02:08 imirkin: karolherbst: i guess i'm not 100% sure. i'm fairly sure it's very slow on kepler and earlier though.
15:50 Lyude: just in time
15:51 Lyude: maxwell2 power usage, blcg/slcg: 23W -> 20W, gonna do more testing soon to see all of the power savings we've gotten
15:51 Lyude: i'm sure we can do even better then that, I wonder if the perf dom stuff that I haven't implemented yet (apparently it has it's own group of power gating registers) would help
15:51 Lyude: also need to make sure that I've confirmed what appears to be another idle filter that got added somewhere in maxwell1/2 to the fb init sequence
15:52 HdkR: Ow, only 3 watts :<
15:53 Lyude: HdkR: that's idling
15:53 Lyude: usually the most power saving happens during load
15:53 Lyude: and after reclocking
15:53 HdkR: I guess so
15:53 Lyude: additionaly; those pm clkgate packs might be keeping some stuff on that normally wouldn't be on, so
15:55 Lyude: mupuf: ^
21:24 ReinUsesLisp: hi, when this value is accessed through S2R (MOV) what is it expected to retrieve?
21:24 ReinUsesLisp: https://github.com/envytools/envytools/blob/1934618b979b6eb3af28380db3b1745eb50e746d/envydis/gm107.c#L134