It's that damn Builder Brush causing my crashes! (Testers needed to verify) - by Hit Deity
Destrier on 6/3/2005 at 14:35
Dell XPS: P4 3.4 ghz - 2 gig RAM
ATI Radeon 9800 pro
ATI Drivers Tried 4.11 and the Cat 5s.
Sun 24" Monitor
My experiance with Running T3Editor has been, like other Radeon users, very problematic when subtracting or modifying brushes. I get complete crashes and occassional hangups consistantly as described here and in several other threads. I also get these crashes/hangs when adjusting brush shapes via vertex editing. First, it works, then you can't grab handles in some windows (notably top) and then crash by the 3rd or 4th try.
It appears (after extensive experimenting) that it does not matter whether or not the 3D window is focused on the objects or not... the crash is inevitably consistant- focusing the 3D window forces it for sure- but it will ultimately happen regardless.
The error traps all point to the ATI drivers in each case.
Now, I have with some success, implemented a strategy of one large subtraction and then many adds, but this is limiting at best and, further, without the ability to modify brush shapes, in the end, somewhat futile.
I should also report that the Unreal Editor for UT2004 also has problems with my Radeon configuration, but it does not crash- it just doesn't work- no error messages are trapped. Admittedly, this is a guess on my part (that it is related to the ATI) but one which I'm fairly certain of at this time.
Downstairs, I have a P3 1ghz HP Pavillion with a GeForce 3 nVidia card and the Unreal Editor works fine (despite a rather slow startup). We are experimenting with loading T3 on it- not sure if it will run. Will provide a report on that activity when complete.
Since I have had two other significant problems with the ATI card (had to give up on KOTOR II) I will probably obtain a nVidia 6800 for this machine in the near future. Near being measured in days...
Hope this information is helpful.
Good Thieving to All.
scumble on 6/3/2005 at 14:48
Destrier - to confirm, did you use the latest ATI drivers?
I'm trying to figure out why my T3ed is bulletproof while doing crazy things with brushes. I've tried in vain to get the thing to crash. I've done subtractions all over the place, intersected cylinders with a bunch of other brushes and subtracted the result. It does seem easier when the Flesh Renderer is off, mind you.
The only odd thing that happened was fixed with a Build All - for a moment my 3d viewport seemed to be exploding with polygons.
However, it does occur to me that I didn't install the AI bug fix patch - this could be important, but I can't see why...
bukary on 6/3/2005 at 17:24
I use Cats 4.9. No crashes so far... I hope it will stay that way. :sweat:
I also know someone who uses Cats 4.4. He also did not experience any crashes...
Destrier on 6/3/2005 at 22:00
I have used 4.11 (6.14.10.6490) and 5.2 (6.14.10.6512 - released in Jan 2005). In this test. Same results on both without fail. Both sets yield the same debug message pointing to the ATI drivers.
We are encountering some difficulty testing this on the other machine due to drive space limitations (working it)- We siitting on an nVidia 6800 GT card right now waiting on the results of my tests downstairs. Will try that if the test passes downstairs and post results. May I ask (scumble and bukary) for comparison, what settings you have your Radeon Direct3D on? Thanks.
Destrier on 7/3/2005 at 02:09
Thank you Scumble: After duplicating all of your settings (all but one- I have a slightly different control panel and was unable to set TRUFORM) and conducting the same tests, the XPS with the Radeon card and drivers crashed as it did prior.
However, after wrestling with the slower machine and a few other items, we managed to mirror the exact same configuration of Thief3/Thief3Edit on the HP Pavillion 1 ghz machine with the nVidia GeForce3 card and drivers: The result was that editor worked in all cases and all tests that had resulted in a crash on the ATI chipset/card worked fine for the nVidia card on the other machine. In fact, its still running and a complex geometry is being worked on.
One last test- tomorrow we will install the nVidia 6800 GT on the XPS- (this machine)- if it runs- its a clear indication to me that my configuation with this machine and the ATI card are the cause of the problem. Unhappilly- it does not provide the specific reason for the incompatability- just the localization of the problem to this card, the 4.11 or 5.2 driver set, and this XPS. In all cases where errors where trapped, the ATI drivers where fingered. These informal tests seem to bear this out.
I would love to know what the problem is- but at this point it may be beyond my ability to drill the problem down to a lower level with out a DAS or some other potent bit trapping piece of test hardware--- that some instruction is not being error handled properly seems the culprit, but I am at a loss to describe the exact conditions causing such nor am I able to explain why some ATI users appear to have no problems. At this point, my observation is that, ATI Radeon users should, to be on the safe side, anticipate experiancing problems with the Thief3 Editor.
