Argh! ATI!

Screet · Post by **Screet** » Mon Feb 23, 2009 10:50 pm

OK, now I'm upset!

I just placed the same installation I have on my primary machine to the development one, which is Windows XP X64 with AMD 3400+, 2GiB RAM at 400Mhz and a 256MiB X800XTPE.

Installed was the ATI 9.1 driver.

Guess what?
- nebulae broken (ATI bug in implementation as Solas wrote)
- NO SLOWDOWN AT ALL when I enter Geteve!

Thus, the old machine with a single processor (same frequency as my Quadcore), half the RAM amount and a graphics card with one quarter outperforms the 4870x2 which has 2GiB RAM (thus 1GiB for each GPU).

Now this appears not to be a driver issue, but to be a hardware issue with the 4870x2!!!!

I'm going to update to 9.2 drivers instantly - if the speed stays, as I expect, I'll be able to ask those ATI and Asus guys some more questions...

Screet

Screet · Post by **Screet** » Mon Feb 23, 2009 11:06 pm

OK, with the 9.2 drivers, the result is just the same on the X800XTPE.

The nebulae bug is still in, as expected due to faulty drivers from ATI.

However, all those graphics bugs (wrong coloring of subentities, missing textures, ...) are GONE.

Screet

Post by **Kaks** » Mon Feb 23, 2009 11:30 pm

If you're compiling from trunk, it might also have something to do with revision 2073...

Screet · Post by **Screet** » Mon Feb 23, 2009 11:38 pm

Kaks wrote:
If you're compiling from trunk, it might also have something to do with revision 2073...

No, haven't compiled yet...should have done several weeks ago, but always something made me stop short of it

I used the same installer for 1.72.2-test.
Then I did copy the whole Addon-folder and all saved games to the development machine.

Works like a charm. Wish that my gaming machine with the 4870x2 would at least be as fast as the X800XTPE.

Unless that's an issue with ATI XP 64 drivers being OK and ATI Vista 32 drivers being buggy, this has to be a hardware issue on the 4870x2 missing OpenGL acceleration. Am I missing some other possibilities?

Screet

Solas · Post by **Solas** » Tue Feb 24, 2009 5:00 am

@ Screet
On your Vista machine did you install Oolite outside of its normal install location
or did you copy / move your Oolite directory to avoid virtualisation.
moving the installation will leave references to the old location in RunOolite.bat
and that may cause problems.

Screet · Post by **Screet** » Tue Feb 24, 2009 5:55 am

Solas wrote:
@ Screet
On your Vista machine did you install Oolite outside of its normal install location
or did you copy / move your Oolite directory to avoid virtualisation.
moving the installation will leave references to the old location in RunOolite.bat
and that may cause problems.

I'm still running it from the program files folder. Once I found out about this nasty virtual store and cleaned it up, only saved games and logs stay there.

Screet
P.S.: After receiving more input, Asus says they've got some of their techs looking at the problem. Hope they can help. ATI still did not respond to technical inquiry, forum and driver feedback.

Post by **Kaks** » Tue Feb 24, 2009 6:50 am

Hope it all goes well on vista! Not that I've got any good experience with vista video drivers..

If it doesn't go to plan, have you considered dual booting with xp? Running anything under xp should be much faster!

Cmd. Cheyd · Post by **Cmd. Cheyd** » Tue Feb 24, 2009 7:18 am

Personally, rather than spend the space for a dual-boot (no guarantee you'll make efficient use of the hard-drive space you alot for it)... Wonder if you could install a VM of XP, and run Oolite from the VM. If you use a dynamically expanding disk, you won't lose any HDD space than you actually require.

Virtual PC2007 SP1 is great, it's free, and it's stable. It's how I still play some of my old DOS / Win9x games.

Solas · Post by **Solas** » Tue Feb 24, 2009 9:53 am

@ Screet

I think your Oolite setup may be suffering from virtualisation.
Can you install a fresh Oolite on Vista as you did on XP.
Except use a new directory such as C:\Oolite or C:\Oo.
The installation of Oolite on Vista should default to C:\Oolite
Leaving anything in Program Files offers no benefit and taints results.

Are you using a non-1.72.2 oolite.exe ?
files in \Resources must be updated to match the dev version.
also from experience this causes strange problems.

the biggest cause of problems for me by far has been OXPs.

the best way to verify a problem is to rule out all others.

you have a dual-card your game slowdown MAY be related to that.

the only ATI driver bug I have found is the star and nebula problem.
--------------------------------------------------------------------
OpenGL EXT_texture_lod_bias bug.
https://bb.oolite.space/viewtopic.ph ... ght=#68305
--------------------------------------------------------------------
if you want to help sort this one just put the previous two lines
in a new topic on ATI forum.

