I wouldn't recommend using the formula, since then you need to calculate $P(T,N=3)$ which isn't a lot easier than the conditional.
Start by imagining N=3. This means that after the sixth roll, there have been three different faces seen. So the expected time to see all 6 conditional on this is $$ 6 + \mbox{(expected time it takes to see the last three faces)}.$$
So we need calculate the additioal time it takes to roll the last 3 faces. When you start attempting this, any one of the 3 faces you haven't seen yet will be a new face. Thus on each attempt, you have a 3/6 = 1/2 probability of getting a new face. The expected number of independent trials to see an event with probability $p$ is $1/p$. Thus, the expected time to roll the fourth new face starting from when you rolled the third is 1/(1/2) = 2. After you roll the fourth face, the probability of rolling a new face goes down to 2/6 = 1/3. So the expected time to roll the 5th face is 3. Finally, after the fifth is roll you only have a 1/6 chance of rolling that final face, so the expected time to do that is 6.
So the expected time to see the last three faces is 2 + 3 + 6 = 11. So the answer is 6 +11 = 17