@AntonVorozhtsov like I said, they are NOT the same thing anymore. One of the C12 changes was to the performance meter. So you cannot compare C11 meter to C12 because they do not show the same thing.
This was the explanation from the Dev at Steinberg: Cubase 12 Upgrade ASIO-Guard Issue - #147 by Chriss
But here it what is says:
Our investigations here refer to audible performance issues with C12, where C12 is failing to playback the projects dropout-free, which played just fine in C11. Please consider when posting 
.
Just some remarks on the Performance Meter:
- the redesigned Performance Meter in C12 is showing four values, aiming to give users a more detailed info about the system’s momentary load:
– Realtime: how much of the available time (in % of the real time latency, set by your audio buffer size) is consumed by real time processing? This value is averaged over several buffers, and therefor moving slowly. It gives an impression of the overall load of time-critical processing in your project.
– Asio Guard: how much of the available time (in % of the Asio Guard latency, set by the Asio Guard level Low/ Normal/ High) is consumed by prefetched (= non-real time) processing? This value is averaged over several Asio Guard buffers, and therefor moving slowly. It gives an impression of the overall prefetch load of your project.
– Peak: this value shows the recent maximum values of the time-critical processing; it gives an impression if there are load peaks, caused by - e.g. - certain instruments, song structure (plenty of Midi notes starting at the same time,…) or FX. This can help you find and omit certain bottle necks.
It is not averaged, and can therefor look jumpy, especially with small real time buffers. It also reflects the jitter of your audio hardware. A heavy jitter can limit the performance with small buffer sizes.
– Disk Cache: showing the read/ write buffer fill level, referred to the max. throughput (you mostly only see it move on cursor relocate or playback start)
and here is the bit that is very different to C11 - It is not averaged, and can therefor look jumpy, whereas C11 was an average and therefore smoother…does not explain performance issues but does help when comparing.