speed-question regarding tesla/geForce

Hi,

some months ago I have changed from my geForce 5xx (don’t remember exactly) to the geForce 770.
It has 7times (or so) more CUDA-cores, so I expected a noticable speedup, when rendering with Cycles via CUDA.
Unfortunately there was no such speedup (a few percent of course).

I found out, that the customers drivers do not support more than x cores to execute CUDA-code.

Now, I am thinking about a Tesla-card, but before I invest some thousand Euro, I would like to be sure.
Does anyone have experience with a Tesla-card? Does it (or the drivers in fact) make use of all the CUDA-cores? How much faster does a Tesla render than, let’s say, a geForce 770?

Regards
Christian

Well for some reason Blender is not recognizing my Quadro 4000 card as well.

Gaming cards like geforce are restricted, while professional cards like quadro support more cores, that is what I was told while buying it in 2011.

I am myself very confused which route to go - cpu rendering or GPU. I do not think a tesla is worthy, you can spend that amount on building a small render farm instead with more processors.

check here

https://cgcookie.com/tutorial/setting-up-a-renderfarm/

Hi, the “Pro” cards are always slower than the gaming cards for Cycles or other render engines like Octane.
Quadro K5000 use the same core as GTX 770 but has reduced core clock and memory bandwidth for 24/7 usage.
The GTX 770 should be faster on complex scenes, need less memory and has no texture limit like the GTX 500.
Because of the much lower power usage you can easy overclock it.
Octane for example is much faster on newer cards, Cycles does not profit from Kepler architecture atm., maybe in the future.

Cheers, mib

Tesla cards are primarily meant for accelerating double-precision float calculations. Thus no big benefits in rendering, which usually (almost always) uses single-precision.

Instead of buying ONE big card, I’d recommend ADDING another 770 (or similar) to your system. Two cards will cut your render times almost in half.