Some things I have picked up from my attempts to reduce the power consumption of my homelab: 1. AMD has processors with G and GE suffix that have greatly reduced TDP and have an integrated GPU. Combine that with an ASRock board with IPMI and ECC RAM and you have a great, low cost, low power base for a homelab server. 2. 10GBE over copper takes a lot of power. If you use SFP+ you have the option to use multimode fibre instead, that usually is lower power AND lower latency. 3. Standby mode for HDDs in a NAS was shunned in the past as it had a negative impact on the longevity of the drives. Modern HDDs however are much more resilient and can tolerate much higher power cycle counts. An inbetween on Truenas is Adv. Power Management that can be set to low power states without spindown. 4. Shuting down unused VMs has a measurable impact on my server. A Win 10 Pro on my system uses 5-7 Watts. Containers tend to use less power, as they use the OS infrastructure of the host. 5. Don't assume, measure. I used some cheap power meters with tasmota to stay on top of my homelab and check the impact of changes on power consumption.
> I used some cheap power meters with tasmota to stay on top of my homelab and check the impact of changes on power consumption. That would be a nice video to watch. How to implement such. It'd be good :) @christianlempa
@@salvationbygracethroughfaith You are right of course. I just am willing to expend the energy for IPMI as it saves me additional hardware, effort and inconvenience. To me it's about reducing the energy needs while still meeting my needs.
This is actually what I have been working on the past month, I am converting all my compute from older DUAL Xeon boards, to a Proxmox Cluster of 8C 16T HP mini PC's, and changing my NAS to a low power system, and way smaller box, from 11 Hard Drives total , to 6 Larger drives, and putting it all in a way smaller rack to conserve energy and space. The whole BIG rack and big PC's is cool, and nice , until you get the complaints from the wife, and the electric / cooling bill. I live in Texas , so any energy I use is basically doubled due to cooling costs.
Good to see a video for homelab that finally addresses power efficiency. Most of my servers have such low activity, I could never justify something with high energy use. Great video 🙂
Год назад+17
It is almost unfair whenever I see US youtubers building homeservers. They are almost always totally overkill or don't show any considerations for power consumotion. Most of the servers in those videos would cost me several thousand dollars of power here in central Europe if I would run them for their expected lifespan. It is nice to see a competent youtuber who considers or even prioritises power consumption.
I have a x86 kubernetes cluster (1 mini computer). It was on 50 - 60% cpu all the time. Then I joined a Raspberry Pi 4 8gb to the cluster helping out the workload. Now the x86 node is on 5% cpu, letting the raspberry pi do most of the work. I am amazed by how far we have come with ARM64 support. Cert Manager, External DNS, Grafana, ArgoCD, Prometheus, Loki, InfluxDB ,Uptime Kuma, Kafka, all works as workloads on ARM64 out of the box, with out a single line of code changed, making it very appealing to create very power efficient ARM64 clusters!
Год назад
I will definitely look into that :D Sounds amazing what can be done with ARM now.
love the timing of this video - over the last 90 days, I've been doing efficiency changes to my homelab, dropping my overall kW commit from 2kW/m to 0.4-0.6kW/m, obviously I did this with money lol but getting rid of very old hardware for some new age 15-75kW servers is amazing. 10th-13th gen infrastructure offers such robust performance packed into tiny packages of power. When it comes to shutting down servers, I try to get things to sleep where I can and yes, WOL is great to instantly wake and resume these servers - unfortunately a lot of users are intimidated by the issues of sleep (automatic wakeup, GPU performance failure, windows poor compatibility since windows7) which is why I think it's not used as much as it can be but is very useful for a desktop PC with high wattage that gets usage randomly throughout the day, sleeping can save atleast 0.2kW/h for my 13th gen i9, that probably doubles or triples for the AMD users. Totally agree with you man, great video.
Great topic. So far I have just two Rpi for my homelab needs, in neat argon cases. Running all on SSD totally efficient and silent in my living room. I would not use them for proxmox but it is crazy how far you can push these little sbc while using very little power.
I know, right? My 14-drive ZFS SAS shelf draws ~133 watts fully loaded - and I'm like, we had LIGHT BULBS that were almost drawing that much (120W) back in the day
I'm in the process of revamping my little home setup and have been working on exactly this. Smart plugs + home assistant (or another logging/monitor app) to monitor power usage has been helpful until I grab some UPS's to figure out how much power I was using where. Also swapped out some older systems with really low power mini pc's where I could (opnsense for instance). Hardest thing has been upgrading my NAS to keep it relatively low power and keep all the features I want but that's probably the last thing I'm going to upgrade so I have time to decide on that one. Only leaves 1 power hungry Epyc server but I'm not getting rid of that one! Funny thing is overall swapping the lightbulbs to LED, replacing some old appliances and turning off idle audio equipment (sub/amp) saved me more power overall but my homelab isn't very big.
I was planning to pickup a used Dell Poweredge R720 with 256GB of ram a few days ago. I had to reschedule. After watching your presentation I am definitely re-thinking my approach. Thank you
I have a bunch of r730, they are idling at around 128 watts. My old r720 is out of the rack since late 2021. I don't recall what were the idle wattage for this one.
I have an R720 with tons of CPU and RAM as testbed of my homelab, which I power on when needed, and for h24 VM and selfhosted services a mini-pc ryzen based
I'm researching currently how to automatically fire-up an additional server (maybe with MAAS) depending on the actual load (k3s/harvester) and powering down when load settles down again.. Not easy, but it should be possible. Would be great if you'd be able to provide your insights if you get to something like this. Keep the vids coming, they are really teaching and entertaining, thank you for this!
