Power implications of disabling USB

[Thomas_W_Intel]

Jeffrey M wrote:

I encountered an interesting issue while collecting socket and memory power data using Intel’s PCM-Power utility.

A. Scenario 1: OS booted, system at idle and all USB enabled
1. pcm-utility.exe reports 14W – 15W DC per socket which complies with Intel’s power specification of 15W DC at C6-state (Intel 442505 Table 4-10)
2. average C6-state per core is approximately 99% and the C6-state per package is approximately 65%
3. the measured AC input power was 70.7W AC for this configuration
4. powercfg –energy report shows composite USB devices NOT entering suspend

B. Scenario 2: OS booted, system at idle, all USB disabled
1. pcm-utility.exe reports 2.5W – 3.5W DC per socket
2. average C6-state per core is approximately 99% and the C6-state per package is approximately 97%
3. the measured AC input power was 64.1W AC for the same configuration above
4. powercfg –energy report shows no warnings or errors

C. Scenario 3: OS booted, system at idle, all USB enabled, Windows registry modified to enable select suspend on composite USB devices
1. pcm-utility.exe reports 2W – 3W DC per socket
2. average C6-state per core is approximately 99% and the C6-state per package is approximately 97%
3. the measured AC input power was 64.1W AC for the same configuration above

This difference in AC power is only 6.6W. Based on the power supply efficiency of 85%, the DC power difference is 5.6W DC or about 2.8W per socket. Thus, I was expecting to see the pcm-power.exe report socket power of 11W – 12W DC with select suspend enabled or ALL USB disabled

Unfortunately, I'm not familliar with the power implications of USB, but I'm posting this to the forum where more people might see it than the comment to the Intel PCM article.

[Patrick_F_Intel1]

Hello Thomas,

When Jeffrey say "Thus, I was expecting to see the pcm-power.exe report socket power of 11W – 12W DC with select suspend enabled or ALL USB disabled", it makes me wonder...

I know I could dig through the pcm-power code and figure it out but... if pcm-power is just reading the RAPL power status MSRs for package power, then USB power won't be reflected in this number right? And it sounds like Jeffrey is expecting USB power to be reflected in package power (if I understand correctly).

Pat