XPost: alt.comp.os.windows-10, alt.comp.os.windows-11   
   From: nospam@needed.invalid   
      
   On Tue, 9/2/2025 3:55 PM, John B. Smith wrote:   
      
   >   
   > I hope I'm not highjacking the thread if I start asking about wattage   
   > measurements? I'll start another thread if that's needed. I'd never   
   > checked wattage on my i6 chip Asus Z790 motherboard before so this   
   > thread got me started.   
   >   
   > I have  53 background processes running in my Win11 Taskmanager. I   
   > don't know how to safely turn off the ones running to make   
   > measurements. Just idling along the KWMeter is reading watts from 53.3   
   > to 136. CoreTemp reading watts from 7.9 to 19.8   
   >   
   > On the boot I see 119 to 214 watts.   
   >   
   > Machine shut down, reads 1.3 to 1.4 watts.   
   >   
   > Can you tell me a safe way to stop those processes in Taskmaster that   
   > aren't needed without crashing the machine?   
   >   
      
   The SVCHOST, most of them use *zero* cycles. They're not   
   scheduled to run and they don't run. They are "on-demand"   
   services.   
      
   Running Sysinternals Process Explorer as administrator, there is   
   an interface that gives a cycle count per process. That's how you   
   can check at that level.   
      
   A few are hopefuls. WUAUSERV for Windows Update, it would run   
   several times a day, and it can run for a bit, preparing   
   WinSxS for maintenance. That's a fairly expensive one, in   
   terms of KwH.   
      
   The SysMain has likely been renamed a number of times. It   
   positions materials on a HDD for faster loading (.pf prefetch files).   
   On an SSD, this is likely a waste of a service. During a Windows   
   update, I run services.msc and I shut that one down. I   
   also try to kill SearchIndexer (a separate process, not   
   a SVCHOST), but that one is the undead and it keeps starting   
   itself.   
      
   But for the most part, the consumption in the OS is going to   
   consist of other maintenance things. These would be forked,   
   run for a while, and close on their own. These like to start   
   when you're trying to benchmark and stuff.   
      
   When you ask for a compute job to be done, the job has   
   "a fixed number of watt-hours" in a sense. Stretching the   
   job out, doesn't make a lot of sense, so for the   
   following, it's better to "rush the job, then turn   
   the PC OFF". But if the machine must run all day, you   
   can try and crank it down a bit. It depends on how   
   "overpowered" the resources are, how much room there is   
   to do this.   
      
   Some motherboards have an "eco mode". My 5950X, that mode   
   is simple -- turn one of the two silicon die off. My other   
   processors have a single-die, and an "eco mode" consists of   
   running reduced frequency (and reduced VCore voltage) on the processor.   
      
   CPUs have closed loop controls, like the video cards got. If   
   you set the BIOS CPU power limiter to "100W", then it starts to   
   throttle when the power gets that high, and that is an   
   automatic way of capping top power. The Intel one goes to 4095   
   watts, which means "unlimited" in a sense. That might help   
   if your cooler isn't very good.   
      
   You can turn the core count down in the BIOS, on some boards.   
   Some CPUs may not have such controls. A Phenom II from AMD,   
   the six die were in two groups, and you could shut one group   
   off (even though they shared a common die).   
      
   One very expensive processor ($1000), if you shut it down to 1 core   
   in the BIOS, right after you save settings... the CPU is   
   destroyed. That's an example of a CPU where you *never* drop   
   to 1 core (setting not blocked in BIOS!). 2 cores would be OK   
   (of six perhaps). These probably aren't power modulated, just   
   a clock thing (reducing clock to a processor, reduces consumption   
   to DC leakage). When the BIOS detects one of your attempts to   
   save power, it still tries to arrange things so the die is as   
   "power balanced" as possible. The ECL designer in my group at   
   work used to do that for the chips he worked on, power-balance   
   the corners so the die heats evenly and doesn't crack at   
   extremely high temps (ECL loves heat).   
      
   But messing with things at that level, ECO modes, power capping,   
   that makes more of an impact than doing a lot of tweaking   
   at individual process levels, especially when a lot of the   
   SVCHOST are drawing zero cycles anyway.   
      
   Microsoft has done a lot already, to optimize things. They have   
   had studies running on machines, to find things they can turn down.   
      
   On some motherboards, it is the *chipset* which is a pig. The   
   CPU gets blamed, but the Idle CPU actually draws less than a   
   Northbridge with 42 PCIe lanes on it. While you can turn   
   chipset voltage down, on those pig chips, fiddle too much   
   and you get memory errors, so it's not really worth   
   chasing that. As it is, the pig chip across the way has   
   a boost of a tenth of a volt to keep it "honest". My X48,   
   that was a badly binned chip, and that needed a boost its   
   whole life and it probably should never have been put in   
   the "good" bin at Intel. That chipset is dead now. The X79   
   is still going strong (and drinking the power like crazy).   
      
   Buying a modern computer can save power, but the payback   
   period would be longer than you will be alive. Like using   
   my 33W computer, instead of the 100W idle computer, that   
   saves money. Pennies a day. It's because of this, the 33W   
   computer tends to run all day, the higher end machines   
   stay OFF. But if I wanted to compress a 3TB Macrium MRIMG,   
   I wouldn't do that on this machine (would take too long).   
      
   On the 5950X, railing one core uses half the CPU power   
   and ramping up all the cores and threads, doubles that...   
      
   200W                ____   
                   ___/   
               ___/   
   100W    ___/   
          /   
         /   
   33W  /   
        0 1               N   
      
   That is why saving power is so hard, as vampire tasks   
   that use a fraction of a core, they tend to climb the   
   steep part of the curve near the "one core railed" end.   
   An ECO setting with an artificially capped Fmax, will   
   likely scale the whole graph. At least some of the load   
   at the low end, is ATX supply inefficiency, chipset   
   static power (power that does not change with usage),   
   you charging your iPhone and so on.   
      
   100W                ____   
                   ___/   
               ___/   
   50W     ___/   
          /   
         /   
   33W  /   
        0 1               N   
      
      Paul   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)