I am looking to get a Framework, but in my looking I have noticed a correlation. The lowest system idle power draw for the Framework 13 appears to be 3.5-4.5W, compared to 2.5-3.5W that some have posted for other comparable systems.
I understand that the expansion cards take a bit more power, and I don’t know how much, but I see that the SN770/SN7100 benchmarks show an idle power draw of ~1W whether the ASPM power-save feature is on or off, indication that it is not implemented. For comparison, the Lexar NM790 idle draw goes almost to zero.
This idle draw appears to match up with the difference in minimum system power between the Framework (which comes with the WD SSD by default) and other systems that probably have a different drive.
So my question is, can anyone (preferably multiple people for more data points) verify in a real usage scenario that the ASPM doesn’t work on this drive? Since there is no sensor to measure this directly you would have to use a tool to measure total battery draw and compare the values of the system in full stable idle, with the drive installed and removed. You would either have to run the OS either from a different drive, or run a live-boot/PE environment.
The readings would probably be more reliable with all power-save features enabled to reduce the noise-floor.
I don´t have a framework myself, but I’m interested in the SN7100. Just want to add a couple points here:
These typical kill-a-watt meter plugs may be inaccurate in the lower end (e.g. below 5 W), since they are typically designed for more and consumer devices optimized for cost. I have a “Brennenstuhl Primera-Line Energiemessgerät PM 231 E” (from Germany) and this goes down to 200mW, which a lot of other ones don’t do, at least according to a German test that I’ve read some years ago. Maybe today the situation is different.
There are reports floating around the internet that WD SSDs can be come unstable with ASPM. I don’t know if it’s true or not.
Quote using Firefox translation: " In the desktop PC environment, ASPM is not usually used, but the function itself can be activated by changing the BIOS settings and Windows settings, so when I measured the idle power at ASPM, the power consumption at idle decreased to 0.03W. In a mobile PC environment where ASPM is used, the idle power consumption of “WD_BLACK SN7100 NVMe SSD” is small to the top class, so it is also ideal as an SSD for replacing and adding mobile PCs."
I bought SN7100, and measured the current consumed.
When ASPM, and APST are working, the current is near 0.002A. The peak is during large discards: 0.8A. Usual intense bulk write: 0.4A. Intermittent heavy work 0.27A. Light work: 0.18A
Regarding the reports of WD SSDs becoming unstable with ASPM enabled (again I don’t know if it’s true, just want to provide a source for my claim), I found this: PSA: Avoid WD NVMe SSDs due to bad ASPM Not sure what to make out of this. Not very convincing (just one guy and two SSDs), but also not very encouraging.
That review is indeed the main source for the concern. I would assume that a passive current measuring device such as the one they use would provide the most accurate and reliable data, especially considering that most of the drives they tested show an obviously functioning ASPM, but there may be some nuance to their setup that causes issues with some drives. Hence the request for normal scenario testing.
I don’t recommend an external physical tool such as a Kill-A-Watt type meter because there are too many unknown or uncontrollable variables that may affect readings. There is, however, a built-in sensor to provide power readings from the battery, and there are a number of programs in Linux and Windows that provide this reading. I also recommend running the test with the machine unplugged, to avoid extra variables that come into play while the charger is connected.
As for the blog post you linked, I am skeptical. I don’t think he shared enough detail to determine with any certainty that the ASPM was at fault. The various measures he took to recover the data (including turning off ASPM and others) helped stabilize it enough to get his data off, which isn’t deterministic enough for me. The SN770’s failure mode sounded like it was due to the bad firmware that was discovered (and patched) on some drives of that model. He doesn’t mention anything about that issue or firmware updates in general.
Yes, I was testing under Linux only. I didn’t test in Windows. I used a raiser cable where I snapped power wires, and inserted Ampere meter in between.
