rng is slow (to control random number generation). Can I go back to seed, or state?

I've been using 'rng' to control sets of random numbers in my large simulation-optimization problem. I was trying to find speed bottlenecks in my codes using the profiler and I realized that rng is taking too much time (more than rand or randn). I then used seed and observed a huge difference between their speed. I know that MATLAB doesn't recommend using seed, state, and all those old ones...

Here's a simple test:

    tic
for i=1:10000
    rng(i)
    rand(10,1);
end
    toc

Elapsed time is 5.380547 seconds.

    tic
for i=1:10000
    rand('state',i)
    rand(10,1);
end
    toc

Elapsed time is 0.202457 seconds.

I'm using seed now and I see that it behaves strangely... I used to use state before. Can I still use that?