Indivision AGA MK3 for A1200

Quote from Jens

We now have reports from a few customers about these screen-dropouts. I'm trying to find what the host computers have in common - you already mentiones your board revision, but that's not enough to really compare the boards. If you can, please list the chip revisions, or even better: Take a picture of the motherboard, ideally from both sides, so we can also see if there are timing mods done on the bottom side of the board.

Please see if the "CCK pull-up" and/or "CCK capacitance" options in the "advanced" menu changes this behaviour. The current theory is that a register access is not working as it should, and access timing is counted along the CCK clock.

Finally some time to try version 1.5 of the firmware/software and to open the amiga. First thing i saw was a factory green patch wire... totally forgot i have put another (recapped) rev 1d1 board in just before the pandamic, any how. I made some images, of the IC's,

I try to make a full image of the board and other components after i have removed everything thats in the way like the rapid road.

On the rev 1d1 i had the folowing issues:

The demo JezesOnEs from LSD is guaranteed to give a lot of black screens, so very easy to reproduce!

The demo Hardwired gave an interlaced effect on the first part displaying the white big text and some blackouts.

Ruff and tumble gave a blackscreen 50% of the time just after the score count af a level.

In work bench, fast switching menus can give blackscreens (not always, hard to reproduce)

Minimizing or scrolling big application windows (also folders) gave black screens (not always, hard to reproduce)

Some times a black screen with just doing nothing on the workbench.

No matter of the setting in the advanced menu, i tried all combinations in version 1.4 and version 1.5 of the indivision software. Results stayed the same.

On the Rev 1d4:

I did the same tests but only with version 1.5 of the indi software. Till now, none of the above happend with the exception some blackscreens in the demo hardwired. but since there is also some other screencorruption i assume this is a software issue.

So it seems the combination Rev 1d1 and IndiMK3 gives the issues. Since you said that with some demo's it can be a timing issue from the software, i didnt pay any more attention to this. But since this is running perfect on my other machine,i have to assume this is not the case.

Oh, and on both boards State of the art is running perfect .

Spoke a bit too soon, after posting and changing my WB screen to 256 colors (was in 16 color), i noticed screenblackouts on the workbench again. But only at the workbench.

What i did :

Open a folder full of whdload games, almost full screensize and then grab the slider button to scroll through the folder. At this point i can reproduce it everytime.

But, when i reduce the color depth to 64 colors or less, this doesnt seem to happen as far as i saw till now. Then i turned back to 256 colors, and doing the same, and black screens again. i changed the color depth about then times to double check.

I also tried some demo's again, and no issues so far. but these are OCS/ECS demo's...

So it's a bit different from the rev 1d1 board, this board almost always give this issue.

Also tried this with CCKline pull up and CCLLine capacitance on/of (4 combinations). Will try PCLK another time .. getting too late now .

Used Amiga mode is hires laced. Normal hires gives the same result. After some more time, closing screens gives black screens too and from then on it gets worse when using 7 or 8 bitplanes.

It seems color depth playes a role here, but after turning the machine of for a minute or ten, i can't reproduce till it heated up again.

before i swapped the indi, i made an image.. just in case this would happen.

THanks for the detaild report. PHotos fromt he top of your boards don't show anything unusual, so my next question is to take a picture of the bottom of the baords, showing the capacitors E123C and E125C: They are in a group of components that you find under Alice - please take a picture of the whole group. Also, please use more light - ideally sunlight, as this will improve picture quality big time.

Quote from Jens

THanks for the detaild report. PHotos fromt he top of your boards don't show anything unusual, so my next question is to take a picture of the bottom of the baords, showing the capacitors E123C and E125C: They are in a group of components that you find under Alice - please take a picture of the whole group. Also, please use more light - ideally sunlight, as this will improve picture quality big time.

Today i got home before sunset for a change.. any how, i hope the images are clear enough.. The are from my rev 1d1 board. Next post wel be from the 1d4 board.

I took one extra foto of alice, the resistor (?not shure in englisch for the name) doesnt seemt e be there from factory.

