMD APUs can still support external memory controllers via the southbridge.

[Bruce Abbott]

Originally Posted by hammer

Wrong. My "full 32-bit 2.5D/3D gaming experience" statement is for a 32-bit gaming PC with at least 386DX-33 CPU and ET4000 or Trident 8900CL or WD90C32 SVGA chipsets.

Er, sorry about that. It's hard to keep track of all the stuff you spew out. I'll have to be more careful parsing it.

SNES delivers a strong late 16-bit 2D gaming experience.

But my point still stands. It doesn't matter how 'strong' the SNES's '2D gaming experience' is, it's not a computer and so is totally uninteresting to me. Same goes for all those consoles that were cut down computers, eg. Sega SC-1000, Commodore 64GS, Amstrad GX4000. The CD32 sneaks in because you only have to plug in a standard Amiga keyboard and mouse to turn it into a computer.

As for your '386DX-33 CPU and ET4000 or Trident 8900CL or WD90C32 SVGA chipsets', such a machine was far more expensive than the A1200 - putting it into another league. But it wasn't long before you could buy an A1200 accelerator card that provided similar performance. Maybe the combination wasn't that much cheaper (why would you expect it to be?), but you could do it as your budget allowed. Personally I was happy to pay a similar price - or even more - to get that performance on an Amiga vs a PC.

Please stop with the silly Amiga vs PC price comparisons. Maybe price was the only important factor for you, but for many of us it wasn't.

Wrong.
[ Show youtube player ]
Ultima Underworld The Stygian Abyss on 80386 @ 25MHz

Wrong back at you. The description says "The 386 is almost a bit too slow for Ultima Underworld.". Comments agree "That looks barely playable."

More importantly, all the magazine reviews suggested that a high-end system was required, putting many PC owners off. It was quite annoying in the PC world when games had 'minimum system requirements' that they barely ran on if at all. Gamers got used to taking the 'recommended' system as the practical minimum.

I had a 386DX-33/ET4000AX-based PC and A3000/030 @ 25Mhz.

I had an A3000 with 25MHz 030 and 386DX-40. The 386 ran Doom quite well, but I didn't use it for much (maybe played Microsoft Windows Solitaire a bit - my favorite PC game!). The A3000 was mostly used for Amiga software development, for which it had the huge advantages of being able to run the code I wrote, run other tools like Deluxe Paint at the same time, and write Amiga disks etc. The 386 was just a toy in comparison.

I run Legends of Valor on a faster A3000 @ 25 Mhz which is like running Wing Commander with 16 colors.

Not sure what you are trying to say here. Wing Commander looks good with OCS graphics. Its only real fault was being a bit slow on a stock A500 (which is hardly surprising considering that the PC version recommended a 16 MHz 386), but it's still playable. Having more colors wouldn't significantly improve the experience. A stock A1200 runs both games well!

[Bruce Abbott]

Originally Posted by hammer

AIBB does this by using data already stored in SRAM registers. The real-world game workload has data being fetched from memory.

You mean CPU registers, right? The most taxing games are typically those that do a lot of calculations. Games that are just moving memory around typically use the blitter to do the 'heavy lifting', but even here the 68020's instruction cache makes memory copy operations 30% faster.

But all this is missing the point - the A1200 was designed to be an enhanced replacement for the A500, not a mid-range or high-end machine. The 020 in a stock A1200 is at least twice as fast as the 68000, and much faster on some stuff like shift/rotate and 32 bit mul/div instructions. So code optimized for the A1200 often runs a lot faster than the clock speed alone would suggest. 2 to 4 times the processing power can make a big difference. Combine this with much lower bitplane DMA contention and the A1200 is a developer's dream compared to the A500.

If the stock A1200 wasn't fast enough then you could add a RAM board - which many owners were expected to do as game requirements increased (just like happened with the A500) - or an accelerator card. Interestingly there were initially more accelerator cards being produced for the A1200 than RAM boards. I guess manufacturers expected that many users would want even more processing power.

[hammer]

Originally Posted by Bruce Abbott

Er, sorry about that. It's hard to keep track of all the stuff you spew out. I'll have to be more careful parsing it.