Screet · Post by **Screet** » Tue Feb 24, 2009 12:10 pm

Solas wrote:
I think your Oolite setup may be suffering from virtualisation.
Are you using a non-1.72.2 oolite.exe ?
files in \Resources must be updated to match the dev version.
also from experience this causes strange problems.

I'm using the 1.72.2-test version - if theres any update required to that one, please let me know. I haven't installed the dev version and I didn't go for the workaround for the bug you found after seeing all these log entries it creates, but then I could have edited the log setup. Maybe I should? I got pretty much used to the buggy display

Since I ran the same oxp's on the development machine for the test as on my playing machine, it's not an oxp issue.

Interestingly, I did not instantly go to full screen, but had it run in the tiny window. It was faster that way, but still jumpy...once I went fullscreen, I had the old slow behaviour. There was a slight improvement as to before, but really not much and annoyingly jumpy. Seems that drawing really proves tough for my card. The CPU almost idle and the GPU entirely idle, slowed down to 300MHz

The exactly same problem others get with other OpenGL software.

The problems with wrong display (subentities of Caduceus, for example) did not go away. The X800XTPE does not have such issues.

Maybe virtualization causes problems only by slowing down the log file updates. At least, I now can safely edit files without having to manually put them back from virtual store and I got a slight improvement, so it was worth doing uninstall/reinstall/movement of folders.

Anyway, that's more improvement than I expected, but still unacceptable and behind that X800XTPE performance.

I would have been astonished if all went away - there were numerous users on the ATI forum complaining about the same symptoms with many other OpenGL apps, both new ones (extreme problems) and very old ones (still bad performance).

I then uninstalled GPU Caps Viewer and reinstalled it also to C:\gpu caps viewer instead of program files. The crash with fur rendering did not go away. The X800XTPE has no problem to run it.

I guess the only thing I could do for further testing would be to buy another SATA disk, uninstall the currently installed disk drives and then illegally install my XPX64 also on that machine. Just for a test. Not so nice.

I did report that texture bug with a link to your thread already before and also gave Asus that link. I did not start a new forum thread for only that bug, maybe I should also do that? Unless they banned me again, then I'm running out of email addresses for new user accounts.

Screet

Solas · Post by **Solas** » Tue Feb 24, 2009 1:58 pm

@ Screet

You can install as many seperate versions of Oolite as you want. they won't interact.
Oolite is self contained in it's directory and currently does not use the registry. [edit: Oolite currently stores uninstall info in the registry. 1.73+ may behave differently]
have you tried installing a fresh Oolite on Vista as you did on XP.
you only need it for checking your slowdown issue.
you haven't said that you've tried this

IMHO Virtualisation can cause all sorts of problems.
I wouldn't trust Vista to virtualise diddly-squat.

The crash with GPU Caps Viewer fur rendering has happened with NVidia and ATI.
that issue may be due to overheating of a particular part of the board.
ATI limited it's performance for Furmark ( another benchmark ) because of this.
http://www.techpowerup.com/index.php?69799
fur rendering crash does not happen for me. so I can't say for sure.

GPU Caps Viewer .. Texture Buffer Size = 0 Mb doesn't mean its not there.
It could mean that its not being detected.
As far as I can guess GL_ATI_meminfo is used for this. ( I am open to correction )
but a search finds no OpenGL Enumerator for GL_ATI_meminfo anywhere.
maybe one of our OpenGL coders may know more.

a new forum thread for the OpenGL EXT_texture_lod_bias bug would be best.
wait and see if they will respond to your post. if not try a new thread

I think this the most widespread ATI bug for Oolite players.

Solas · Post by **Solas** » Tue Feb 24, 2009 3:08 pm

@ Screet

I ran GPU Caps Viewer fur rendering and let it run ( normally ran only briefly )
observed temperatures with RivaTuner ..
Main > Target Adaptor > small button pointing bottom-right > Hardware Monitoring.

