Nvidia

[Devin Johnston]

Does anyone know if Nvidia releases cards at a certain time every year, I'm asking because I don't want to purchase a bunch of their top of the line cards and two months later find out they just released new ones.

[AJLynn]

Interesting question. I'm looking here: http://en.wikipedia.org/wiki/Comparison_of_Nvidia_graphics_processing_units and there doesn't seem to be a really well defined pattern. A couple of consistent thing for the last few generations: new lines are launched about a year and a half apart (the 9000 series doesn't count, it wasn't really a new line) and they start with a high end card, then fill in the lower tiers, then half way through the cycle slip in something a bit over (but not a hell of a lot over) the card that started out at the high end, and some whack job dual GPU high end card.

 

If they come out with a GTX 490 or 495 dual GPU you won't have to feel bad about it because it will burn too much power and cost more than the 480.

[Devin Johnston]

That makes me feel a little better but I think I'm going to wait until the official release of Vray 2.0 before buying anything, maybe ATI will come out with something better. My main concern is the limited memory on these cards and the Tesla's are too expensive.

[Jeff Mottle]

Generally you'll see new technology release to the consumer products first and then 6-12 months later it comes out on the professionals series products. That's how Fermi came out.

[AJLynn]

Is the memory an issue because you have CUDA software? The 2GB version of the GTX 460 seems appealing.

[Devin Johnston]

Yes I think memory is going to be a big deal which is why the Tesla cards are so appealing but their price is just way to high, I've looked on the Nvidia site and they only show two 460's one is 1GB and the other is like 800MB.

[AJLynn]

You can find the 2GB version here: - I'm pretty confident that's the memory you can get on a reasonably priced card with CUDA. The Quadro 4000 is a lot more expensive and also has 2GB... whether it's faster in CUDA would be hard to say. Compare:

 

Geforce 460 GF104 GPU 336 cores 675 MHz

Quadro 4000 GF100 GPU 256 cores 475 MHZ

 

Now direct comparisons of cores and MHz are tricky because the layout of the "cores" (and the use of that term is... well, I don't love it but it's the closest thing... maybe "threads" is better) into units and sub-units is different and it all effects the efficiency, but until some real testing data is available I'm going to have to guess that the Geforce 460 is more powerful. The GF104 supports all the CUDA versions currently used so that's not going to be an issue either.