I have run this board with : ACA1233n @40Mhz and, IDE speeder at the maximum speed and Rapidroad USB ,before the Indi MKIII there was an MKII installed. No crashes or any other issues. Before the new CA-PSU a a500 4.5 amp psu was used.

I also recheckt my findings on the 1d4 board. Results where the same.

The first picture in your posting shows that E123C and E125C are installed. Please remove them - they can only cause trouble, no matter what accelerator you are using.

Quote from Jens

The first picture in your posting shows that E123C and E125C are installed. Please remove them - they can only cause trouble, no matter what accelerator you are using.

Couldn't ask a hardware engenire at work (were making medical hard/software, too busy at the moment). So i managed to find a tool to remove SMD resistors my self.

The amiga servived, its running now for 30 minutes the demo "Jezus on Es" with no black screens. And this was a garanty on this machine to reproduce black screens.

So the timing fix is not only for accelerators i gues (?).

Till now i didnt have any issue without this fix. No lockups, crashes or what so ever. And since i dont what to fix what doesnt seem to be broken i didnt feel the need to apply it.

Afcource 30 minutes testing is to short, the amiga is hardly warmed up. I wil do some more testing for longer time with everything i have tested before and let you know.

Quote from mbruines197703

So the timing fix is not only for accelerators i gues (?).

Absolutely not, you're right. There is a memo from Commodore published on amiga-wiki that essentially says "remove E123C and E125C for any version of Budgie we have ever made".

Still, 30 minutes of operation may be too short to call it a successful fix. The weekend just started

Quote from Jens

Absolutely not, you're right. There is a memo from Commodore published on amiga-wiki that essentially says "remove E123C and E125C for any version of Budgie we have ever made".

Still, 30 minutes of operation may be too short to call it a successful fix. The weekend just started

Well so it seemed... I was testing using my ACA1233@40 (from 2014).

now i have put in my 1233n@40Mhz, no fpu with IDE speeder, and the black screens where instand back.

Tried using without the IDE speeder too.

When using the 1233 it seems ok, but using the 1233n. Afcource i have swapped the acceletors several times to confirm.

I hope there is a fix for this too, i realy like the speeder and the possibility to turn off to play some older games. For now i use the aca1233, dont want to swap the Indi too much. So i jumper will have to do for now to turn of the card when needed..

Now that's an interesting observation - and I don't have any explanation for this, as the two cards are mostly equal. The ACA1233n does consume a bit more power due to it's bigger and faster CPLD, but that should not affect any timing on the board (unless the power supply is not up to the task).

Quote from Jens

(unless the power supply is not up to the task).

I think we can rule out the PSU, i'm using the CA-PSU. The only PC PSU's i have are in an actual PC where they belong.

So i think its best to test for longer time with the ACA1233. This way i can verify if this is is also true after an hour or 2.

There is another tiny difference, i have an RTC on the 1233n, not on the 1233. So i will check if this gives any difference without RTC.

Just a double check, the 1d4 needs the same timing fix? If so then the next steps can be to test this with this board to, but first without the timing fix, maybe later with timing fix.

Edit :

I just checked my other ACA1233n (@40Mhz with FPU, no IDE speeder), also gives black screens

RTC is ruled out, no diference

I just checked the schematics of the old ACA1233 vs. the ACA1233n, but there is no difference that would explain the different behaviour you're describing. However, I'll send an ACA1233n to Peter, so he has the same setup as you have - let's see if he can reproduce it.

Quote from Jens

I just checked the schematics of the old ACA1233 vs. the ACA1233n, but there is no difference that would explain the different behaviour you're describing. However, I'll send an ACA1233n to Peter, so he has the same setup as you have - let's see if he can reproduce it.

Well, i did some more testing last night on my 1d1 board, maybe this gives some more info. I still have to recheck to see if i can replicate all results again

What i have checked with the demo Jezes on Es wich is garantied to give the most black screens:

- Cold amiga, cold aca1233 : Ok

-Warm amiga cold aca1333 : Ok

-warm amiga, warm aca1233, atleast running 60 minutes : Not ok!

