Guru ramlib 80000004 when using Zorro cards. Hardware issue?

[grelbfarlk]

Originally Posted by jbenam

Both my Mediator and Prometheus exhibit the same symptoms.

I am just testing with the Voodoo 3, right now. If I just use RTG, everything is fine. If I use VoodooMemOS, and I actually try to use it as Fast (executing code from it) the ramlib error happens.

I first became aware of this issue when testing the Sonnet library with my Killer M1. It allocates RAM correctly, then probably tries to execute something from it and it fails. Still, I can reproduce the same exact issue with my Prometheus (it fails when it loads the Prometheus driver) and I *think* also with my A2090. Sadly I don't have an A2091 otherwise I could put some RAM on it and see if this issue also happens in Zorro II mode.

I have tried CacheInhibit'ing everything to no avail - I have also tried disabling CPU caches entirely just to be sure.

Sadly I don't have access to a BigRAM/ZorRAM, but I suspect I would see exactly the same issues with them.

If the Voodoo is working fine including with Warp3D, I find it really unlikely your Zorro-3 is malfunctioning. The VoodooMemOS thing can happen for a variety of reasons, do you run Workbench in a BE or LE mode?

[jbenam]

Sorry for the late replies everyone, thanks to another kind EAB member, I have finally been able to get another A4000 to test my setup on and... The same exact stuff happens.

The Zorro side of this A4000 was fixed by hese years ago, so I am confident it was fixed properly.

I am at a loss here - I wonder if it's the Mediator at fault here? Maybe it was always faulty? And what about the Prometheus, why do I get the same issues with it? Just a weird coincidence?
This is a Prometheus reproduction, so I am not 100% sure that it was ever working.

Originally Posted by DarrenHD

Have you looked at the Bridgette chip? Inspect it's connections and ensure it's ok. I had an issue with Zorro 3 and it was due to a bad solder joint on Bridgette.

Also it wouldn't hurt to swap out Fat Gary. The gary chip surprisingly has it's fingers in lots of system functions.

I did reflow Bridgette suspecting the same issues - so that should be covered.

I haven't got another Fat Gary around, so I will keep that option open.

Originally Posted by grelbfarlk

If the Voodoo is working fine including with Warp3D, I find it really unlikely your Zorro-3 is malfunctioning. The VoodooMemOS thing can happen for a variety of reasons, do you run Workbench in a BE or LE mode?

Mostly 8-bit mode. Should I try 16-bit LE or BE?

[trixster]

So just to confirm - you have a second machine now, you put the mediator (or the prometheus) into the new machine, use the harddrive/sd card from the old setup... and exactly the same crashes occur when trying to use zorro III ram? Or any pci card? or any zorro 3 card?

When exactly do the crashes occur? is it just when trying to copy stuff to/from Z3 ram? Can any Warp3D games work properly? Does the machine just hang once the K1 is initilised?

Is workbench stable with just the voodoo installed and used as an RTG card?

Have you tried any other voodoo or radeon cards in conjunction with the K1?

[trixster]

How are the mediator logic board jumpers set?

Winsize should be closed, master should be closed and Config should be open

[jbenam]

Originally Posted by trixster

So just to confirm - you have a second machine now, you put the mediator (or the prometheus) into the new machine, use the harddrive/sd card from the old setup... and exactly the same crashes occur when trying to use zorro III ram? Or any pci card? or any zorro 3 card?

When exactly do the crashes occur? is it just when trying to copy stuff to/from Z3 ram? Can any Warp3D games work properly? Does the machine just hang once the K1 is initilised?

Is workbench stable with just the voodoo installed and used as an RTG card?

Have you tried any other voodoo or radeon cards in conjunction with the K1?

That's correct. I can reproduce the issue with either VoodooMemOS (when running stuff copied to it) or just by starting the M1 with StartPPC.

I have only tried the Warp3D demos and those work properly.

Yes, rock stable. Never had an issue.

