Raspberry Pi 5: Hot takes and cooler mistakes

We've used a pair of Raspberry Pi 5s as daily drivers and have some hot takes. Some really hot takes. Why two devices? The first exhibited some mysterious lock-ups as it heated up – we ran it without any cooling solution, so you don't have to – and it was subsequently returned to base for investigation. The second has …

[Roj Blake]

Thermodynamics

"With the approach of winter in the UK, running a Pi 5 without a cooler might have a certain... appeal for those with central heating issues at home"

A cooler removes heat from inside the Pi to just outside the Pi. I don't see how this is a problem if your home is a little chilly.

If anything, running a cooler would help with the heating at home, as it uses electricity.

[Flocke Kroes]

Re: Thermodynamics

A cooler on a Pi can increase the heat output. Without one, the CPU could throttle and cause it to use less power. It depends on whether the work load is a fixed size (re-coding one video uses less power for more time with throttling) or infinite (bitcon mining).

[42656e4d203239]

Re: Thermodynamics

>>A cooler on a Pi can increase the heat output.

Surely the chip will largely generate the same quantity of heat, for the same task, regardless of clock speed?

Adding a cooler doesn't change the work required, it changes the rate at which waste heat can be moved out of the CPU.

/mines the one with "Thermodynamics for beginners" in the pocket

[ajekb78]

Re: Thermodynamics

Heat output refers to the rate of production of heat, not the total. That's consistent across CPUs and central heating systems. For a given task the total quantity of heat will be more-or-less the same regardless of clock speed, but if the chip hits its TDP then it throttles to reduce its thermal output, so the wattage going into the atmosphere in the cold house will be higher and thus the net heat outflow from the house will be reduced. (Plus, adding a fan-operated cooler will also increase the total thermal output by the power consumed by the fan which, whether as direct waste heat from motor inefficiencies or indirectly through the inevitable conversion of energy from kinetic energy of the air molecules it pushes around to heat.

It's not going to be enough to keep anyone warm though, you still need a GPU for that XD

[robinsonb5]

Re: Thermodynamics

> Surely the chip will largely generate the same quantity of heat, for the same task, regardless of clock speed?

For the same task, yes - but if it can complete the task sooner and start working on another task in the same time frame, then the total heat output during that time frame will be higher.

[TDog]

Re: Thermodynamics

Worked in an office in the early 90's in mid winter. CFO was an idiot who wanted to save money by cutting heating bills. I didn't, but some one genius wrote a script that ran infinite calculations on the PC's. As the CFO was not replacing staff whilst retaining the workload we were not his friends. So all the spare PC's were running at 100% cpu. It was between 1KW and 1/2 KW per PC. Not fan heaters deliberately but they all had "cooling" (heat redistribution fans). It was quite cosy. Still left.

[carl0s]

Re: Thermodynamics

Agent you assuming an infinite task queue vs the usual 96% idle time?

[robinsonb5]

Re: Thermodynamics

Well yes, I'm assuming the CPU is busy enough that throttling is a likely scenario. (And by "task" I simply mean "chunk of work" rather than a thread or process.)

[Nick Ryan]

Re: Thermodynamics

Largely, although I would guess that a CPU running faster, particularly with voltage control, will be running less efficiently with more electrons going to generate heat than switching circuits therefore running the CPU at a lower speed should, overall, use less energy for a given task.

Probably not a lot in it for a fixed voltage CPU but I suspect more identifiable for a variable voltage CPU, probably depends on the task and so on...

[katrinab]

Re: Thermodynamics

No, because if you double the clock speed, I think you more than double the power consumption?

[robinsonb5]

Re: Thermodynamics

Interesting question - is that actually true if all else remains constant, or does it only apply if the core voltage has to raise to achieve the faster clock speed?

[doublelayer]

Re: Thermodynamics

"Surely the chip will largely generate the same quantity of heat, for the same task, regardless of clock speed?"

As others have said, that's thermal energy (joules), not heat output (watts). How hot you tend to find things depends a lot more on watts. Also, you're not necessarily correct about the clock speed, because the heat output depends on the efficiency of the chip. If it uses a lot more power for a little more performance, that makes it less efficient and therefore generates more joules and watts. If you clocked it down, it would take longer to run a task but be more efficient, which could mean the same amount of joules, but not necessarily, and especially not for a Pi. For the typical CPU of today, this is closer to the truth, because when they are not running they put themselves into a sleep mode which reduces the power usage significantly, although not anywhere near zero. The Pi, at least all previous ones, have had a minimum level of power consumption to stay online. If you compute with that power level or leave it idle, it produces the same amount of thermal energy. This makes a task that can run at the idle clock rate of the chip as effectively free from a thermal energy perspective since the chip would have produced the same amount anyway. Keeping it at that underclocked frequency is more difficult and usually unnecessary, though.

