570091D
Xtreme Addict

Actually I was referring to your comment about cryengine 2. You were complaining about someone lacking basic knowledge before posting when a quick glance at wikipedia would have let you know that cryengine 2 = crysis 1 and cryengine 3 = crysis 2. Far Cry was cryengine 1. Maybe you just assumed that because it was the second Crysis game that cryengine 2 would be the name of the engine, but the "cry" in cryengine is from the name of the company (crytek) and not the game (crysis). So it actually is logical that crysis 2 = cryengine 3.

ahhh, i see your point, i can't imagine trying to program for 240 threads at once.... so i suppose it's at it's development stages and hopefully we can see some sort of cuda2.0? or perhaps nvidia will find it in their interest to help with opencl development....

when i used avivo i was going from divx (mpeg-4?) to mpeg-2, and it was horrible. i'm really not lying, it was as though it had completely hashed the entire movie and forgotten about the sound.... perhaps it was an early version and it's better now, but i already bought badaboom and it works perfect for me. another thing, when i opened windows performance viewer, it showed the cpu as completely loaded while running avivo. so is avivo a cpu bound program with the gpu helping, or is it really a gpu based program? finally, i guess i really can't compare the speed of a 9800gx2 to a 4850, but badaboom will transcode a movie about 4x faster that avivo would. maybe the 320 shader limitation?

good idea to move it to profiles. what im saying about cuda is.

1) if u try to code for it its some sort of magic, it uses native C and u have to program each thread manually so that means that u have to find a method to use a bunch of threads, so i dont see it catching on in its native form as a dev platform for full GPGPU apps. but using 16 shaders or so would be a nice supplement.

2) the AVIVO transcoder, it dose nothing more than dx and cpu operations like i said and according to expreview it is faster than the pay NV client, and its higher quality. and from what ive messed with it worked flawlessly at a good IQ. im not sure about the problems but did u rip batman or get a pre transcode, and what input codec did u use

3) cuda can use openCL on teh quatros right now, so CUDA has no point since the server apps are already there in openCL and thats what was intended for the GPGPU


and the folding isnt exactly cuda its uses some cuda but mostly conventional gpu just like the ati client but the ati client still uses 320 shaders or so and need a complete overhaul