I am pretty sure I have tested everything with my Radeon 9250, but I am going to try again with it in place of the Voodoo 3.

Originally Posted by trixster

How are the mediator logic board jumpers set?

Winsize should be closed, master should be closed and Config should be open

Yes, that's how it's set right now.

[trixster]

It now sounds to me like the issue we had with A4000 and the apocalypse board, where having the cards in the wrong pci slot order caused insta-crashes like you are experiencing.

What order to you have the ppc card and voodoo? From top to bottom.

[trixster]

Also, try with a jumper on swap Config. I have to do that to get Deneb working with mediator/ppc, so it’s worth a try

[jbenam]

Originally Posted by trixster

It now sounds to me like the issue we had with A4000 and the apocalypse board, where having the cards in the wrong pci slot order caused insta-crashes like you are experiencing.

What order to you have the ppc card and voodoo? From top to bottom.

I have tried every order possible. Sometimes the card doesn't get detected, but when it does, it crashes like that.

I have also tried filling the other two slots with a PCI128 and a Realtek LAN card.

Anyway, my usual config is top slot empty, second slot Voodoo3, third slot M1, fourth slot empty.

But I have also tried every other combination possible.

Originally Posted by trixster

Also, try with a jumper on swap Config. I have to do that to get Deneb working with mediator/ppc, so it’s worth a try

Also tried that, nothing changes

My next step is to ask for my friend's Mediator 4000D (not 4000Di, but it should be the same - my Amiga is on a bench right now) and see if the same thing happens. If it does, my Mediator is borked.

[trixster]

Yeah, quite possibly. As I said in post 7, wrangler had to send a logic board back to elbox to be reworked to fix inexplicable crashes, so maybe yours needs a fix too. Doesn’t quite explain Prometheus issues though.

[jbenam]

Originally Posted by trixster

Yeah, quite possibly. As I said in post 7, wrangler had to send a logic board back to elbox to be reworked to fix inexplicable crashes, so maybe yours needs a fix too. Doesn’t quite explain Prometheus issues though.

The Prometheus might be borked. I can't reprogram the CPLDs. Or maybe it's just my clone USB Blaster. Not sure. For now I am going to exclude it from my testings.

I have got another Mediator (another Mk. I, though), and the same stuff happens.

I have also tested with a THIRD Amiga 4000 and I still get the same issues.

So right now I have tested with:
1) two A3640
2) three A4000 mobos
3) numerous power supplies
4) two different Mediators (both Mk. I)
5) two different GPUs (Voodoo 3 and Radeon 9250)

Literally this is the thing I am testing right now:

1) copy stuff to RAM, make sure that mobo RAM is exhausted (first one to be used in terms of priority), with VoodooMemOS or RadeonMemOS enabled
2) run something you just copied into RAM

Or you can remove enough fast RAM for AmigaOS to start putting boot stuff into it and #8000 0004 errors should happen.

Can someone please do this quick test with their system and see if you can reproduce this?

Maybe it's expected behaviour (can't see how, but who knows) and I am just losing my mind for no reason.

Thanks

[jbenam]

Originally Posted by trixster

what happens if you run PPC-MemTest?

https://eab.abime.net/showpost.php?p...postcount=1414

I don't think I can run it - my system falls short way before I can run PPC executables.

The problems is that these error happen even without the M1 - just with a bare Voodoo3. If I try to run something from the Voodoo3 memory, I get #8000 0004.

I am now not even trying with the M1 in since it's clearly trying to run some code from Zorro III space and it crashes because of that.

[trixster]

I think if I were you I'd try and borrow a bigram or a zorram (or other zorro 3 ram board) and try your testing with one of those on a standard A4000 busboard and again on your mediator busboard.

If using Z3 ram then works ok then that suggests it's the mediator that's at fault when it's acting as the method to add graphics card ram into Z3 space.