But my point still stands. It doesn't matter how 'strong' the SNES's '2D gaming experience' is, it's not a computer and so is totally uninteresting to me. Same goes for all those consoles that were cut down computers, eg. Sega SC-1000, Commodore 64GS, Amstrad GX4000. The CD32 sneaks in because you only have to plug in a standard Amiga keyboard and mouse to turn it into a computer.

For the 1990s, your argument alone didn't sustain Commodore.

Originally Posted by Bruce Abbott

As for your '386DX-33 CPU and ET4000 or Trident 8900CL or WD90C32 SVGA chipsets', such a machine was far more expensive than the A1200 - putting it into another league.

For Australia's 1992-1993 market, that's NOT true.

For Australia
June 1992's PC 386DX


VS

June 1993's Amiga 1200


vs

Sep 1993's 486SX25 (near Xmas 1993)


In Australia, the Amiga wasn't price competitive in Q4 1992 and 1993.

"Power without the price" comes back to bite A1200 with a 32-bit CPU accelerator platform.

Originally Posted by Bruce Abbott

But it wasn't long before you could buy an A1200 accelerator card that provided similar performance.

For 1993 money, an A1200+030 @ 50Mhz accelerator is like buying a Gateway 2000 486SX-33 PC.

"Power without the price" comes back to bite A1200 with a 32-bit CPU accelerator platform.

[Bruce Abbott]

Originally Posted by hammer

68000 requires 4 clock cycles to access RAM comparing only 2 cycles for 65C816.

Irrelevant. We're talking about 68020, not 68000.

On most systems in those days memory cycle speed was limited by the RAM, not the CPU. If you put a 65C816 into the Amiga it would have to run at 3.58 MHz and wouldn't be any faster in general. The 65C816 is more efficient at doing some stuff than the 68000, but only has a few registers and is not good on 32 bit. It doesn't scale well. A system using it would be stuck in 16-bit land.

We are fortunate that Jay Miner wanted a 68000 in the Amiga, rather that the 65C02 the team was originally going to use in their game console (65C816 wasn't an option because it wasn't fully released until 1985).

PS5 has 512 MB DDR4 and 16 GDDR6 memory pools.
PS4 Pro has 1 GB DDR3 and 8 GDDR5 memory pools.
AMD APUs can still support external memory controllers via the southbridge.

More off-topic garbage.

[hammer]

Originally Posted by Bruce Abbott

You mean CPU registers, right?

CPU registers are small SRAM data storage, the fastest known data storage method.

Modern GPUs' SRAM registers in the multi-megabyte storage range as an extreme implementation.

Originally Posted by Bruce Abbott

The most taxing games are typically those that do a lot of calculations. Games that are just moving memory around typically use the blitter to do the 'heavy lifting', but even here the 68020's instruction cache makes memory copy operations 30% faster.

Turrcian 2 AGA's Fast RAM recommendation says hi.

Originally Posted by Bruce Abbott

But all this is missing the point - the A1200 was designed to be an enhanced replacement for the A500, not a mid-range or high-end machine. The 020 in a stock A1200 is at least twice as fast as the 68000, and much faster on some stuff like shift/rotate and 32 bit mul/div instructions.

A1200's $599 USD Q4 1992 wasn't a replacement for 1987's A500's $699 which is $872.33 in 1992.

68EC020 has a barrel shifter, but the CPU operates 32-bit 7 Mhz 68EC020 effective in stock A1200.

AGA Alice's' Blitter is still 16-bit, hence wasted the "32-bit" memory bus.

The "new 32-bit 3D" general market expectation from 1992-to-1994 needs to be more than 2X over A500's 68000 implementation.

Deliver a "new 32-bit 2.5D/3D" gaming experience or compete against SNES's late 16-bit strong 2D gaming experience.

Originally Posted by Bruce Abbott

So code optimized for the A1200 often runs a lot faster than the clock speed alone would suggest.
2 to 4 times the processing power can make a big difference. Combine this with much lower bitplane DMA contention and the A1200 is a developer's dream compared to the A500.

AGA Alice's' Blitter is still 16-bit, hence wasted the "32-bit" memory bus.

AGA Alice has 2 clock cycle improvements and AGA Lisa has 32-bit bus improvements.

The Blitter needs to be improved by 4X for fast 640x400p 256 colors.

