Individual Computers - New accelerator cards for A600 & A1200

[techguruuk]

Originally Posted by desiv

Does it do any of that if you remove the ACA1230?
Are you using an Indivision/SD/FF or just a DB23 to VGA/SCART cable or ?

How is the 1200 Power Supply? Is it an original 1200 PSU or an A500 PSU?

desiv

It does not happen without the ACA1230 present, no. I do have an Indivision, but it is not being used here (it is attached, but this is with the RGB output on scart from the Amiga).

It is an A500 PSU I use, but I have just ordered a Goliath to see if it helps.

Originally Posted by Schoenfeld

Heat is not an issue with a PGA processor @28MHz. However, A1200 motherboard issues are. Look back further in this thread for a description of how to correct the assembly errors: check E123C, E125C and if they are gone, the second "more complicated" mod, described about a year ago.

Jens

Thanks Jens. Yeah E123C and E125C are both gone. I'll have to see if I can find that other mod.

Should the 030 CPU be so hot that I cannot touch it though? It is VERY hot. The fact that the problems only appear after an hour or so of use, then go when it's been off and cooled down make me wonder if cooling issues are present?

[Mequa]

I wonder if a new accelerator board for classics featuring a fast FPGA-based 68060 clone (and perhaps FPGA-based 24-bit RTG) would be a better idea? It would probably have a nice niche market. Some may question its authenticity though.

[Mr B]

Softcores have already been argued about. I think we'll see em in a few years. Simply put, as performance gets better, and price drops it will become the only option, and quite honestly, you wont be able to tell the difference in use, by that time.
B!

[Schoenfeld]

Originally Posted by techguruuk

Thanks Jens. Yeah E123C and E125C are both gone. I'll have to see if I can find that other mod.

The "easy" mod is here:
http://eab.abime.net/showpost.php?p=...postcount=1359


The "more complicated one" is here:
http://eab.abime.net/showpost.php?p=...postcount=2065

Originally Posted by techguruuk

Should the 030 CPU be so hot that I cannot touch it though? It is VERY hot. The fact that the problems only appear after an hour or so of use, then go when it's been off and cooled down make me wonder if cooling issues are present?

PGA CPUs don't really have heat issues if they're operated within specifications. Since the 5V rail of the Amiga is only controlled by ageing power supplies, you might want to check how precise the 5V rail is gegulated. Some PSUs deliver over 5.5V on that rail, which results in more heat. I already had to fight with this topic on Micromys V3, where the right mouse button would not work when the Amiga was operated with over-voltage. You'd be surprised how many PSUs deliver too-high voltage.

In any case, I would not expect the CPU to be the problem. Like I wrote a number of times (in other places), Commodore didn't know all the details about their own chipsets, and it's still a trial-and-error job to get a specific motherboard up to date. I could have designed the new accelerators in a way that they are as slow as the old Blizzard boards on chipmem access, but that would have been pointless if mods become necessary on other occasions, such as using higher-resolution and higher-colour screenmodes.

Jens

[desiv]

Originally Posted by techguruuk

It does not happen without the ACA1230 present, no. I do have an Indivision, but it is not being used here (it is attached, but this is with the RGB output on scart from the Amiga).

It is an A500 PSU I use, but I have just ordered a Goliath to see if it helps.

Is this something that used to work well, and you just started having problems?

Is this ACA "new" to this Amiga.

Is the Indivision "new" to this Amiga.

I realize you're not using the Indivision but it is using power, so maybe that's related???

Anything else "new" recently.

Just throwing out ideas.. ;-)

Good Luck

desiv

[techguruuk]

Originally Posted by desiv

Is this something that used to work well, and you just started having problems?

Is this ACA "new" to this Amiga.

Is the Indivision "new" to this Amiga.

I realize you're not using the Indivision but it is using power, so maybe that's related???

Anything else "new" recently.

Just throwing out ideas.. ;-)

Good Luck

desiv

I've had the ACA1230 for about a year and the problem has been there before, but seems worse now. It has been worse since the Indivision was slipped over the Lisa Chip, but as I said, the Invision is not being used until I can get the config tool, but it could be adding extra heat to Lisa by covering it?

I have a Goliath PSU on order so we'll try that.

[techguruuk]

Originally Posted by Schoenfeld

The "easy" mod is here:
http://eab.abime.net/showpost.php?p=...postcount=1359


The "more complicated one" is here:
http://eab.abime.net/showpost.php?p=...postcount=2065



PGA CPUs don't really have heat issues if they're operated within specifications. Since the 5V rail of the Amiga is only controlled by ageing power supplies, you might want to check how precise the 5V rail is gegulated. Some PSUs deliver over 5.5V on that rail, which results in more heat. I already had to fight with this topic on Micromys V3, where the right mouse button would not work when the Amiga was operated with over-voltage. You'd be surprised how many PSUs deliver too-high voltage.

In any case, I would not expect the CPU to be the problem. Like I wrote a number of times (in other places), Commodore didn't know all the details about their own chipsets, and it's still a trial-and-error job to get a specific motherboard up to date. I could have designed the new accelerators in a way that they are as slow as the old Blizzard boards on chipmem access, but that would have been pointless if mods become necessary on other occasions, such as using higher-resolution and higher-colour screenmodes.

Jens

Thanks for the reply Jens. The easy mod is not needed as those components are already missing from my 1D4 board, but the more "complex" fix looks a bit beyond my electronics capabilities, I fear I'd end up damaging my board.

Could the issue be averted by adding cooling to Lisa then if hot Lisa shows the problem?

[Schoenfeld]

Originally Posted by techguruuk

Thanks for the reply Jens. The easy mod is not needed as those components are already missing from my 1D4 board, but the more "complex" fix looks a bit beyond my electronics capabilities, I fear I'd end up damaging my board.

AmigaKit provides the service of applying the patches.

Originally Posted by techguruuk

Could the issue be averted by adding cooling to Alice then if hot Alice shows the problem?

Not all issues are heat issues. Also, you haven't seriously places Indivision AGA MK2 over Alice, have you? The unit must be placed on Lisa!

Jens

[techguruuk]

Originally Posted by Schoenfeld

AmigaKit provides the service of applying the patches.


Not all issues are heat issues. Also, you haven't seriously places Indivision AGA MK2 over Alice, have you? The unit must be placed on Lisa!

Jens

Whoops, sorry I meant Lisa chip, my mistake. Would adding a fan to Lisa help do you think? The problems only appear when the system has been on for a while.

I will contact Amigakit too then.

[alenppc]

Ok, so finally yesterday I have received the ACA-1231 from Vesalia after two months waiting! (Also the first shipment from them that got taxed at the customs... $40, dammit. )

Wow, what a beast! This card is seriously fast. I did the AIBB tests, and it beats the stock A4000 in many of them (obviously excluding those where the 040's excellent FPU made the difference).

Here's a curiosity. The PCMCIA CNet ethernet card now works only when using the standard 020 driver, while the "Turboio" driver (which made a lot of difference with my previous M-Tec 1230), freezes the system. Not a big deal however. The regular driver works fine and I don't mind that it achieves a slightly lower throughput.

Other than that I have not experienced any other freezes with my system which leads me to believe the mobo timing is correct.

A question for Jens. The $C0000 mapped memory is 32-bit (obviously) but according to AIBB it is just as fast as the rest of the expansion RAM (4.0 index!). I thought this was in Agnus' address space, so wouldn't the access have to be shared with the the custom chips, thus slowing it to the same speed as the Chip RAM?

What is the main reason all of the memory is only added when using Kick 3.1 but not 3.0? Is there a new function that is part of 3.1 but absent in 3.0?

I can't even begin to imagine what kind of a throughput could Jens squeeze out of a 68040 with this kind of design...

EDIT: Forgot to mention. Despite having inserted the battery, the RTC doesn't seem to work. A "setclock load" gives me a 1978 date. Tried also "setclock reset", date changing, "setclock save", then "setclock load". This leads to a correct date while the system is on, but it resets again when the system is off. This makes me think that I didn't insert the battery properly. Does it need to be pushed really hard in? It's being held in place by the plastic tabs now... not sure if that's enough.

[Bamiga2002]

Originally Posted by alenppc

EDIT: Forgot to mention. Despite having inserted the battery, the RTC doesn't seem to work. A "setclock load" gives me a 1978 date. Tried also "setclock reset", date changing, "setclock save", then "setclock load". This leads to a correct date while the system is on, but it resets again when the system is off. This makes me think that I didn't insert the battery properly. Does it need to be pushed really hard in? It's being held in place by the plastic tabs now... not sure if that's enough.

Hey I got this same problem too! How should the battery be inserted? Some dummy instructions please

[Schoenfeld]

Originally Posted by alenppc

A question for Jens. The $C0000 mapped memory is 32-bit (obviously) but according to AIBB it is just as fast as the rest of the expansion RAM (4.0 index!). I thought this was in Agnus' address space, so wouldn't the access have to be shared with the the custom chips, thus slowing it to the same speed as the Chip RAM?

The CPU+RAM subsystem is totally separated from the Amiga. I can take any address space I want and map it to fastmem in my SD-Ram controller. While for some addresses (chipmem and IO) it would not make sense to "steal" it from the rest of the system, it makes perfect sense for ram-only and rom-only space, such as kickstart and $c0 mem space.

The trick is to stay on the accelerator with that specific access. In layman's terms: I'm just not telling the Amiga that I'm accessing $c0 space, but do it silently in my own universe while the chipset minds it's own business. The beauty of the Amiga architecture.

Originally Posted by alenppc

What is the main reason all of the memory is only added when using Kick 3.1 but not 3.0? Is there a new function that is part of 3.1 but absent in 3.0?

Kick3.0 adds memory in $c0 and chipmem (obviously), plus any zorro cards. Kick3.1 has a new routine that looks for memory starting at $0800.0000 in 1MByte-increments.

I am mapping $0be0.0000 to $0bef.ffff to $00c0.0000, so there's a 1MByte gap that keeps kickstart from adding memory after 62MBytes.

As a result, memory at $0bf0.0000 to $0bff.ffff will be usable if you do NOT use Maprom. You can use the addmem command, should you have an application that requires 1MByte more. However, be careful not to use the Maprom option after adding that mem to the free memory pool, because ACAtune does not check if you've added that mem.

Originally Posted by alenppc

EDIT: Forgot to mention. Despite having inserted the battery, the RTC doesn't seem to work. A "setclock load" gives me a 1978 date. Tried also "setclock reset", date changing, "setclock save", then "setclock load". This leads to a correct date while the system is on, but it resets again when the system is off. This makes me think that I didn't insert the battery properly. Does it need to be pushed really hard in? It's being held in place by the plastic tabs now... not sure if that's enough.

It should be enough, but I've had a few support cases where the battery holder's contacts had to be cleaned, for example with a Q-tip and alcohol. After inserting the battery, you should be able to measure about 3V at the two solder contacts of the battery holder on the bottom side of the accelerator.

Originally Posted by Bamiga2002

Hey I got this same problem too! How should the battery be inserted? Some dummy instructions please

"how" is described in the manual: With the "plus" side up.

Jens

[fitzsteve]

Originally Posted by alenppc

Here's a curiosity. The PCMCIA CNet ethernet card now works only when using the standard 020 driver, while the "Turboio" driver (which made a lot of difference with my previous M-Tec 1230), freezes the system. Not a big deal however. The regular driver works fine and I don't mind that it achieves a slightly lower throughput.

Maybe it requires a FPU which the ACA does not have? You will find a few apps like this where you need to use the non FPU varients like AWEB or IBrowse

[alenppc]

Originally Posted by Schoenfeld

The trick is to stay on the accelerator with that specific access. In layman's terms: I'm just not telling the Amiga that I'm accessing $c0 space, but do it silently in my own universe while the chipset minds it's own business. The beauty of the Amiga architecture.

Aha! Awesome. Thanks for the insider info.

Originally Posted by Schoenfeld

Kick3.0 adds memory in $c0 and chipmem (obviously), plus any zorro cards. Kick3.1 has a new routine that looks for memory starting at $0800.0000 in 1MByte-increments.

Excellent, thanks. Most A1200 accelerators identify themselves as Zorro cards, which is how, I suspect, the memory gets added to the system by any kickstart. Did you decide not to use this technique for cost reasons or does it have other benefits?

Regarding battery, I am having the same issue with the A604. I am gonna look into this in the coming days.

Originally Posted by fitzsteve

Maybe it requires a FPU which the ACA does not have? You will find a few apps like this where you need to use the non FPU varients like AWEB or IBrowse

Actually CNet "turboio" mode is a hack in terms of PCMCIA access, it's documented in the readme, so I guess it disagrees with this aritecture. The standard version runs fine, however.