You can use a linux script in a cron that mesure the cpu usage or memory during an interval and following a condition you put, you awake one or more server with WOL. A bash script can do it or python , depending on your taste. #!/bin/bash # Configuration CPU_THRESHOLD=80 MEMORY_THRESHOLD=90 INTERVAL=300 # Check every 5 minutes MAX_SERVERS=3 CURRENT_SERVERS=1 # Function to get CPU usage get_cpu_usage() { top -bn1 | grep "Cpu(s)" | sed "s/.*, *\([0-9.]*\)%* id.*/\1/" | awk '{print 100 - $1}' } # Function to get memory usage get_memory_usage() { free | grep Mem | awk '{print $3/$2 * 100.0}' } # Function to wake up a server (replace with actual WOL command) wake_server() { echo "Waking up additional server" # Add your Wake-on-LAN command here # Example: etherwake -i eth0 00:11:22:33:44:55 } # Function to power down a server (replace with actual shutdown command) power_down_server() { echo "Powering down excess server" # Add your shutdown command here # Example: ssh user@server-ip "sudo shutdown -h now" } while true; do CPU_USAGE=$(get_cpu_usage) MEMORY_USAGE=$(get_memory_usage) echo "Current CPU usage: $CPU_USAGE%" echo "Current Memory usage: $MEMORY_USAGE%" if (( $(echo "$CPU_USAGE > $CPU_THRESHOLD" | bc -l) )) || (( $(echo "$MEMORY_USAGE > $MEMORY_THRESHOLD" | bc -l) )); then if [ $CURRENT_SERVERS -lt $MAX_SERVERS ]; then wake_server ((CURRENT_SERVERS++)) echo "Servers now active: $CURRENT_SERVERS" else echo "Maximum number of servers already active" fi elif (( $(echo "$CPU_USAGE < $((CPU_THRESHOLD - 20))" | bc -l) )) && (( $(echo "$MEMORY_USAGE < $((MEMORY_THRESHOLD - 20))" | bc -l) )); then if [ $CURRENT_SERVERS -gt 1 ]; then power_down_server ((CURRENT_SERVERS--)) echo "Servers now active: $CURRENT_SERVERS" else echo "Minimum number of servers already active" fi else echo "No action needed" fi sleep $INTERVAL done
One thing that I don't see enough coverage about: LXC Containers. Especially if you run Proxmox, using LXCs instead of full VMs saves quite some resources. You can still run stuff like Docker and if you don't need hardware passthrough. Also these containers start a lot quicker than some VMs. Basically anything that doesn't require Windows runs as an LXC in my homelab, and I run these on a cluster of 2 NUCs with 1TB NVMe SSDs and 64GB RAM each. The only power-guzler at this time is my Storage server, which is an enterprise-grade 1U server with 4 HDD bays, which sits at around 90W. But everything I store is on that NAS, so I don't see a reasonable way to turn it off at night or something.
On my primary server I ended up having some simple cronjobs to automatically switch cpu governor between schedutil and powersave depending on the time of day. To be more specific, electricity at night is slightly cheaper here, so I run it as schedutil (default) during the night and powersave during the day. This will cause the cpu to basically stay at the lowest frequency, but for my workload that's fine. Some cpu heavy tasks just take slightly longer. In my grafana dashboard where I see the wattage used by this server through the PDU I can see exactly where this script triggers every day.
As others have said in the comments, set up energy usage monitoring. If you absolutely have to, you can report usage in units of Raspberry Pi, but it's optional. You can find the advertised energy consumption of components before you buy or install them. If you decide to install them, please check your own measurements (discrepancies can be configuration options and are not always bad faith marketing). Home automation components can help shut down containers, VMS, servers, etc., when not in use and be sure startup and shutdown work flawlessly while you save power. Run components in the public cloud where sensible; the hyperscalers are putting a lot of effort into reducing power usage and we can make use of that while adding some "hybrid cloud" to our home labs.
I'm building my first homelab, and I've had efficiency in mind all along, I plan on having orange pi's doing the heavy lifitng of whatever I can fit on them basically. This saves power, and this also being my gaming rig, I'd like to save resources for when I want to try new games, or also ai workloads, or video editing.
For me, switching from a Nas, where the hard drives ran 24/7 to an SSD cached Unraid server has made a huge difference. I am able to save about 80-100W by that. Unraid is optimized to let the drives sleep as long and often as possible. The server is still on 24/7, but needs WAAAAY less power because only the cpu, ram and SSDs are running. And yeah, less drives bigger size.
With power efficiency in mind i decided to go with Unraid. The way Unraid works allows that only the one drive wich contains the actual data is active. All other drives are in stand by mode.
Have this power reduction project under way too. Hardware Firewall, Dell r415 AMD Server, Synology NAS, ISP bridged Modem, PoE Switch (2 APs, 2 cameras) and UPS drawing about 250W. Pretty sure I can drop this down with a Tuofun 4 port router for the firewall and swapping out some spinning disks for SSD's. Great video, lots to think about!
So partially agree. If you run servers, with a 750w PSU, that’s what most desktops have. I do agree that virtualization is the key, plus changing from performance mode to balanced also works. Most server hardware will also let you run in redundant mode so only one PSU is taking the load while the other is idle.
The best way of "saving energy" for me is to buy and install 4 windmills and solar panels with the appropiate bank of Energy. Depending on the size that could save a LOT of energy not just for all my hardware stuff, but for the full home for several years. Is an investment.
Good video! I built a TrueNAS scale machine last years with a HDD pool and a SSD pool, thinking I could make the HDDs go to sleep when I'm not making backups or watching some video. I practice, they wake up too often for it to make a real difference, while lowering the lifetime of the disks. Also, getting a server board with features like IPMI was cool, it it consumes around 8W even when everything else is off. So 8W on top of whatever the server uses to do useful stuff.
Modern drives park the head so putting them to sleep has minimal lifetime degradation. One article estimated you could park the head every 20 minutes for 10 years and still not reach the manufactures estimated sleep cycle expectancy.
Personally I feel like switching the 30+ lights in my house that used to be 100watt bulbs to 7w led more than makes up for any computers I would run all the time.
Basically did the same, removed 3 HP servers, 1 g8 micro, 1 g8 380p and 1 g7 380. Replaced all of them with a amd system built with consumer hardware with 3 16tb drives and some nvme drives for faster storage... It is SUCH a difference in power consumption 😅 and since the electricity prices in Sweden went crazy it was just in time... Absolutely not as fun, but the wife approval was much higher with the new server :-D probably because she have no idea what I spent on it... But hey! Low power consumption! 🎉🥳
1- TrueNAS in Jonsbo N3 - 50W 2- Proxmox on Beelink SER5 - 60W 3 -Raspberry Pi 4 as DAC/Media Server - 10W 4- 2x Switch + 3x AP = 20W 5- Firewall/Router = 10W Total: 150W 24/7 I am not able to get lower but that is ok
in my experience even with a big NAS, 24bay in my case, with proper CPU hardware, new HBA SAS adapter and last but not least a proper sized PSU the power consumption is on check. PSU wattage and efficiency rating is really important.
I have 2x Lenovo ThinkCentre M900 Tiny PCs with Proxmox. Most of the stuff I use are running in docker containers that are running inside LXC containers. But included in my cluster are things such as Plex Server (Direct Play Only), Two instances of Pihole, Wireguard VPN, Bitwarden, etc... I put a power meter on the PDU that both of these Tiny PCs are connected to and together they're averaging about 25 Watts. Makes me sleep so much better at night knowing that my entire home lab just sips power.
Thanks for the video.. Recently bought some KASA Matter Smart Plugs for power monitoring; need to measure the watt draw on a few of my components.. I'm probably going to hate my electric bill over the next few months.