On the upside, I do recall working on the ATI Radeon problems that Flight Simulator 2004 (FS9) had right out of the box- and it took us several weeks before we hit upon a solution. I do hope that something similar happens here.
Will report my findings with the nVidia 6800 GT when they are in. In summery- I can, at least with the hardware I have available here, confirm that there are issues with the ATI Radeon 9800 Pro (XT) whereas the nVidia GeForce3 card seems to operate without the crash issue related to the subtraction of brushes- certainly in the cases I've tested here.
Again, hopes this helps... new video cards are anything but a... steal...
Good Thieving All:)
jay pettitt on 7/3/2005 at 03:19
You don't need a new video card.
Simply switch to wireframe view (there are little buttons of a wireframe boxes on the top of each viewport) for performing brush operations, switch back to textured view modes for doing other things.
Scumble's settings don't work magic for me either..? Radeon 9000s must have extra goodness ;)
Durinda D'Bry on 7/3/2005 at 07:01
Well, about drivers: at least for my config (Radeon 9600XT) I tried Catalyst 4.1, 4.4, 4.12 and 5.2. And problem appears for all these versions. Perhaps it also relates chip family (Radeon 9000 and 9600 are very different chips). Generally this bug seems as some incorrect data is transferred via DirectX from editor to driver and driver error handling is incorrectly implemented: I never heard about similar problem with other Unreal Editors and never had the similar behaviour with any games on my config.
About workaround with wireframe view usage I would add also: it is possible to see "good looking" preview from this mode using Flesh Renderer (note that Flesh Renderer enabled from Wireframe view and from Textured view are 2 different things!). In Flesh Renderer it is possible to use Fullbright Mode. But for texturing we still need to switch to textured mode. Before this switching geometry should be built to avoid possible lockup!
scumble on 7/3/2005 at 09:58
Is anyone else running a 9000 Pro then? I can easily believe it has something to do with the different GPUs - most of you who are having trouble have "better" Radeon cards than me.
This also means that the next card I buy won't have an ATI chip ;)
I think we are rapidly coming to the conclusion that the drivers
specifically for 9600, 9800 and so on are causing trouble, so those who don't want to buy nVidia chipset cards have to make do with workarounds.
If someone could work out exactly what was happening, it may be possible to petition ATI for a patch (a vain hope I know).
Quote Posted by Destrier
Will report my findings with the nVidia 6800 GT when they are in.
Do. I think it might be useful to have a page on this issue for people looking to buy new cards. Much misery could be saved with a little advice...
I put up a page (
http://www.ttlg.com/wiki/wikka.php?wakka=GraphicsCardIssues) here with a preliminary list of cards. I'd encourage anyone who has problems to add to the page, or list. There's a link to a (
http://www.ttlg.com/wiki/wikka.php?wakka=ATIWorkarounds/edit) workarounds page, which is ready to be created by someone who has had to work with them. jay perhaps :)
Destrier on 7/3/2005 at 14:17
We did, in fact, test with wireframe and with the 3D window changed to T, F, S, etc. Although the results were slightly different for each on the configuration we tested with, ultimately the error occured and the traps pointed to the ATI drivers. In fact, we tried quiet a few settings and all with the same results. I will say that, observationally, increasing the virtual memory from approximately 2 gig to 3 gig did have a postive effect on basic box-brush subtractions- but when we modified the brushes with any angular lines- the crash occurred, and occured repeatedly.
Due to a large film project currently underway- I have extra video cards that I have been using to try to solve an editing effects problem. Hence why I have the 6800 GT on hand to try this so that test is in the offing for tonight.
We did run on the PIII HP with the editor for four hours solid doing all sorts of brush transformations and volumn add-subtracts and the editor did not crash once from that- (it did crash once but that was wholly user error).
We have not repeated any tests using UT2004's editor on the ATI P4 XPS.
I concur that it would be good to provide a consolodated set of findings for others to gauge video cards with. Personally, I have been happy with the Radeon but begining with several recent releases, I have noticed an instability or, perhaps, simply, an incompatability between the card and the latest apps. It cumulated with serious crashes on KOTOR2 and issues with our film editing. We are conducting separate tests to look at that problem, which is with video capture- a problem which also seems localized to the ATI card. Its dumping several frame buffers worth of data during transfers from the editor- significant data loss- about 2.5 NTSC frames worth (very noticable) which is why we have the other cards on hand for testing.
Will post results when they are in :)
