You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Modern laptops have thermal protection, so they tone down the cpufreq when it gets too hot. This effects benchmarks, or makes them vary a lot, as other things affect the CPU on a multi-core machine.
The bench11 and bench12 are short runs with gyrfalcON and hackcode1 resp, usually taking 10-20 sec.
The bench11a and bench12a are the same, except they run "forever". This way tools that monitor the cpufreq (KDE has a nice graphical tool hidden in their System Monitor tool, but you have to activate it) can be seen how they react to a CPU heating up.
It was observed that bench11a would stay at the peak freq, whereas bench12a would drop a notch. For example on my k2 laptop the peak is 4.7GHz, but bench12a would drop to 4.4GHz.. This does depend on power settings, in particular performance mode would also attract lots of other tasks to steal cpu cycles, and the cpufreq curve looks erratic. In powersave mode it will achive this clear distinction that gyrfalcON is able to hover at peak cpufreq, but hackcode1 drops.
To note, this can depend a lot on the linux kernel version, and the CPU model of course.
This observation was made on a i7-1260P with 6.8.0-51-kfocus kernel. YMMV as they say.
The text was updated successfully, but these errors were encountered:
Modern laptops have thermal protection, so they tone down the cpufreq when it gets too hot. This effects benchmarks, or makes them vary a lot, as other things affect the CPU on a multi-core machine.
The bench11 and bench12 are short runs with gyrfalcON and hackcode1 resp, usually taking 10-20 sec.
The bench11a and bench12a are the same, except they run "forever". This way tools that monitor the cpufreq (KDE has a nice graphical tool hidden in their System Monitor tool, but you have to activate it) can be seen how they react to a CPU heating up.
It was observed that bench11a would stay at the peak freq, whereas bench12a would drop a notch. For example on my k2 laptop the peak is 4.7GHz, but bench12a would drop to 4.4GHz.. This does depend on power settings, in particular performance mode would also attract lots of other tasks to steal cpu cycles, and the cpufreq curve looks erratic. In powersave mode it will achive this clear distinction that gyrfalcON is able to hover at peak cpufreq, but hackcode1 drops.
To note, this can depend a lot on the linux kernel version, and the CPU model of course.
This observation was made on a i7-1260P with 6.8.0-51-kfocus kernel. YMMV as they say.
The text was updated successfully, but these errors were encountered: