From: nospam@needed.invalid   
      
   On Wed, 6/25/2025 4:51 PM, Mike Easter wrote:   
   > Mike Easter wrote:   
   >> I see an Optiplex 9020 USFF (ultrasmall formfactor) 320G hdd 8G ram W10   
   refurb for $111.   
   >   
   > They should've made that w/ an external PS.   
   >   
   > I saw a guy comparing a mini PC w/ similar CPU comparing it w/ RPi and he   
   went to the trouble of tearing it down in his YT vid so he could get   
   consistent benchmarks because of heat.  Not Dell's.   
   >   
   > I heard Dell (allegedly) sells their replacement proprietary PS in different   
   watts for that model, but I don't see it on Dell's page. I think the 9020 usff   
   is 200W.   
   >   
      
   Using a Kill-A-Watt meter, you can measure the actual power consumed.   
   Mine for example, on a machine with a 620W Seasonic supply, is currently   
   drawing 34.3W. That's a processor and a cheap video card, rather than   
   just a processor (with iGPU). I had to turn off the iGPU because of   
   some machine-wide malfunctioning, and the video card takes its place.   
      
   https://upload.wikimedia.org/wikipedia/commons/0/04/P3-Kill-a-watt.jpg   
      
   Your 200W is the nominal MAX of the unit, rather than the actual consumption.   
      
   A 200W measurement, that's closer to Prime95 on 16 cores.   
      
   One of the machines here, idles at 33W, another idles at 22W, and   
   the ten year old 4930K idles at 100W. The only thing that beats   
   all of them at idle, is my Celeron 300 with 440BX chipset, which   
   idles at 150W. Lord knows where that electricity goes on   
   that old thing.   
      
   The older the machine, the higher the opportunity for gluttony.   
      
   At one time, video cards had no power-save at all. Clock rates   
   were constant on the "frame buffer" cards of the day. Around the   
   time of the 8800, the high:low ratio was 2:1 . Some savings then,   
   when the video card was not playing a game. The HD6450 draws 13W   
   doing something, and 3W doing nothing. A ratio of over 4, and   
   not the best number either. There are higher power cards   
   that managed to idle at 3W too. But these things change over   
   the years, and current video cards are back to slightly higher   
   idle. Leakage in CMOS N-channel/P-channel pairs can be reduced   
   by the usage of a third transistor. This was used to throttle   
   the leakage path, after Prescott. Performance gates continue   
   to use the classical gate design (like in the Intel ALU).   
   Non-performance paths can use a good percentage of three transistor   
   gates, with one gate to throttle leakage.   
      
   The reduction in idle clock speed, reduces "toggle losses", but   
   also makes things like GPUs "slower".   
      
   The N150 I was looking at, appears to stretch all the way from   
   x1 multiplier, to x36 multiplier, with a BCLK of 100MHz. Running   
   at 100MHz in the core, must be barely enough to keep the   
   embedded DRAM charged. While its iGPU does not state clock   
   range, it too is likely to use downclocking in a similar way.   
      
      Paul   
      
   --- SoupGate-DOS v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)