I have automation in home assistant to pause not needed VMs on proxmox after 6PM (leaving only 2 Domain Controllers up). This is saving quite a lot of watts. My proxmox cluster is 3x miniPCs (intel 2x gen10 and 1x gen12) with 64gb ram each. I can just press "wake" button anytime in HomeAss :)
Great video. But I should say that first advice should be that always think about what use case you build your server for, for light load your advice is exelent, in my case for example it should be expensive and waste time. My home server can often have full load for days, use them as render farm for blender and other 3D applications. Then old (5 years or so) is most efficient
One thing I learned the hard way is consumer hardware can be tempting but can also be limiting. I ran into a issue where my MOBO/CPU doesn't have enough PCI lanes/slots to plug it in . I have a 10g nic which takes up the only free non 1x PCI slot which I wanted to put a HBA card in(I have a Tesla P40 to handle video work/Stabil diffusion). Since I do use my server as a NVR so it has to run 24/7 so turning off my server is not a option.
Lenovo Thinkcentre mini Pc with Proxmox ,Vm , Cluster, intern 2x 2Tb SSD and 2x 4Tb Toshiba extern USB HDD only 8W in Idle. Coming soon again 2x 4Tb on USB. Runing 24/7 since 2 years now . Great for what i need as Homelab
Is it possible that Christian can sell some of his unused hardware for his fans and/or subscribers as refurbished hardware by putting them on Ebay or by answering the questions with correct answers with some luck or selling them to server campus like IONOS or Hetzner so that these conpanies can test and re-setup the hardware with their own structure to optimize the power and to save the bills?
Crhis! Amazong that you are doing this and inspiring other to be efficient. My home lab is already scaled down to low power server (desktop) machines. However, my truenas core uses 6x4tb in raidz-2. As you suggested small number of disks would be a better way to go in future. However, I don't have a plethora of sata ports to add another zpool and migrate. What could be a solution in such situation. Remember most of homelabs are not big server with 8 or 10 hot-swaps. I hope you can share some guidance here.
I dropped VMware in 2-3 Dell servers and tranformed all to 1x Proxmox + 2x Synology/Xpenology NAS. Its always about Intel "L" version of CPUs for servers - sure, very bad Turbo 2-2,3Ghz - but my observation with VMware /10 VMs/ - most of the time CPUs sit at 10-15%. Only Win-Lin update and Chia shytcoin farm rise it to 30-50%. I disconnected 30x HDDs from Chia farm too, as that coin is dead and electricity is a luxury now. Finaly I am down from 500W to 270W alltogether including 12-port full 10GB Mikrotik switch.
The exception on the modern cpu lower power is the 13th gen intel. Power usage went through the roof... 14 gen improvement was to start lowering that power usage again
Saving money is the big thing. I built my storage and VM as best as I could. My main NAS is a truenas setup running on a Xeon Bronze 3204 CPU with 12 HDDs which are in 3 pools. My backups run in the night, so powering down is not an option. Also starting up and shutting down would cause a lot of wear and tear on the system. I have my main Proxmox running on a desktop MB with 64GB of RAM and Ryzen 5950X CPU and a 2nd Proxmox system running on a older Micro HP G3 800 computer. My backup NAS is powered by an Atom based MB also running TrueNAS. I lastly have another TrueNAS system which is my off-line archive. It get connected to power and network only when I backup to it and then turned off and disconnected. Will have to look into what settings I can tweak on my NAS system to save some power. Besides all the regular data on the NAS, I have my movies, TV show and music collection. Would love to have SSDs instead, but the sizes I require are not available yet and even if they were, I could not afford it.
It's a great opportunity to automate with HA. I am shutting some stuff during the night and when I am away (I don't need hot water in the boiler when I am not home). It automatically turns on when I get back home. I am running NUC with Docker and Kodi, I might try to turn some VMs off automatically.
a good way to save power would also be to generalize your server, dont buy many things for servers but just make 1 good server that runs everything, virtualize truenas, virtualize your windows servers and virtualize your other stuff, ive been doing this now for a couple years and power consumption isnt too bad when you do it like this.
My docker host is an older NUC (i3-6100U) that draws 4.x W at idle. I also have a 4x2TB DIY NAS (25-40W idle), that is only used for backups. Since that backup machine is not needed 24/7, it starts via RTC schedule, runs all the backup job and shuts down again.
The first rule of hardware for tight-money people: try to get a motherboard with a lot of RAM slots. compromise a little to get a slightly pricier mobo, and get more breathing room in the future.. with that, you can add RAM as you go in the future. Of course, first, you need to set your target, intel or AMD, and which gen you want to get.
@christianlempa I think I have another idea...You can take mobo with OC possiibility, plus CPU with larger computing power then u need and use CPU OC functionality in opisite way - underclock and undervolt it, eventiually modify it to work at idle at lower frequency, turbo boost to base/stock frequency. ;)
I'm using a Ryzen 5 5600G in my Proxmox host. I have 14 VMs (1x Windows Server, 1x HomeAssistant OS, 12x Debian) running 24/7 and my server only has a power consummption of 40-50W/h. (it can go up to 60W when backing up my VMs). That is about 8-15€ per month here in Germany.
Hi, for everyone posting exactly how much power each individual system draws, do have some power strip that measures each port individually (and constlantly)or do you measure for one system at a time? Looking for good solution that measure for each plug that don't cost a fortune
Main problem with point 1 is that not many consumer boards/cpus support ECC, so that eliminates the ability to use truenas, which is what i run on my t320 currently.
Now would be a great time to buy used intel Nucs. They just discontinued them, and you might want to get them before the scalpers. There are also other brands too if you prefer something other than intel.
@@christianlempa, if you are going for cheap and refurbished, I would go for a 1 liter PC like the HP Elite Mini 600 G9 PC. I have a bunch of Dell OptiPlex 7080 micros. They are great, especially if they are enterprise or business-grade. They also offer upgradable solutions similar to a tower PC. I also like Beelink mini PCs. You can get the older ones cheaper, but the new ones provide incredible performance. The RUclips channel Servethehome offers great reviews for specific projects. (thank you for commenting. It made my day from a cyber attack)
How will i know which add-in cards stop me from using better c states? how do i find out which c states are supported by a mobo or cpu before i make the purchase?
One thing to note for truenas scale, is that you can't just add one or two more hard disks and expect it to work. You have to 2x/3x/4x... the number of disks for it to work. For this reason I prefer unraid.
