It's been almost three weeks since AMD launched the new Ryzen 3000 series CPUs and our extensive coverage of the new parts. One of the things that did not go so smoothly was AMD's BIOS and software situation, in which things were still very much in the air after startup, with the new CPU's idle behavior in terms of both temperature and voltage. In particular, the new parts appeared to look as if they rarely idled at lower power states and instead stayed at high frequencies, even though they did not do much of the side effect of AMD's new CPPC2 boot-up behavior. For poorly-programmed monitoring applications, the monitoring loop is usually too strong, resulting in a load on the CPU. This results in a frequency increase when the CPU is subjected to a higher load. Given the sub-1
AMD has now resolved this issue and optimized the CPPC2 behavior in the new Ryzen energy plans by releasing a new chipset driver package.
As AMD points out in its community letter on this topic, part of the new behavior change is that the new scheduler settings now have a much slower startup time than previous versions. In particular, when the chip is at its base frequency and open circuit voltage, it now takes significantly longer for the chip to reach its boost frequencies.
In our quick A / B test between the two driver versions We can see that before the update, the CPU would rise to its boost clock in about 840 microseconds, while the new power plan in that record took 17.5 milliseconds longer Has.
The new behavior should make it so The CPU startup is less prone to smaller transient loads. The new boost duration is still quite sufficient and extremely fast – sustained CPU utilization will see barely noticeable differences, while intermittent loadings like games will also not be affected as the CPU, once it exceeds the initial fundamental frequency ramp threshold, will Utilization maintains -1ms frequency change behavior.
I also looked at the Windows Power Plans, if they were changed, and it seemed they did. While in the old version the CPU was idle at ~ 2.2 GHz, the new driver was idle at 3GHz. Apparently, the increase in frequency was slowed down a bit, because I noticed in my rapid tests a half as fast increase in frequency.
AMD also received concerns over the reported high temperatures of the chip. The company explains that the value that most applications read is generally the maximum of multiple sensors on the chip. Essentially, this is the junction temperature of the chip – while most of the chip actually has a different / lower temperature.
A new version of Ryzen Master now includes a different temperature readout algorithm to better represent the total value. The temperature of the tool is not the absolute maximum reported by a sensor. According to AMD, this is a better representation of the CPU temperature. In addition to averaging over various sensors, readings are also averaged over a small time window. In my tests, no-load and low-load scenarios are hardest hit, and the new temperature behavior is far from being so irregular and prickly.