I wont even switch drivers to try. Sorry AMD, no trust there.
Although I would be surprised if drivers could change the amount of voltage it took for my core to operate at a given speed without erroring out. I would be interested in what IDC had to say about this.
Heh, purely educated speculation here based on my experience with other hardware/drivers, but it is possible if you consider that one of the purposes of new driver revs is to improve the performance by finding ways to optimize and make better use of the hardware.
In other words, if the new drivers contain "improvements" that have optimizations which make the existing hardware work harder to deliver better performance (lowered latency executions, less stalls, higher effective IPC, etc) then the GPU is being taxed and stressed in ways with the new drivers that the old drivers never enabled.
Think of an extreme example analogy of LinX/IBT (or Prime95) for Intel cpus and the change in stress brought about by enabling AVX. One version of IBT/prime and your OC'ed CPU is just fine, try the next version up and all of a sudden your OC rig is crashing from the extra stress brought on by the software-side of the optimizations.
So this could be in play with 12.3 vs 12.4 for those folks who were already right on the hairy edge of stability with 12.3. Optimizations in 12.4 push GPU IPC up just enough that the stability margin is compromised.