I just conducted a test on a 16 core OSX machine with a nVidia Quadro4000.
The results are in…using the 16 CPUs is faster than using the 256 cuda cores.
Time with GPU: 1:27.
Time with CPU: 1.07.
What a pathetic video card.
All world knows (or should by now) Quadro have a poor CUDA performance.
They are not designed for it, and their lower clocks do the rest.
An Intel Core i7 4770K 3.50 GHz has around 100 GFLOPS
A Quadro 4000 has a single precision performance of 486.4 GFLOPS.
A GeForce GTX 780 has 3977 GFLOPS.
I’m just crying because that is what are tech department bought. Now I’ll never get a new card…
Don’t cry, you just have to properly present your case to the one in charge of the moneyspending.
Without knowing what CPU you got, I simply “times 4” the Ci7 preformance for your 16 cores, resulting in 400 GFLOPS.
With a GTX 780 you have ~10 times the performance.
Now present them with the cost of such a card, your hourly rate and the productivity increase as rendering takes 1/10th of the time and how soon the cost of the card is amortized and how much money is wasted if you have to work with the slower card for a year.
Easy as pie - assuming the iBox properly supports your chosen card -.-