@cdimauro

Quote:

cdimauro wrote: Quote: Tseng Labs ET4000AX has a 14.31818 MHz crystal oscillator with an ICS2494N video clock generator, which runs at 89.8 MHz. The clocks generated are 25.175 MHz (for VGA mode), 28.322 MHz (for VGA and CGA modes), 32.514 MHz (for EGA mode), and 40.0 MHz (for "Extended" modes). ET4000AX supports 50 to 70 Hz vertical refresh rates and screen resolutions up to 1024 x 768. 25 Mhz partly covers VGA mode. And? What's the point here?

Point is Hammer's favorite ISA bus VGA card was an excellent performer on the PC, so the Amiga should have been able to match it - according to him. But the Amiga chipset did a lot of things the ET4000 didn't - like sprites, copper, blitter, sound etc. - that made it far more complicated.

The ET4000AX achieves its high bus speed using 2 fifo buffers and parallel RAM banks. To get best speed it has to be fully populated with RAM. Using 256kx4 DRAMs (the largest it can handle) requires 8 chips, providing a total of 1MB RAM. Using a smaller number of bigger chips won't work, making the system cost higher than it could be.

Another issue is that the RAM is only accessible to the CPU in banks of 256k max. This can be a problem if you want to copy graphics data from one place to another, or if you want to use it as general purpose memory. The Amiga had shared memory with a linear address space, so it couldn't apply some of the tricks that the ET4000 (and other VGA chips) used.

The end result is - to get similar performance to the ET4000 would require extensive modifications to the Amiga chipset. Maintaining compatibility with OCS would be a big headache. It might be necessary to have dedicated Video RAM that can't be used for programs etc., which would raise the system cost dramatically and introduce more incompatibility.

I do think a simple fifo buffer could have been implemented to make CPU writes faster. This could be implemented on the CPU side, independently of the graphics chipset. However the need for this did not become apparent until textured-mapped 3D games appeared in 1994 (Doom et al).

Even today new CPU accelerator cards are not including write buffering. The PiStorm for example is much slower writing to ChipRAM than it should be. If it was so easy and Commodore was just too incompetent to do the job properly, you would think modern cards would have this feature, yet they don't. Does this mean today's Amiga hardware developers are also incompetent, or is it not as easy as it sounds? Being an armchair chip designer is easy, actually doing it not so much...

@bhabbott

Quote:

bhabbott wrote: @cdimauro Quote: cdimauro wrote: And? What's the point here? Point is Hammer's favorite ISA bus VGA card was an excellent performer on the PC, so the Amiga should have been able to match it - according to him. But the Amiga chipset did a lot of things the ET4000 didn't - like sprites, copper, blitter, sound etc. - that made it far more complicated. The ET4000AX achieves its high bus speed using 2 fifo buffers and parallel RAM banks. To get best speed it has to be fully populated with RAM. Using 256kx4 DRAMs (the largest it can handle) requires 8 chips, providing a total of 1MB RAM. Using a smaller number of bigger chips won't work, making the system cost higher than it could be. Another issue is that the RAM is only accessible to the CPU in banks of 256k max. This can be a problem if you want to copy graphics data from one place to another, or if you want to use it as general purpose memory. The Amiga had shared memory with a linear address space, so it couldn't apply some of the tricks that the ET4000 (and other VGA chips) used.

I think that you never searched for the ET4000 specs or, even better, opened its technical manual (Data Book), right? Because you reported a lot of completely wrong data.

Here are the more relevant specs for this graphics card:
- fully linear address space for accessing its memory (from 512kB to 4MB);
- hardware cursor (64x64@2 bits. 2 colors or transparency or invert colour);
- advances BitBlt blitter which worked with any pixel depth (1, 2, 4, 8, 15, 16, 24 and 32 bit) and formats (planar or packed);
- picture-in-picture;
- dual display mode with two independent linear addresses for the two displayed windows;
- up to 640x480@24 bit (interlaced) or up to 1024x768@16bit (interlaced) or up to 1024x768@8bit (non-interlaced) or up to 1280x1024@4bit (non-interlaced).

It doesn't have the sprites (besides one usually used for the mouse) and the Copper, but I think that it speaks for itself, right?