Originally Posted by Bruce Abbott

If the stock A1200 wasn't fast enough then you could add a RAM board - which many owners were expected to do as game requirements increased (just like happened with the A500) - or an accelerator card. Interestingly there were initially more accelerator cards being produced for the A1200 than RAM boards. I guess manufacturers expected that many users would want even more processing power.

Tell that to the major game developer who attended the meeting with Mehdi Ali.

[hammer]

Originally Posted by Bruce Abbott

Irrelevant. We're talking about 68020, not 68000.

It's relevant. Refer to 68000 (without factoring four clock cycles for memory access) being faster than the 65C816 (two clock cycles for memory access) argument.

A1200 has two cycles for memory access improvements.

Originally Posted by Bruce Abbott

On most systems in those days memory cycle speed was limited by the RAM, not the CPU. If you put a 65C816 into the Amiga it would have to run at 3.58 MHz and wouldn't be any faster in general. The 65C816 is more efficient at doing some stuff than the 68000, but only has a few registers and is not good on 32 bit. It doesn't scale well. A system using it would be stuck in 16-bit land.

SNES wasn't designed for 32-bit OS. Ricoh 5A22 CPU is customized with DMA engines, extra registers, and other game console-related hardware.

For Xenix 286, 16-bit 80286 runs at a higher clock speed, 16-bit front side bus, and includes MMU.

65C816 remains as a toy CPU.

Originally Posted by Bruce Abbott

We are fortunate that Jay Miner wanted a 68000 in the Amiga, rather that the 65C02 the team was originally going to use in their game console (65C816 wasn't an option because it wasn't fully released until 1985).

32-bit OS requirement needs a 32-bit programming model. 68000 allows for a 32-bit programming model for 32-bit OS. In the early 1980s, Unix workstations have custom MMUs with 68000/68010.

1985 had 32-bit with MMU i386's release which is too late for Amiga Lorraine.

65C816's R&D road map is inferior. For a 32-bit CPU, Bill Mensch recommends the ARM CPU family instead.

Originally Posted by Bruce Abbott

More off-topic garbage.

You can't handle the truth with shared memory's downside.

[Bruce Abbott]

Originally Posted by hammer

For Australia's 1992-1993 market, that's NOT true.

Like a cracked record you are.

For Australia
June 1992's PC 386DX

No hard drive, no sound card, no mouse, no operating system. Who knows what that VGA card was, but based on their attempt to make the price look as cheap as possible I'm betting it was a crappy one.

Comparing this clone crap to a name brand computer is not fair. Show us the price of a full 386DX system (not a few no-name clone parts thrown together) in a high-street shop.

VS

June 1993's Amiga 1200

I remember Don Quixote Software. His A1200 price was inexplicably a lot higher than mine. And my 386DX system prices were much higher than the A1200 - but then we did only sell reasonable quality clones and they did come with everything the customer expected.

I wish I had kept some price lists and flyers from that era. Unfortunately I threw away all the paperwork after closing the business and only have a bit of stuff from 1996 on (hiding on my 1200's hard drive).

For 1993 money, an A1200+030 @ 50Mhz accelerator is like buying a Gateway 2000 486SX-33 PC.

No, it's like an A1200 with a 50MHz accelerator added. How many times do I have to tell you wasn't just about the price!

"Power without the price" comes back to bite A1200 with a 32-bit CPU accelerator platform.

That was Jack Tramiel's mantra. It made the ST sound cheap and nasty - which it was.

[hammer]

Originally Posted by Bruce Abbott

Like a cracked record you are.
No hard drive, no sound card, no mouse, no operating system. Who knows what that VGA card was, but based on their attempt to make the price look as cheap as possible I'm betting it was a crappy one.
.

Hint: Look in the 1992 Australian Amiga Review magazine for A1200 ads.

A1200 is in play in Australia's 1993 market.

The year 1993 was Commodore's critical year and its main bulk for A1200 shipments with Commodore's strongest markets.

During Australia's October 1992,

386DX-40 PC with 128 KB cache, 2 MB RAM, 80 MB HDD, 1.44 FDD, Keyboard, Super IDE card, 512 KB SVGA card, SVGA monitor, 2 serial, 1 printer port, 1 game port, mini-tower) has $1570 AUD

386SX-33 PC with 2 MB RAM, 42 MB HDD, 1.44 FDD, Keyboard, Super IDE card, 512 KB SVGA card, SVGA monitor, 2 serial, 1 printer port, 1 game port, mini-tower) has $1260 AUD

Australian market's November 1993 for the state of Victoria. From Australian Personal Computer magazine


486SX-33 (4MB RAM, 125 MB HDD, SVGA monitor, SVGA card, 2 serial, 1 printer port, 1 game port, 1.44 FDD) has $1595.

