Struggling to heat your home? How about 500 Raspberry Pi units?

Reusing heat from servers has gained momentum recent years, but UK Power Networks (UKPN) is taking an unusual approach: installing mini datacenters powered by Raspberry Pi hardware in customers homes to provide heating for families struggling with energy costs. UKPN, which manages the "last mile" of cables and substations …

[that one in the corner]

Less cynical than an older offering[1]

Can anyone help with a reference for the following (early morning Duckducking the web only leads me back to what TFA already mentions)

IIRC there was (is?) a US company that was pushing a similar idea - heat your home from a server - except that it was a more calculated approach: they sold you the kit outright, you connected it to your power and broadband then made money by selling the compute back to company "at market rates". So only for the already rich (beautiful homes used to illustrate the system in use, no painted brick utilty room) and the "calculated" part was, of course, that the company had absolutely no skin in the game at all once they'd shipped the package off.

Anyway, fingers crossed that this new scheme pans out. At least the server farms are using well-known components: if Thermify does quietly vanish away then we can always draft Jeff Geerling to get the boxes running and talking to each other again.

[1] unless I'm suffering from LLM-style hallucinations - maybe I'll step out of the shower and find that it was all a terrible dream

[Vikingforties]

Re: Less cynical than an older offering[1]

Not USA but this has been done in France for a while by a company called Qarnot: https://qarnot.com/en

They use 4kW "boilers" out at various sites with heat exchangers.

They're more in the business of directly selling their cloud compute / HPC now.

[Anonymous Coward]

Re: Less cynical than an older offering[1]

There have been several companies around the world that had similar offerings more than 10 years ago - I remember there was a Dutch or Danish company that had domestic wall-mounted radiators containing Intel Xeon CPUs for grid/cloud computing and there was a German company that a slightly different solution for businesses.

This is nothing new.

[The Dogs Meevonks]

I use my home server and PC's as mini heaters, as well as the conservatory.

My home server (Ryzen 5600G, 32GB DDR4, 60TB of HDD) is on 24/7, it's not a high power machine and idles at around 30ºC as it's slightly underclocked for sound/thermal reasons... and even under load transcoding/streaming media around the house... Barely ever gets above 42ºC thanks to the very large cooler that's more suited to a 200w load than the 65w max of the current CPU (probably more like 15-20w on avg).

But that's still 30ºC or so of heat that needs dissipating. So with some nice large 140 & 200mm noctua fans installed on the case to intake and exhaust the air as efficiently and quietly as possible... it's left idling away in the corner of the office and the door is normally left open during the day. It's not a huge amount of heat, but it keeps the office nice and cosy a few degrees warmer than other rooms in the house... so I let that heat spread through the upstairs to all of the bedrooms. Throw in my main work and gaming PC's that are often on for 6-12hrs a day between them pumping out a reasonable amount of heat too (work rig Ryzen 3 3300X, RX550GPU) and gaming rig (5800X3D, 64GB DDR4, RX6900XT)... it can get quite toasty in the office in the warmer weather.

Throw in the upgrades to insulation I've done to the house (upgraded loft from 100mm to 350mm, and installed 75mm acoustic/thermal insulation in upstairs stud walls, as well as mineral wool under some floors when lifting boards for pipework and repairs)... then fixing draught old double glazing in a few places and replacing tired old external draught doors with modern more efficient ones... and the house is getting quite cosy with less reliance of gas central heating. Winter gas bills are down about 15-20% over the first 2 winters (a saving of around £300 a year) with an outlay of some £800 for the insulation thanks to some special offers at B&Q last year for rolls of mineral wool insulation for an absolute bargain... 175mm rolls for £25 each... so I bought 14 of them. 4 used in the garage conversion roof and the other 10 in the main loft. With each roll long enough to do lengthways across the whole loft 1.5 times and 4 rolls doing one layer across it all, so we did 2 layers in opposing directions. and used the rest around the vaulted ceiling of the master bedroom when 350mm still wasn't enough to cover all of the exposed sides.

Downstairs, we use the tired old conservatory as a heater... when the sun comes out, no matter how cold it is outside... because it's South facing, it warms it up like an oven. Open the doors into it and let the warmth spread into the house. It can warm the downstairs 1-2ºC extra and avoid the need for the heating to come on at all.

The downside to the old conservatory with draughty sliding door and blown panes... it's too hot to use from Around May to Sept, so we're looking to replace it next year with a more modern one with warm roof. So we can actually use it as a room rather than the junk and clothes dryer we do now.