Yes, it has no sound, because... it was a graphics card. But on 1990 you had plenty of choices here as well for PCs, some much better than the Amiga.
Quote:

The end result is - to get similar performance to the ET4000 would require extensive modifications to the Amiga chipset. Maintaining compatibility with OCS would be a big headache. It might be necessary to have dedicated Video RAM that can't be used for programs etc., which would raise the system cost dramatically and introduce more incompatibility. I do think a simple fifo buffer could have been implemented to make CPU writes faster. This could be implemented on the CPU side, independently of the graphics chipset. However the need for this did not become apparent until textured-mapped 3D games appeared in 1994 (Doom et al). Even today new CPU accelerator cards are not including write buffering. The PiStorm for example is much slower writing to ChipRAM than it should be. If it was so easy and Commodore was just too incompetent to do the job properly, you would think modern cards would have this feature, yet they don't. Does this mean today's Amiga hardware developers are also incompetent, or is it not as easy as it sounds? Being an armchair chip designer is easy, actually doing it not so much...

We know that the Amiga chipset was too much obsolete on 1990 and it required a major redesign. The PCs were already very ahead, as you can see above.

However games didn't required high specs graphics card, and usually 320x200 was the resolutions used. Here an ECS like what I've proposed would have been very very competitive.

Last edited by cdimauro on 25-Nov-2023 at 07:01 AM.
Last edited by cdimauro on 25-Nov-2023 at 07:01 AM.

@cdimauro

Quote:

cdimauro wrote: Here an ECS like what I've proposed would have been very very competitive.

No it wouldn't (regardless of the exact specs you've dreamed up).

- any changes would mean "opening up" a design already at the limit of what the process could do. Hence even small changes took forever to implement (see how late ECS and AGA were).

- updating MOS was no option as the A500 had some volume but no margin, while Ax000 had the margin but no volume

- outsourcing was also no option as there were no margins where there was volume

So a A500++ in 1990 would have had the same fate as the A1200 in 92, not enough of an upgrade to justify the higher price for most of the buyers, hence selling in lower numbers and developers not being enticed to support it (see how few proper AGA games were released in 93).

So you are not discussing a possible past, but an impossible one. Unless you go further back to make C= position itself in a more upmarket and bleeding edge way.
A way that would have them not be desperate enough to have to buy Amiga in the 1st place.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Kronos

We can only wonder what Amiga had been without commode, CIAA/CIAB chips comes directly from the C64.

MOS was on other hand a large player in the semiconductor business, was rescued by Commodore, had they not, the C64 might never have existed at all. This how Commodore was able to make custom chips while Atari was not able to.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

MOS was rescued by C= the same way the "rescued" Amiga.

The bought it, made sure it ran and never invested a single ct beyond that.

Thats how C= operated, thats why the C64 and A500 was so cheap and successful and thats why the A1200 was so little so late....



As for "256 color low res games" on the Amiga:

An 8bit mode would have saturated the bus on an A500++ crippling the 68k to a standstill.

Make the bus run faster with better RAM chip -> complete redesign + higher BOM -> too expensive.
Add extra FastRam to give the CPU some breathing room -> too expensive
Make the bus wider -> you need an 68ec20 or better to pull that of -> too expensive (in 1990)

But that is only one part of the equation as those PC games did often require more compute (which the 68000 did not have even when running in FastRAM) and an HD (which C= never made standard).

So to really compete C= would have to replace(yes replace not just an optional higher spec) the A500 with an A1200-HD light (omitting 8bit modes for high res) and some FastRAM.
Even if they could have done that in 1990 the result would have costed about as much as decent "gaming" PC at that time, making it a non starter.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Kronos

Quote:

Kronos wrote: @cdimauro Quote: cdimauro wrote: Here an ECS like what I've proposed would have been very very competitive. No it wouldn't (regardless of the exact specs you've dreamed up).

I beg to differ (a lot!).
Quote:

- any changes would mean "opening up" a design already at the limit of what the process could do. Hence even small changes took forever to implement (see how late ECS and AGA were).

The proposed changes are minimal and inline with was possible (and was done) with the ECS chipset.
Quote:

- updating MOS was no option as the A500 had some volume but no margin, while Ax000 had the margin but no volume - outsourcing was also no option as there were no margins where there was volume

There was no need: wasn't MOS producing the ECS?
Quote:

So a A500++ in 1990 would have had the same fate as the A1200 in 92, not enough of an upgrade to justify the higher price for most of the buyers, hence selling in lower numbers and developers not being enticed to support it (see how few proper AGA games were released in 93).

With my proposal, ECS would have been much much better, performance-wise, than AGA.
Quote:

So you are not discussing a possible past, but an impossible one.

