Is there actually some statement regarding maximum possible optimization of Disk Access and the patterns involved? I could imagine the chia team has a lot of ideas what could be done to reduce energy consumption further.
Not really. The drive needs to be all the time in idle mode, as such drawing somewhere between 5-10 Watts (for bigger drives). The plot reads are really brief and sparse, so don’t make much increase over the idle base line. Sure, small drives draw as little as 2W, but you need to have several of those to hold for instance 18TB, what negates the power saving. So, if your farm is only one 2 TB 2.5" HD or SSD, it is really efficient, but your chances of getting paid are basically zilch. To make that $2/day you need around 100TB farm, and the best efficiency is to use the biggest drives.
So, there is an attempt in that paper to say that drive manufacturers may modify F/W for Chia. However, the problem is that for those high-density drives, you need high RPMs, as otherwise drive heads will not be able to read the data. Power efficiency for HDs comes through smaller cells on the platter, so the same spindle motor can be used, when cells get smaller. So, I think this is just wishful thinking (kind of investors targeting paper).
If you could use SSD, the idle power is almost zero which means for chia the total power use is zero too. Then you get into power per sata port…
Yes, you are absolutely right. However, do you remember the last Nov/Dec posts about 18 TB drives for $300 or so? Folks were looking to buy drives for about $15+ / TB, as they wanted to add some massive storage. What such cost would be for SSDs? When do you think we will use SSDs to run PB farms? Currently, with the drive price of about $20/TB, the ROI is around 2 years or so. What would that be for SSDs? So, those are just far fetched speculations in that paper (assuming conditions we see today).
It is funny but I worked it out as 25 years! This is with higher EU energy prices and guessing 5 years life for a 2.5 hdd. Needless to say, but I got 18tb 3.5s instead. In 25 years I’ll regret than choice!
1 Like
I didn’t guess that. Sometimes I think there are only about 1000 of us in the whole country
Deleted my post, beers getting to me, thought you had meant 25 yrs to roi on hdd, then realised you meant ssd.
Not many of us, I have very few chia uk nodes to connect to.
2 Likes
We kind of touched it in another thread. This is done on purpose by Chia (a bad choice in my opinion). They want to prevent nodes from grouping (i.e., sitting on the same ISP). However, as you have noticed, that implies that your node mainly draws data from Elbonia or Wakanda instead from UK. It should be better, if they rather classify those IPs into two groups - local ones, and bridges/remote and assign some percentages to those (e.g., 80% local / 20% remote). This way, the local groups would update really fast without causing much of international traffic, where those bridge connections would still allow good connections for remote regions.
The current approach is kind of going from one extreme (only local) to another (only remote).
1 Like
Comparing energy cost per dollar earned is totally useless. What is the energy cost per transaction? BTC is around $100 USD per transaction in power cost. What’s the equivalent cost per transaction with Chia? Energy cost per transaction is all that matters especially when Chia is FAR more decentralized and secure than BTC already.
Can I see the math? ( 20 chrs )
1 Like
Google it. (20 characters)
Found it, they assumed an avg price for elec, that’s really shitty math so means nothing.
Take the price out. What is the cost per transaction in watts? Jesus…
No body knows without making assumptions do they.
No one knows what gear is running and how efficient it is.
You clearly have no interest in knowing. Cognitive dissonance is a hell of a thing. 
1 Like
I do, but its you stating it, so the onus is on you to prove it, which you have not.
Nothing to do with cognitive dissonance, something I’m well aware of.
I’m not doing the work for you. All the answers you seek are fairly easy to calculate, and as referenced before, have already been demonstrated. All things being equal Chia power consumption is orders of magnitude less than PoW networks while being more secure than PoS ones.
It’s not for me, it’s for you , your statement in question not mine.
As ive stated, can’t be calculated without assumptions, and those assumptions wreck the result.
Are you saying that when the energy cost will approach dollars earned people will still be mining / farming? We have already lost 10% of net space when XCH went down into ~$100 range. There is plenty of people asking whether to shut down their farms or not.
Not sure how you compare “$100” to “Energy cost per transaction is all that matters especially when Chia is FAR more decentralized and secure than BTC.” Are you trying to say that those are the same units?
Which cryptocurrency is the most environmentally friendly? (trgdatacenters.com)
Kilowatt hour (KWh) consumed per transaction
Chia = 0.023
Ethereum = 62.56
Bitcoin = 707
My local cost per KWh average $0.1079
Dollar cost per transaction
Chia = $ 0.0025
Ethereum = $6.86
Bitcoin = $77.56
Orders of magnitude less power utilization, greater decentralization, same or better functionality than BTC or ETH.
Hopefully that is clear enough and my basic math holds up.
1 Like