October 15, 20223 yr Hello all! My tower suck 300W on idle (no vm on, no docker on, only 1 sata ssd and 1 nvme ssd on, Unraid up-to-date, plugins too), which is insane in my opinion... And then, on the dashboard, the CPU stays at 3800MHz, so I dug... this is what I have, right now: root@ValiLab:~# cpufreq-info -d acpi-cpufreq root@ValiLab:~# cat /sys/devices/system/cpu/cpuidle/current_driver acpi_idle root@ValiLab:~# cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave powersave root@ValiLab:~# cat /proc/cpuinfo |egrep -i mhz cpu MHz : 2200.234 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2199.484 cpu MHz : 2200.000 cpu MHz : 2199.498 cpu MHz : 2200.112 cpu MHz : 2198.911 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.181 cpu MHz : 2200.010 cpu MHz : 2200.000 cpu MHz : 2198.322 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2199.078 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.157 cpu MHz : 2200.199 cpu MHz : 2200.000 cpu MHz : 2199.946 cpu MHz : 2199.926 cpu MHz : 2199.492 cpu MHz : 2199.892 cpu MHz : 2199.275 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2200.000 cpu MHz : 2199.939 I seems like my CPU is "powersaving", but 2200MHz is still huge, and then I run that: root@ValiLab:~# cat /sys/devices/system/cpu/cpufreq/policy0/scaling_available_frequencies 3800000 2800000 2200000 So my CPU CAN'T go down 2200MHz. And when I see reviews on the web measuring something like 96W on idle (less here https://hexus.net/tech/reviews/cpu/136796-amd-ryzen-threadripper-3960x-3970x/?page=10), I think I need to add more "scaling_available_frequencies" but I can't find those frequencies on the web, I don't know how to add them, and to maintain them when Unraid updates comes Some one can help please? 🕵️♂️
October 16, 20223 yr 8 hours ago, Valiran said: My tower suck 300W on idle (no vm on, no docker on, only 1 sata ssd and 1 nvme ssd on Are you sure not metering issue ? Pls try at command prompt type "sensors" and focus on CPU actual voltage, then check does BIOS setting CPU in performance mode. Edited October 16, 20223 yr by Vr2Io
October 16, 20223 yr Author Hello! Thanks for your answer This is what I get when I type "sensors": root@ValiLab:~# sensors nvme-pci-0100 Adapter: PCI adapter Composite: +46.9°C (low = -273.1°C, high = +80.8°C) (crit = +81.8°C) Sensor 1: +46.9°C (low = -273.1°C, high = +65261.8°C) Sensor 2: +47.9°C (low = -273.1°C, high = +65261.8°C) acpitz-acpi-0 Adapter: ACPI interface temp1: +16.8°C (crit = +20.8°C) eth0-pci-4400 Adapter: PCI adapter PHY Temperature: +38.1°C MAC Temperature: +37.5°C k10temp-pci-00c3 Adapter: PCI adapter Tctl: +34.8°C Tccd1: +35.2°C Tccd3: +32.0°C Tccd5: +34.0°C Tccd7: +34.5°C I'm gonna check the BIOS and come back with parameters I have! EDIT: I've disabled gaming Mode, removed the XMP profile (even if I never had a memory issue in the past 2 years), I forced C-States to Enabled (was on AUTO). Everything else is on auto, except enable virtualization and the boot sequence And when I watch my CPU using "watch grep \"cpu MHz\" /proc/cpuinfo", it stays on 2200MHz (range from 2197 to 2200,910)... Edited October 16, 20223 yr by Valiran add info, correct typo.
October 16, 20223 yr 6 hours ago, Valiran said: eth0-pci-4400 Adapter: PCI adapter PHY Temperature: +38.1°C MAC Temperature: +37.5°C k10temp-pci-00c3 Adapter: PCI adapter Tctl: +34.8°C Tccd1: +35.2°C Tccd3: +32.0°C Tccd5: +34.0°C Tccd7: +34.5°C By observation, CPU even lower temperature then PCI adapter, I would assume CPU haven't draw huge power. Pls further check metering, if power down tower, does meter reading almost zero. Also, check GPU too. Edited October 16, 20223 yr by Vr2Io
October 16, 20223 yr Author Already done that, with the tower OFF, it consumes ~6w (dual nic on the motherboard that still light green when tower off). If I shut off the power supply with the physical button rear, it's 0w. And the Nvidia cards: I can see these Nvidia gpus in your server GPU 0: NVIDIA GeForce RTX 3080 (UUID: GPU-0a86b645-636f-2446-2cfa-ccec26192a1b) GPU 1: Quadro P2200 (UUID: GPU-90cc5a04-37ec-52d0-a94e-267e664cc883) ------------------------------------------------------------- Persistence mode is already Enabled for GPU 00000000:21:00.0. Persistence mode is already Enabled for GPU 00000000:4A:00.0. All done. Power state is P8 P8 ------------------------------------------------------------- Power draw is now power.draw [W] 13.86 W 7.44 W
October 16, 20223 yr Author I use this: https://www.amazon.fr/gp/product/B087M8LF63/ref=ppx_yo_dt_b_search_asin_title?ie=UTF8&psc=1
October 16, 20223 yr Pls also check does relate smart plug firmware. https://forum.futurehome.io/t/aqara-smart-plug-power-usage/3242 Pls use a well know power rating water boiler to verify the measurement accuracy or not.