As I've said, it was possibile with the ECS technology.
Quote:

Unless you go further back to make C= position itself in a more upmarket and bleeding edge way. A way that would have them not be desperate enough to have to buy Amiga in the 1st place.

Markting is another issue which Commodore had, as we know. But I wasn't talking about it with on my articles.
Quote:

Kronos wrote: As for "256 color low res games" on the Amiga: An 8bit mode would have saturated the bus on an A500++ crippling the 68k to a standstill.

Not with my proposals.
Quote:

Make the bus run faster with better RAM chip -> complete redesign + higher BOM -> too expensive.

Here is evident that you haven't read the article: no, there was NO need for a complete redesign!

The changes were feasible using the same ECS technology.
Quote:

Add extra FastRam to give the CPU some breathing room -> too expensive

NOT needed!
Quote:

Make the bus wider -> you need an 68ec20 or better to pull that of -> too expensive (in 1990)

NOT needed!
Quote:

But that is only one part of the equation as those PC games did often require more compute (which the 68000 did not have even when running in FastRAM) and an HD (which C= never made standard).

If you mean 3D games using texture mapping, yes.

But it was still 1990, right? Do you remember that we were talking about THIS period of time (and not later)?
Quote:

So to really compete C= would have to replace(yes replace not just an optional higher spec) the A500 with an A1200-HD light (omitting 8bit modes for high res) and some FastRAM.

Which is much more expensive and, as I've already said, NOT needed (and much less powerful compared to my suggested ECS).
Quote:

Even if they could have done that in 1990 the result would have costed about as much as decent "gaming" PC at that time, making it a non starter.

Same as above: no! More expensive compared to a regular A500, yes (because you need a 16Mhz 68000 or a 12.5Mhz one clocked at 14Mhz. And 1MB of DRAM @ 14Mhz), but NOT to the same ballpark.

@Kronos

I remember always wonder if Amiga can have been clocked higher with active cooling, it was strange that there were no overclocking kits for it back then. You always had to insert a new accelerator if wanted more speed, and that was costly $$, the different CPU sockets was big issue.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@cdimauro

Your free to live in whatever world you wanna do, but in this one OCS was pushing the limits of what MOS could do and any chance no matter how minor took a lot of effort to get it fully working.

Thats why ECS was all we get after 5 years and AGA required C= to bite into the sour apple of outsourcing a chip.

8bit LowRes would saturate the 7MHz/16Bit bus just like 4Bit HighRes does.
Everything else would require a massive update to gain more bandwidth.


And no I was not talking bout 3D games as even an A1200HD wasn't capable of running them (beyond proof of concept, postal stamp sized screens).

Most Amiga games were simple 2D action titles were everything was preprogrammed and most AGA games were the same with slightly better GFX.

You may want to look what kinda "grown up" games existed for the OC in the late 80s and why so few of them ever got ported to Amiga.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@NutsAboutAmiga

OCing the Amiga meant OCing everything including the chipset.

Floppy going out of whack would be the 1st issue and soon after you would hit the wall of the chipset timing breaking down.

I think you could go as far as applying a 30MHz clock running the CPU a a whooping 7.5MHz but even that wasn't reliable.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Kronos

Perhaps you need to cut clock to the CIAA/B chips as they control input and output, and put counter on CPU clock signal, to scale it back down,

I guess you don’t want to run Paula at too high frequency as well at least not for audio playback.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

Everything needs to be on the same clock or

To be more precise, some parts run at 1/4 of the base clock (7.09 or 7.14MHz) and others at 1/8.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Kronos

I once fitted a 14 MHz 68000 in my A500, derived a 14Mhz signal from the two 7 MHz clocks
which were 1/4 cycle out of phase.
It worked well, but the E clock was twice the frequency and could not write to floppy or HDD
A590 normally.
However it did double the storage capacity on the floppy and corrupt the HDD.

Wol...

_________________
It is my conviction that killing under the cloak of war is nothing but an act of murder.~Albert Einstein

@Wol

Well it was still somewhat synced to the rest of the system, but actual performance gains on the CPU must have been minimal since the 68000 has noch cache and the RAM was still as slow as before.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

The 68EC020 was worth considering by 1990 even for the low end. Data Quest gives the ASP of all 68020 CPUs as $80 in 1990 and $55 in 1991 but a 68EC020@14MHz is cheaper than average and C= had a deep volume discount. The C= bankruptcy docs have a 68EC020 cost of $8 but not a specific date although this data is likely late 1992 or 1993 corresponding to other data. Using this data I can estimate the 68EC020 in 1990 would have cost C= $18-$27 (other circuitry for 32 bit would have increased but the address bus is the same size as the 68000). By 1991, I estimate the C= 68EC020 cost would have dropped to $12-$18. If the chipset could have been changed to allow twice as many memory access, this may have been the cheapest way to 4 times the chipset memory bandwidth (8 times if increasing to 32 bit in chipset which was worth considering too). Of course C= upper management loved the idea that the Amiga chipset offloaded the CPU so they could buy the cheapest possible CPU but that was a fatal mistake they realized too late.

