HOT HOT HOT!

As some have noticed, we've been pretty much living in the forums.  The support team has been keeping statistical track of how many users run into problem X, Y, Z.

In our non-game stuff, any serious problem that affects more than 1% of the user base is considered serious. Anything that affects more than 5% of the user base is considered critical.

One problem that has fit just over the serious but just below the critical has been game crashes.  And the logs have confused us because they don't make sense.  Why would these people have the game crash? Crashing should be reproduceable to some extent.

And yet, here we are with about 2.5% of users running into interittent crashing. 

The update (1.0D1) made the game better for most people because it fixed bugs that were found. But for some people, it actually crashes on them more than 1.0D - digital release).  Why? We had done some performance optimizations but it's really straight forward stuff.

But then..

Hmm...

During development, one of our mantras was to force everyone to run and test on lower end hardware so that we could "know their pain".  Too many games come out these days that require state of the art hardware to run.  So we mostly played and tested on lower end hardware. (not exclusively by any means, I did all my AI testing on a state of the art machine with an ATI X1800 in it).

But we noticed something in the logs -- most of the people who had these random crashing issues had really REALLY good computers.  A statistical anomaly.  A couple of people noted that the game runs their GPU very hot. VERY hot.

This week I've been playing the game and monitored the temperature on my ATI Control Panel applet (which, coincidentally, is built using licensed Stardock technology from the other side of the company <g>).  My GPU's temperature was 85C!  That's really hot!

But it's no biggie for me.  I have a Dell, it's got 50 fans on it. But what would happen on a machine that wasn't as well vented? I know what would happen: The game would start crashing -- randomly.  And it would happen relatively quickly because it reaches that high temperature within 10 minutes give or take.

So why don't other games do this?  That depends on the game of course.  Some games, like Halo, throttle their FPS. Other games simply don't use nearly as many video card features do.

I know strategy games are notorious for crappy graphics, but GalCiv II uses it all. I mean ALL OF IT.  Specular Lighting? Check. Bump Mapping? Check. Zillions of lighting passes? Check. Stuff I don't even begin to understand? Check.

If you read the entry that showed a screenshot of that Bird of Prey in teh game, that should tell you somethign about the engine.  One of my graphics guys told me that model may have over a million polygons.

The art team found a whole bunch of Star Trek models that were designed for video production (i.e. NOT for being displayed in real time).  They literally had 20 different ships in the game, each with an absurd number of polies. They ran.  To put it in perspective, a typical ship in GalCiv II only has around 600 polygons.  So there is a lot of room to grow.

Which brings us back to our heating problem.  What can be done?  The answer it seems to me is frame rate throttling.  So that's what I did:

if(ulElapsedMilliSeconds < 10 && bThrrotleOoption == TRUE)
    Sleep(5);
EnterCriticalSection(&g_csD3DDevice);
Draw3D();
LeaveCriticalSection(&g_csD3DDevice);

By doing this, my temperature stayed at around a more normal 65C.   We could probably be even more stringent.  I mean, if the time between the last frame is less than 10 that means the game's running at >100fps.  Realistically, anything over say 40 to 50fps is probably pointless.

But this way, people's machines will stay cooler.  And there will be another benefit: Laptop users again.  I'd like to put in a thing that lets us REALLY throttle it for low power usage.  I.e. >25fps is throttled (optionally).  This way, I could play the game on my ThinkPad all the way from Michigan to California on a plane trip.