I've been looking to replace my current homelab with newer more efficient hardware. But it's too expensive or nearly impossible for me at the moment. My NAS has a Xeon E5 2676v3 with 256GB ECC RAM, and my vSphere Cluster has 4 servers with Xeon E5 2696v3 and 256GB ECC. Finally the router is a Core i5 8500T with 32GB, which is really efficient (12VO Power Supply, it idles at ~15W). Everything adds up to 940W, with 10GbE switch, AP PoE, ISP routers (dual WAN), 1GbE switch. So, in total each Xeon requires between 160-180W. The issue is that I would need to go from DDR3 to DDR4 or DDR5 and hardware that supports 256GB is really expensive.
Год назад
May I ask what your power bill is? Doing a rough calculation, your setup would cost me about 4000 USD per year just in power. Living in middle Europe.
@@christianlempa Awesome thanks, your videos have been extremely helpful as I work on getting the skills to transition into IT. Appreciate the timely response,this is the faster I've gotten a response ever lol.
What do you guys thin about E-Cores in a Homelab/NAS? Are they worth it? I am thinking about getting a lattepanda sigma for my first homelab and it's a bit expensive, but offers almost everything I need in quite small dimensions. Also as a normal x86 cpu it would be capable of running a remote windows machine so I don't need to bring my laptop everywhere I go to create some CAD designs when I am not at home.
Very very interesting :) I'm in the same boat and all my purchases are done with energy consumption in mind. I'm just struggling a little bit with my TrueNAS scale server that has an intel 12100T and HBA 9300 : I canno't do more that C3 for C-States. Anyone would know how to find out the reason ? Thanks !
My first home lab was on the 1950x thread ripper.. When it was new.. Yes. I vowed never again, so I have a few extra tips. Use old gaming LAPTOPS if you need a gpu. Way more efficient and bonus ups. The m.2 ports can act as pci express ports btw. Power supplies use more power than you would expect. If your going for efficiency look into a pico psu. Otherwise do your research Along the same vein, aliexpress router boards are awesome. I have one with 4 2.5g ethernet ports and a pci express port. I virtualise my router and one of my nas here. While using a pico psu because I use ssd. Along with what Christian said, inbuilt on motherboard is more efficient then addon cards. SSD ARE MORE EffICIENT THAN SPINNING RUST. Containerise everything, or if you can't Virtualise everything... Including your gaming PC. It is 100 percent possible. (I used to virtualise windows for this, but these days I virtualise a Linux gaming vm... It's more efficient). Don't put your router and gaming PC on the same hypervisor though 😅, learned that the hard way. A Raspberry Pi with a ethernet port is all you need to automate the power for these 😁. Finally, find a way to automate turning off the power at your home.. A managed powerboard.. Or a ups.. Or even add automation to your switchboard. That way you can turn everything you need off when you aren't home.. It will literally save you thousands.
Good video for the most part, but *please don't* recommend RAIDZ1 with drives over ~2TB. The chance of a 2nd drive failing during a rebuild is NOT zero, especially with larger multi-TB drives. And a lot of people apparently don't have backups.
how do people recommend c series chipsets for "cheap and has ecc" when all used boards are neither c236/c246 boards start at $300-400 and come with standard ddr4 slots and apparently ecc modules that would fit in there just don't exist
Saved 150 watt per hour by using hdparm for my HDDs. Use command hdparm -y and ps aux | grep hd-idle | grep -v grep | cut -c 66- ; for f in [a-d] ; do hdparm -C /dev/sd$f | grep -v "^$" ; done command to check the state. Saved me over 30% power consumption. This box has several HDDs that only being used a few times per week. The run ProxMox v8.
there always gotta be paradox on finish compute faster and rest early versus just do lazier compute on old 2nd hand cheaper peace of crape that save money now
![[FREE] DON TOLIVER TYPE BEAT ~ "SHIBUYA"](http://i.ytimg.com/vi/YW6KRkZmOg8/mqdefault.jpg)
Some things I have picked up from my attempts to reduce the power consumption of my homelab:
1. AMD has processors with G and GE suffix that have greatly reduced TDP and have an integrated GPU. Combine that with an ASRock board with IPMI and ECC RAM and you have a great, low cost, low power base for a homelab server.
2. 10GBE over copper takes a lot of power. If you use SFP+ you have the option to use multimode fibre instead, that usually is lower power AND lower latency.
3. Standby mode for HDDs in a NAS was shunned in the past as it had a negative impact on the longevity of the drives. Modern HDDs however are much more resilient and can tolerate much higher power cycle counts. An inbetween on Truenas is Adv. Power Management that can be set to low power states without spindown.
4. Shuting down unused VMs has a measurable impact on my server. A Win 10 Pro on my system uses 5-7 Watts. Containers tend to use less power, as they use the OS infrastructure of the host.
5. Don't assume, measure. I used some cheap power meters with tasmota to stay on top of my homelab and check the impact of changes on power consumption.
> I used some cheap power meters with tasmota to stay on top of my homelab and check the impact of changes on power consumption.
That would be a nice video to watch. How to implement such. It'd be good :) @christianlempa
@@salvationbygracethroughfaith You are right of course. I just am willing to expend the energy for IPMI as it saves me additional hardware, effort and inconvenience. To me it's about reducing the energy needs while still meeting my needs.
From my understanding, the suffix G and GE does NOT reduce the idle consumption, but only limits the maximum possible consumption.
@@mcmormus You are correct. From what I gather, some Intel CPUs have lower idle power than AMD, but that quickly negates under load.
A 6th gen Core processor can easily idle around 10W or less with the right PSU. I'm seeing far less desktop AMD CPUs idling this low.
This is actually what I have been working on the past month, I am converting all my compute from older DUAL Xeon boards, to a Proxmox Cluster of 8C 16T HP mini PC's, and changing my NAS to a low power system, and way smaller box, from 11 Hard Drives total , to 6 Larger drives, and putting it all in a way smaller rack to conserve energy and space. The whole BIG rack and big PC's is cool, and nice , until you get the complaints from the wife, and the electric / cooling bill. I live in Texas , so any energy I use is basically doubled due to cooling costs.
Good to see a video for homelab that finally addresses power efficiency. Most of my servers have such low activity, I could never justify something with high energy use. Great video 🙂
It is almost unfair whenever I see US youtubers building homeservers. They are almost always totally overkill or don't show any considerations for power consumotion. Most of the servers in those videos would cost me several thousand dollars of power here in central Europe if I would run them for their expected lifespan.
It is nice to see a competent youtuber who considers or even prioritises power consumption.
Thank you! I totally agree with you, so I’m excited to create more Homelab content for the „small“ people :D
Soon enough we in the US will realize that power consumption needs to be factored in here.
I have a x86 kubernetes cluster (1 mini computer). It was on 50 - 60% cpu all the time.
Then I joined a Raspberry Pi 4 8gb to the cluster helping out the workload. Now the x86 node is on 5% cpu, letting the raspberry pi do most of the work.
