For testing, I downloaded a bios file with the same date and time as the modified file that you are using. I forgot to mention last night that there is an ATI Catalyst Control Profiles.xml file hidden many layers deep that controls the 2D and 3D clocks you end up with. I edited this file to eliminate the silly 157 MHz / 300 MHz (core/memory) 2D limitation my card has. Great for saving power I guess but these super low speeds can lead to other problems. Here's an example of how I edited this file and where it is hiding. You can see that I boosted the 2D memory speed up from 300 MHz to 1200 MHz and bumped the intermediate and 3D memory clocks up to 1300 MHz and bumped up the core speeds as well. I'm thinking that the problem you are having might be because the bios clock settings that you have created are fighting against what CCC is trying to set the clocks too. Based on your back and forth screen shot, it's like two different pieces of software are both taking turns at trying to control your GPU clocks so they are bouncing back and forth. I know this can happen with Intel CPUs when two different power saving features are not in agreement but I don't have any experience with this problem possibly being in GPUs too. You should consider flashing back to the original bios and then try editing the CCC Profiles.xml file and see if you can equalize your memory speeds for 2D and 3D that way. I found a problem with the ATI CCC polling scheme earlier this year where it was taking huge amounts of time to read the sensor and would oscillate in a continuous pattern from normal, to high, to very high to a crazy high amount of time to read a simple sensor. That bug is gone now in the latest CCC release but I can't remember which version it was solved in. This wasn't a problem for everyone but I was able to notice in 3DMark06 that it directly corresponded with a jerk / micro stutter in the frame rates whenever the temperature sensor was sampled. That's why I created the Polling Interval option in RealTemp so I could try and figure out what was going on. I originally thought that maybe I was doing something wrong but I didn't change my code and with one of the Catalyst releases, this problem went away. Then I knew who was to blame. Here's the little testing program I wrote for this issue. http://www.fileden.com/files/2008/3/3/1794507/ATITemps.zip It shows the amount of time that it takes for code to get temperature data from the ATI driver. This is a lot more consistent in 10.12 than it was when I first installed my 5770 card last year. ATI just redesigned the CCC so that's what's in the super sized download now. I heard it works OK and starts up quickly but I haven't tried it yet. I'm still using the previous CCC. I'm always afraid to try any new ATI code before they have 6 months to get the bugs out.