[Schoenfeld]

Originally Posted by alenppc

Excellent, thanks. Most A1200 accelerators identify themselves as Zorro cards, which is how, I suspect, the memory gets added to the system by any kickstart. Did you decide not to use this technique for cost reasons or does it have other benefits?

Most previous accelerators have SIMM slots where the user can add different sizes of memory, and that needs to be identified. The Zorro-autoconfig that you see on many other accelerators is not a memory card, but a ROM that runs a bit of code with memory-test/add-routines. Since my 1200 accelerators always have the same size memory, I found it to be a waste to use a ROM, and I also saved a bunch of pins on the CPLD by not doing autoconfig.

The A600 accel is slightly different, because it must be compatible with Kick2.0. I'm avoiding the ROM, but still give this thing 32-bit autoconfig-mem by pretending to be a Z3 memory card. It's almost like the Kick3.1 accelerator-mem routine: Looks for memory and adds whatever it finds in 1-MByte increments. I could have done the same on the 1200 accelerator, but that would have cost me more pins on the CPLD, thus requiring a larger, more expensive chip. I know it looks like a compromise, but the trick with the same-speed-memory at $c0 will locate any OS stuff in true fastmem, so you don't have any drawback under kick3.0. It's just a different approach: Fewer components, less power consumption, same (or even better) effect:

You'll find that many OCS/ECS programs that were previously incompatible with the A1200, will now work with the ACA1231. That's because some of those old programs expect memory at $c0 without checking if it's actually there. It was *the* standard address for the 512k expansion of the A500, and the majority of the market back in the days was A500 users with a trapdoor mem expansion.

Jens

[alenppc]

Again, thanks for replies Jens, very appreciated.

I have solved the battery issue with both the 1231 and A604. Measuring the voltage at the pins of the battery holder, there was indeed none. So I experimented taking out and replacing the battery a couple of times, cleaning the holder with 99% alcohol and so on. It seems that the battery should be put into the holder VERY lightly (so that it feels very lose and almost falls out). This will ensure contact with the + tab, it seems that's the troublemaker. At this point the RTC works just fine.

A few more questions, if I may. How did you get the necessary signals from the A600 trapdoor slot for Indivision ECS to work? This is really bewildering to me.

How did you manage to find A1200 and A600 trapdoor slot connectors? I thought they were quite hard to find nowadays.

Any chances of a production rerun of ACA630?

Thanks

[Schoenfeld]

This is drifting slightly off-topic, so just very short answers.

Originally Posted by alenppc

A few more questions, if I may. How did you get the necessary signals from the A600 trapdoor slot for Indivision ECS to work? This is really bewildering to me.

The trapdoor slot has the full chip bus, including register address lines. Indivision ECS does NOT watch the Denise output, but sniffs the Chip bus and feeds it's own implementation of Denise with that data. The output picture is pushed into the framebuffer, and then the flickerfixer does it's thing.

Originally Posted by alenppc

How did you manage to find A1200 and A600 trapdoor slot connectors? I thought they were quite hard to find nowadays.

I have my own tooling, so I can make any quantity and almost any number of pins in that Raster.

Originally Posted by alenppc

Any chances of a production rerun of ACA630?

Nope, the ACA630 will not be produced any more. I do have a lower-cost, lower-performance design for the A600 in the CAD system, but need to find the time to give that a test run. I hope that this new card will be available for this year's xmas business. I want this to be cheap enough to compete with bare fastmem expansions for the A600.

Jens

[Methanoid]

Originally Posted by Schoenfeld

Nope, the ACA630 will not be produced any more. I do have a lower-cost, lower-performance design for the A600 in the CAD system, but need to find the time to give that a test run. I hope that this new card will be available for this year's xmas business. I want this to be cheap enough to compete with bare fastmem expansions for the A600.

Jens

More info please... now you have wetted appetites.. Sounds like a 68020 to me.

What I would like to see is some cheap networking solutions for Amiga.

[ptyerman]

What I would like to see is some cheap networking solutions for Amiga.

Networking's fine (and cheap) if you have a A600/1200, I know what you mean for big box miggys though!

[Schoenfeld]

Originally Posted by Methanoid

More info please... now you have wetted appetites.. Sounds like a 68020 to me.

Correct. 68EC020@16.6MHz with all the 0-waitstate fastmem the thing can address, plus Maprom - that is the plan. However, that's just a theoretical thing in my CAD system.

Jens