A query on "Errors in memory testing" (ACA 1234 + ACA 500 Plus)

I’m with you… my machines are rev 5 with ace2b’s

The firmware hot fix let me play some whdload games that did not work prior to it. I posted the specific ones if you feel like … I think Gods was one of them

Quote from ljmarent

The firmware hot fix let me play some whdload games that did not work prior to it. I posted the specific ones if you feel like … I think Gods was one of them

Ah right - unfortunately I didn't test them before flashing the hotfix. So, there might indeed be some improvement there.

Quote from kot_behemot53

Maybe it's a silly idea: I have 2 MB chip RAM (thanks to rev 8.1A mobo with fatter Agnus and appropriate RAM expansion from icomp) - could it have anything to do with my problems? It's not a "typical A500" config after all.

This is equivalent to the A500+ or an A500 with ACE2b. We could play a little with the chip ram acceleration that the ACA1234 uses - may<be the 8375 Agnus doesn't like that, although it never showed any problems here during testing. I could make a special CPLD version that will only activate this acceleration on A1200s, but not on ACA500plus (although that's where it's needed the most).

Quote from Jens

This is equivalent to the A500+ or an A500 with ACE2b. We could play a little with the chip ram acceleration that the ACA1234 uses - may<be the 8375 Agnus doesn't like that, although it never showed any problems here during testing. I could make a special CPLD version that will only activate this acceleration on A1200s, but not on ACA500plus (although that's where it's needed the most).

I happen to have a second A500, "standard" fat Agnus, rev 6A. It's disassembled right now, but I'll put it back together and try the ACA500Plus + ACA1234 combo on her for more insight.

Quote from ljmarent

did you patch the CPLD ?

Because that bug still affects 1200 users until patched.

Not yet. I just wanted to confirm that the problem, even without CPLD-patch, seems restricted to the combination ACA1234 and ACA500+.

Better for Jens to answer of course, but your ACA1234 on A1200 without the CPLD patch is affected still by memory corruption.

if you mean to ask about the WHDLoad and Amiga Test Kit problems that I’ve brought up more recently in this thread, then those do seem specific to the ACA500+ & ACA1234 combination

I am a little confused here reading about all this:-) I have downloaded the latest firmware and clicked on the icon for update. No errors occured. When I test the memory just as in previous comment's I get the exact same memory error messages as everyone else. I do not seem to have the same problem with software as the rest. Can I trust that it's upgraded correctly? And will the CPLD update sort the rest?

A1200 with the 1234.

thank you

The firmware upgrade procedure is now painless, quick and not very dangerous, and if it says "No errors occured." you can be sure that at least no additional harm is done. I cannot totally rule out (I'm confused too) that perhaps more than just one issue is at play here. But the general answer is: Yes, the CPLD update is supposed to sort out the rest. The hotfix blocks the offending 1mb of memory, that's all. We have tested this against MBRTest-2, but I did this only on an ACA1234 in an A1200.

I’ve been confined to bed for a few days, so I’ve not been running any additional experiments. Is it safe to assume that someone at iComp has repeated the ACA500+ & ACA1234 experiment with WHDLoad software and the JP0 jumper moved in and out of 25MHz safe mode. And has seen the errors stop happening in 25MHz safe mode ?

I’ve posted screen shots of the errors in an earlier note in this thread.

Another case was opened in the German part of this forum (also on the ACA500plus), where the mere change from 33MHz to 25MHz has solved problems, so at this point, the software appears to be not the culprit, and we're looking for something that works better at 25MHz than at higher speeds. I haven't been able to reproduce any of this.

Interesting...I was at 33MHz

Then the jumper dropped me to 25MHz

Glad to hear it's still being chased

Quote from ljmarent

Final data point for the evening, I need to rest too

I pulled the ACA500+ and ACA1234 out and popped a different 68030 50MHz accelerator into my computer.

AND...All of my WHDLoad games work.

I even tried my backup machine with the 'other' accelerator.
Mean Well RT-50B PSU, No shielding, sitting on a wooden table
All of my WHDLoad games work....

