GeForce vs. Quadro for CUDA

[phaxmohdem]

I was wondering if there is any difference, or benefit that would make purchasing a Quadro card more desirable than a cheaper, similar GeForce card for CUDA programming, and other GPGPU shenanigans.

 

[Crusty]

You need to look at the specs of the GPU chips on there and compare what resources you have available. I'm not familiar with the differences between them, but im sure they have different amounts stream processors and probably has a different memory architecture.

 

[phaxmohdem]

Granted. Lets for this thread use as an example the following cards then:

Quadro FX3700 512MB & Geforce 8800GT 512MB

Both cards have nearly the same specs: G92 core, 112 Stream Processors, 512MB GDDR3 RAM, clock speeds etc.
Quadro card = ~$850
GeForce card = ~$135

I know that Quadros have better support and special drivers that graphical programs such as CAD, 3D Modelers, Simulation software... etc use. But do these drivers/benefits carry over to the realm of non-graphical, GPGPU computations?

 

[aka1nas]

I don't think you get any extra benefits with Quadros, but the newer Teslas have double-precision floating point enabled, while all of the "regular" GPUs only have single-precision enabled.

 

[wwswimming]

the FireGL & Quadro are not exactly a scam. they just charge a lot extra
to guarantee the drivers work. which you can test yourself.

suggest exercising the 8800 as soon as you get it, so if there is a driver
or compatibility problem, you can return it & just pay a 15% restock fee.

example -
http://features.cgsociety.org/...stom.php?story_id=4671

that's a review of the Quadro by George Maestri. as much as i admire
him as an animation instructor - he teaches Max & Maya, and does a
damn good job - i think he's neglecting to mention a very important
detail. you can sometimes match Quadro's & "regular" video cards
by eyeballing the heatsinks, and that heat-sink on the Quadro looks
a lot like the 8800GT heatsink.

but Quadro is advertised on the CGSociety website.

mostly, Quadro & ATI FireGL take advantage of managers who are
nervous about their schedules, have been bit in the @$$ by hardware
glitches, and have equipment budgets.

10 years ago, the workstations i used cost $50K. now, they cost
$5K, and you can get equivalent performance for $1.5K.

managers in the field are used to spending the bigger bucks.

 

[Modelworks]

Originally posted by: phaxmohdem
Granted. Lets for this thread use as an example the following cards then:

Quadro FX3700 512MB & Geforce 8800GT 512MB

Both cards have nearly the same specs: G92 core, 112 Stream Processors, 512MB GDDR3 RAM, clock speeds etc.
Quadro card = ~$850
GeForce card = ~$135

I know that Quadros have better support and special drivers that graphical programs such as CAD, 3D Modelers, Simulation software... etc use. But do these drivers/benefits carry over to the realm of non-graphical, GPGPU computations?

No advantage whatsoever to Cuda.

However I do think OPenCL would be something to look at if your just getting started with Cuda as I think it is going to see a much wider usage.