Re:Now that computers have more than 4MB of memory, can we g...
by throwawaylinux on Tuesday April 12, 2022 @02:59AM
(#30999503)
> Any periodic activity with a rate faster than one minute incurs the scrutiny of the Windows performance team, because periodic activity prevents the CPU from entering a low-power state.
The power savings available in very low latency sleep states has improved a great deal in CPUs (even desktop and server ones since power saving ~= performance with DVFS and turbo modes). Such that 60s should be well beyond the knee of the curve. A few seconds possibly gets you the vast majority of saving (I'm less sure about PCI/GPU implications).
That said if it's an aspirational goal that's great. But I'm very surprised that windows would not have a lot of periodic activity at intervals below 60s. Even when you're talking about an unrealistic system that has no network connection or activity.
The power savings available in very low latency sleep states has improved a great deal in CPUs (even desktop and server ones since power saving ~= performance with DVFS and turbo modes). Such that 60s should be well beyond the knee of the curve. A few seconds possibly gets you the vast majority of saving (I'm less sure about PCI/GPU implications).
That said if it's an aspirational goal that's great. But I'm very surprised that windows would not have a lot of periodic activity at intervals below 60s. Even when you're talking about an unrealistic system that has no network connection or activity.