October 16, 20223 yr Author The measures are reliable. Moreover, I don't use Futurehome Hub so issues doesn't apply here
October 16, 20223 yr In my experience AMD desktop/gaming/server CPUs are much worse than Intel for real world power consumption especially at idle. Adding two power hungry graphics cards just makes your problem worse. I don't think there's an easy fix besides moving to more power efficient hardware. Depending on electricity rates in your area you could spend more on power in a year or two than it would cost to switch to much more power efficient hardware. I have two Unraid systems. One idles at 15 watts with the drives sleeping and the other idles at 23 watts with the drives sleeping. That's power consumption as measured from the wall with a Kill-A-Watt meter. Both are recent generation Intel with no graphics cards. Neither of them has been CPU limited even with many dockers and a few VMs running. So, depending on your use cases 300 watts, is just a huge waste of energy. If you want to game then use a gaming rig that's only powered up when you're actually gaming. That's much more power efficient than using a power hungry gaming grade rig powered 7x24 for an Unraid server.
October 16, 20223 yr 2 minutes ago, Valiran said: The RTX3080 use 13,86W at idle, so that's not the issue Well it's still part of the issue. I have a gaming rig with an RTX3060. It idles at around 60 watts and about 15% of that is the graphics card. But you're basically trying to use a gaming rig as a 24x7 Unraid server. That's just not going to be efficient no matter what you do.
October 16, 20223 yr Author I'm using Unraid as it's intended to be: a powerfull machine that can handle multiple VM and/or docckers. My power consumption at idle is double comparing to the same rig without unraid. The problem is not the GFX as I posted earlier, they cost roughly 25w on idle. The problem is not storage, because I have only 2 SSD (1sata 1 nvme). But when I look at the CPU frequencies, what a magic, the lower frequency is still 2,2Ghz and can't be lowered because of ACPI driver configuration. So I'm looking to find a way to lower that. I think you misunderstood the issue. Consumming watts doesn't bother me when I use thoose watts. But staying at 220W with a machine that do nothing, this is bothering me EDIT: by the way: reinstalling the BIOS + restoring my flash drive make my rig use 203,6w on idle. So I went from 300w yesterday, to 220w today and now 203w, just by modifying parameters. That's a good way EDIT2: the dashboard always report "AMD Ryzen Threadripper 3960X 24-Core @ 3800 MHz" and the 3800MHz never move Edited October 16, 20223 yr by Valiran
October 16, 20223 yr 36 minutes ago, Valiran said: So I'm looking to find a way to lower that. To be honest, I don't think it is cause, 2.2Ghz really normal in my point of view and won't got much different even you lower it. I agree 200w idle power is abnormal, but need you found out why that. My desktop was 30w, 7*24 unraid 40w, main Unraid HEDT 9800X is 110w, TR 1920x also in this range too. ( all in idle power ) Edited October 16, 20223 yr by Vr2Io
October 16, 20223 yr 1 hour ago, Valiran said: I'm using Unraid as it's intended to be: a powerfull machine that can handle multiple VM and/or docckers. But my Unraid machines also handle multiple VMs and dockers and they idle at 15 watts and 23 watts. Your server is using around 10 times more power doing nothing. It's not the fault of Unraid it's using the wrong hardware. It's like buying a V12 Lamborghini and complaining it uses way more fuel than a Toyota Prius. Edited October 16, 20223 yr by dev_guy
October 16, 20223 yr Author 37 minutes ago, Vr2Io said: To be honest, I don't think it is cause, 2.2Ghz really normal in my point of view and won't got much different even you lower it. I agree 200w idle power is abnormal, but need you found out why that. My desktop was 30w, 7*24 unraid 40w, main Unraid HEDT 9800X is 110w, TR 1920x also in this range too. ( all in idle power ) I'm gonna dig in the bios (tomorrow, that 2am in France now :) ) and if I find nothing, I will try tu put some BIOS screenshots, maybe you will say "that's obvious, turn this off!"
October 17, 20223 yr 3 hours ago, Valiran said: I'm gonna dig in the bios (tomorrow, that 2am in France now :) ) and if I find nothing, I will try tu put some BIOS screenshots, maybe you will say "that's obvious, turn this off!" That's fine. Huge power draw should only happen in CPU or GPU, if possible suggest you try to unplug all GPU if you think this also easy. 4 hours ago, Valiran said: My power consumption at idle is double comparing to the same rig without unraid. Same rig within 110-120w idle? Windows ? 4 hours ago, Valiran said: the dashboard always report "AMD Ryzen Threadripper 3960X 24-Core @ 3800 MHz" and the 3800MHz never move This normal, dashboard won't show actual CPU clock. Below are the TR 1920x boot up until shutdown power usage log, max 129w and idle in 100w. Clock also 2.2Ghz. ( 3 spinup HDD, NVMe, 10GNIC, HBA , GT710 ..... ) Edited October 17, 20223 yr by Vr2Io
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.