Topic: Running a node dumb question (Read 334 times)

Even checking watt during idle, it's not small. That GPU use 13.2W[1] while i7-10700k (probably most popular version of i7 10th gen) use 50W during idle. And that exclude other desktop component and PSU efficiency.

[1] https://www.tomshardware.com/reviews/nvidia-geforce-gtx-1080-ti,4972-6.html
[2] https://www.techpowerup.com/review/intel-core-i7-10700k/18.html

I guess your gaming computer will likely consume around 1kWh/day running only a Bitcoin Core node. This is not really a power efficient setup for a node. If you turn off the monitor your GPU doesn't have much to do, your water cooling eats likely more energy than a simple CPU cooler fan, at typically low load by the node software your PSU doesn't operate in an efficient power region, too.

If you plan to operate a node for longer time and 24/7 I'd recomment something more power efficient than your gaming desktop.

You should also reconsider if you use your gaming gear for crypto coin stuff, like a node and your wallet(s). Just saying from a security point of view. Don't manage your crypto stuff on a computer for gaming and other daily internet shit.

I hope that I can tell her this to her face but I’m afraid that I might reduce my computer time if I made a comment about his electricity usage. I’m very serious about my wife getting crazy about our electricity bill even though I’m the one pays it using my salary.

Anyway, I have my previous electricity consumption in average to check as reference on how much electricity it adds during my node challenge period. It just calming if I can show her some hard proof about the actual consumption measurement.

I’m using a gaming desktop with 16GB ram, GTX 1080i GPU and I7 10th gen with water cooling setup. It usually consume too much electricity when I’m playing games on high resolution but I knew that running a node will not consume much power based on what I read.

Such software only as accurate as sensor to measure certain part of the computer. It's not available for certain part and it doesn't consider how efficient is PSU of the computer.

We can't know what components your computer consists of. Turn off any internal or external monitor when not in use.

I have nodes (RaspiBlitz, Umbrel, RaspiBolt; don't ask why ) running on Raspi 4B with 8GB RAM, those consume on average 5-6W each, all measured with good watt-meters. A Lenovo ThinkPad T540 (don't know how old this thing is, I bought it refurbished) with 8GB RAM, a Crucial 1TB SSD runs 24/7 with Ubuntu 22.04 LTS and a Bitcoin Core daemon headless. This laptop consumes on average somewhere in the ballpark of 12-14W. Lowest power consumption is little less than 10W for shorter moments but mostly between 11-12W. When a new block is seen power consumption spikes up to 26-27W as long as the new block is verified. This doesn't take long for a block, maybe around 10s or so (never measured precisely).

Internal screen of this ThinkPad is all the time off, lid is closed. In the evening I do some browsing the forum and other stuff with a larger external monitor. During those periods this laptop consumes somewhere between 12-18W depending on what I'm doing and of course with the block verification spikes when bitcoind is busy with new blocks.

Desktop computers are usually worse than laptops or power efficient small form factor modern office gear. As Pmalek said: get a decent power-meter (Kill-a-watt or similar) and measure your stuff yourself.

If your water is heated by electricity you can almost forget about the consumption of such devices like above. Cooking (to lesser extend), hot water, electric heating, washing machine, dryers and air condition are the power hogs.

The HDD is the problem. Raspberry Pi takes more than a month to finish if you're downloading and verifying on an HDD. That's why it's recommended to finish syncing on your main computer (that has SSD), and transfer everything back to Raspberry later, if you don't want to buy an extra SSD.

Depends a lot on your setup. If you're running a Raspberry Pi, give this link to your wife to calm her down: https://raspberrypi.stackexchange.com/questions/5033/how-much-energy-does-the-raspberry-pi-consume-in-a-day/31826#31826.

Tell her not to use the hair dryer so often or empty the water heater every time she takes a shower if you don't mind sleeping on the couch for a few night.

I don't think it's that easy to calculate. If you are serious about it, you might need something like a watt meter. There are also third-party software you can run to measure the consumption for you. Like with anything, use them at your own risk.

Here is just one article I found on the subject:
https://www.makeuseof.com/windows-pc-power-usage/

@Cricktor,

I already apply all the suggested tweaks here and it improves a lot my syncing progress on an insane level. I'm already at 98% syncing progress since I started running a node with a prolonged start. Everything went smoothly after I applied all the suggestions.

Does anyone here know how to get my computer's approximate power consumption while running a node? My wife keeps asking me why my computer is always open 24/7. She is already threatening me about our electricity bill. 

You have been given a few meaningful optimization tweaks. Did you apply those and now wonder why sync speed improved?

We can't know how your computer is utilized and thus what all affects the sync process of your Core node.

