Update:

>> Please blame Aeon first for not allowing to have two PCIe cards in X5000 :)

> Dual-booting works on the X5000 with a PCIe x16 GCN-based Radeon card
> in the PEG slot for OS4 and a PCIe x1 R500/R600-based FirePro/FireMV
> card in the PCIe x1 slot for Linux. The question is why MorphOS cannot
> work in this setup when Linux can.

Seems the cause was finally found and fixed within MorphOS:

https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=13303&forum=3&start=4


Fix released:

"X5000 hardware support was improved to handle two Radeon cards"
https://morph.zone/modules/news/article_storyid_2574.html

"Tweaked the PCI address mapping on the X5000 to accommodate a Radeon on the PCIE1 bus"
https://morphos-team.net/releasenotes/3.17

[ Edited by Andreas_Wolf 02.05.2022 - 09:50 ]

I am the OP.
I just stumbled upon the great news that the MorphOS bug preventing a 2nd PCI-E GPU being used has been fixed in 3.17.
Thank you to the developers for fixing this. Now I can have the best of both worlds.
I will be looking for a 1950 XT for MOS to complement the 580 for AOS41.
The 1950 will fit in the Define 7's vertical GPU slot, and with riser should not block other AIC slots on the 5000 motherboard.

I will say my perception from this experience is some of the MOS team comes across as hyper sensitive cry babies. Instead of getting angry at users asking questions and blaming others but not looking in the mirror until now is really sad. Luckily this MOS bug has been fixed and I am personally grateful to the developer(s) who fixed this. Was the bug more complicated than hardcoded expectation of the display adapter being in the x16 slot? Without sources I will never know, but regardless now all x5000 owners in my shoes can now use MorphOS. Well done to whomever fixed it.

Quote:
No good deed goes unpunished.

Quote:

What an unnecessary insulting comment. Totally uncalled for. Where on earth did that come from?

OP gets the fix they wanted and then goes on a rampage attacking the developers.

Boggles the mind...

> Was the bug more complicated than hardcoded expectation
> of the display adapter being in the x16 slot?

At least my impression, which may very well be plain wrong, from the relevant thread (which fell victim to the recent data loss unfortunately) was that bigfoot was easily able to identify the nature of the bug (as his initial hunch in that thread proved true). To me it seemed that it wasn't that hard for him to incorporate the actual fix also.
Assuming that this impression isn't too far from the truth, I must admit I've been somewhat puzzled that it took 4 years of reporting this issue here on MorphZone before it was eventually taken on.

I seem to remember that I did write a somewhat detailed explanation of what the problems were, but it got lost in said data loss. What I don't understand is why some people seem to jump to conclusions, that it was just a matter of flipping a switch, a tiny change to somehow make MorphOS "search" for a graphics card in the right slot, and the MorphOS developers were just too lazy to do this tiny change.

I honestly don't want to go into too many technical details once again, but let me just say that that is not how MorphOS (or any other operating system really) works. Nobody looks for a specific card in a specific slot. The PCI (express) bus(es) get scanned, and then the operating system makes an API available that allows drivers to find cards, usually based on the card's device and vendor IDs. At no point does any specific slot come into this equation.

When it comes to fixing this issue, first you gotta realise that it involves are quite uncommon hardware setup, and not one that is readily available. Initially there was no MorphOS developer who had such a setup, or could create such a setup, without investing in additional hardware.

Fixing this issue spanned a duration of 2.5 weeks. Of course not 2.5 weeks of non-stop work on this issue, but it definitely wasn't a simple matter of just flipping a switch. I did have an initial idea of what the problem was, one I believed I could fix without having access to this setup myself. So I went ahead and did just that, and I enlisted the help of a person here on MorphZone with the prerequisite setup for testing.

It turned out that while my "blind" fix was indeed correct, it wasn't the only problem preventing MorphOS from booting with multiple Radeons installed, and figuring out the remaining problems would take too long without having such a setup myself, so I went out and bought a PCI Express riser card so I could run my own X5000 with two graphics cards. From there on, I identified multiple different problems that all contributed to this problem, and I fixed them all.