Did not make me a better player...but Agony plays just fine

So does Supremacy

I didn't find one that didn't work...

I think I can rule out that:

- WHDLoad doesn't work with 68030 as it is setup.

- The games might have been corrupt.

- My setup of the games could have been wrong.

Display More

Do you live in a self storage container? We see more reflection than anything else.

someone using an a1200 can tell me if the upgdrage fix somes demos issues?

I mean,i tried 2 demos,nexus 7 and wit premium

Both are lagging, artefacts on somes moving graphics,3d slower tant a1200 stock without aca1234

Thanks for your answer

Quote from shan2752

Do you live in a self storage container? We see more reflection than anything else.

I was in the basement lab, half storage half test gear on the walls reflected in the pictures,

Those were shown only as proof the games now loaded. You should not need more from them than that.

It’s a 46” monitor with a plexiglass shield in front so there’s going to be reflection. Just happens to be where the stock A500 lives. 😀

My basement lab is allowed to be messy and crafty

So, a follow-up on my tests. I reassembled my other Amiga 500 (rev 6A + 1 MB trapdoor expansion), connected ACA500Plus with ACA1234 to it and again checked:

- MBRTest-2,
- AmigaTestKit Memory Test,

- Supremacy,

- Unreal,

- Agony.

As mentioned before, I have the firmware patch installed.

Previously, on my A500 rev 8.A1 with 2 MB chip RAM, all these failed on me (except Agony which worked) with the ACA1234 and worked correctly without it (with just ACA500Plus or ACA500Plus with ACA1221lc).

With the other Amiga (rev 6A, 1 MB chip RAM) the results are the same - everything fails at some point in similar or the same ways. Interestingly, with this config Agony failed too - maybe I was lucky previously because these fails happen at a bit random moments (although typically quite early). The only 100% consistent failure profile is the MBRTest-2 which always freezes at exactly 40080000 - 4008ffff. Others sometimes result in a freeze, sometimes in some sort of exception (various) and at slightly differing moments. Sample exceptions are in the attachments.

At least it looks like it's ruled out that "Fatter Agnus" or 2 MB Chip RAM contributed to the problem.

So I probably need to continue to check the interface between ACA1234 and ACA500plus - try to find a method to replicate it here to capture what's happening on the logic analyzer.

Those of you who experience the problems on ACA500plus, do you also have an A1200 to test the card with? I mean, MBRtest-2 failing at 0x4008.0000 is not chip ram, but fastmem, which is all within the card.

Quote from Infirmier

someone using an a1200 can tell me if the upgdrage fix somes demos issues?

I mean,i tried 2 demos,nexus 7 and wit premium

Both are lagging, artefacts on somes moving graphics,3d slower tant a1200 stock without aca1234

This is expected behaviour and will not be changed. I've explained this in the German part of this forum: It's because these demos put code into chip ram, and expect that caches of the CPU will take care of the slow access speed.

However, switching caches on in chip ram is not allower, because there is DMA going on that may change memory contents without the CPU knowing about it. On modern systems, caches are invalidated when a DMA is happening, but there is no method on the Amiga, and also none on the 68030 processor that would even allow me to implement such a selective cache-invalidation. I therefore decided to switch off all caches on chip ram with my 68030 accelerators.

I did verify this by making a temporary CPLD firmware that allows caches on chip ram. Nexus7 works without lag then, but you must not allow data caches, otherwise WB does not work properly. I therefore decided to not publish this firmware.

Demos are artwork, and especially Nexus7 (released 1994) was released at a time when the rules specified "bare A1200", which means 68ec020 CPU and 2M chip ram only. You can achieve this with the ACA1234 by switching the card off - the demo will then run without trouble. In 68030 mode, I don't want to allow cache on chip ram, as this will introduce more problems that it will solve. Instead, I encourage demo coders to use fastmem for code when available - fastmem is not only cache-able, but it's significantly faster with it's burst capability.

Thanks for your explications jens

But it does mean that games would bugs too?

If demos are art works,like games too😉, games were designed for bare a500 or a1200?

Best choice to run it all is to use whdload?