-Cold amiga, cold aca 1233n : Not ok

-Warm amiga (30 minuts+) + cold ACA1233n : not ok

-Warm amiga (30 minuts+) + warm ACA1233n : Ok

-Cold amiga, ACA1233 off with fastram : Ok

-Warm amiga, ACA1233 off with fastram : Ok

But i didnt have any sudden blackscreens in workbech. I could not replicate the issue when scrolling a big window, but this was on my 1d4 board (also no time fix).

At 4:30 i didnt feel like playing "ruff and tumble" for some time so didnt test the score count to see if this has improved

So the removal of the resistors did make a big improvement on the 1d1 board.

For the 1d4 board revision, do i need to apply the same fix? (if any extra needs to be done i dont dare to do this myself)

Quote from mbruines197703

For the 1d4 board revision, do i need to apply the same fix?

The Commodore memo is pretty clear that no matter what Budgie revision you have, the capacitors E123C and E125C need to go.

Ok. just wanted to be 101% sure. Some places they tell some other things.

I have the "tweezer" ready to remove them.

Quote from Jens

The Commodore memo is pretty clear that no matter what Budgie revision you have, the capacitors E123C and E125C need to go.

I have removed them and retested with the 1d4 board. It didnt make any difference. The warmer the machine gets, the more black screens in workbench on a 256 color screen when scrolling or opening/closing big screen area's, but also at seemingly random.

So i have put everything back on the 1d1 board. On the 1d1 when cold there are some issues in for example Jezus on es, but not games like Turrican or in workbench, and the black screens seem to disappear when the machine is warmed up complete (including the aca1233n).

My conclusion till now is that the 1d1 gets less issues when warmed up and the 1d4 board gets more issues when warmed up.

Did you also try the different settings for CCK capacitance and/or pull-up resistor? These are different for different combinations of Alice, Budgie and Lisa. I do see that you've done a lot of hardware swapping, but you didn't continue with these settings, which are much more promising.

Quote from Jens

Did you also try the different settings for CCK capacitance and/or pull-up resistor? These are different for different combinations of Alice, Budgie and Lisa. I do see that you've done a lot of hardware swapping, but you didn't continue with these settings, which are much more promising.

Re tried all the settings on my 1d1 board, It seems to help for the 'sparkling pixel' issue in super hires when warmed up, but none of the settings help to get rid of the black screens.

On this board i can only reproduce this with Ruff and Tumble (At coin time, fadein/out) Jezus on es and HardWired when the system is cold.

Peter is back from holidays this week. We have "product mantenance" on the agenda for the whole week - both Chameleon and Indivision AGA MK3. There may be a thing or two we can do on the register access timing, which is most likely what's failing every once in a while.

Quote from mbruines197703

On the rev 1d1 i had the folowing issues:

The demo JezesOnEs from LSD is guaranteed to give a lot of black screens, so very easy to reproduce!

The demo Hardwired gave an interlaced effect on the first part displaying the white big text and some blackouts.

I may be missing something from earlier posts in this thread, but I tested on an HDM! TV, an unexpanded A1200, firmware 20200824. I selected 'original chipset' and 'disable CPU caches' in the boot menu. The VGA mode selected was 'EDID 1280x720@50', audio mode 96kHz.

I sat about 15 minutes into the Jesus on E's (before it crashed). As far as I can tell, the show was totally flawless visually and sound-wise. I didn't get black screens and I didn't see the OSD.

How did you get Hardwired to work on an A1200? It says 'this demo don't like axel' on CPUs > 68000. (I expected this, it is what I remember.)

Timm, it appears to be board- and accelerator specific. mbruines reported that the old ACA1233 works fine, while the ACA1233n doesn't, and it's worse on one main board than on another. If you want to replicate the problem, you should use the ACA1233n-55. IHave sent one of these accelerators to Peter as well, hoping that the problem shows on his machine.