Yes, it is speculation. I'm quite willing to drop it as soon as I get any evidence that this is not the case. Like someone encountering the problem while using NVidia GPU or older Catalyst drivers. So far, we haven't had such a case.

So, what is the cause of said error then? InnoSetup? Would have seen more cases. Xenonauts files? The error pops up during installation, not execution. I'm quite willing to accept another possibility, but I haven't seen any so far.

This is the usual method of investigation. Make a hypothesis. Try to disprove it (proving also works, but is usually harder). If hypothesis is proved wrong, look for different possibilities. If hypothesis hasn't been proven wrong, but seems highly unlikely, look for more probable causes, and see if those can be proved wrong or right.

So please, do give me another possibility for the floating point errors. One that fits the facts so far. I'll gladly accept one, and apologize for my mistake.

Well, Xenonauts only has this problem on gog.com, that's where I'd build my personal hypothesis. It happens only with the gog.com installer, evidently because this is not an issue outside gog.com.

Those are the only two facts I'm aware of that are 100% consistent.

Inconsistent facts:
- FP error on machines with AMD GPU when using gog.com installer, frequency of issue can't be determined, but seems rare, looking by the number of people who have encountered it. This has happened on more than one game, Xenonauts, AvP, Witcher 2, et al.

- No FP error on machines with AMD GPU when using gog.com installer, frequency of this scenario can't be determined either, but people usually don't post when things work without problem. Anecdotally, this is the case for me, I've never had an FP error on gog.com install and have an AMD R9 280 GPU with the accused C14.9 WHQL drivers.

Without more data, it's impossible to nail down a likely culprit, but it is easy to point out the inconsistencies in the AMD-hypothesis. Namely, that these games install problem-free on AMD systems. They also don't. Sometimes, for some people.

Also, the gog.com installer for these games that have been mentioned, don't do anything other than unpack, copy and quit. GPU drivers don't enter into the operation of the installer. If you can show where and how the GPU drivers enter into the installation, well that would make it more of a hypothesis.

Where's the inconsistency? See any car callback. Model X is recalled because some cars exhibit problem Y. I have a model X that doesn't exhibit problem Y, thus problem Y doesn't exist.
Is the above statement correct, or does the problem exist, even if the majority of cars don't exhibit it?

I was under the impression that the LMZA used OpenCL if available, but I can't verify that part anywhere, other than a mention in this page. If the unpacking does use OpenCL, then the drivers do play a role. And too tired atm to go through Inno Setup's code to see if it does use OpenCL or not.


P.S. Once again, I don't say the AMD drivers are universally bad, or that you shouldn't use them. I say that it appears to be a case of a combination of things, including a specific version of the drivers causing errors. Similar to a car model being recalled. And the fact that changing drivers fixed it for most of the people who had this problem (or at least all of those that reported back) is another indication that the drivers are part of the problem, and most likely the easiest one to change.

But again, please offer a reason that one would get the FP problem and changing drivers would fix it, if the drivers are not at fault.

I see the problem, you think that this is an actual fact. It is not. It is just some person who claimed this. This person made the claim about AMD drivers, that doing "something" to them made a difference.

It's not for me to prove ghosts don't exist, it's for others to prove that they do. You just assume it is a fact ghosts exist. Ghosts being supernatural beings, etheral in their nature, much like random vooodoo tech claims on the interwebs.

The inconsistency is this: using the same drivers claimed to be faulty, installing the same software that is supposedly conflicting with this driver, nothing happens out of the ordinary.

With the same fundamentals, doing the same thing, the result must be the same - but it isn't. But then again, what can one expect from people who don't explain at all what they did, what kind of equipment they have, what possible conflicts they could be experiencing, what software they've installed and so on.

Anyway, this isn't a problem, and I am starting to think that you want to push it as such ... the only people debating this are people who have never experienced this. It isn't a problem for you, me or anyone except the OP and seemingly less than 10 people of all those who post on the gog.com forums.

Much ado about nothing, especially since it's all based on an unsubstantiated claim. At least I can run the same equipment (that the claim is based on anyway) to test this, and there are no problems with the driver. If there was anything to it, a lot of people would be experiencing this error.

Almost no one is.

I'm going by the posts by getlogan and hellmonster67 in this thread, mostly the one that reads "I could always make the problem go away by uninstalling the display driver".

To give a better example, I'm willing to accept Rutherford's claim of nucleus and electrons instead of claiming that the plum pudding model is correct. And even though Rutherford's model proved wrong, the evidence did support it for a while.

Using the same handgun to shoot at the same target causes different holes. Because there are more parameters at play, like wind, shooter, bullet used etc. Or, in our case, GPU manufacturer, chip batch etc

I'm not pushing it as a problem. I don't say "Don't use AMD drivers because they cause problems", I say "If you encounter a floating point error while installing a GOG.com game, see if changing your drivers fixes it". Similar to suggesting to disable your antivirus/firewall to troubleshoot something. We don't claim that you shouldn't use antivirus or firewall, we ask to troubleshoot.

And that is why we suggest changing the drivers, to see if it's really the cause of the problem or not, to be able to troubleshoot it better.

Yes, I totally agree some guy made a claim on the internet, I just don't take it seriously because setting aside his claim, there's nothing to back it up. No actual consistency, no masses of people experiencing the same, even though they use the same drivers, nothing.

And computers are not handguns, they execute code in a uniform and predictable way, meaning every time software X is used, it is the exact same binary for me as it is for you or anyone else. Which means the pure interaction between the AMD drivers and the gog.com installer is 100% problem free, because it works demonstrably.

Perhaps the users in question have installed some other third party program that interferes in the FPU call and conflicts with the AMD driver because of it, which would explain why this is so rare.

It would also exhonorate the AMD driver, as well as explain why the gog.com installer "works" for those users when they remove the AMD driver. A third party app conflicting with the AMD driver, for instance, is not a buggy AMD driver, rather a buggy third party app.

I find that hypothesis supporting the observed effects quite nicely. It explains why so few are seeing this issue, it explains why gog.com has no idea what is causing this, it explains why uninstalling the AMD drivers helped those individuals and it explains why the gog.com installer works just as expected for me and most people who use AMD GPUs.

Yes. If everything is the same. Different GPU manufacturer (Saphire vs MSI) is a different machine, thus different outcome is possible (even though unlikely). Different chip batches may cause errors (see car recalls examples). Different soldering agents may cause connections to fail. MTBF is also a thing, so your computer may run for 10000 hours while my identical one crashes after 100. Each and every component may cause trouble. If changing a specific component solves the problem, said component is the most likely culprit.


Yes. Drivers + program X cause the installers to fail. It's program X's fault, not the drivers. Why? Someone on the internet claims it's so ;)


Yeap. Assuming there is an app all users are using. If there isn't such an app, then it's either a family of apps, or the drivers. Insufficient data for the app, though that is a valid theory.


Yes. Something is causing problems. So far, the most reliable "solution" is to change the display drivers, similar to the "please disable your antivirus and try installing again" method. While that doesn't prove that the drivers are at fault, it has (so far) been the best short term solution.


P.S. Once again, I'll direct you to this thread which seems to have a few users that had said trouble. Feel free to ask them for a list of the third party programs, and see if there is a connection. If a connection can be found, quite a few people will be happy to know what is causing said error, and who to notify for a fix.