I just wanted to post some unscientific power draw results from running my Framework 13 AMD Ryzen AI 9 HX 370 with 96GB (kit from Framework directly).
I recently started wondering if having the 96GB of RAM was increasing my idle and standby power draw, as compared to just having less RAM, say 64GB.
Experiment 1)
From memory.
I played around with Linux memory offlining to see if I could offline an entire bank of RAM, when not in use, to reduce power draw.
This was a while ago, so I don’t have exact figures, but I recall it made zero impact on power draw, even when offlining a full DIMM + most of the remaining one.
It was easy enough to offline large sections of memory, but to offline the full DIMM, I believe I had to change some startup parameters (maybe kernel cmdline or module params), so that it would offline the full DIMM on boot, before some unmoveable page was allocated on the second DIMM. From some additional reading, it seemed like only servers would implement the logic to actually power down a bank of RAM + mem controller, when offlined in the kernel.
Experiment 2)
From memory.
I pulled one of the 48GB RAM DIMMs. This seemed to reduce the idle power draw somewhere between 0.5W to 1W (it was hard to get a good eyeball average) out of about 4.5W nominal power draw.
I monitored the power draw while plugged into AC through Anker Prime Charger (160W, 3 Ports, Smart Display) - Anker US, and then tested on battery and monitored through Power Statistics.
I have no clue if this extra idle power draw is simply the overhead for using an additional DIMM (with extra mem controller power draw) OR if this is related to the capacity increase, and thus switching to a 64GB kit would reduce power draw.
I also checked standby/sleep power draw using the same Anker power adapter. There appeared to be no human-eye detectable power difference when sleeping. The Anker power adapter would simply cycle between saying 0W power draw and 2W power draw (0.1A @ 20V). I could have analyzed the frequency/cycle of this, but it seemed like a bad indicator of power draw.
Would love to hear other people’s info about the subject.