NV40 to support FP24?

I'll be home in fifteen minutes, Snookums.

 

Just in time for me to have to go pick up my son from school...kind of weird/scary, eh?

BTW-I checked out nVnews today and almost fell over laughing, there's a thread where some of the most raving nVidia fanboys seem to have found a rumor they believe telling 'em that the nV40 will be 8x2 and the R420 will be a true 16x1 and they're panicking and moaning and already giving the crown to ATi! I almost peed myself laughing, I'll hunt up a link.

EDITED BITS: Found it! "R423 will have 16 pipes(16x1) and will own nv40".

Sorry for the OT, I blame it on a lack-o-new-rumors.

 

Well actually you don't have to create new vertices on fly. Creating shadow volumes entirely on GPU is possible. The only question is cost.
http://www.mpi-sb.mpg.de/units/ag4/gallery03/stefan.brabec/
http://www.gamedev.net/columns/hardcore/shadowvolume/page5.asp

 

Ah! That's true, though in this case speed would be entirely related to vertex processing capabilities, so it's kinda off topic.

 

One reason for FP16 is filtering and blending in FP16 is cheaper than FP32.

 

Slightly less than a fourth of the size, correct, since multipliers grow a little faster than quadratically as the bit width increases?

 

What about integer support on NV40?

It is there, and what format FX12, FX16?

 

Excuse my ignorance, but what is the big deal about floating point texture filtering?

 

Atleast one GI approximatino needs float blend it involes placing a whole bunch of particles around or something I've seen ppl asking for it several times around places.

 

well if there aren't fx12 and fx16 i believe everyone that bought a 5200-5800 are pretty much screwed . I nvidia kills that off and starts replacing the current fx line up with nv4x based cards quickly (before the end of the year ) the devs wont spend as much time if any making special paths for these cards

 

What you say seem most likely jvd. But still perhaps there is still a little possibility that they'll support the nv3x generation for at least 6 more months. But then again, this has been mentioned as a high probability situation :?

 

Hey Tim!
Have they dropped fx12 and fx16?

 

I cant see nvidia supporting fp24 with all the r & d they've thrown into the nv3x generation. It would be going backwards as opposed to furthering future development. Ati have stated down the track they wish to increase precision, but chip size/speed/yield issues have limited current chips to fp24.

I wonder if nv40 will remove the integer/legacy pipes and go for 16xfp16 pipes which could equate to 8xfp32 pipes for workstation/full precision work (if they have some way of using 2x16=32 performance.. unlikely??). This would give the good quality and max speed for <dx9 and partial-precision pixel shader programs(doom3??), and offer the quality when needed??

FX16?? Does this even exist?
FX12 was lower than r2xx tech shaders in precision and thus a technology that was probably outdated before it even existed. If nVidia concentrated on fp16 from the start im sure that they would have been better off.

 

FX12 is right out. It does support FX16, though, according to stuff we heard earlier on that hasn't been contradicted.

 

Err... what is a "GI approximation"

 

AFAIK they've already dropped fixed FX from NV35...

 

IMO long time ago.

 

I'm not sure...Global Ilumination? - there was one approach to GI made on R3xx cards but i guess it didn't required FP blending

 

Yeah global ilumination but I dunno how this bazillion of particle method works I just remeber seeing ppl asking for FP blends to do it.