Pontifex makes Baby Jesus (and my Video Card) cry

Started by beaujob, October 24, 2001, 10:48:16 PM

Previous topic - Next topic

mendel

I commited an eror in my previous post because the score I referred to did not relate to a TNT2. I only have TNT2 scores for a P233 and K6-2/350 at this time, and the bench that is "constant" with the geforce2go is closely CPU-bound on these TNT2 scores, at these machine levels. This affirms Entroper's guess, but I'd like to have the 800Mhz TNT2 score for a more accurate analysis.

beaujob

I don't know if this has been mentioned, but wouldn't it be nice if for us neanderthals with TNT2's and Voodoos you could have a rendering option to just draw a single tube for each cluster of links.  This would reduce the polygon count 96-fold.  http://www.pontifex2.com/iB_html/non-cgi/emoticons/smile.gif" border="0" valign="absmiddle" alt=':)'>

The individual links could still break and stuff, and just have triangles start falling off the tube, or something.  Or, perhaps just have a construction option for tubular steel vs. box trusses.

"And once again, the day is saved thanks to the Powerpuff Girls."

You can check if your cpu or video card is slowing it down by pausing the game, this will stop all physics calculations.  If you can move the viewpoint around at full speed the cpu is slowing it down.  With most computers the cpu will be the bottleneck.  Also the lighting calculations can be slow on large bridges, turning off the lighting can speed it up.  The terrain on most levels is 128x8, which is 2048 triangles, but not that much compared to the number of triangles needed to render an average size bridge.

My cpu or video card  has no trouble with the game but that is a good idea.

I think it would increase the popularity of the game by making it run faster on slow machines


mendel

beaujob, going from high detail to low replaces each tube with a line, as far as I can tell. If that isn't enough to reduce polygon count, I don't know what is.

But... a 16x8 level has (at least?) 256 triangles of ground texture.... those big levels have even more ... must investigate this ... email mailto:pfxbench@mendelsohn.de">pfxbench@mendelsohn.de if you want to help with the experimental pfx bench ....


thedoc

I think Pontifex is more reliant on the CPU speed and not the graphics card. I only have a TNT2 card but I am running it on a 1.4Ghz AMD CPU and I have no problems at all at the highest graphics setting. (It runs very sweet indeed!)

beaujob

I guess I didn't describe my problem fully.  I've got an 800 MHz athalon and a TNT2 Ultra, and my friend here has an Intel 550 and a Geforce 2 Pro and his performance on large bridges with higher numbers of links (like, 3000+) is markedly improved, so my suspicion is that it's a video card issue.  Though, I have win2k and he's got win98, so there could be some processor choke from w2k.  Then again, maybe the game is optimized for Intel special instructions?  Any feedback from other people with similar configurations to mine would be appreciated.
"And once again, the day is saved thanks to the Powerpuff Girls."

Entroper

I'm the friend with the Intel 550 and the GF2 Pro.  http://www.pontifex2.com/iB_html/non-cgi/emoticons/smile.gif" border="0" valign="absmiddle" alt=':)'>  Thedoc, I'm guessing that your Athlon 1.4GHz doesn't have a problem handling all the transform and lighting that your TNT2 can't do, but his Athlon 800 does have issues with it.  Handling the physics engine for thousands of links, plus doing the polygon transforms for them is just too much.  You can tell that your computer isn't fast enough when "fast" doesn't speed up the train at all.

It's my feeling that a lot of optimization of the graphics engine could be done; for instance, optimizing the terrain rendering as mendel mentioned.  I'd be happy to offer any sort of assistance in getting the game to run faster.

Entroper
Programmer, Serious Fortress Team
http://www.seriousfortress.com

mendel

The xperimental Pfx bench (send mail to mailto:pfxbench@mendelsohn.de">pfxbench@mendelsohn.de for info) has a measure (result #2, zooming around) that is hardly CPU-bound (checked running the same laptop with 900/700 Mhz, geforce2go). Result#1 (rotating bridge) and #3 (test run) scaled proportional to CPU speed in this case.

(Edited by mendel at 3:40 pm on Oct. 24, 2001)