Tom wrote on Nov 6, 2024, 18:57:
Simon Says wrote on Nov 6, 2024, 16:12:
That old myth is dying hard I see.
Total system idle power consumption is the same VS an equivalent Intel system.
Has been for generations now.
Sources:
Source 1
Source 2
And you can lower it further with some easy undervolt tweaks to the Vsoc voltage ( my reported idle CPU consumption on 7800X3D is around 12-15W after the undervolt ).
The 35W number came from p.4 of the Tom's Hardware 9800X3D review:
Note that AMD's chips generally have a higher idle power draw. I've measured this on a few motherboards, so this might be applicable across the AM5 ecosystem. AMD's AM5 chips can draw up to ~35W of power during idle compared to 12W with Intel's Core Ultra, a significant disadvantage. This is certainly something AMD should look to address with future firmware updates, as there's little reason for a modern CPU architecture to idle at such high power levels.
Are you saying that's wrong?
No I'm saying both can be right. Can you not figure out that simple problem by yourself?
Hints:
One system indicates 35W on the CPU while idle, the other indicates 12W.
Both systems indicate the same total idle system power draw from the wall, say 75W ( 7800X3D number from source #2 ).
Both systems have similar motherboards, same RAM, same of the rest of the components. What is the variable?
Answer:
One platform segregates the measured powered draw of the CPU differently from the other platform.
You can also see this phenomenon on GPUs with AMD and Nvidia dies having different TDPs and reported power draw ( say 125W with one board, the AMD one, 150W with the other, Nvidia's ), but the same total board power ( say 175W ) when you include the VRMs losses and all the other components that feed the VRMs. Nvidia includes more of the related/peripheral power draw in their TDP, AMD doesn't ( they act more like Intel with CPUs ). Both use the same on total board power.
In other words, the Intel CPU has similar idle power draw, but segregates some of this so you only see the "TDP" not the "total board power" when including all the other components and losses needed to feed the CPU. AMD doesn't segregate ( they act more like Nvidia for GPUs, including more of the related/peripheral power draw in their total ), making it look like the CPU uses more power than Intel while idle.
Ironically, they do the opposite with GPUs ( they report 125W, but total board power is 175W, when Nvidia reports 150W with the same 175W total board power ).
Different teams, different platforms, same total system power draw while idle.
TL&DR: Reported ( even measured at the EPS plugs ) power draw can lie ( some power can come from other motherboard rails and not be measurable directly nor be self-reported ), a kill-a-watt plugged to the wall that measures the whole system doesn't and doesn't miss anything either.This comment was edited on Nov 6, 2024, 19:55.