[Mishak]

Interesting idea

But a couple of questions:

1) Where does the heat go if it's not needed? I'm guessing there must be some way to reject it to the environment.

2) How hot does it make the water? I'm guessing not very, as the aim is to keep the servers reasonably cool.

[RSW]

Re: Interesting idea

This was also my thoughts, there must be a thermal store somewhere as the demand for both heating and compute would never be at the same time. We have solar so still have to have a hot water tank which is used to store as heat the excess power generated. This system can't replace a standard combi boiler

[Neil Barnes]

Re: Interesting idea

My concern also. A combi boiler is heat on demand, so works well for both instant heat (shower, washing etc) and load heat into the house. But it's not generating much load heat in the middle of summer (even in Yorkshire!) so there's a lot less heat being generated other than at bathtime.

The area heating we have here near Potsdam might work better; that has circulating hot water full time, and heat exchangers in the basement. I suspect that the overall thermal load is managed at the heater (possibly by something as simple as a thermostat on the return temperature, but more likely with input from the day's predicted weather and other such frills). That might be something that could better use a server as a heater.

[Wellyboot]

Re: Interesting idea

I agree. Back of envelope numbers, a busy CM4 is around 20 Watt - so a box of 500 CMs needs about 10KW and at a high running temp (lets go with 75C operational) this should be enough to keep a couple of hundred litres at 40 degrees (after a good few hours running). That'd certainly reduce the heating costs for an average UK house.

* the box in photo looks to be about 125x55x70 (using a guesstimate brick size) so the oil fill is possibly not far off 200L.

[Mookster]

Re: Interesting idea

40 Degrees Celsius. Is that a legionella farm you're planning?

[The man with a spanner]

Re: Interesting idea

Certainly an interesting idea. Paired with one of these you only need to run at 57 degC

https://sunamp.com/en-gb/hot-water-solutions-thermino-range/

[druck]

Re: Interesting idea

CM4 is nowhere near 20W, it is at most 7W and usually 2W-3W.

[Anonymous Coward]

Re: Interesting idea

"How hot does it make the water? I'm guessing not very, as the aim is to keep the servers reasonably cool."

I suppose you coud interpose a heat pump to take the low grade heat from the cluster's oil coolant and get the +40°C for hot water.

With increasingly hotter summers in the UK I might wonder how these boxes would hande 35°C days. In AU I have had a 40°C day knock out the refrigeration unit for the water cooled racks in a server room.

An oil filled box of 500 CM4s does make the whole exercise appear to be something out of Uncle Lubin's adventures†.

Any odds on the oil being flammable ?

_{† W.Heath Robinson (1902)}

[Marty McFly]

Re: Interesting idea

"Any odds on the oil being flammable ?"

100% odds

Electrical cooling is a very common use case for mineral oil. The power transform on the pole next to your house is full of it.

[Anonymous Coward]

Re: Interesting idea

What power transformer on the pole? Where I live cabling is in the ground, and even if there are local transformers, they are built into buildings. Also, not all oils used for cooling are flammable, and certainly not at temps these things are expected to see.

[david 12]

Re: Interesting idea

They gave up on using PCB transformer oil, like they gave up on using non-flammable propellants in spray cans, because non-flammable = persistent. I think that they are unlikely to be using a non-flammable oil.

Not volatile like petrol, or even like diesel, but adding to the fire load in your home, like storage of cooking oil -- slightly more dangerous than wood or cotton, because as well as being flammable, those are self-protective insulating materials.

[Marty McFly]

Re: Interesting idea

Ah, but when does it stop being "oil" and start being a non-flammable heat transfer fluid?

[Ian Johnston]

Re: Interesting idea

I had the joy of watching a pole-mounted transformer in the field opposite my house explode in flames a few years back. Very impressive. The two day wait in the dark for Scottish Power to replace it was less fun.

[werdsmith]

Re: Interesting idea

I want to find and look at a transformer on a pole now, I can remember them from decades ago, it would give me a nice nostalgic feeling.

[jdiebdhidbsusbvwbsidnsoskebid]

Re: Interesting idea

Both good points, but they could answer each other. If there is no need for heat, then the water could become very hot (stored for when there is a need).

If in the long-term, more heat is generated than can be used, maybe a thermoelectric generator can convert excess heat back into electricity, which I'm sure would be used. Lots of questions there about process efficiency of course.

[Wellyboot]

Re: Interesting idea

If there are enough (hundreds?) of these installed by the cloud operator and customer-tasks are being handed blocks of CM time (randomly scattered across the cloud) they'd have 24/7 monitoring and control in place, any units running hot would be throttled down to avoid damage. there's also nothing stopping the operator from having a large shed filled with several dozen of these and some really complicated heat transfer equipment for hot summer day capacity buffering.

[I ain't Spartacus]

Re: Interesting idea

With some very complicated controls, you could do quite a lot. The article says this is fitted with solar as well. I used to sell solar hot water systems, many moons ago. One of the problems with those, is that once your primary circuit with the solar cells goes about 200°C you melt the pump. Once it goes above temperature you have to stop circulating - and you can't turn the system back on until the sun has gone in. So the larger your thermal store, the more panels you can have - which means you can still get hot water in Winter (in the UK latitudes) - without having your system go offline in Summer. We used to have a 1,000L thermal store acting as a sort of pretend boiler - and then a smaller calorifier (hot water vessel) of about 300-400L. If the sun hasn't been out, then you have a second heating coil in the samller vessel to a backup heating system, like a gas boiler or heat pump.

There's a way to keep the system running longer, which is to have a radiator with cooling fans underneath the eaves of the roof - which can reject some heat - so the system can stay on longer - and you could use something similar to reject some heat from your server? Plus they're probably only fitting photo-voltaic, not solar-thermal.

So just build a swimming pool.

This is all getting a bit big though. That unit isn't going to fit in the kitchen or bathroom, where most domestic boilers are in the UK.

[Wellyboot]

Re: Interesting idea

It also doesn't fit in very well with the intended market - low income areas.

[I ain't Spartacus]

Re: Interesting idea

Especially if you do fit a thermal store. Which is usually 1,000L. So a 2m high cylinder about 800m diameter. Plus controls, wiring, pipework and manifolds.

It will work well with underfloor heating, which only wants water at 40°C. But again, that's not something you get in cheap housing and is horribly expensive to retrofit.

[Almost Me]

Re: Interesting idea

I really hope you meant 800mm..

[I ain't Spartacus]

Re: Interesting idea

I really hope you meant 800mm..

Almost Me,

Oops. 800mm, 800m - what's the difference? Maybe it's just really thick insulation?

It's big, but not that big.

[Anonymous Coward]

Re: Interesting idea

I fully expect some American to read that as 800 miles.

[midgepad]

Re: Interesting idea

Well, my shower doesn't need water above 50°C

And my CPU doesn't seem to mind running at that.

The oil presumably cools better than air would, I.E. takes the heat away with a smaller temperature differential.

I expect you open a window in summer, and the unit does a quick zap s bit warmer weekly with an immersion heater in the water tank.

[druck]

Re: Interesting idea

All hot water systems need to reach at least 60C to prevent legionella, you the mix with cold water to your desired shower temperature.

[Adrian Harvey]

Re: Interesting idea

All tank-based, at least. Continuous flow systems don’t.

Pedant icon for obvious reasons.

[druck]

Re: Interesting idea

Even with those most heat a small amount of water to a much higher temperature and mix it down.

Low grade heat really isn't much use, except for large volumes at even lower constant temperatures, such as heated swimming pools.

[Anonymous Coward]

Re: Interesting idea

Summer?

[Marty McFly]

Re: A hacker's dream

Meh.

There are a zillion workloads that do not require the high security common in modern data centers. With the exponentially increasing use of AI by the common folk, it does draw the question of whether those workloads need to processed in a high security data center.

Sure, I want my bank to use a secure data center. Even my email provider. Does the CPU which renders my AI request for a cute kitten meme need to be in the same high-security DC location? What about the servers hosting the latest on-line games or helping write my resume? The value of those data assets does not warrant the expenses currently being incurred to have them hosted in the same physical DC as high security assets.

Assuming there is job tasking which breaks out secure workloads for different DC's, I don't see this as being a hacker's opportunity at all. It actually looks like a brilliant opportunity to reduce data center costs for mundane low security workloads.

[mirachu]

Re: A hacker's dream

Tons of BOINC!

[doublelayer]

Re: A hacker's dream

Not so useful, especially as the hacker would have no guarantee that the people buying compute from that box are doing anything they care about. Network traffic into the box can be encrypted using keys that get set before it is installed. To mess with the computers directly, they have to open up the box full of oil and start pulling modules off which sounds like a messy, easily botched, and relatively detectable action.

[Tron]

It's going to have to be a slot in replacement for a boiler.

In terms of size and power. It's going to have to cope with the heatwaves we get, and not require any new pipe runs (which is why we are keeping our 25 year old boiler - builders regs would require new pipes for a new one, which would mean ripping out a fitted kitchen).

The problem with small alternative energy projects in the UK, is that the companies often go bust, leaving the homeowner screwed. They bag a load of VC, crack on, and then the govt. change the rules/subsidies, and they just fade away. In this case, that would leave you with an oily Pi cemetery that your boiler guy wouldn't touch.

[Vikingforties]

Re: It's going to have to be a slot in replacement for a boiler.

"oily Pi cemetery"

Just for the pedantry, surely it would be a Pi mausoleum, since it's above ground?

[goblinski]

Our oldest brewery in the old country had a loop going under the street to the buildings across the brewery, providing them with free heat and hot water. Apparently it didn't make much of a difference.

Then Interbrew bought it out and cut that loop quite quickly. Apparently it didn't make much of a difference but it was still difference enough.

[A-nonCoward]

What about summer?

uh?

OK, simple, we add a cooling unit, right?

Or maybe we don't use this computing power when it's hot outside?

Meseems like a case of chasing the silly grant... Not a real anything. Bah, wasting our time, let me see what else is around ElReg

[abufrejoval]

Most stupid use of RPs from A to Z

One of the reasons I like The Register so much is that common sense still seems to be a thing there.

But with articles like this one, I wonder...

The PIs are terrible in terms of computing efficiency, units of compute per Watt at peak, but also rather wasteful on anything down to idle.

Sure an ECL VAX or Cray-1 might be still worse (and slower), but the main design goal of PIs is putting a reasonable amount of computing power into some spot for a minimal price.

Once you start aggregating them, e.g. to scale out computing workloads, their inefficiency and the overhead of any interconnect very quickly kills their economy and you'd be better off putting those workloads into containers or even VMs running on something both more powerful and more efficient per Watt.

I can't think of a single application, where 500 PIs in a single place might be able to do a useful workload, that couldn't be done better by just about anything else built from PC hardware, up to a single server, or perhaps three, if you want some fault resilience as well.

Sure, you might not be able to heat a home with those, but that's sort of the point: that heat ist pure waste!

And unfortunately neither the PIs nor the potentially hazardous cooling liquid just decompose peacefully at the end of their life cycle, the cost of disposing of all this might ruin whoever that job will fall to, after the provider has left the scene (likely insolvent).

Putting datacenter heat output to secondary use may be a good idea when carefully planned for the full life cycle of both the producing and the consuming side.

But datacenters and housing rarely align in terms of life cycle times.

[Fido]

Re: Most stupid use of RPs from A to Z

I had a friend who for a few years heated his home doing contract work involving image processing of flyover data using a cluster of x86 servers. It was effective because the heat and computing were both needed in the same place.

I'm a Raspberry Pi enthusiast since the first model was released more than 10 years ago. I think the questions are interesting what sort of cloud computing could be done with a cluster of 500 compute modules, how is the network fabric is arranged inside the boiler, what are the failure modes, redundancy and what kind of maintenance is possible.

As an experiment I set up a tiny cluster of 12 Pi computers each with no local storage. They mount their filesystems over iSCSI from an x86 server. It works well, but I have trouble imagining that scaling to 500 devices would result in anything but reliability problems.

On the other hand, maybe some magic is possible with compute modules where one can use more of the PCIe on the SOC for high-speed communications than possible on the standalone Pi. From an engineering perspective it would be interesting to know how much local storage is available inside the boiler and what kind of bandwidth there is in and out of a single node.

[K555]

Re: Most stupid use of RPs from A to Z

Good points. This isn't intended as a counter argument, just a different angle on the efficiency side.

Whilst these might not be the best in terms of compute per watt, they're aiming to replace a boiler so part of their job is to produce enough heat to be useful. My gas boiler does exactly zero compute, so it's even less efficient than a dedicated PC.

[abufrejoval]

Re: Most stupid use of RPs from A to Z

I just can't even see this launching or lasting longer than a blink...

One underlying assumption is that the compute power from those 500 PIs can actually be sold in a meaningful manner: I very much doubt it can, because it's neither competive nor attractive on the ISP side of things for those "business workloads customers": PIs can't compete in absolute performance nor performance/Watt with current server kit.

Now, you might not care if Thermify can't sell the compute and UKPN doesn't get paid for the electricity, but just how long are they going to supply those 500 PIs in your house with power, when Thermify goes belly up?

That leaves you out in the cold because if electric heating was economically viable, you could have just gotten a heating coil installed: much cheaper and less e-waste than PIs.

[doublelayer]

Re: Most stupid use of RPs from A to Z

The alternative is using a computer that consumes the same amount of power, thus producing the same amount of heat, but gets more computing done using it. That might make for a more expensive box, but depending on which version of the module they're using, they're already spending $22.5k-$65k on Pis alone (not sure how much of a bulk discount you can get on those). If the computer was faster, they could make more money selling its performance to people who want compute, whereas tasks are going to have to be tailored for Pi performance to work on this. You can get either much more single-threaded performance from X64 or much more parallel performance from GPUs.

The price difference might not be that high after all. That lower bound of $22.5k is for 500 base modules with 2 GB ram and no storage. No storage could probably work with images pulled from something else at boot time, but that RAM is probably going to limit the tasks. True, in total there's a terabyte of RAM in that box, but it's too split up for anything memory hungry to use it that way and it will be partially taken up with 500 separate running kernels. They've probably opted for more RAM in their modules, increasing the price.

[Aaronage]

500 Pi units?

A couple old Intel MacBooks would do the trick

[lordminty]

Back to the Future. Again

Back in the middle to late 1980s our office building was heated by our water cooled IBM 3081 mainframe.

Like lots of things in IT, we are going around in circles and reinventing the wheel!

[PRR]

I am sure the last time ElReg published this idea (good journalism even if poor thermodynamics), I did the numbers and analysis for MY house in the Maine woods. Firstly my maximum electric load is small compared to my typical need for heat, so I'd need new wires, and I'm a long way off the road/mainline. Next, presumably these servers need repair or upgrade, which will seriously annoy my dog-pack. I also have poor data connectivity, but it has got better, and things like more digits of pi don't need heavy data. Last time I thot the company paid the whole electric bill and now they want to be paid for the box and part of the electric. So like many "savings" plans it doesn't age well.

[Filippo]

The last time I could be bothered to run some back-of-the-envelope calculations on this idea, it turned out that you would almost always be better off using a heat pump to heat your house, and a data center to process your data. The totals would come up better, both for cost of heating and cost of computing.

Things might have changed, but I wouldn't be sure how. Data processing has massive economy of scale benefits, and heat pump positive coefficients are also very large. The only advantage for the box-o'-pies here is that it generates the heat right where it's useful, but I'm skeptical on whether that's enough to win out.

[David Newall]

who pays

for the electricity?

[MachDiamond]

Too much load

It sounds like a novel idea on the surface, but I would suppose it will depend on how the power company has calculated loads on a residential distribution grid. I don't think they have figured on having too many people drawing close to maximum service current very often.

[anonymous boring coward]

If electricity was cheap I would consider it.

Sadly electricity is several times more expensive than gas per energy unit in this (UK) country.

[werdsmith]

If you are not planning on building a new home then there is no question that this would be a good choice. These are about new builds where a standalone gas boiler is no longer an option due to Part L.

[Fursty Ferret]

This is a terrible idea in every way imaginable

I mean, where do you begin?

1. The heat output is completely dependent on the compute load. When will this be highest? Likely when electricity prices are cheapest, so overnight. Exactly when you don't want your house that warm.

2. Unless they're maintaining the entire unit at 60C or more, you're not going to get meaningful hot water without a huge heat exchanger inside, and being oil you have a limited thermal mass compared to water.

3. For heating a home you'd be much better off using a heat pump with 400-500% efficiency than waste heat from a bunch of Raspberry Pis.

4. What happens in a fault? You're reliant on the provider of the unit for service.

5. What happens in summer months? You're going to have a very warm house, or a huge fan blasting heat into the garden. You can't just turn it off, you need hot water.

6. Theoretical security issues from data off-site.

7. Synchronising compute load and power use with our ropey smart meter network will make billing a nightmare.

8. Do you have to give the provider 24/7 access to your home for troubleshooting?

9. What if there's a leak from your heating supply into the unit? Who's responsible for repair / replacement?

Honestly I can't think of a single pro.

[Auntie Dix]

Back in the 80s, a co-worker had a similar story about where he used to work.

He previously worked in the office in the corner of a large, unheated warehouse. The building had two electric meters, one for operating the warehouse and another for providing heat for the offices. The office heat was charged at a higher rate, but that meter never showed usage during winter.

The electric company came back several times to figure out why. They replaced the meter multiple times and then they carefully examined the wiring, but they didn't see a fault. It turned out the office was heated by the mainframe which was plugged into the cheaper electric meter.