All in all, the diff for the cumulative fixes that made MorphOS boot with multiple Radeons in the X5000 is almost 1000 lines long. It was not a simple fix.

> I seem to remember that I did write a somewhat detailed explanation
> of what the problems were, but it got lost in said data loss.

Seems the data loss occurred too soon after you wrote that detailed explanation for me to read it. If I had, my previous comment would surely have come out differently.

> […] It turned out that while my "blind" fix was indeed correct, it wasn't the
> only problem preventing MorphOS from booting with multiple Radeons
> installed […]. From there on, I identified multiple different problems that all
> contributed to this problem, and I fixed them all. All in all, the diff for
> the cumulative fixes that made MorphOS boot with multiple Radeons in the
> X5000 is almost 1000 lines long. It was not a simple fix.

Thank you for the insight.

So can we now boot an X5000 with the two cards and OS4 doesnt choke on having two? I havent tried this and just saw the update supports two cards now. Pretty impressive solution.

What about a 2nd HDD installed with MorphOS on it, will this work out of the box if I put in a 2nd card and replaced the DVD with my MorphOS SATA drive?

[ Edited by Acill 29.05.2022 - 08:58 ]

> can we now boot an X5000 with the two cards and OS4 doesnt choke
> on having two?

Yes, but OS4 still requires its card in the primary slot, else there wouldn't have been any problem as the MorphOS card has always worked in the primary slot and OS4 would have simply used its card in the secondary slot (which it cannot, at least with the primary slot in use).
As is known from running dual-boot OS4/Linux on X5000, OS4 running on the primary card isn't disturbed by a secondary card.

Quote:

The logic is slightly faulty here, and it might be caused by the previous (wrong) speculation. I have no idea if AmigaOS 4 can or cannot boot with its graphics card in a non-primary slot, but in the case of MorphOS, it couldn't boot at all with a Radeon in anything but the primary (x16) slot. At all. It wouldn't even get to the point where it would try to load the Radeon driver.

> I have no idea if AmigaOS 4 can or cannot boot with
> its graphics card in a non-primary slot

IIRC from the discussions over the years, if there're two OS4-compatible graphics cards installed in the X5000, OS4 will always use the one in the primary slot. There's no way to make OS4 ignore the card in the X5000's primary slot and use the (better supported) one in the secondary slot instead (or in addition). This was stated as an unfortunate regression from the X1000 where OS4 can even use two graphics cards concurrently in a dual-GPU setup. I have no idea though about OS4-incompatible graphics cards in the X5000's primary slot as all MorphOS-compatible graphics cards are also OS4-compatible to the extent that OS4 can boot with them.

> MorphOS [...] couldn't boot at all with a Radeon in anything
> but the primary (x16) slot. At all. It wouldn't even get to
> the point where it would try to load the Radeon driver.

Thanks for the clarification. Do I understand the situation right that pre-3.17, MorphOS couldn't even boot with a MorphOS-compatible card in the X5000's primary slot as long as another Radeon (MorphOS-compatible or not) was inserted into the secondary slot?

Yes, you're understanding it right. And technically it didn't specifically have to be a Radeon, but any set of PCI devices with the right set of PCI memory bases could trigger this bug. Radeons in general just do that reliably.

Quote:

Thank you bigfoot for fixing the bug that was preventing myself (and perhaps other x5000 owners) from using MorphOS.

I ordered a 1950 XTX and will order a riser adapter soon.
It is a nice benefit that the 1950 XTX PCB is red, as is the semi-transparent wind tunnel matching the x5000 PCB. My Define has a tinted glass side panel.
Luckily I checked and have a Corsair 750 PSU, as I thought I had put a SF450 in the machine.
Reading up on the cards they both seem to be power hungry.

This is the best of both worlds for me. A ~2006 GPU for MOS, and a ~2018 GPU for AOS-maximizing capabilities on each OS.

Some may not like statement of my impression(s) of the MOS teams response to potential new users and that is OK.
I think it is important the team be told how they interact with their own user base.
Do what you will with this feedback.

Now that this bug has been fixed all x5000 users stand to benefit.