386DX-40 (4MB RAM, 125 MB HDD, SVGA monitor, SVGA card, 2 serial, 1 printer port, 1 game port, 1.44 FDD) has $1450

386SX-33 (4MB RAM, 125 MB HDD, SVGA monitor, SVGA card, 2 serial, 1 printer port, 1 game port, 1.44 FDD) has $1385


The next page has ET4000 for $145.
Sound Blaster 2.0 for $135.



------------------


https://archive.org/details/Australi...ge/n3/mode/2up
Australian Commodore and Amiga Review, October 1993.
Page 4 of 84
A1200 barebone = $799
A1200 with 40 MB HDD = $995 AUD
A1200 with 85 MB HDD = $1349 AUD
A1200 RAM card with 0 MB = $249 AUD
From the state of Western Australia.

Page 12 of 84
GVP A1230 030 with No Co-Pro and 0 MB RAM =$ 876
GVP A1230/030 & 68882 40Mhz with 4 MB RAM = $1176

Page 42 of 84
GVP A1200 SCSI with 4 MB RAM = $895
GVP A1200/030 at 40 Mhz and 4MB RAM = $1195.00

A1200 with 40 MB HDD + GVP A1200 SCSI with 4 MB RAM = $1,871

Page 56 of 84
Phase 5 Blizzard A1200/4 with 4MB RAM and clock = $499

A1200 with 40 MB HDD + Blizzard A1200/4 (4MB RAM card) = $1,494.
A1200 with 85 MB HDD + Blizzard A1200/4 (4MB RAM card) = $1,848.

A1200 barebone + Blizzard A1200/4 (4MB RAM card)= $1,298. No hard disk. A1200 with 32-bit Fast RAM is about a PC with 386DX-16 to 20 and ET4000AX or Trident 8900CL.

Page 64 of 84
Seagate 2.5 inch, 128 MB IDE = $245.

A1200 barebone + Blizzard A1200/4 (4MB RAM card) + Seagate 128 MB HDD = $1,543

Originally Posted by Bruce Abbott

Comparing this clone crap to a name brand computer is not fair. Show us the price of a full 386DX system (not a few no-name clone parts thrown together) in a high-street shop.
.

Before my 386DX-33 PC clone, after owning the IBM PS/2 Model 55SX, your statement is trash.

My 1st C= A500 Rev6A
1. A fault with the Agnus socket and it was sent to Commodore for repairs.
2. A fault with the PSU and it was sent to Commodore for repairs. This is the same hardware flaw as my friend's A500 Rev5 PSU. This A500 Rev5 was abandoned when his family purchased a 486DX-33 PC and I kept this A500 Rev5 in storage since 1993 until the COVID-19 lockdown and I repaired this PSU. Without the internet, diagnosing the hardware fault was a problem.

My 1st A500 Rev6A was traded in for an ex-corporate A3000 (AmigaOS 2.04, 1MB Chip RAM, 1 MB Fast RAM, 25Mhz CPU/FPU, Amiga Vision software package) in early 1992.

I purchased my 2nd A500 Rev6A with chips from the mentioned A500 Rev5, ECS Agnus 1 MB, and a UK keyboard from A500Plus during the COVID-19 lockdown.

My C= A1200 Rev1d4, I repaired its timing bugs.

PC's brand names like Dell are for corporations with corporate support contracts.

Originally Posted by Bruce Abbott

I wish I had kept some price lists and flyers from that era. Unfortunately I threw away all the paperwork after closing the business and only have a bit of stuff from 1996 on (hiding on my 1200's hard drive).

No, it's like an A1200 with a 50MHz accelerator added. How many times do I have to tell you wasn't just about the price!
.

Keep failing.

Originally Posted by Bruce Abbott

That was Jack Tramiel's mantra. It made the ST sound cheap and nasty - which it was.
.

Jack Tramiel's "power without the price" is against Apple Mac's B/W release. Amiga OCS's "power without the price" is against other 4096 color palette competitors. e.g. IBM PGC, NEC PC-98.

