There is no hard and fast answer, and there is no such thing as without any risk. There is always risk, but if you take your time and find the limits slowly you will expose your GPU to less chance of overheating and being damaged.
The way you do it is get a piece of paper and pencil and some software (furmark is good) for stress-testing the GPU. Find out the maximum temp your GPU is rated for (if I remember correctly it should be somewhere in the mid 80 degree C range).
Start at stock, write down the stock settings, stress test, record max temp.
Increase the mhz by a few mhz. Write it down. Stress test. Record max temp under stress. Repeat this until you get a crash of the video system (or perhaps the entire system which is why you write it down) OR until you hit your maximum rated temp for the GPU.
If you get a GPU crash, you can possibly bump the voltage IF you haven't hit the max GPU rated temperature, this might bring it back to stable. If you keep crashing, you're done, back it off a few mhz and stress test it again to verify it's stable.