Ryzen has always had a bit of a clockspeed problem. First-gen Ryzen parts hit a 4.0 GHz clockspeed ceiling: they just didn’t like going any higher. The situation gradually improved with second-gen and third-gen Ryzen parts. But even after Ryzen 3000 caught up to Intel in terms of IPC, those lower clockspeeds meant that team Blue could still cling to the single-threaded performance crown.
But a new custom power profile from AMD developer 1usmus seems to indicate that at least some of Ryzen’s clockspeed issues are down to poor power management. 1usmus’ tweaked profile, remarkably, allows Ryzen 3000 parts to run 200-250 MHz faster on average. That’s an extra 7 percent or so of extra performance just from a power plan. We’ll emphasize here that this is not overclocking. This is literally just changing the power plan your processor uses. How exactly does this work?
It has more to do with how inefficient AMD’s stock power plan is, to be honest. The stock power plan evidently stacks loads on “bad cores,” which aren’t able to stably boost as high. 1usmus’ power plan, meanwhile, seems to draw a leaf from Intel’s book. Intel’s Turbo implementation intelligent boosts the “best cores,” the ones most able to hit high frequencies. This is what we see here as well. Considering how the entire business model behind Ryzen is AMD figuring out how to utilize defective cores, this has important implications: few Ryzen chips will have every core capable of running at a high frequency. The flip side to this is that most have one or two really good cores as well. By shifting the load to these good cores, you get higher frequencies since they’re able to boost higher. We’re not sure as yet whether 1usmus’ power plan has an impact on all-core Turbo, but it’d certainly be nice to get an all core clock that’s a few hundred megahertz higher.