[Bruce Abbott]

Originally Posted by hammer

It's relevant. Refer to 68000 (without factoring four clock cycles for memory access) being faster than the 65C816 (two clock cycles for memory access) argument.

This thread is about the A1200. Does it have a 68000 in it? No.

SNES wasn't designed for 32-bit OS.

It wasn't designed for any OS. It was just a box that was needed to use the real product - games cartridges.

For Xenix 286, 16-bit 80286 runs at a higher clock speed, 16-bit front side bus, and includes MMU.

More irrelevant PC crap.

1985 had 32-bit with MMU i386's release which is too late for Amiga Lorraine.

Yep, and the Amiga wasn't designed to be a multi-user business system so it didn't need one.

You can't handle the truth with shared memory's downside.

Since a mouse piddled on my workshop PC's graphics card all my PCs are now using shared memory. It doesn't seem to affect them for the stuff I do (not playing the latest 3D games). But that's today. Let's see which home computers used shared memory back in the 80's and 90's. Oh look, almost all of them!

BTW in 1981 I designed and built my own home computer from scratch using shared memory. Upside - much simpler circuit using fewer parts. Downside - none. I also used the latest 2k static RAM chips. They cost more but were cheaper once you factored in the reduced support circuitry and design effort. Someday I hope to reproduce this design - I still have the original 6802 CPU!

[hammer]

Originally Posted by Bruce Abbott

I had an A3000 with 25MHz 030 and 386DX-40. The 386 ran Doom quite well, but I didn't use it for much (maybe played Microsoft Windows Solitaire a bit - my favorite PC game!). The A3000 was mostly used for Amiga software development, for which it had the huge advantages of being able to run the code I wrote, run other tools like Deluxe Paint at the same time, and write Amiga disks etc. The 386 was just a toy in comparison.

My 386DX PC clone had Lotus 123/AmiPro for Windows for back-office work, Borland Turbo Pascal, and Visual Basic. A3000 was used for my high school's technical drawing subjects, Deluxe Music 2/MIDI (for non-profit work), Deluxe Paint/CANDO presentations (e.g. biology, hydro-carbon chemistry, physical education), and a fake Macintosh (e.g. MS Office, Adobe Acrobat).

Originally Posted by Bruce Abbott

Not sure what you are trying to say here. Wing Commander looks good with OCS graphics. Its only real fault was being a bit slow on a stock A500 (which is hardly surprising considering that the PC version recommended a 16 MHz 386), but it's still playable. Having more colors wouldn't significantly improve the experience. A stock A1200 runs both games well!

Wing Commander VGA is slow on the IBM PS/2 Model 55SX (386SX-16, 16-bit MCA, IBM VGA).

[ Show youtube player ]
Wing Commander VGA on 386SX-33 (16-bit front side bus). Slightly jerky frame rates.

vs
[ Show youtube player ]
Wing Commander VGA on 386DX-25 (32-bit front side bus). 386DX-25 version has smooth frame rates.

[ Show youtube player ]
Wing Commander CD32 with Akiko wasn't smooth on stock CD32. Fast RAM is not enough. Needs about 25 Mhz 68020/68030 CPU. I have 8MB Fast RAM with my A1200 before TF1260 and PiStorm32.

[Locutus]

Originally Posted by hammer

https://en.wikipedia.org/wiki/Pilotwings_(video_game)

Pilotwings didn't use SuperFX, it's effectively a game tech demo for SNES's Mode 7.

It uses the DSP-1 chip.

[hammer]

Originally Posted by Bruce Abbott

This thread is about the A1200. Does it have a 68000 in it? No.

Follow the thread.

Originally Posted by Bruce Abbott

More irrelevant PC crap.

It's relevant when this topic is about comparing A1200 against the world.

Originally Posted by Bruce Abbott

Since a mouse piddled on my workshop PC's graphics card all my PCs are now using shared memory.

Modern PC CPUs and IGP have larger caches.

PS4's Liverpool GPU's ROPS has a very tiny render cache directly connected to the memory controller. GCN's texture units are connected to the GPU's L2 cache.

VEGA GPU generation was AMD's 1st GPU ROPS (raster operations) connected to the GPU's multi-MB L2 cache design.

