Problems running on ATI Mobility Radeon HD 4670 graphics

[jeffmack]

Fenerit: I had already tried disabling the anti-aliasing options within the Catalyst Control Panel, with no apparent effect either way. But thanks for the suggestion anyway. Thanks for the suggestion regarding the "AntialiasingSamples" setting within "celsesti.cfg" as well.

Selden: Thanks for the info - especially regarding the various render paths. Considering your explanation, it seems especially odd that I experienced immediate lock-ups under the "Basic" and "Multitexture" render paths, whereas things seem to be very stable under the "OpenGL vertex program" render path! But perhaps this has more to do with the fact that I have also changed the texture resolution from "Hi" to "Medium." I've kind of lost track of the sequence of my experiments! By the way, my graphics adapter OpenGL driver reports a maximum OpenGL texture size of 8192 x 8192.

As for now, things look pretty stable. I am using the "OpenGL vertext program" render path; "Medium" texture resolution; and anti-aliasing turned on in Catalyst and in the Celestia "render / Anti-aliasing" checkbox. Celestia has only locked-up once since then, and this was toward the end of running the included Demo script ("demo.cel"). Yet, another run of this script a day later ran smoothly to completion! Also, on one occasion, I once again saw some flickering colored bands in the background (my original problem before updating the drivers! Perhaps a future driver update will resolve some of these remaining problems. And your suggestion of submitting a bug report to ATI/AMD is a good one.

I really don't understand the relationship between the various anti-aliasing settings in Catalyst (driver settings); "celestia.cfg;" and within Celestia under the "render / Anti-aliasing" checkbox setting. And which have precedence? What overrides what? For example, I have left the "celestia.cfg" "AntialiasingSamples" commented out, meaning (according to the comments within "celestia.cfg") the default is "1," in turn meaning no anti-aliasing. Yet toggling the Celestia "render / Anti-aliasing" checkbox clearly is having the desired visual effect! Also, when this checkbox is unchecked, why don't the Catalyst anti-aliasing settings do the job?

One more question): You (and the comments within "celestia.cfg") mention that I can specify a list of "IgnoreGLExtensions." But the example only shows one ignored extension. What is the syntax for multiple entries in this list?

A)

IgnoreGLExtensions [ "extension1, extension2, ..." ]

B)

IgnoreGLExtensions [ "extension1", "extension2", "..." ]

or,
C)

IgnoreGLExtensions [ "extension1" ]
IgnoreGLExtensions [ "extension2" ]
IgnoreGLExtensions [ "..." ]

Many thanks again!

[jeffmack]

Ahhh! On the "IgnoreGLExtensions" syntax question - I just spotted an example with the various object catalogs near the top of "celestia.cfg." Should be the same. Sorry.

[selden]

The Render anti-alias option only affects how lines (orbital paths and coordinate grids) are drawn. It doesn't affect the drawing of 3D objects.

System level settings normally take priority over application settings. Which are used is explicitly specified in Nvidia's control panel. I don't know how Catalyst handles them.

[jeffmack]

OK, thanks.

[Fenerit]

There is another setting you can try to play around a bit to see if your system became more or less stable: the anisotropic filter. Even this filter should be handled from the Catalyst Control Center (also myself has a Nvidia card, so I do not for sure). High samples values can do either jerks or put zigzag artifacts on screen. Notice that there are not settings for this filter in Celestia.

[jeffmack]

Thanks Fenerit. I think that I changed it at the same time that I changed the anti-aliasing settings in Catalyst (ATI/AMD). What does the anisotropic filter do?

[Fenerit]

Naively speaking, the anisotropic filtering allows to constraints the texture mapping in its places for oblique views. It remove the polar pinch effect because forces to constraint "in position" the area of the texture's cylindrical projection which would tends toward infinite (poles). High values for old/broken graphic drivers can makes artifacts, though; by freezing the views and crashing the program.

[jeffmack]

OK, thanks again Fenerit.

[jeffmack]

ATI/AMD just released a new update of my graphics adapter driver - just one month after their previous update! Although it's still too soon to be sure, it's quite possible that they have resolved the problems that I was experiencing while running Celestia. After installing the new update, I turned back on both the "OpenGL 2.0 Render Path," and "High Resolution Textures." So far, so good - no strange artifacts, and no lock-ups of Celestia! I have not yet tried re-enabling some of the optional features within the Catalyst Control Panel, however - (such as "Adaptive Anti-aliasing" and the "Anisotropic Filter"). Perhaps later.

[jeffmack]

Well, it now appears that the updated graphics adapter driver did not solve all of my problems in Celestia. However, the most serious of them (the lock-ups of Celestia) does seem to have been fixed by the driver update.

The two remaining problems (the occasional flashing streaks of light in the background - and the strange gridwork of spiked, star-shaped, black, flashing artifacts in the textures of objects) still occur. I should add that I have recently noticed that, when the black, flashing artifacts are present, I can see the background stars through the object's texture. I still have been unable to screen capture these two problems.

But I now know why I thought the driver update solved these two problems! The problems disappeared for awhile after the update - not because of the update itself - but because of the reboot after the update!

