Synthetic Ryzen Benchmark Results Accurate To 10%ish, 15 Times Out Of 20?

An Old Bug And A New, One Fixable The Other Less So

It is a bit of a bad day for AMD as their Ryzen 3 chips are misbehaving in the wild.  Der8auer has exhaustively research the actual frequencies the CPUs reach during Boost and found that only a tiny percentage of users are seeing their silicon reach those peaks.  AMD has admitted this is an issue and is working on a BIOS update to resolve the issue, promising to “provide an update on September 10 to the community regarding the availability of the BIOS.”  In the mean time enabling only a single core will allow you to hit your Boost Clock, which somewhat defeats the point of the high core counts on Ryzen, and it is possible that updating Windows 10 and your BIOS as well as using the Ryzen scheduler software may help mitigate the issue.

The second issue will bring a sense of déjà vu, as it involves Windows 10’s troublesome Real Time Clock.  The RTC bug will be familiar to overclockers, it is when the Windows system timer exhibits some strange temporal behaviour when you change your bclock.  Clocking your bclock below 100MHz will result in some synthetic benchmarks reporting higher performance while paradoxically raising it above 100MHz will lower your results.  Windows interprets that change in frequency as also being a change in time elapsed thus a lower bclock shows the benchmark as occurring in a shorter time span than the higher frequency test.

Originally this bug occurred on many Intel chips and essentially all Ryzen chips and workarounds were eventually implemented to mitigate the issue, but not solving the root cause as evidenced by the fact that it has reared it’s head again on the newest generation of Ryzen chips.  Microsoft will be working on this, and hopefully we will have an update soon but in the mean time there are a few resolutions to consider, though they come with their own troubles.