Get a benchmark like Unigine Valley, Unigine Heaven, or some other -demanding- graphical benchmark. Use either MSI Afterburner or EVGA Precision as the program, then (with stock voltage) increase your core clock by 10/20 and memory by 20/40 per increment. For each rise in the clocks you will run your benchmark of choice, recording the score in a file (assuming you don't crash). I name my successful Unigine Valley runs with the following format, using the below score as an example:[Intel] 2306 (30.400) (Extreme HD)
The format is easy to follow.
-[Intel] is what I use to differentiate my scores between my old AMD CPU and my newer i5
-2306 is the score
-(30.400) is telling me that my current GPU overclock was +30 on the core and +400 on the memory
-(Extreme HD) Tells me what preset I was running. This can be left out, as running anything other than Extreme HD is a waste of time in my opinion. I continue to use this solely for the fact that my first 30 recorded benchmarks with my old AMD CPU had this in the name and I wanted to keep the fidelity.. because reasons.
Keep in mind that Unigine will not detect your CPU overclock, as shown here:
As you can see, Unigine Valley is reporting my i5 running at 3.545Ghz, when it was actually running at 4.5GHz. If you wish, you can add in your CPU overclock to the format mentioned above, looking something like: [Intel] 2306 (30.400) (4.5GHz) (Extreme HD)
After you have peaked at your stock voltage you can begin considering overvolting, which (at the cost of increased thermal output) will allow you to push your GPU farther with its overclocking capabilities. I haven't overvolted my 770 yet, as I've seen no need to do so with the games I'm currently playing, nor have I gone on a Unigine Valley score binge.
Once you start crashing, decrease your overclock gradually by smaller increments than 10/20 and 20/40 until you can find your maximum stable overclock.
That's all I can think of.
Edit1: I completely ignored your question. In layman's terms, the core clock will determine how fast your GPU itself is running, whereas the memory will determine how fast the memory will run. For performance it is generally desirable to focus on the core clock more than the memory, but Unigine Valley seems to favor memory overclocks. In my personal experience, with the EVGA Superclocked 770 (4GB), I can only get my core clock to boost +30 on the core, regardless of if my memory overclock is at 0 or 400, anything over 30 on the core causes instability and crashing, whereas the memory can go as high as +500-550, but I keep it at +400. This is without
voltage modifications, mind you, if I were to overvolt I would be able to get much higher overclocks, but the GPU came pre overclocked. I wouldn't -ever- buy a GPU that comes pre overclocked unless I was getting a good deal on it, which was the case in this situation.Edited by Youown - 10/22/14 at 9:28am