Efficient use of 2 GPUs?

Are there any simple examples that use two GPUs efficiently?

I tried matrix multiplication, with a parfor loop of two iterations, but the 2 GPU time is horrible.

Time for the multiplication of two 4096x4096 matrices (including data transfers):

CPU : 0.81 sec
1 GPU : 0.31 sec
2 GPU : 3.18 sec

I am using 8.2.0.701 (R2013b) on a host with 64 AMD cores and 2 K20c GPUs.

By the way, the 8192 product returns wrong results silently!