[Charlie Clark]

Re: Thermodynamics

I think the key measure is whether the energy management system kicks into throttle the CPU if it gets too hot.

As for heating the local environment: electricity is usally the most expensive way to heat and this only makes sense when it's cold. So best avoided where possible.

[JulieM]

Re: Thermodynamics

What we have here is a perfect example of the n+1th Law of Thermodynamics:

For every simple energy calculation, there will be another scientist along shortly to explain why one of the simplifying assumptions is invalid.

[FIA]

Re: Thermodynamics

If only you'd changed your handle to 'JoulieM' that comment would be just about perfect. Have an upvote.

[deadlockvictim]

Names for childer

You mean, in the same way that chemists should give their daughters names like 'Esther' & 'Heidi' (short for 'Aldeheid'), physicists should name their childer 'Joules' or 'Joulie' (as mentioned above)?

[TRT]

Re: Names for childer

I always thought the Emperor of Technicians, Bryson Gore, should name his daughter Kensington.

[RobDog]

Re: Names for childer

Yep, just like Dr Emmet Brown named his kids Hules and Verne

[MrDamage]

Re: Thermodynamics

Oversize wings on a heatsink, plonk the Pi5 under your desk, and you have a heated foot rest.

[Tom 7]

Re: Thermodynamics

You can make a copper heat sink that lives on the desk and the hot air is fed to a chimbley which provides some draw. But there will always be a bit of lag in the requirements of chucking heat out of the cpu /gpu unless load is continuous.

[TRT]

Re: Thermodynamics

Didn't Apple try this once?

[Roland6]

Re: Thermodynamics

Surely, just swap that plastic case for a case iron or aluminium one.

Perhaps a good cabinet design for the home datacentre cabinet is something along the lines of this: https://www.stovesareus.co.uk/esse-ironheart-eco-multifuel-wood-burning-range-cooker.html

[Stoneshop]

Re: Thermodynamics

Argon should start making a version of their One for the Pi5: metal, slightly sloped, all connections at the rear.

One under each foot looks ideal to me.

[ecofeco]

Re: Thermodynamics

Uhm, just pop the cover off?

[Stuart Castle]

Re: Thermodynamics

During lockdown, I had to start working from home. Thankfully, I'd just rebuilt my PC, and had installed a new M/B, new CPU and more memory. I went with stock cooling.

The PC was in our dining room, and for the first couple of months, I was OK using this as an office. Then, we needed the room back, so I moved my "office" to my bedroom. I noticed the PC was running a little wam, and potentially throttling under load, so I bought an all in one liquid cooler.

Even under heavy load, my PC never throttles now, but as soon as the PC is under heavy load, the room heats quickly, so I spend most of the summer sweating.

[Neil Barnes]

Got to be an improvement on my company Lenovo

Which is loaded with so much company crapware and spyware and monitoring and such that it spins up the fan - sounding a bit like a 747 spooling up - before the boot is complete.

[Anonymous Coward]

Re: Got to be an improvement on my company Lenovo

I had a couple of Dell servers that definitely sounded like engines spooling up at boot. Until the OS took over from the BIOS and set them to a sensible level

[andy gibson]

Re: Got to be an improvement on my company Lenovo

They certainly blow up plenty of dust!

[A Non e-mouse]

Re: Got to be an improvement on my company Lenovo

Many devices do something similar. At initial power on, fans run a full throttle. Once either the firmware or an OS is in a sane state, they take a more intelligent control of the fans.

[TimMaher]

Re: Many devices.

Yeah. Try a Drobo 800i.

[My-Handle]

Re: Got to be an improvement on my company Lenovo

I ran an HP Proliant server as a "desktop" machine for a little while. That did the same. It was a shouty beast at the best of times, but on boot it sounded like it was about to hit the ceiling.

[Anonymous Coward]

Re: Got to be an improvement on my company Lenovo

One of my ProLiants used to detect any non-HP HW and then always run the fans at 100%. The internal "justification" was that they didn't know the thermal characteristics on non-HP parts so they had to make a lot of noise just to be on the safe side. A new FW fixed this. I guess some significant customer (not me) told them to sick their heads up a ducks bottom.

[Anonymous Coward]

Re: Got to be an improvement on my company Lenovo

The DL380 (2U) are less noisy than the 360 (1U).

Also depends on the installed hardware, mine got much less noisy when I removed an unused PCI-E SmartArray.

[collinsl]

Re: Got to be an improvement on my company Lenovo

Well of course, the 2U fans are twice as tall so they have to spin half as fast to move the same airflow.

Those tiny 1U fans run at incredibly high speeds in order to get the required airflow through the server, hence ending up sounding like a drone taking off in a heavy downdraft

[Alan Brown]

Re: Got to be an improvement on my company Lenovo

pi(r^2)

If they're twice the diameter it's 4 times the swept area and significantly less than half the rpm required

There's a lot to be said for centrifigual fans in 1U boxes

[My-Handle]

Re: Got to be an improvement on my company Lenovo

It was a gen 7 DL385 (AMD Opteron version of the 380). If it was less noisy than the 1U units, I don't think I'd want to be in the same house as them :)

[Anonymous Coward]

Re: Got to be an improvement on my company Lenovo

Except my rack. I use a golf cart PWM controller to run 8 fans...four on the front, four on the back. At 40% those fuckers screeeeeeeeam.

When I open the door to my server room I hear a reassuring "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAHHHHH!" and I get gently pulled in with the pressure. Hell yeah.

Your racks are 80% air...gotta move that air like a room full of ravers on ecstasy.

Even with ear protection in, when I leave that room, I feel like I've left a KoRn concert at 2am.

[scaley]

Crimbo

Wife asked what I wanted for Christmas... Ordered from PiHut and hopefully arrive any day now

She called me a sad B*****d :-)

[Anonymous Coward]

Re: Crimbo

Mine said "why do you need another one?" She's probably right in the short term though, but when the ancient PC I'm using to run Home Assistant eventually dies, I'll get whatever the latest Pi is at that point. I've already migrated from a Pi 3A+ to the PC to make it a bit more snappy, so going back again should be no trouble. (The migration was surprisingly easy, keeping z-wave and zigbee network configurations with no problems.)

[Lon24]

Overclocked & Overstocked

The really annoying thing about RaspberryPis is that they don't die. I have a drawer of assorted RPi1 & RPi2 in perfect condition gathering dust. Their tasks now handled by RPi3 &RPi4 most running 24/7/365 without cooling in open racks. So trying to justify a RPi5 means I need a bigger drawer for the displaced kit. Trying to get that past my senior management team is going to be challenging.

Throwing them in the bin would be sacrilege. Putting them on eBay is a bit of a faff. But I really want a new one, or maybe two ;-)

[A.P. Veening]

Re: Overclocked & Overstocked

Make that three ;)

[spireite]

Re: Crimbo

To be honest, I don't need to buy anything new in tech for that to be said.

My wife uses that description with no purchase necessary

[GlenP]

Re: Crimbo

As I don't have a wife I had to order my own Christmas present (and call myself a sad B*****d!)

[Atomic Duetto]

Re: Crimbo

Dude,

We’re here for you … Sad Bastard

(No wife either, but survived a Brazilian ex… so far)

[quxinot]

Re: Crimbo

Wow. How many is a Brazilian?

[Number6]

Re: Crimbo

If done properly, pretty much zero after the initial scream.

[ecofeco]

Re: Crimbo

I am going to hell for laughing at this.

[Dagg]

Re: Crimbo

No, but you owe me a new keyboard...

[Greybearded old scrote]

So what was the passive solution?

With my Pi 3 some heat sinks small enough to fit in the official case didn't achieve much. Switching to an all aluminium case took 10 C off the running temperature. Would one of those be good enough?

[m4r35n357]

Re: So what was the passive solution?

Yep, the relative silence on passive cooling is very noticeable by now . . . no sign yet of aluminium cases.

c.f. "Raspberry Pi 5 has been designed to handle typical client workloads, uncased, with no active cooling" - https://www.raspberrypi.com/news/introducing-raspberry-pi-5/

[Ken G]

Re: So what was the passive solution?

Yes, I've seen some fairly clever variations on heat sinks for the Pi4, the universal comment being that you pay extra for the quality of the thermal tape/paste supplied, not for the design of the sink.

[Dan 55]

Re: So what was the passive solution?

You really need a heat sink stuck on top of the CPU, at least.

[Greybearded old scrote]

Re: So what was the passive solution?

Yes. I have the RP3 version of the Flirc Case, and it sticks to the chip with a standard thermal pad. Others are available.

[werdsmith]

Re: So what was the passive solution?

I've been using Flircs on the 4 and found it to be very effective for my workloads.

The Pi 5 version is already on their website for shipping early December.

[Anonymous Coward]

I'm going to create a novelty cooling system for the new Pi...

It'll be based on the design of that famous statue of the little boy peeing- in this case, onto the overheating CPU.

I'm calling it...The Manneken Pi5.

[Anonymous Coward]

Re: I'm going to create a novelty cooling system for the new Pi...

...and if you use actual piss you won't need one of those expensive Adafruit LED strips...it'll create it's own rainbow.

[Anonymous Coward]

Re: I'm going to create a novelty cooling system for the new Pi...

I don't want to ask how you know that…!

[JulieM]

Re: I'm going to create a novelty cooling system for the new Pi...

Any clear liquid with a refractive index in the right range will cast a rainbow. There is nothing special about urine.

[My other car WAS an IAV Stryker]

El Reg: "We're special like that."

We know, and we love you for it. That's exactly why we -- at least some of us -- keep coming back (and some of us might also be "special"; just ask our families/coworkers).

[nautica]

"...let me count the ways..."

"...Raspberry Pi CEO Eben Upton told The Register: "This is the first time, across our internal testing and a pretty broad beta, that we've seen someone lock up a unit in this way." "

Eben Upton has a very long history of double-talk, dissembling, and equivocation when it comes to trying to explain away any of the many design faults with the Raspberry Pi...all of which were the results of lack of planning and/or poor, or no, engineering design expertise on the part of the Raspberry Pi organization.

[Eponymous Bastard]

Re: "...let me count the ways..."

All the skills required to be a politician then.

I think double-talk, dissembling, and equivocation is rather harsh.

[Anonymous Coward]

Re: "...let me count the ways..."

...but if they perfectly executed everything and could implement everything flawlessly the device would cost a lot more money and you'd be complaining that it costs too much.

I don't think the Raspberry Pi foundation has ever stated that producing the best SBC on Earth, period, is one of their goals...they've always been straight about building the best they can to a given price point and to keep that price point as low as possible.

I don't think anyone is under the allusion that the Raspberry Pi is anything else...except maybe you.

Personally, I don't use Raspberry Pi boards because they are the best. I could double my budget and get a much better board...or I could go the other way and halve my budget for something that will do the job that isn't as good.

What Raspberry Pi represents is the "sane" middle ground. Does the job, not too pricey, decent support. If you go either side of the Pi you get a great board, expensive price, shit support or crap board, great price, shit support...the best place to be is decent board, reasonable price, great support.

[Ozzard]

Re: "...let me count the ways..."

Let's just say these boards aren't going out into the field yet, where we expect them to survive a decade or more with no visits and total data traffic of under a megabyte a month. Six months or a year from now, once the firmware starts to settle down? Sure. They're good enough for what we need, and have an excellent ecosystem around them. That ecosystem is the difference for us.

[werdsmith]

Re: "...let me count the ways..."

I don't think the Raspberry Pi foundation has ever stated that producing the best SBC on Earth, period, is one of their goals...they've always been straight about building the best they can to a given price point and to keep that price point as low as possible.

I don't think the Raspberry Pi foundation has had anything to do with making SBCs for many years.

[carl0s]

Must admit the fans do worry me.

My environment is dusty (tissue factory). Plus fans fail even when there isn't tissue dust everywhere to block everything.

We've used Rk3399-T based SBCs due to unavailability of Raspberries. I'm at about a 10% hardware failure rate over the last 18 months though and might try some PI 5s. I think 4 of 35 have failed and thow up uBoot errors mentioning ddr controller stuff that I don't understand

I've been using a large ish heatsink on the rock pi's (rock 4+ / plus). Same heatsink fits a pi 4. Not sure about 5. It's 40x40x20mm and I have to cut a couple of notches into the corners

The rk3399s do get warm but seem to be ok. The heatsink helps even inside a warm industrial machine with no fan.

Something fishy going on with boards dying every now and then though, suspect overvolting ram or something. No response on radxa forums so a return to Raspberry would be comforting.

[David 132]

Something fishy going on with boards dying every now and then though, suspect overvolting ram or something

Just spitballing here, but have you tested the power supplies on the failing units to see if their output voltage is within spec?

[carl0s]

I haven't, but they are official Raspberry Pi 4 USB-C adapters, 3 amp.. RS were selling them for peanuts (£4.88 + vat) at one point. RS Stock No.:

187-3416

[Anonymous Coward]

@ "Must admit the fans do worry me."

Yes, I know what you mean about the official RPI forums.

Rabid zealots that L Ron would be proud of

[werdsmith]

Re: @ "Must admit the fans do worry me."

Yes, Jeff Albertson style sarcastic responses, curtness and downright rudeness is the way to treat genuine innocent new entrants to the official RPi forums.

[Anonymous Coward]

Re: @ "Must admit the fans do worry me."

The forums are reactively moderated - if you have a problem with a post, report it and the mods will deal with it.

[werdsmith]

Re: @ "Must admit the fans do worry me."

Yeah that’ll make a difference…..

[anonymous boring coward]

I was going to explain that a cooler would actually increase the total heat output of the RPi, but others have already done that.

[Bitbeisser]

When I saw the announcement of the Pi5 (I think it was some quotes from Upton) is said that the Pi5 DOES need a cooler, that it is no longer optional....