OpenUru.org

rarified wrote: How do you tell which version of PhysX installed? Open the Control Panel on Windows, select "Programs and Features" (from Win 7, other versions may have slightly different wording). That panel will give you a list of all programs installed on the machine. PhysX will show up as either "Aegia PhysX" or "NVIDIA PhysX" along with it's version number

On my 64-bits Windows 7 system PhysX does not show up in the Control Panel.

Anyway, you can check the installed PhysX engines in the folder "c:\Program Files (x86)\NVIDIA Corporation\PhysX\Engine\" (on Windows 64-bits) or "c:\Program Files\NVIDIA Corporation\PhysX\Engine\" (on Windows 32-bits). If a folder with the name "v.2.6.0" exists and contains the files PhysXCooking.dll and PhysXCore.dll you're good.

D'Lanor wrote:On my 64-bits Windows 7 system PhysX does not show up in the Control Panel.

Correct. But you can also look at the Control Panel -> Programs -> Uninstall a program listing to see what's installed. My Physx is listed as " NVIDIA PhysX System Software 9.12.0604, with the "9.12.0604" also listed in the "Version" column. I don't know if all versions will report version numbers, especially the older Ageia ones. But either way, the "acid test" is whether or not the "v.2.6.0" folder is populated as D'Lanor describes.

Mac_Fife wrote:It's not so much a case of "overlooking something"

Well, I think it is, in the sense that there was a solution out there that I wasn't aware of. I had been thinking in terms of MOULa failing to recognize that the new PhysX was installed, rather than of a significant change to PhysX itself. But my approach to troubleshooting is to put two questions up top: 'What am I doing wrong?' and 'What am I missing?'. If I can't manage to get past those myself, I ask around. I was able to rule out user error, and then realized I didn't have--and wasn't finding--the right information to get any further on my own. So, me saying I was overlooking something is just me being practical rather than self-critical.

But you're right, it was unexpected, although I suspect that the real motive is just a desire on Nvidia's part is just to be able to stop supporting older software. That's a flaw in the profit motive: if there's no money in it, they won't do it.

I think charura is right that the need for the legacy installer should be put out there in some way that's more obvious to MOULa users. I don't know how to do that, either.

Well, once we get the MOUL forums back, we can post an advisory note in the "Technical Discussion" section there, maybe on a sticky or in the existing Common Problems thread.

I don't know if this is the kind of thing that falls into GoMe territory, as another channel for disseminating new knowledge?

If you post that note be sure to have the facts straight: PhysX versions from 9.12.1031 onwards require the legacy installer.
PhysX 9.12.0613 was the last version which installs all engines. The release notes for this version on the nVidia website are just plain wrong. I have this version installed and it definitely includes all engines from 2.3.1 to 2.8.3.

Mac_Fife wrote:I don't know if this is the kind of thing that falls into GoMe territory, as another channel for disseminating new knowledge?

I wouldn't have anything against it myself; what exactly should we tell users?

Simply this:

MOULa users (and users of most MOUL-based shards) that get an update to PhysX through the NVIDIA Updater service or perform an install of PhysX at version 9.12.1031 or later, either from the NVIDIA website or a supplied driver CD, should also run the optional legacy installer. This is not necessary if PhysX version 9.12.0613 or earlier is installed.

The symptom will be getting a message similar to: "PhysX is not installed, or an older version is installed. Install new version? (Game will exit if you click "No")".

Download locations:
PhysX 9.12.1031: http://www.nvidia.com/object/physx-9.12 ... river.html
Legacy Installer: http://www.nvidia.com/object/physx-9.12 ... river.html

As D'Lanor has observed, the release notes for PhysX have been (at best) ambiguous about what SDK versions are supported by each release of the System Software for some time (at least since CWE was released) since they've tended to make statements like "Supports NVIDIA PhysX acceleration on GeForce via CUDA 3.0 for SDK versions 2.7.1, 2.7.3, ..." which is mainly a reference to NVIDIA hardware based acceleration and takes no account of the fact that PhysX actually works with Intel and ATi/AMD graphics too.

Voilà : http://www.guildofmessengers.com/en/con ... discovered

OK, another observation or two on the Nvidia installer:

I got a pop-up to say there was an updated driver for my video card, so as I'd skipped the last couple (if it ain't broke, don't fix it) I decided I'd take this one - It also means that I'd side-stepped the PhysX issue described here until now. I chose to run the graphics driver update with the advanced options so I could choose what was or was not installed. One of the items was PhysX and as I had the last version that didn't need the legacy installer I unticked the box, expecting to keep the current version that was listed. Nope, that caused the graphics driver installer to remove PhysX altogether

So, back to the Nvidia site to get the PhysX installer (and by this point I might as well get the newest version). That decided to abort about 90% of the way through reporting nothing more helpful than "Error 2738", and a couple of retries after shutting some things down didn't help. After a bit of Googling I found that (at least on Win 7 x64) you need to go into the registry editor and delete this key to get the installer to work: After that PhysX installed perfectly OK.

It may be that the install works fine if it's run as part of the graphics driver update, but as a "stand alone" install it looks flawed - I found several threads of people complaining about "Error 2738".

And then we take my suggestion to replace the bundled PhysX_Setup.exe with the legacy installer such that there's no need for manual fiddling with PhysX that can deliver us into the maw of gotchas such as the one described above