I am amazed by how far we have come with ARM64 support.
Cert Manager, External DNS, Grafana, ArgoCD, Prometheus, Loki, InfluxDB ,Uptime Kuma, Kafka, all works as workloads on ARM64 out of the box, with out a single line of code changed, making it very appealing to create very power efficient ARM64 clusters!
I will definitely look into that :D Sounds amazing what can be done with ARM now.
love the timing of this video - over the last 90 days, I've been doing efficiency changes to my homelab, dropping my overall kW commit from 2kW/m to 0.4-0.6kW/m, obviously I did this with money lol but getting rid of very old hardware for some new age 15-75kW servers is amazing. 10th-13th gen infrastructure offers such robust performance packed into tiny packages of power.
When it comes to shutting down servers, I try to get things to sleep where I can and yes, WOL is great to instantly wake and resume these servers - unfortunately a lot of users are intimidated by the issues of sleep (automatic wakeup, GPU performance failure, windows poor compatibility since windows7) which is why I think it's not used as much as it can be but is very useful for a desktop PC with high wattage that gets usage randomly throughout the day, sleeping can save atleast 0.2kW/h for my 13th gen i9, that probably doubles or triples for the AMD users.
Totally agree with you man, great video.
Thank you! We’re in the same boat 😊
Great topic. So far I have just two Rpi for my homelab needs, in neat argon cases. Running all on SSD totally efficient and silent in my living room. I would not use them for proxmox but it is crazy how far you can push these little sbc while using very little power.
Is everyone here too young to remember when single lightbulbs used 60 or 100w? 😅
Your spouse because you compare your server consumption to “just 2 or 3 lightbulbs on all the time” 😂
Yes, and also when leaving one of those 100w bulbs on wasn't likely to affect your credit rating 😂
Lol
I know, right? My 14-drive ZFS SAS shelf draws ~133 watts fully loaded - and I'm like, we had LIGHT BULBS that were almost drawing that much (120W) back in the day
Noticed this weekend I missed one in the attic....
I'm in the process of revamping my little home setup and have been working on exactly this. Smart plugs + home assistant (or another logging/monitor app) to monitor power usage has been helpful until I grab some UPS's to figure out how much power I was using where. Also swapped out some older systems with really low power mini pc's where I could (opnsense for instance). Hardest thing has been upgrading my NAS to keep it relatively low power and keep all the features I want but that's probably the last thing I'm going to upgrade so I have time to decide on that one. Only leaves 1 power hungry Epyc server but I'm not getting rid of that one!
Funny thing is overall swapping the lightbulbs to LED, replacing some old appliances and turning off idle audio equipment (sub/amp) saved me more power overall but my homelab isn't very big.
I was planning to pickup a used Dell Poweredge R720 with 256GB of ram a few days ago. I had to reschedule. After watching your presentation I am definitely re-thinking my approach. Thank you
I have a bunch of r730, they are idling at around 128 watts. My old r720 is out of the rack since late 2021. I don't recall what were the idle wattage for this one.
I have an R720 with tons of CPU and RAM as testbed of my homelab, which I power on when needed, and for h24 VM and selfhosted services a mini-pc ryzen based
Really liking this trend in the last year where labbers are reevaluating their compute needs and power consumption
I'm researching currently how to automatically fire-up an additional server (maybe with MAAS) depending on the actual load (k3s/harvester) and powering down when load settles down again.. Not easy, but it should be possible. Would be great if you'd be able to provide your insights if you get to something like this. Keep the vids coming, they are really teaching and entertaining, thank you for this!
You can use a linux script in a cron that mesure the cpu usage or memory during an interval and following a condition you put, you awake one or more server with WOL. A bash script can do it or python , depending on your taste.
#!/bin/bash
# Configuration
CPU_THRESHOLD=80
MEMORY_THRESHOLD=90
INTERVAL=300 # Check every 5 minutes
MAX_SERVERS=3
CURRENT_SERVERS=1
# Function to get CPU usage
get_cpu_usage() {
top -bn1 | grep "Cpu(s)" | sed "s/.*, *\([0-9.]*\)%* id.*/\1/" | awk '{print 100 - $1}'
}
# Function to get memory usage
get_memory_usage() {
free | grep Mem | awk '{print $3/$2 * 100.0}'
}
# Function to wake up a server (replace with actual WOL command)
wake_server() {
echo "Waking up additional server"
# Add your Wake-on-LAN command here
# Example: etherwake -i eth0 00:11:22:33:44:55
}
# Function to power down a server (replace with actual shutdown command)
power_down_server() {
echo "Powering down excess server"
# Add your shutdown command here
# Example: ssh user@server-ip "sudo shutdown -h now"
}
while true; do
CPU_USAGE=$(get_cpu_usage)
MEMORY_USAGE=$(get_memory_usage)
echo "Current CPU usage: $CPU_USAGE%"
echo "Current Memory usage: $MEMORY_USAGE%"
if (( $(echo "$CPU_USAGE > $CPU_THRESHOLD" | bc -l) )) || (( $(echo "$MEMORY_USAGE > $MEMORY_THRESHOLD" | bc -l) )); then
if [ $CURRENT_SERVERS -lt $MAX_SERVERS ]; then
wake_server
((CURRENT_SERVERS++))
echo "Servers now active: $CURRENT_SERVERS"
else
echo "Maximum number of servers already active"
fi
elif (( $(echo "$CPU_USAGE < $((CPU_THRESHOLD - 20))" | bc -l) )) && (( $(echo "$MEMORY_USAGE < $((MEMORY_THRESHOLD - 20))" | bc -l) )); then
if [ $CURRENT_SERVERS -gt 1 ]; then
power_down_server
((CURRENT_SERVERS--))
echo "Servers now active: $CURRENT_SERVERS"
else
echo "Minimum number of servers already active"
fi
else
echo "No action needed"
fi
sleep $INTERVAL
done
Great Video Christian! Also like the fact that you are honest with your NAS Build
One thing that I don't see enough coverage about: LXC Containers. Especially if you run Proxmox, using LXCs instead of full VMs saves quite some resources. You can still run stuff like Docker and if you don't need hardware passthrough. Also these containers start a lot quicker than some VMs. Basically anything that doesn't require Windows runs as an LXC in my homelab, and I run these on a cluster of 2 NUCs with 1TB NVMe SSDs and 64GB RAM each. The only power-guzler at this time is my Storage server, which is an enterprise-grade 1U server with 4 HDD bays, which sits at around 90W. But everything I store is on that NAS, so I don't see a reasonable way to turn it off at night or something.