NVIDIA's Maxwell GPU family has texture and ROPS units connected to the GPU's L2 cache.

Intel IGP has access to the L3 cache via the ring bus.

Intel/AMD/NVIDIA are aware of the shared memory bus issues and have mitigated against them by throwing transistors at the problem.

AMD and NVIDIA have improved Delta Color Compression to maximize GPU's cache storage.

Originally Posted by Bruce Abbott

It doesn't seem to affect them for the stuff I do (not playing the latest 3D games). But that's today. Let's see which home computers used shared memory back in the 80's and 90's. Oh look, almost all of them!

You failed to factor in the tiny details.

3DO and PS1 have two discrete memory pools. There was a design flaw with the PS4.

Commodore can't survive with your non-gaming Amiga use case.

[paul1981]

Originally Posted by bruce abbott

like a cracked record you are.

Some would say like artificial intelligence.

[Promilus]

Originally Posted by hammer

Is this applicable to SNES's CPU?

Sure, it was a design which was not centered on high performance of 16bit apps, it was a hack on existing 8bit 6502 to address more memory and maybe improve performance slightly. It was hardly 16bit much like 68000 was hardly 32bit. But at least 68000 had fairly large amount of registers and pretty wide ones. Didn't suffer penalty from working on 16bit data (65C816 does due to limited performance of 8bit data bus). That's why I wrote MIPS ain't everything. How powerful instruction set is affects performance more than just MIPS count. That's why I did apply your mythical MIPS count as indicator of performance. 8bit RISC core can't outperform 16bit CISC processor easily and hardly has a chance against 32bit one. And you kind of did such silly comparison of 5A22@3.58 vs 14MHz 68EC020.

That's meaningless when Turrcian 2 AGA (with VGA 256 color artwork) needs Fast RAM to reduce A1200's shared bus problem. MIPS shows how many instructions are being completed per second.

Sure MIPS shows how many instructions are being completed. It doesn't show how much work is done... That's why ppl invented something called Dhrystone MIPS. It's not perfect but it at least shows something relevant to performance. Also - SNES needs coprocessors in titles which runs great on stock amiga 1200 so I find that argument amusing.

68020 has a tiny instruction cache and real-world game workload has data being fetched from memory.

Yet it can manipulate data of sound memory or graphics memory pretty easily while 5A22 cannot.

SNES has a Wolfenstein 3D port.
Mega Drive (68000@ 7.6 MHz) has a Wolfenstein 3D port with line skips like on A500's Dread/Grind.

[ Show youtube player ]
[ Show youtube player ]
Yeah...right.

A1200's CPU is almost 2X faster with 32-bit Fast RAM.

Sure, and if you drive your DMA and WRAM at 3.58MHz SNES is also faster, but somehow by default it goes to 2.68MHz iirc.

68000 requires 4 clock cycles to access RAM comparing only 2 cycles for 65C816.

And again - has 16x32b registers excluding PC and status register. Why you keep omitting the obvious limitations of basically 3 register 6502 architecture?

Again, there is a discrete memory bus for the CPU's system RAM and S-PPU's video RAM.

Entirely separate memory means CPU can't easily access PPU video RAM. I will shortly prove you that design was seriously flawed.

For proper acceleration, the CPU and GPU should remain in their lanes.

And they are, and they were with Amiga, only CPU exclusive RAM didn't come as standard for home amiga computers.

So what. The host CPU shouldn't be involved with S-PPU's graphics processing.

And that's why every enhancement chip had to use own RAM and do magic tricks to copy over results back to PPU. That's absolutely horrible design.

CPU's IO interference with the GPU's operations can disportationally degrade raster performance.

Sure, but that design costs less and both CPU and GPU have several levels of cache and uses zero copy memory so they don't starve.

AMD APUs can still support external memory controllers via the southbridge.

And it generally is unused from the moment it was invented (I think you refer to "sideport memory"). And again - PS3 design was kind of like Amiga 1200 with fast. CPU (Cell) was capable of supplementing relatively weak RSX when it comes to things like transformations and shaders. If you look at PC nowadays you'll see strong PCIe interface and IOMMUs. Guess why... Because GPUs are perfectly comfortable with having own, private memory? Not really, no.

[Cris1997XX]

Damn, seeing you all bite into hammer's bait is really pitiful...