As the temperature passed 100 C I imagined all the accessories ATI could sell ..
Espresso makers, poptart toasters and mini Pizza bakers on the high end models !
the card crashed at about 130 C.
restarted computer ( phew ) .. I won't be doing that again

Screet · Post by **Screet** » Tue Feb 24, 2009 3:47 pm

Solas wrote:
have you tried installing a fresh Oolite on Vista as you did on XP.
you only need it for checking your slowdown issue.
you haven't said that you've tried this

I would still need my playership and assassins - as I do need a situation with enough ships to cause this slowdown. It does happen in normal flight too, but that's more rare. It's really linked to the amount of textured objects being displayed.

I did not put RS/OSE in for those test scenarios, as it's a BIG package. But I used the exact same newly installed oolite and addons to test both X800XTPE (works fine) and 4870x2 (slows down). I cannot see how an oxp could introduce a bug that's only experienced on more powerful cards or vista machines.

Solas wrote:
IMHO Virtualisation can cause all sorts of problems.
I wouldn't trust Vista to virtualise diddly-squat.

Even after that install to C:\oolite the program has "virtualization" flag under task manager.

Funny thing is that I finally noticed that my razer copperhead mouse drivers do the same! I've got an ongoing issue with them since I have this vista machine: Upon booting and resuming the mouse has a chance of approx. 90% to first be initialized and then go playing dead. Solution is to unplug it and then plug it in again. I had the same problem with another computer running XP32 (old machine with intel chipset) but my XPX64 machine did not run into such problems. They could not reproduce...but now I have been able to ask them for a rebuild of their drivers to get rid of virtualization, in the hope that this solves my issue with them. Since the mouse is first initialized and then goes defunct, I can very well imagine that this is beause vista reaching a boot state where the mouse driver would want to load it's files but fails because the files are in the users virtual store instead of the program files folder.

There really should be some random hits contract to punish someone for creating virtualization! Even razer did fall for it and did not notice for over a year!

Solas wrote:
The crash with GPU Caps Viewer fur rendering has happened with NVidia and ATI.
that issue may be due to overheating of a particular part of the board.

It's definitely no heat issue. If I run the card in auto mode, it tries to keep as close to 85C as possible, not going over that point, but also not cooling below 80C. If I enable manual fan control, I can get the card to below 30C.

I only once have been able to get rid of several graphics display issues and got FurRendering to work - that was by using the outdated HIS driver package for their ATI 4870x2 2GiB card. I had the same driver version from ATI and it did not work that way, it's as if HIS is somehow modifying (fixing) the ATI drivers. That would explain why they lag a few months behind. It did not give me the speed of the card, though.

The really stupid thing is that it doesn't even matter if I run the card in one monitor mode (both GPUs should be used for acceleration) or two monitor mode (one GPU only). It also doesn't affect performance if I launch other applications like webbrowsers or play back a movie while attempting to play.

The card just goes to it's lowest speed stepping and then Oolite slows down when there's much to display. Everything else runs fine and fluent, while oolite begins to miss keypresses because it cannot scan the keyboard in time due to the slowdown!

Screet

Solas · Post by **Solas** » Tue Feb 24, 2009 7:15 pm

I re-read an article I mentioned before ..
http://technet.microsoft.com/en-us/maga ... 6.uac.aspx
turning off UAC should de-virtualise Oolite files.
Control Panel - User Accounts - Turn User Account Control Off (deselect)
this will hopefully have you flying fast again.

It seems stupid that Vista must be entirely down-configured to XP level
( - UAC - Aero - Compositing ) to allow Oolite to run properly.

btw .. well spotted with the Copperhead .. I had the same problem ..
its in a box somewhere

Screet · Post by **Screet** » Tue Feb 24, 2009 7:42 pm

Solas wrote:
It seems stupid that Vista must be entirely down-configured to XP level
( - UAC - Aero - Compositing ) to allow Oolite to run properly.

btw .. well spotted with the Copperhead .. I had the same problem ..
its in a box somewhere

Aero turns itself on and off quite nicely here...?!?

Concerning the idea to turn off UAC: As you might guess I had to entirely reconfigure my mouse. The razer driver suffers seriously from this problem. As to the question wether this will allow proper detection of the mouse by the driver during startup (before logon) I've yet to see...but this time I did not have to unplug it!

Oolite didn't get any further benefits, though.
The crash of fur rendering wasn't addressed by this either.

Thus, if the razer driver works better now, I can use this as a workaround, but if razer finally manages to improve their vista driver, I guess I can and should switch UAC on again.

Did you report your troubles to razer? Maybe they did never fix it because I was one of some few who did report that problem?

Thank you anyway for all your suggestions!!!

Screet