Wetterberg

I am considering getting a new dual core 2 machine with a beefy GeForce 

Now, since I am moving from a lowly laptop to a fully fledged graphics 

machine I am expecting some improvement in performance running the 

openGL things, but is there any idea of how vast the change will be? The 

card has 640 megs of onboard ram and fits a PCI-express slot. The blurb 

even says: "Crank up 16x full-screen anti-aliasing without missing a 

beat - no jaggies, no worries", which of course is music to my ears.

I really want a system that can handle the shader end of things without 

problems, but just how much punishment can an upscale "enthusiast" gpu 

handle? Is it a case of "more ram and flops = better openGL in Jitter" 

or does the cpu factor in a lot when you reach that stage - will some 

amount of the shader/slab/polygon/etc processing always be handled on 

Oh, and are there any circumstances where heavy-duty gpus can help boost 

I know this must be along the lines of questions asked a million times 

over, but if anyone has some insights into this or examples, that would 

I am considering getting a new dual core 2 machine with a beefy GeForce 
8800 GTS in it.

Now, since I am moving from a lowly laptop to a fully fledged graphics 
machine I am expecting some improvement in performance running the 
openGL things, but is there any idea of how vast the change will be? The 
card has 640 megs of onboard ram and fits a PCI-express slot. The blurb 
even says: "Crank up 16x full-screen anti-aliasing without missing a 
beat - no jaggies, no worries", which of course is music to my ears.

I really want a system that can handle the shader end of things without 
problems, but just how much punishment can an upscale "enthusiast" gpu 
handle? Is it a case of "more ram and flops = better openGL in Jitter" 
or does the cpu factor in a lot when you reach that stage - will some 
amount of the shader/slab/polygon/etc processing always be handled on 
the cpu?

Oh, and are there any circumstances where heavy-duty gpus can help boost 
performance of matrix processing?
I know this must be along the lines of questions asked a million times 
over, but if anyone has some insights into this or examples, that would 
be great.

geforce-8800-gts-overkill

I haven't gotten my hands on an 8800 yet, but it will definitely give

you a massive performance boost when using shaders. It has a unified

shader architecture meaning that vertex, geometry, and vertex shaders

all run on the same chunk of hardware in the GPU. You won't be able

to use the geometry shader end of things from within Jitter until

those hooks are in the library, but for your more standard vertex and

fragment shading, it will be very fast. As for speeding up matrix

processing, if you're talking about jit.matrix, jit.xfade,

jit.alphablend, etc. then no. These are CPU operations, having

nothing to do with your graphics card. You can however do more of

this stuff on the GPU with an 8800 if that helps.

Hi Andreas,
I haven't gotten my hands on an 8800 yet, but it will definitely give
you a massive performance boost when using shaders.  It has a unified
shader architecture meaning that vertex, geometry, and vertex shaders
all run on the same chunk of hardware in the GPU.  You won't be able
to use the geometry shader end of things from within Jitter until
those hooks are in the library, but for your more standard vertex and
fragment shading, it will be very fast.  As for speeding up matrix
processing, if you're talking about jit.matrix, jit.xfade,
jit.alphablend, etc. then no.  These are CPU operations, having
nothing to do with your graphics card.  You can however do more of
this stuff on the GPU with an 8800 if that helps.

> I haven't gotten my hands on an 8800 yet, but it will definitely give

> you a massive performance boost when using shaders. 

Oh Wes, once again you make my day (or night, as it were). "Massive" is 

> shader architecture meaning that vertex, geometry, and vertex shaders

> all run on the same chunk of hardware in the GPU. You won't be able

> to use the geometry shader end of things from within Jitter until

> those hooks are in the library, but for your more standard vertex and

> fragment shading, it will be very fast.

Again, good news. Will the unified shadar architecture be supported 

within the foreseeable future? Do you know?

> processing, if you're talking about jit.matrix, jit.xfade,

> jit.alphablend, etc. then no. These are CPU operations, having

> nothing to do with your graphics card. You can however do more of

> this stuff on the GPU with an 8800 if that helps. 

Clumsily explained by me, but yes, that was what I meant. Haven't been 

able to do much of these things (yet), but I have studied up on 

videoplane mixing and so on. Seems very very capable, yet simple, I must 

say, and Vade has also provided some invaluable tips so far. I am truly 

psyched about this, and I hope to do some far-out vj stuff this year.

I hope I won't mail-bomb this list too much with the slew of shader 

Wesley Smith skrev:
> Hi Andreas,
> I haven't gotten my hands on an 8800 yet, but it will definitely give
> you a massive performance boost when using shaders.  
Oh Wes, once again you make my day (or night, as it were). "Massive" is 
a word I like to hear!
> It has a unified
> shader architecture meaning that vertex, geometry, and vertex shaders
> all run on the same chunk of hardware in the GPU.  You won't be able
> to use the geometry shader end of things from within Jitter until
> those hooks are in the library, but for your more standard vertex and
> fragment shading, it will be very fast.
Again, good news. Will the unified shadar architecture be supported 
within the foreseeable future? Do you know?
> As for speeding up matrix
> processing, if you're talking about jit.matrix, jit.xfade,
> jit.alphablend, etc. then no.  These are CPU operations, having
> nothing to do with your graphics card.  You can however do more of
> this stuff on the GPU with an 8800 if that helps. 
Clumsily explained by me, but yes, that was what I meant. Haven't been 
able to do much of these things (yet), but  I have studied up on 
videoplane mixing and so on. Seems very very capable, yet simple, I must 
say, and Vade has also provided some invaluable tips so far. I am truly 
psyched about this, and I hope to do some far-out vj stuff this year.

I hope I won't mail-bomb this list too much with the slew of shader 
related questions I'll have :-)
Cheers!
Andreas.


> Again, good news. Will the unified shadar architecture be supported

> within the foreseeable future? Do you know?

This is just some engineering mumbo jumbo to describe how the GPU chip

is designed. I think most GPUs are heading toward this type of

architecture. The only implication is has from a software standpoint

is more flexibility in how shaders are processed and thus a potential

speed improvement as well as less limitations on how many instructions

a shader can execute. Currently ATI hardware and some Nvidia chips

are limited in the number of instructions a shader can have. So,

there really are no issues with whether OpenGL supports it because

asking that question doesn't make sense since the architecture is

abstacted away from the OpenGL interface.

> Again, good news. Will the unified shadar architecture be supported
> within the foreseeable future? Do you know?

This is just some engineering mumbo jumbo to describe how the GPU chip
is designed.  I think most GPUs are heading toward this type of
architecture.  The only implication is has from a software standpoint
is more flexibility in how shaders are processed and thus a potential
speed improvement as well as less limitations on how many instructions
a shader can execute.  Currently ATI hardware and  some Nvidia chips
are limited in the number of instructions a shader can have.  So,
there really are no issues with whether OpenGL supports it because
asking that question doesn't make sense since the architecture is
abstacted away from the OpenGL interface.

GeForce 8800 GTS - overkill?