Wow, I have 5 minutes of Internet fame. My car hit Wykop's main page [PL] (Polish version of Digg). Pretty impressive.
I was curious why my FPS had dropped to unacceptable values. I have a lot of optimization coded, so I've checked them all. And that's what I've found:
if (object->culled)
{
// continue;
}
Nice hack, isn't it? And I don't have slightest idea why was this line commented out.
I wonder how many games have a multiplayer that doesn't really suit them. And why? Does every publisher really require it (especially in PS3/X360 games)? I’m aware that most developers are afraid of game’s reception (about both gamers and critics), but is “shitty multiplayer” really much better than no multiplayer at all? Yeah, I can see possible opinions like “dude, this game is only 8 hours long and has no MP”, but… aren’t two additional hours of SP more attractive than multiplayer nobody would play anyway?
My final point is that there are games that are simply multiplayer friendly or unfriendly. For example most story-based FPS have very poor MP. FEAR2 is a great game if we talk about “story mode”, but what about its online mode? Yeah, it’s just totally average, and has no relation with original game whatsoever. Gears of War, Crysis and other similar games - same story. Stop. GOW has great co-op mode. Yeah, but it’s double single-player mode, not multiplayer one. But co-op is a good option for story-based FPS. On the other hand we have action-based shooters like Unreal Tournament that feature poor single-player mode.
Finally, there are “bonus” multiplayer modes, like in Red Faction: Guerilla. The team demolition mode is something completely different, but at least fun.
My simple raytracer is 10x faster than NVidia OptiX!
21.10.2009 05:20 in raytracing, 3D graphics
It's frustrating. I coded this sample about month ago. But the very next day I've upgraded my system to 64-bit Windows 7 and my raytracer magically refused to work. Finally I've got some time to fix it and post.
Basically it's the same demo that NVidia's using to promote OptiX. First of all, you can run OptiX samples on GeForce cards (not only QuadroFX) by disabling video card name check (very sneaky, NVidia!). So did I. And NVidia sample was running at 3-4 FPS, but it needs a lot of frames to produce final results (it's simillar to progressive JPEG). On the other hand, my very simple raytracer coded in CUDA easily hits VSync (60 FPS) using higher resolution and instant multisampling. Sure, my raytraces is not pixel accurate with OptiX, but its a matter of setting lights, objects positions and other stuff. And why does OptiX perform so bad? I have no idea.
You will probably have a lot of problem running these demos (I have no luck with release builds). But if you have Visual Studio and CUDA SDK you have chances. x86 build runs fine on 32bit Windows XP. x64 build runs on 64-bit Windows 7 (and probably Vista). Visuals in x64 versions are somewhat different from image in my post, because I needed to backport some code.
PS. I love 7 driver model. NVidia drivers (191.07 in my case) are still very buggy (or rather unstable) when running CUDA code, and automatic driver restart is much better than total system crash.
PS2. I'll try to find some more time and finally post my OpenGL 3.2 multisample demo source code.
