Originally posted by: Rollo
Gee, Pete, excuse me for breathing.
And excuse me for going off on you, but don't tell me I'm not the tiniest bit justified for doing so.
I thought running the game in SM3 with VTF might be faster than running it in SM2 without, I've gotten some sizable speed gains from geometry instancing in the past.
Seem like a reasonable question to me?
GI is meant to increase performance by reducing (driver?) overhead. VTF potentially reduces performance b/c it introduces more complexity (geometry) to the scene. I say potentially b/c the idea behind VTF, at least according to NV, is to take advantage of unused VS cycles when a game is bound by PS or other performance. But, IIRC, using VTF in PF reduces performance.
I never said SM2 had VTF, so don't put words in my mouth.
I didn't. I know you know VTF is a SM3 feature, but Xbit pretty clearly said the ATI cards were running SM2 mode. So how did we arrive at your thread title?
One thing is for sure PF runs like ass on ATI cards, so whether it's poor GL drivers or not being able to run in SM3 and take advantage of geometry instancing, it's noteworthy to people who play flight sims?
It runs slower, but not all benchmarks I've seen show it running like ass.
It keeps up fairly well without AA+AF (mostly w/o AF, from R3D's benches), but bombs with AA+AF. I'm not even sure it uses GI. The terrible AA+AF performance is noteworthy, yes, but the issue seems to be ATI's typically poor OGL performance, not missing SM3 features. The fact that AA &/ AF cause the biggest drops might mean some "legit" memory controller tweaking can fix things.
You're just turning into another person who believes they're in some non existent struggle, with a "duty" to call out the "Anti-ATI Infidels", which is pretty sad given what you used to be.
Eh, hyperbole aside, maybe. But you've yet to apologize or otherwise acknowledge that you're technically wrong, which is what I'm after. I guess we have to wait for Xbit to reply to you.
I created this thread to discuss ATI leaving a SM3 feature off the R520 when they claimed it would have it originally.
Nit-picking again, but did they?
Anyway, just to end on a constructive note,
there're some new .dlls out from the Maddox crew that makes water look better with ATI cards. The interesting thing is that ATI still can't run the two highest water IQ levels b/c they lack texture pre-fetch. I'm not sure if that's another term for VTF, a separate (PS) feature, or simply the mod misspeaking (b/c the NV pdf below makes it pretty clear it's using VTF: "They combine multiple dynamic normal maps to calculate the appropriate geometric displacement at each vertex.").
It'd be nice if we could see some comparison screenshots for both ATI and NV on all levels.
BTW, NV's Vertex_Textures.pdf says that "DirectX 9 also supports vertex textures in software vertex processing mode, so an application could use vertex textures even on hardware without native vertex texture support." I wonder if ATI could get this PF water effect to work with their dual-core-optimized drivers, or if NV can speed things up with theirs.