I normally leave my system running for days at a time, and apparently after running for some amount of time, these two problems begin recurring. A reboot clears the problem for, say, a few days! It's as if the driver has a memory leak or memory corruption problem of some kind.

Has anyone ever seen (or heard of) a case where graphics problems occur only after a system has been running for awhile, or because of a graphics driver memory corruption problem? Or possibly after changing graphics driver settings without rebooting?

[selden]

Have you been monitoring the internal temperatures?

Problems due to overheating are common, due to vent blockage or dust. Dust builds up internally over time, blocking airflow and makes laptops overheat. Many people add external ventilation systems to help the airflow. Laptops are intended to run at higher temperatures than desktops, but excessively high temperatures can cause lots of intermittent and long-term problems.. Memory errors in particular are hard to detect since most systems don't use ECC memory.

FWIW, I use SpeedFan, but there are lots of other free temperature and fan control programs.

[Fenerit]

To the Selden's remarks, which indeed are rights, my personal suggestion is whether you runs programs DirectX based for long before to start Celestia (OpenGL based); being their graphic card's response very different and can do the drivers upsets when such environments are loaded in the same session. The computer's restart "polish" the supposed interferences, expecially whether after restart the first program you launch is Celestia; that is, the OpenGl driver is the first and the unique to be handled by the graphic card.

[jeffmack]

Thank you Selden and Fenerit for your helpful suggestions.

The laptop is a new one, and I've had the Celestia graphics problems from the beginning. It's located in a dust-free and smoke-free environment, with an ambient temperature of 65-68 degrees Farenheit. It also sits on a running external laptop fan. So I'm thinking that it's probably not a temperature problem. Also, the graphics problems seem to be more related to how long the system has been up without rebooting as opposed to how long Celestia has been running (Celestia is the only application that I run that kicks the fan into an audibly higher speed). Once I begin seeing the problems (say after a few days since rebooting), they will occur immediately upon launching Celestia. Rebooting clears the problem, and a reboot doesn't give the laptop time to cool down.

However, keeping an eye on temperature is certainly always a good idea. I have an application which can monitor temperatures and fan speeds (PC Wizard). I had SpeedFan on an older system; perhaps I'll download and install it again.

Selden, any idea what are acceptable temperatures, and what are not? I don't have a good feel for that.

I also don't think that DirectX is the problem as Celestia is the only graphics intensive program that I run (certainly the only 3D application); I don't run any games (other than PopCap's "Dynomite" - which is not 3D). So I don't think I'm using DirectX much, if at all. I do sometimes run Google Earth, but it's been several weeks since I've run it - so that can't be the problem either.

However, Fenerit's comments do lend credence to my suspicion that graphics adapters' memories can get corrupted by usage, so I'm still thinking that a buggy graphics driver is at the root of the problem!

[selden]

It sounds to me like the problem could be caused either by a software problem or a hardware problem -- something causing stored information to get corrupted. Unfortunately, I have never been able to find any diagnostic software which does a serious test of modern 3D graphics hardware. (Dell's firmware diagnostics only do rudimentary 2D VGA tests.) You might want to consider investigating some of the 3D graphics benchmark programs to see if letting any of them run for a long time reveals similar problems. Some are based on DirectX, some on OpenGL and some use both. If both types of graphics have problems, I'd tend to suspect a hardware problem, although that wouldn't really be conclusive. If only OpenGL (what Celestia uses) has problems, I'd tend to suspect the graphics drivers.

It could be a CPU/main memory problem, I suppose. Running an extensive memory diagnostic for a long time should find something, in that case, though.

I don't know how hot your graphics chipset should get. There may be some mention in its spec sheet or you might ask on an appropriate graphics hardware Web forum. The Nvidia 3100M chipset in my laptop typically gets up to over 70C, which supposedly isn't a problem.

Extensive use of 3D graphics drastically raises the temperature of graphics chipsets, since it uses much more of the circuitry than does simple 2D or even Vista & Win7's Aero graphics. Just how hot they get depends on the chipset, the design of the associated heatsink hardware, and how well the thermal coupling compound was applied. (Celestia is single-threaded, so it'll stress only one core at a time of modern multi-core CPUs, although that also lets i7s raise their clock speed to the maximum. Running only one copy at a time is unlikely to overheat a CPU.)

p.s. for what it's worth, my system was plagued by graphics/mouse/typing pauses and graphics driver restarts until I reinstalled Win7 from scratch, leaving out those proprietary Dell device drivers that I really didn't need. Since then it's been rock solid. Apparently there are problems in some of their factory system images.

[jeffmack]

Thanks again for all your help, Selden.

I'll continue keeping my graphics drivers up-to-date as well.

[selden]

Rebooting clears the problem, and a reboot doesn't give the laptop time to cool down.

That statement seems to be incorrect, at least for some laptops. I just now updated to the most recent stable version of SpeedFan (v4.42), noted the initial temperatures (about 55C), started up several copies of Celestia, ran them until the temperatures all stabilized (at just under 70C), then shut down all the copies of Celestia. The CPU and GPU temperatures returned to 55 in about 30 seconds, although the fans continued to run at high speed for somewhat longer.

[jeffmack]

Good to know. Who'd have "thunk" it?

I'm going to download and install SpeedFan soon.