On my primary server I ended up having some simple cronjobs to automatically switch cpu governor between schedutil and powersave depending on the time of day. To be more specific, electricity at night is slightly cheaper here, so I run it as schedutil (default) during the night and powersave during the day. This will cause the cpu to basically stay at the lowest frequency, but for my workload that's fine. Some cpu heavy tasks just take slightly longer. In my grafana dashboard where I see the wattage used by this server through the PDU I can see exactly where this script triggers every day.
Hey, do you mind sharing the script? I could use this. Thanks!
As others have said in the comments, set up energy usage monitoring. If you absolutely have to, you can report usage in units of Raspberry Pi, but it's optional. You can find the advertised energy consumption of components before you buy or install them. If you decide to install them, please check your own measurements (discrepancies can be configuration options and are not always bad faith marketing). Home automation components can help shut down containers, VMS, servers, etc., when not in use and be sure startup and shutdown work flawlessly while you save power. Run components in the public cloud where sensible; the hyperscalers are putting a lot of effort into reducing power usage and we can make use of that while adding some "hybrid cloud" to our home labs.
Keep it up man, love your channel for this topic, saving money and power, welcome back looking good
Thanks :)
I'm building my first homelab, and I've had efficiency in mind all along, I plan on having orange pi's doing the heavy lifitng of whatever I can fit on them basically. This saves power, and this also being my gaming rig, I'd like to save resources for when I want to try new games, or also ai workloads, or video editing.
Agewise, I have a rule that anything older than Skylake is too old. 6th gen was the dealbreaker in terms of idle power optimization.
For me, switching from a Nas, where the hard drives ran 24/7 to an SSD cached Unraid server has made a huge difference. I am able to save about 80-100W by that. Unraid is optimized to let the drives sleep as long and often as possible. The server is still on 24/7, but needs WAAAAY less power because only the cpu, ram and SSDs are running.
And yeah, less drives bigger size.
Haven’t tried unraid for a long time, never got used to their weird usb stick setup ;D but seems like a good option too
With power efficiency in mind i decided to go with Unraid. The way Unraid works allows that only the one drive wich contains the actual data is active. All other drives are in stand by mode.
Have this power reduction project under way too. Hardware Firewall, Dell r415 AMD Server, Synology NAS, ISP bridged Modem, PoE Switch (2 APs, 2 cameras) and UPS drawing about 250W. Pretty sure I can drop this down with a Tuofun 4 port router for the firewall and swapping out some spinning disks for SSD's. Great video, lots to think about!
So partially agree. If you run servers, with a 750w PSU, that’s what most desktops have. I do agree that virtualization is the key, plus changing from performance mode to balanced also works.
Most server hardware will also let you run in redundant mode so only one PSU is taking the load while the other is idle.
True
The best way of "saving energy" for me is to buy and install 4 windmills and solar panels with the appropiate bank of Energy. Depending on the size that could save a LOT of energy not just for all my hardware stuff, but for the full home for several years. Is an investment.
Good video! I built a TrueNAS scale machine last years with a HDD pool and a SSD pool, thinking I could make the HDDs go to sleep when I'm not making backups or watching some video. I practice, they wake up too often for it to make a real difference, while lowering the lifetime of the disks.
Also, getting a server board with features like IPMI was cool, it it consumes around 8W even when everything else is off. So 8W on top of whatever the server uses to do useful stuff.
Modern drives park the head so putting them to sleep has minimal lifetime degradation. One article estimated you could park the head every 20 minutes for 10 years and still not reach the manufactures estimated sleep cycle expectancy.
Personally I feel like switching the 30+ lights in my house that used to be 100watt bulbs to 7w led more than makes up for any computers I would run all the time.
Basically did the same, removed 3 HP servers, 1 g8 micro, 1 g8 380p and 1 g7 380.
Replaced all of them with a amd system built with consumer hardware with 3 16tb drives and some nvme drives for faster storage... It is SUCH a difference in power consumption 😅 and since the electricity prices in Sweden went crazy it was just in time...
Absolutely not as fun, but the wife approval was much higher with the new server :-D probably because she have no idea what I spent on it... But hey! Low power consumption! 🎉🥳
Whenever a new cpu generation was released I used to never care about the "efficiency" improvements until I got a home lab
1- TrueNAS in Jonsbo N3 - 50W
2- Proxmox on Beelink SER5 - 60W
3 -Raspberry Pi 4 as DAC/Media Server - 10W
4- 2x Switch + 3x AP = 20W
5- Firewall/Router = 10W
Total: 150W 24/7
I am not able to get lower but that is ok
Love watching your videos ! Keep up the great work
in my experience even with a big NAS, 24bay in my case, with proper CPU hardware, new HBA SAS adapter and last but not least a proper sized PSU the power consumption is on check. PSU wattage and efficiency rating is really important.
I have 2x Lenovo ThinkCentre M900 Tiny PCs with Proxmox.
Most of the stuff I use are running in docker containers that are running inside LXC containers.
But included in my cluster are things such as Plex Server (Direct Play Only), Two instances of Pihole, Wireguard VPN, Bitwarden, etc...
I put a power meter on the PDU that both of these Tiny PCs are connected to and together they're averaging about 25 Watts.
Makes me sleep so much better at night knowing that my entire home lab just sips power.
Sounds like a good setup 👍
Thanks for the video.. Recently bought some KASA Matter Smart Plugs for power monitoring; need to measure the watt draw on a few of my components.. I'm probably going to hate my electric bill over the next few months.
Your welcome
the elephant in the room is the blue matrix wallpaper... where did you get that?!
It's about making the right compromises. And of corse, getting the best from the hardware you have, But mostly compromises.
I have automation in home assistant to pause not needed VMs on proxmox after 6PM (leaving only 2 Domain Controllers up). This is saving quite a lot of watts. My proxmox cluster is 3x miniPCs (intel 2x gen10 and 1x gen12) with 64gb ram each. I can just press "wake" button anytime in HomeAss :)
Great video. But I should say that first advice should be that always think about what use case you build your server for, for light load your advice is exelent, in my case for example it should be expensive and waste time. My home server can often have full load for days, use them as render farm for blender and other 3D applications. Then old (5 years or so) is most efficient
Thank you! :)
One thing I learned the hard way is consumer hardware can be tempting but can also be limiting. I ran into a issue where my MOBO/CPU doesn't have enough PCI lanes/slots to plug it in . I have a 10g nic which takes up the only free non 1x PCI slot which I wanted to put a HBA card in(I have a Tesla P40 to handle video work/Stabil diffusion). Since I do use my server as a NVR so it has to run 24/7 so turning off my server is not a option.