I once made a test with my Raspi 4B 8GB RAM with a 1TB SATA-USB3 SSD (UASP active) running a Umbrel node: I had to uninstall any app in that Umbrel and thus wiped all Bitcoin Core blockchain data on purpose. I then installed only the Bitcoin Core app and tweaked dbcache to use about half of RAM (4883MB were accepted as dbcache value). The initial blockchain download and sync took (Tor-only for outgoing connections) to my surprise only about 95h on my Raspi 4B (I have stable ~100MBit downstream for my internet connection, you hardly get this speed sustained over Tor all the time; but it's possible this Umbrel node talked to any of my other three local Core nodes that are running in my local network, I haven't excluded this possibility to keep things simple).

All those who complain a Raspi 4B takes weeks to sync, I can reply that with optimized parameters a Raspi 4B is capable of executing an IBD in less than one week with a normal 1TB SATA-USB3 SSD enclosure for data storage.


P.S.
I still have the debug.log of this IBD experiment. I started it on approx. 2023-06-22T18:00:00Z and finished sync to chaintip at block 796033 on 2023-06-26T16:46:31Z. Later bloating of the UTXO set won't make the sync process any easier for low RAM devices.

I have a follow-up question to this, why did blocks during 2016 give me a slower sync while Bitcoin transactions during 2017 onwards were much higher due to the demand? I sync faster now after I finished syncing blocks in 2016. I'm already at 85% progress compared to when I posted this which is just 15% sync stuck for a long period.

I'm curious what's with 2016 blocks that cause slow sync?

You should create your bitcoin.conf not .txt  and once it's setted you should see changes
The # makes it a comment jut like writing comments for codes
Comments  don't get to be read in a code just for readability and reference mssg
On

# is a commenting character. Putting it before the command disables this command.

Regarding the value of dbcache. As a rule of thumb dbcache=1/4 RAM. Thus in your case the optimal would be dbcache=4096

Read more on dbcache which   essentially acts as write buffer rather than a cache.

How can I know if the "dbcache=8000" setup is already working? The Bitcoind windows process too much data quickly which makes it hard to read the initial summary when I open it.

why this code doesn't have "#" in the beginning compared to other commands listed on my bitcoin.conf. Lastly, what's the difference between Bitcoin.conf file located on the file path in which I installed my Bitcoin core and Bitcoin.conf that I created on C:\Users\YourUserName\AppData\Roaming\Bitcoin path when I input the pruned command? Thanks for your guidance sir.

You are correct,  even though it is  not 88% nowadays but  "absolute minimum"  matters. In  fact , I had a chance to appreciate blocksonly=1 setting   a  few months ago at my node's IBD.

88% was obtained from Bitcoin Core 0.12 though. Since then, Bitcoin Core implement compact block which allow faster block propagation while reducing node's bandwidth usage (since node doesn't have to send transaction data (standalone and in block). And i recall node which in sync progress doesn't obtain TX mempool, although i can't find source of that statement.

Set blocksonly=1 in bitcoin.config file. As it was mentioned by gmaxwell  it may result in  "reduced the node's bandwidth usage by 88%." After completing IBD you may set default , i.e.   blocksonly=0

Last I checked Electrum wallet file didn't have any extension and it was not a dat file either. It's plain text using JSON format.

You should be able to increase your cache size and speed up the verification process. Set the dbcache in your bitcoin.conf file to 8 GB (since you have 16 GB memory, I believe half is good).

Gotcha. This makes sense a lot. Thank you, I'm just a little bit carried away by how fast it was synced on the first day it almost completed 15% so I thought that it would be inconsistent speed. So this means that the 2016 onwards block content is much heavier in file size due to the volume of transactions.

Is there a way to speed this up a little bit using the remaining unused RAM and CPU?

Follow-up question, What is the data shown by the Bitcoind window? Is it related to the syncing process on Bitcoin core because my Bitcoin core is progressing even though Bitcoind is closed?

There is no problem and syncing becoming slower is completely normal.
As you know, your node verifies the validity of each block it downloads and the transactions in it. Currently, your node is synced up to 2016 (according to the screenshot you posted) and prior to that date bitcoin wasn't so popular so blocks back then we're either empty or had few transactions on them therefore verifying their validity (syncing) doesn't take too long.

I appreciate all the replies and learnings to this uncharted territory for me.

@NotATether I will use this thread to inquire about my problem with your running a node challenge.

Bitcoin core syncs very slowly when it reaches 15% progress while Bitcoind performs at a usual speed. I'm not sure what is the problem because my RAM and CPU usage were still the same when I started this challenge including the storage on the hard disk since I used the prune command that you suggested on the challenge thread.

the screenshot below is my current stats


it doesn't have any good progress even though I left it overnight running.