Impact of RAM density on suspend power consumption

Thanks for this thorough analysis, @Miles_Frain!

It so happens that I have a 64 GB kit, so I did the measurements on my system.

Specs:

  • Framework 13, AMD 7840
  • BIOS 3.05
  • Opensuse Tumbleweed (Kernel 6.11.3)
  • Solidigm P44 Pro NVMe (2 TB)
  • Kingston Fury Impact 64GB-5600 (KF556S40IBK2-64)

For the expansion cards I tested two configurations:

Config 1:

  • 2 USB-C in the rear slots
  • USB-A front left with Yubikey attached
  • HDMI front right

Config 2:

  • No expansion cards at all

Each test was performed with the system in single user mode with the battery charged to 80% (the max set on my system in the BIOS) at the beginning of the test. And each test was performed twice, once with s2idle, once with rtcwake as per the description in the first post. Idle time was one hour, and the average current drained in mA is listed in the table.

Results for Config 1:
(results from the 10h idle test in parentheses - see EDIT below for details)

RAM rtcwake s2idle
Both populated 33 mA (32 mA) 33 mA
One populated 27 mA 28 mA
Per stick 6 mA (5 mA) 5 mA
Base (extrap.) 21 mA (22 mA) 23 mA

Results for Config 2:

RAM rtcwake s2idle
Both populated 27 mA 28 mA
One populated 23 mA 24 mA
Per stick 4 mA 4 mA
Base (extrap.) 19 mA 20 mA

Somehow, the power consumption of my 32 GB memory sticks seems to be much closer to your 8 GB sticks and quite a bit lower than what you find with your 48 GB. And my base consumption seems to be higher. That may be due to the SOC (7840 vs 7640).

EDIT:
I also did a longer test to see if 1h of sleep is sufficient to get results of a reasonable precision. This was done in Config 1, both RAM slots populated, setting the rtcwake timer to 10h.

10h: 320 mAh battery drain (32 mA average)
1h: 33 mAh battery drain (33 mA average)

So the results seem to be OK ± 1 mA, which corresponds to the resolution of the reporting of the current batter charge (which is to within 1 mAh). I updated the table with the results for Config 1 to include this measurement point.

EDIT2: typos everywhere…

4 Likes