"This is the way"
My home lab consists of 12th Gen i7 NUC and Synology 4 bay nas. That is all i need :)
sehr nice! Danke dir! Grüße aus Österreich 🇦🇹
Grüße zurück :)
Lenovo Thinkcentre mini Pc with Proxmox ,Vm , Cluster, intern 2x 2Tb SSD and 2x 4Tb Toshiba extern USB HDD only 8W in Idle. Coming soon again 2x 4Tb on USB. Runing 24/7 since 2 years now . Great for what i need as Homelab
Sounds like a great option!
Is it possible that Christian can sell some of his unused hardware for his fans and/or subscribers as refurbished hardware by putting them on Ebay or by answering the questions with correct answers with some luck or selling them to server campus like IONOS or Hetzner so that these conpanies can test and re-setup the hardware with their own structure to optimize the power and to save the bills?
i've been thinking about getting an i5-13500T for my virtualization server, 20 threads at 35w TDP seems like a good balance of power and energy usage
1st Step to save energy; Do you need your homelab to be on 24/24??? Program sleep mode when you sleep just that simple.😅
Good honest Video Sir !
Thanks
Crhis! Amazong that you are doing this and inspiring other to be efficient. My home lab is already scaled down to low power server (desktop) machines. However, my truenas core uses 6x4tb in raidz-2. As you suggested small number of disks would be a better way to go in future. However, I don't have a plethora of sata ports to add another zpool and migrate. What could be a solution in such situation. Remember most of homelabs are not big server with 8 or 10 hot-swaps. I hope you can share some guidance here.
Thanks. Any bios setting suggestion for X570 + 3700x to saving energy?
I dropped VMware in 2-3 Dell servers and tranformed all to 1x Proxmox + 2x Synology/Xpenology NAS. Its always about Intel "L" version of CPUs for servers - sure, very bad Turbo 2-2,3Ghz - but my observation with VMware /10 VMs/ - most of the time CPUs sit at 10-15%. Only Win-Lin update and Chia shytcoin farm rise it to 30-50%. I disconnected 30x HDDs from Chia farm too, as that coin is dead and electricity is a luxury now. Finaly I am down from 500W to 270W alltogether including 12-port full 10GB Mikrotik switch.
you really have a trash hardware to be above 100w
The exception on the modern cpu lower power is the 13th gen intel. Power usage went through the roof...
14 gen improvement was to start lowering that power usage again
Saving money is the big thing. I built my storage and VM as best as I could. My main NAS is a truenas setup running on a Xeon Bronze 3204 CPU with 12 HDDs which are in 3 pools. My backups run in the night, so powering down is not an option. Also starting up and shutting down would cause a lot of wear and tear on the system. I have my main Proxmox running on a desktop MB with 64GB of RAM and Ryzen 5950X CPU and a 2nd Proxmox system running on a older Micro HP G3 800 computer. My backup NAS is powered by an Atom based MB also running TrueNAS. I lastly have another TrueNAS system which is my off-line archive. It get connected to power and network only when I backup to it and then turned off and disconnected. Will have to look into what settings I can tweak on my NAS system to save some power. Besides all the regular data on the NAS, I have my movies, TV show and music collection. Would love to have SSDs instead, but the sizes I require are not available yet and even if they were, I could not afford it.
It's a great opportunity to automate with HA. I am shutting some stuff during the night and when I am away (I don't need hot water in the boiler when I am not home). It automatically turns on when I get back home.
I am running NUC with Docker and Kodi, I might try to turn some VMs off automatically.
a good way to save power would also be to generalize your server, dont buy many things for servers but just make 1 good server that runs everything, virtualize truenas, virtualize your windows servers and virtualize your other stuff, ive been doing this now for a couple years and power consumption isnt too bad when you do it like this.
Absolutely 👍
My docker host is an older NUC (i3-6100U) that draws 4.x W at idle.
I also have a 4x2TB DIY NAS (25-40W idle), that is only used for backups. Since that backup machine is not needed 24/7, it starts via RTC schedule, runs all the backup job and shuts down again.
The first rule of hardware for tight-money people: try to get a motherboard with a lot of RAM slots. compromise a little to get a slightly pricier mobo, and get more breathing room in the future.. with that, you can add RAM as you go in the future.
Of course, first, you need to set your target, intel or AMD, and which gen you want to get.
this is a great topic, thank you
@christianlempa I think I have another idea...You can take mobo with OC possiibility, plus CPU with larger computing power then u need and use CPU OC functionality in opisite way - underclock and undervolt it, eventiually modify it to work at idle at lower frequency, turbo boost to base/stock frequency. ;)
Thank you 🙏 this video is amazing!!
Glad you enjoyed it :)
Thanks Christian
@christian Thank you for the video. I think the best option is to go virtual.
Thank you! :) True!
I'm using a Ryzen 5 5600G in my Proxmox host. I have 14 VMs (1x Windows Server, 1x HomeAssistant OS, 12x Debian) running 24/7 and my server only has a power consummption of 40-50W/h. (it can go up to 60W when backing up my VMs). That is about 8-15€ per month here in Germany.
I’ve seen similar results with AMD Ryzen. I feel like Intel is a bit ahead of idle power consumption, but it’s still in a good range I agree.
Great! Do you think about cloud cost And self-host cost? Im in this before renew my homelab. Thank!s
Hi, for everyone posting exactly how much power each individual system draws, do have some power strip that measures each port individually (and constlantly)or do you measure for one system at a time? Looking for good solution that measure for each plug that don't cost a fortune
Main problem with point 1 is that not many consumer boards/cpus support ECC, so that eliminates the ability to use truenas, which is what i run on my t320 currently.
But why do you need ECC?
Now would be a great time to buy used intel Nucs. They just discontinued them, and you might want to get them before the scalpers. There are also other brands too if you prefer something other than intel.
Oh no, what about some alternatives?
@@christianlempa, if you are going for cheap and refurbished, I would go for a 1 liter PC like the HP Elite Mini 600 G9 PC. I have a bunch of Dell OptiPlex 7080 micros. They are great, especially if they are enterprise or business-grade. They also offer upgradable solutions similar to a tower PC. I also like Beelink mini PCs. You can get the older ones cheaper, but the new ones provide incredible performance. The RUclips channel Servethehome offers great reviews for specific projects.
(thank you for commenting. It made my day from a cyber attack)
NUCs are quite good for power consuption
unfortunately discontinued :(
How will i know which add-in cards stop me from using better c states?
how do i find out which c states are supported by a mobo or cpu before i make the purchase?
I think there's no way other than just try it out :(
@christianlempa oof that's unfortunate
Nested virtualization is also a nice trick.
