Previous-Gen Ryzen, Testing 3000-Series Without the Ryzen-Aware Scheduler
How Do Zen and Zen+ Perform?
We've determined there are faster and slower cores on the Ryzen 3000-series processors, but why does that matter? It denotes a shift in AMD's chip binning practices and what the company deems as fit for shipping. That means we need to see how the company handled its binning with prior-gen chips.
We tested the Ryzen 5 2600X, the 3600X's prior-gen counterpart, which all cores active, on the latest revision of Windows. As you can see, all cores, at one point or another, reach this chip's 4.2 GHz boost clock. In fact, they often exceed that rating due to the company's XFR2 (eXtended Frequency Range) feature, which allows the chip to boost even higher if it is equipped with beefy power delivery and cooling solutions.
AMD doesn't specify an XFR2 ceiling, but we can see that the chip reaches up to 4.25 GHz on any given core.
Here we plot the utilization of each core on the previous-gen Ryzen, and we can see that the workloads move around seemingly at random, just like we have been trained to expect given the past decade of multiple core processors: every CPU core should theoretically hit the turbo frequency.
This is most pronounced during the early stages of our benchmark, where the five ~35-second LAME benchmarks are assigned to different cores without any apparent targeting to maximize boost clock rates. Although the CPPC2 feature is included in our chipset drivers and the Ryzen-aware scheduler is also present, it doesn't appear to impact first-gen Ryzen chips in regards to thread targeting.
Testing Ryzen 5 3600X Without the Ryzen-Aware Scheduler
Finally, to see how important the impact of the combination of the new Ryzen-aware scheduler and CPCC2 features are, we tested the current-gen Ryzen 5 3600X with Windows 10 1703. We do have the CPPC2 drivers installed via the chipset driver package, but it appears thread targeting doesn't function as efficiently without the new scheduler directing traffic.
The workloads move among the cores, but in a limited fashion. Core 2, Core 4, and Core 6 yield measurements above 4.3 GHz.
Surprisingly, Core 5 takes the brunt of the workload even though it doesn't have the highest boost capability. As we have seen in the prior album of charts, that core largely sticks at a 4.25 GHz clock rate but has peaks of 4.3 GHz. In either case, Core 4 has a boost of 4.35 GHz, indicating it is the faster core, but the workload isn't landing in that core.
That illustrates that the combination of the new scheduler and CPPC2 is a clever implementation to get the most out of the silicon, but it also means that enthusiasts measuring clock rates with older versions of Windows, or without the correct chipset drivers installed, will see reduced performance. It also explains why some are not seeing the peak clock speeds during some workloads.
MORE: Best CPUs
MORE: All CPUs Content