We could estimate the base cost of color registers as being the same as SRAM just to hold the data although additional logic would be needed.

128 color registers 9,216 transistors (+96 color registers)
256 color registers 21,504 transistors (+224 color registers)

EHB mode with 128 color registers requires less than half the logic compared to 256 color registers. How absurd was it that C= couldn't afford 10k transistors on their 5000nm chip process? Even the poor Atari Jaguar in 1993 used 2 chips of 700k transistors and 600k transistors on a 500nm process. The competition was pushing 1 million transistor chips in 1993 and consoles with 3 million transistor chips were being designed in the mid 1990s.

https://www.atariarchives.org/cfn/09/03/01.php
https://websrv.cecs.uci.edu/~papers/mpr/MPR/ARTICLES/090704.pdf

3D chipsets using 1 million transistors were placed on a single chip for the desktop market circa 1993 too. C= not only could have had a single chip SoC 68k Amiga but they should have. The whole Amiga chipset in a single chip was likely possible in 1990. Combining chips saves transistors and pins!

There was not much discussion when bhabbott suggested synthesized sound for the Amiga. It wasn't just the desktop market which had it but also some consoles like the Sega Mega Drive/Genesis. C= had the much loved C64 SID chip but also the logic for it. The filter is analog but a digital filter that sounded acceptable could have been developed as several FPGA reproductions have done. I couldn't find any estimates of how many transistors the logic used but it would have been low at that time. If it was 10k transistors then that is cheap enough that stereo SID chips could have been included in the Amiga chipset. This would have appealed to C64 owners encouraging them to upgrade to the Amiga allowing the more expensive C128 and C65 development mistakes to be discontinued and not developed. C= also had access to the 6502/6510 logic which uses as few as 3,218 transistors and could have been used as an I/O processor (some consoles had a Z80 or 68000 on board for I/O). Would dual SIDs and a 6510 even have increased the chipset cost $1 in 1990? What percentage of C64 owners upgraded to the Amiga? How much did C= increase the price of the Amiga 2000 adding PC compatibility with all the ISA slots that weren't even connected?

Last edited by matthey on 27-Nov-2023 at 05:34 AM.

Quote:

Wol wrote: I once fitted a 14 MHz 68000 in my A500, derived a 14Mhz signal from the two 7 MHz clocks which were 1/4 cycle out of phase.

It wouldn't lead to much speed increase, memory access would still have to be the same speed only some internal calculations would have been accelerated.

There was a an 16MHz accelerator for the AtariST that had a tiny 16 word cache on board, this would have increased the speed by maybe 50%, it could have been something that would have made the A2000 more interesting but more incompatible.

Last edited by OneTimer1 on 25-Nov-2023 at 11:59 PM.

@matthey

Your are throwing around things from different timeframes and add some wishful thinking on top of that.

A $20 68ec20 in 1990 doesn't sound bad, but even without touching anything else that would have been a $50 price increase which just wasn't feasible for a low budget machine.

Whether other vendors were able to make bigger 500nm chips and have them run at higher clocks is irrelevant as MOS was running at it's limit with the chips how they were.

Integrating the CPU, even if C= could have done it (in house or not) they would still have to get a license for it.

The reality is that the A500 was the only Amiga selling in real numbers and that was mostly due to it getting cheaper and cheaper to the point where margin were ... marginal.... prohibiting any mayor upgrades, forcing it to get cheaper in order to still ship in numbers.....

The only real errors I see here is that given that AGA was done early both the A4000 and A1200 should have been 1991 releases and that the A1200 should have had a small 3.5" HD by default. Sure would have upped the prices but would have been real value.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

I like ECS. It was good enough.
Games on Amiga in ECS times was better than on pc.

@ppcamiga1

It was AGA that made Commodore bankrupt.
Escpecially lack of chunky pixels.

@NutsAboutAmiga

Quote:

NutsAboutAmiga wrote: I remember always wonder if Amiga can have been clocked higher with active cooling, it was strange that there were no overclocking kits for it back then. You always had to insert a new accelerator if wanted more speed, and that was costly $$, the different CPU sockets was big issue.