One thing to note for truenas scale, is that you can't just add one or two more hard disks and expect it to work. You have to 2x/3x/4x... the number of disks for it to work. For this reason I prefer unraid.
yeah I also power off my home server when not in used and also have a raspberry pi that my wife uses daily for web development that runs 24/7
So you put your server and your gaming pc in the same machine? Is that correct?
I've been looking to replace my current homelab with newer more efficient hardware. But it's too expensive or nearly impossible for me at the moment. My NAS has a Xeon E5 2676v3 with 256GB ECC RAM, and my vSphere Cluster has 4 servers with Xeon E5 2696v3 and 256GB ECC. Finally the router is a Core i5 8500T with 32GB, which is really efficient (12VO Power Supply, it idles at ~15W). Everything adds up to 940W, with 10GbE switch, AP PoE, ISP routers (dual WAN), 1GbE switch. So, in total each Xeon requires between 160-180W. The issue is that I would need to go from DDR3 to DDR4 or DDR5 and hardware that supports 256GB is really expensive.
May I ask what your power bill is? Doing a rough calculation, your setup would cost me about 4000 USD per year just in power. Living in middle Europe.
@ It costs me between 350-400 per year (Argentina).
@@andibiront2316 I am not familiar with argentinias currency, inflation and averge wages but that sounds like quite a lot too.
Where can i get that empty server rack that you used to turn the old gaming pc into a server .
It’s an open frame server rack from startech, which you can get in different sizes. It’s superb
@@christianlempa Awesome thanks, your videos have been extremely helpful as I work on getting the skills to transition into IT. Appreciate the timely response,this is the faster I've gotten a response ever lol.
The problem is not power usage and power efficiency of home lab equipment but electricity prices in Europe.
Both is a problem :/
More nuclear reactors, solar farms and wind power would definitely help you all there
You missed out , connect to your neighbors supply . 😜 But really good tips
lol :D true !!
between proxmox and ESXi, which one would you think/prefer to go with for homelab server?
Both of them 😉
@@Prime0pt I guess I'm going to try proxmox this time since I've tried ESXi
11:00 how much power did the storage server consumed in total? Are you willing to sell it?
ow 4TB only per disk? yeah, that is not smart.
Couldn't you use home assistant to automatically shutdown and boot up your server, instead of having to activate some control on your phone?
That’s exactly what I’m using on my phone ☺️
What do you guys thin about E-Cores in a Homelab/NAS? Are they worth it? I am thinking about getting a lattepanda sigma for my first homelab and it's a bit expensive, but offers almost everything I need in quite small dimensions. Also as a normal x86 cpu it would be capable of running a remote windows machine so I don't need to bring my laptop everywhere I go to create some CAD designs when I am not at home.
No idea, I’m using a CPU with E-cores, but never recognized a significant difference when running proxmox, I think I need to dig further…
Very very interesting :) I'm in the same boat and all my purchases are done with energy consumption in mind. I'm just struggling a little bit with my TrueNAS scale server that has an intel 12100T and HBA 9300 : I canno't do more that C3 for C-States. Anyone would know how to find out the reason ? Thanks !
I think it’s just because of the HBA, but it also could be an old or cheap NVME, NIC… it’s sometimes very confusing
this is the reason i'm using unraid with a big cache ssd
it spins down the drives that i don't use
12x raspberry pi cluster = 1 standard light bulb in your home as of 10 years ago
:D
My first home lab was on the 1950x thread ripper.. When it was new.. Yes. I vowed never again, so I have a few extra tips.
Use old gaming LAPTOPS if you need a gpu. Way more efficient and bonus ups. The m.2 ports can act as pci express ports btw.
Power supplies use more power than you would expect. If your going for efficiency look into a pico psu. Otherwise do your research
Along the same vein, aliexpress router boards are awesome. I have one with 4 2.5g ethernet ports and a pci express port. I virtualise my router and one of my nas here. While using a pico psu because I use ssd. Along with what Christian said, inbuilt on motherboard is more efficient then addon cards.
SSD ARE MORE EffICIENT THAN SPINNING RUST.
Containerise everything, or if you can't Virtualise everything... Including your gaming PC. It is 100 percent possible. (I used to virtualise windows for this, but these days I virtualise a Linux gaming vm... It's more efficient). Don't put your router and gaming PC on the same hypervisor though 😅, learned that the hard way.
A Raspberry Pi with a ethernet port is all you need to automate the power for these 😁.
Finally, find a way to automate turning off the power at your home.. A managed powerboard.. Or a ups.. Or even add automation to your switchboard. That way you can turn everything you need off when you aren't home.. It will literally save you thousands.
MY EX WORKSTATION WITH xEON 1275v5 32GB, 5X2.5 HDD ~20W
But what mini pc can give me gaming specs, and vm specs
?
My servers warm my office. So I turn off the radiators, crack open the windows and it’s all good!
Good video for the most part, but *please don't* recommend RAIDZ1 with drives over ~2TB. The chance of a 2nd drive failing during a rebuild is NOT zero, especially with larger multi-TB drives. And a lot of people apparently don't have backups.
my TrueNas server consumes approx. 20W more with a 10Gbit card than without... that's almost 50% of the idel consumption.
Wow that's crazy
how do people recommend c series chipsets for "cheap and has ecc" when all used boards are neither
c236/c246 boards start at $300-400 and come with standard ddr4 slots
and apparently ecc modules that would fit in there just don't exist
Dont use ECC and your life will be a lot easier
My server running tons of services are NOT running idle, that is for sure.
This is very important especially in Germany. I currently pay 32 €ct/kWh! 🤮
True :(
Saved 150 watt per hour by using hdparm for my HDDs. Use command hdparm -y
and ps aux | grep hd-idle | grep -v grep | cut -c 66- ; for f in [a-d] ; do hdparm -C /dev/sd$f | grep -v "^$" ; done
command to check the state. Saved me over 30% power consumption. This box has several HDDs that only being used a few times per week. The run ProxMox v8.
Interesting 🤔
Me listening to all this with a 7 years old gaming desktop repurposed as a server : 😮
Hi
Wie heißt der wallpaper ? 🥰
Du meinst „cmatrix“? :D
JO DANKE DIR Xdd
@@christianlempa
I think 90% of homelabs are more power efficient than mine with 16 blade servers and some rack servers 😂
And much quieter🤣
@@Prime0pt true 🤣
I use cron to shit down and wake in lan script from a raspberry pi to wake them up when I wake up.
Ich glaub da gab es Ärger von der besseren Hälfte was die Stromrechnung angeht 😁💸
lol 😂 no comment
дякую
there always gotta be paradox on finish compute faster and rest early versus just do lazier compute on old 2nd hand cheaper peace of crape that save money now
after watching this video, his recommendations are costly and should be reasoned before following his projects.
Just run your homelab in the cloud. Simple.