[jbenam]

Originally Posted by trixster

I think if I were you I'd try and borrow a bigram or a zorram (or other zorro 3 ram board) and try your testing with one of those on a standard A4000 busboard and again on your mediator busboard.

If using Z3 ram then works ok then that suggests it's the mediator that's at fault when it's acting as the method to add graphics card ram into Z3 space.

I will see if I can get one. But how can the same stuff happen on TWO different Mediators? Maybe it's because they both are Mk. I?

[trixster]

I am afraid I don’t know! I am at an impasse now as what to suggest next.

I would offer to send you my zorram to test with, but the current state of shipping stuff to the EU from Britain and the fact that zorram is now rare is an issue, sadly.

[jbenam]

I am starting to think it might have something to do with the A3640s. I found out mine has the same issue with the oversized IC Turran had. If that's the QC these things went through who knows what else might be wrong with it.

The other A3640 I am testing with is not exactly good-looking either.

It might be something like that... I need to have a look at the schematics and see if I find something blatantly wrong.

[DarrenHD]

Originally Posted by jbenam

I am starting to think it might have something to do with the A3640s. I found out mine has the same issue with the oversized IC Turran had. If that's the QC these things went through who knows what else might be wrong with it.

The other A3640 I am testing with is not exactly good-looking either.

It might be something like that... I need to have a look at the schematics and see if I find something blatantly wrong.

Yeah it wouldn't hurt to try out the C= 030 card in that A4000. Or an 060 board.

[jbenam]

I have since discovered that the Prometheus repro could have never worked - the two CPLDs were horribly soldered and I could verify them only if I kept them pressed in a very specific way. I have since then resoldered one of the two CPLD and I am waiting for replacement on the other one (which apparently gets very hot very fast, couldn't remember if that was the case earlier as well or not).

Which in turn is making me think that the Mediator I was sold has something lacking (MACH upgrade?) and won't work with the PowerPC card (as the other one I have tested is a Mk. I as well and it exhibits the same issues) and as such I might have spent the last three years resoldering my Amiga for no reason at all basing everything on the wrong premise that the bridgeboards are okay

The only surefire way of finally settling this would be to test my Amiga with a known-working PowerPC capable Mediator, but I can't justify spending 350€ without even knowing if it will work on my Amiga.

Any kind soul that could lend me their known-working Mediator?

[trixster]

Were it not for brexit causing delivery chaos and the worry of dismantling a known working system, I would send you my logic board, but sadly it is too risky these days. I am sorry.

[jbenam]

Originally Posted by trixster

Were it not for brexit causing delivery chaos and the worry of dismantling a known working system, I would send you my logic board, but sadly it is too risky these days. I am sorry.

No worries, that's perfectly understandable.

So it turns out that the Buster 11 I was using was busted (*rimshot*) as the Fusion emulator couldn't properly work in MMU mode (since it maps in Zorro III land) even without the Mediator.

I have since then started using my other Buster 11, which works fine in MMU mode on Fusion.

Now the Amiga "gracefully" freezes instead of giving out the ramlib 8000 0004 error.

As reported in the Sonnet thread, Elbox told me that the A3640 is not compatible with VoodooMemOS/RadeonMemOS, at least when it comes to my revision. I have asked if they can sell me replacement MACHs but I haven't heard back yet.

[jbenam]

Sorry in advance for the bump, but I think this might be useful for other users as we finally have a conclusion for this VERY long-winded story.

I have finally tested everything with a CyberstormPPC, lent by the great @reflex (many thanks again!) and RadeonMemOS worked wonderfully.

I could finally use Fusion on the ZorroIII-mapped memory.

It seems that Mediator 4000Di Mk. Is are actually INCOMPATIBLE with executing from Zorro III spaces. Since other users have A3640s working with their Mk. IIIs, I guess this bug must have been fixed in a subsequent revision of the Mediator 4000Di.

Thanks again to @reflex, without his help I would have never found out that my Amiga 4000 is actually perfectly working