What was good about NV3x?

There been so mutch bashing of the NV3x series(and to most of it i agree)
but if you were to name a few "good" things about it, what would it be and do you still think it´s good?
 
As a gamer or as a developer?

[edit] I'm talking about the original NV30. As a gamer, it just plain sucked. As a developer, it means 1.5+ years of experimentation with FP32, experimentations that hopefully will show real benefits something like a year from today.
 
overclocked said:
There been so mutch bashing of the NV3x series(and to most of it i agree)
but if you were to name a few "good" things about it, what would it be and do you still think it´s good?

Take a look at this comparison in terms of performance:

http://www.beyond3d.com/previews/nvidia/nv43agp/index.php?p=6

Reverend said:
As a developer, it means 1.5+ years of experimentation with FP32, experimentations that hopefully will show real benefits something like a year from today.

"Experimentation with FP32" is meaningless, since its just an internal precision format. Had you said "Experimentation with longer shaders" then you may have a valid point, but the performance of NV3x with FP precision shaders was probably counter productive to that.
 
I should have been more precise, for me atleast i think more of NV35 as NV30 was cancelled so fast and replaced with NV35.
 
I personally liked its non angle dependent AF, I am not a fan of the AF implementation in the NV4x and r300 +. However much of that AF edge it had was lost when they started implementing texture stage optimisations and trilinear optimisations. It also had pure super sampling modes (Which the Nv4x does as well) which were good at its time.
 
DaveBaumann said:
"Experimentation with FP32" is meaningless, since its just an internal precision format. Had you said "Experimentation with longer shaders" then you may have a valid point, but the performance of NV3x with FP precision shaders was probably counter productive to that.

There was surely still something to be taken from being able to work with 32 bits of precision in the pixelshader, research if you will, which competing hardware didn't give you?

Usable or not in terms of performance, it still let you have a peek at effects that required more than FP24 (of which there are no real-world examples maybe, but that doesn't mean meaningful work on them for future hardware couldn't have been done, using NV30/35/38 ), imho.
 
There was surely still something to be taken from being able to work with 32 bits of precision in the pixelshader, research if you will, which competing hardware didn't give you?

Realistically the only "research" along those line would be "what shader attributes require FP32 to not be lossy", which in itself doesn't really tell you much.
 
DaveBaumann said:
Non-angle dependant AF was already a feature of NV2x.

But couldnt realistically do it either. The NV3x was the first card that had realistic performance factors with it. I know you're not going to tell me the Nv2x had usable AF. Lets not forget the notorious bilinear AF bug in the NV25.

Besides, your point is irrelevent since he asked what was "good" about the card. Not what exclusive good features it had.
 
DaveBaumann said:
There was surely still something to be taken from being able to work with 32 bits of precision in the pixelshader, research if you will, which competing hardware didn't give you?

Realistically the only "research" along those line would be "what shader attributes require FP32 to not be lossy", which in itself doesn't really tell you much.

Not even, since FP24 might have been fine. More like lots of painful practical research in the limits of FP16 (i.e. all that 'try it at _pp first' coding)
 
if you were to name a few "good" things about it, what would it be and do you still think it´s good?
I had an absolute field day going to town on fanboys trying to defend the undefendable! :D

It wasn't a complete failure, it provided me with oodles of entertainment!
 
ChrisRay said:
Besides, your point is irrelevent since he asked what was "good" about the card. Not what exclusive good features it had.

If thats the way you took the question then why not just include things like texture mapping, filtering, Z buffering, alpha blending, triangle setup, transformation, lighting, etc., etc. ;)
 
Couldn't any 32pp expriementation be done in software?? If so would that reduce the need for hardware in 32pp? I can see good points to having the hardware however.....
 
DaveBaumann said:
"Experimentation with FP32" is meaningless, since its just an internal precision format.
Um, you could render to and read from FP32 buffers in OpenGL with the NV30, and I believe can now in Direct3D.
 
jb said:
Couldn't any 32pp expriementation be done in software?? If so would that reduce the need for hardware in 32pp? I can see good points to having the hardware however.....
You mean FP32? No. GPU's don't support the bitwise operations that would be required to support a custom floating-point format.
 
Back
Top