Not strange at all. It was the same with all machines of the day that had shared memory. The CPU had to run synchronously with the graphics controller in order to access memory without corrupting the screen display, and the chips used were almost always running close to the maximum speed possible. Overclocking the whole motherboard wouldn't be useful even if the chips could take it, because all the timing would be off. Overclocking was a mad idea that unsurprisingly didn't work.

You could put in a faster CPU with some circuitry to split clock domains and slow it down when accessing the motherboard. But to get a useful speedup you also needed faster RAM because the 68000 didn't have internal caches. So it was going to be expensive anyway. In fact there was no shortage of such solutions on the Amiga, they just cost more than the average user was willing to pay. Why? All commercial software was designed to work on a stock machine, and few applications would benefit much from the extra CPU speed. For most users the cost didn't justify the limited benefit.

In 1990 I started doing commercial software development on the Amiga. Assembling large programs was taking a while on the A1000, so I hankered for a faster machine. When the A3000 was announced I ordered one. 5 times faster, woo hoo! In the mean time I bought Hisoft Devpac as a replacement for Assempro (which was a nice integrated editor/assembler/debugger, but sometimes crashed which was annoying). I was shocked to find that Devpac assembled code 5 times faster on the A1000 than Assempro did. I didn't need the A3000, just more efficient software! NZ$7,200 dollars wasted!

Unfortunately my A1000 was stolen shortly after I received the A3000 (I hadn't even unpacked it, which was lucky because the burglars missed it), so I no longer had a stock Amiga. A lot of games wouldn't work properly on the A3000 due to the faster CPU. Since the burglars took all my software too I didn't bother playing games for a few years. But if I had the incompatibility would have been quite disappointing.

The Amiga had heaps of great games that naturally worked perfectly on a stock machine, so why did we need 'new' chips? For most stuff we didn't. What we did need was software that made best use of the Amiga's existing hardware. That was a bit harder than just banging out inefficient code and/or being over-ambitious and relying on faster hardware to make up for it, but once you went down that lazy path there was no end to it.

IMO the Amiga didn't need a radically improved chipset when ECS was introduced, except to permit easier porting of VGA games from the PC. At this time PC games were starting to switch to VGA, which was certainly a big improvement for PC gamers but not so much on the Amiga - which already had awesome graphics. In fact even in 1993 and 1994 - after AGA came out - over 300 OCS games were released per year (about one per day), some of which were quite stunning.

It wasn't until 1994 - when Doom came out - that 256 colors became 'essential' (though Doom would still have looked look pretty good in 64 or even 32 colors if tuned to it). But Doom also needed a very fast CPU so it was never going to be suitable for low-end Amigas - no matter what graphics chipset they had.

If the PC didn't have VGA and ever-increasing CPU speeds, nobody in 1990 would have been clamoring for a new chipset to stop the Amiga 'falling behind' in graphics capabilities. It was way better than every other popular home computer out there, so why change it? Only to assuage PC envy. Somehow it was unacceptable that the Amiga couldn't beat PCs which were 2-3 times the price (even though in many cases it could - just not in 256 colors).

Amiga fans bleat about Commodore not changing the Amiga's chipset for 7 years. But this is a similar time period that the hugely popular Sega Genesis / Mega drive and Sony PlayStation were produced for, as well as many other home computers. It was not unusual to keep the same hardware standard for so long - for good reason.

38 years later we are still exploring the capabilities of the OCS chipset, with games like Dread showing what it was capable of from the start. Imagine that game on the A1000 in 1985 - it would have blown our minds! Imagine if Commodore had insisted that the Amiga was capable of realistic texture-mapped 3D back then - we wouldn't have believed them! Yet the hardware was capable - it just took the right software to realize it.

Last edited by bhabbott on 26-Nov-2023 at 12:36 PM.

@bhabbott

Quote:

That was a bit harder than just banging out inefficient code and/or being over-ambitious and relying on faster hardware to make up for it, but once you went down that lazy path there was no end to it.

I think that’s a myth, not everyone can write in assembler, and not everyone has passions for looking at Guru Mediations all day long. As for quality of code, no know what they were really doing, I remember trying to use what learned in math class, regarding Pythagoras. I had no cluse about vectors how they work, how to manage 3 axes, I tried my best make some lines, look like a 3D objects. Putting a simple bob on the screen is 200% more annoying in assembler then in AMOS.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS