89
points
One would assume after 10 Gen chipsets, and 10nm or shorter designs large heat sinks would suffice. But most of fanless devices are Celeron/Pentium-M chipsets in Chromebooks or Windows laptops with 4/8GB RAM and eMMC. Is it impossible to passively cool i5, i7 (integrated graphics)? Due to OS needs I cannot go for Apple M1.
If we move for a minute beyond the direct "hard" technical constraints which have already been expressed here (i.e. small chassis, lots of chips etc.).
You have a second aspect, the old PBCAK (Problem Between Chair And Keyboard) one.
Users will use their laptops in less than optimal (from a thermal perspective) ways. They will use it on their laps. They will use it in bed, resting on duvets. They will use it on pool-side sun loungers with the mid-summer sun blasting down. Yes you can tell them in the manual that they shouldn't, but they still will.
So even if you can magically cool the chips passively, the possible use-cases of the product might ultimately constrain your ability to forgo a fan entirely.
I note that some people here are seeking to blame the "race to thin" for the requirement for fans. To them I would merely point them at the Panasonic Toughbook. Modern laptop but built like an old-shool 90's brick ... it still has a fan.[1]
Apple have done a stunning job with the M1 (and previously with the i5/i7). But even with the unibody chassis of a MacBook (i.e. the whole device is a massive heatsink) the necessity of forced cooling is still there (edit to add: with the exception of the MacBook Air M1 ... thanks for the correction guys !) . But with the non-Air M1s you really have to work super hard to make the fans ramp up at all, they've done a stunning job.
[1] https://www.youtube.com/watch?v=vgWZwP28trI
The MacBook Air M1 is passively cooled. I used an M1 Air for a while (switched to M1 Pro) and it would only throttle slightly during very long builds. There have also been many tests where the MacBook Pro M1 (which does have a fan), the fan only spins up after minutes of loading all cores.
The MacBook Air M1 shows that it is possible to have a fast [1], passively cooled laptop.
[1] Beat my 3700X in most Rust/C/C++ project builds.
I remember my first question on Quora was whether there was a way to sidestep the (then) 30MB size limit for installing apps on the iPad while on cellular. A guy with several 'signal engineer' tags explained that that would completely overwhelm the mobile network. I responded that the SIM card I was using was actually marketed to be used as your main internet for your home and came with a Wifi router for that purpose (which I also used sometimes, transferring the SIM). His only response was that I obviously lacked the mental capacity to apprehend how physically impossible this was (something I was using regularly mind you, for much larger downloads and data-intensive apps), and he reported my profile.
I can get the fans to spin up easily on my MacBook Pro 14" (M1 Pro), e.g. when building a large project like PyTorch. Luckily, the fans are not very loud. I can't hear it at the office (plenty of ambient noise), though I can hear it sometimes in our home office.
Passive cooling is one of the reasons I recommend people to get the M1 Air, unless they need 8 performance cores or > 16GB RAM. A completely quiet laptop is such a nice feature. I hope that they keep the Air passively cooled in the future.
Edit: >= corrected to >
My M1 Air has 16GB of RAM.
This used to be how I found runaway processes on my previous Intel based MBP.
Thanks for the correction
My mind was still on the old pre-M1 Air which did have a fan IIRC.
Thanks to you my credit card is trying to tempt me to pull it out of my wallet. ;-)
And the fan on the Intel Air was loud and the CPU performance not great :(. The Intel Air and the M1 Air are like night and day. Even though I have the 14" Pro now, IMO the MacBook Air M1 is the most revolutionary Mac of the past decade, if not more.
I believe the entry-level MBA is the best laptop you can buy when it comes to price-performance ratio.
I would not be surprised if Apple eventually brings out a new laptop that addresses the ultra-portable market. With new screens it might be a 13” screen in a tight & light case.
I had a Docker build that would take 4 hours on mine when it completed in under 15 minutes on my Mac Mini.
But it's not just laptops that lose their fanlessness (which I appreciate both due to noise and reliability concerns), but also NUCs - almost every single NUC-like device that has entered the market and managed to come to my attention over the last months has had a fan installed. I wonder why that is - TDPs do not seem to have gone up, at least not when looking at data sheets... maybe it's just cheaper to make stuff that way.
This is also why it's generally recommended to go with a classic air cooler that just has the fan speed set to the lowest level over passive air coolers in desktop PC-s. It's still quiet, but you get the cooling performance while using much fewer resources.
Data sheet TDPs seem to have less connection to reality with each generation.
But with all of the boost clocks and what not, the more heat you can disipate, the faster the chip will go, and the power usage/temperatures can ramp really quickly too. Fans are so effective in most situations, that it's hard to leave them off and accept the performance difference. Personally, I'm ok with most fans, although I had to replace the PSU fan in my most recent build cause the included one had a noise profile I couldn't stand. For things like NUCs and Desktops, running longer cables and hiding the noise source is probably a good option too.
- Copper is one of the most expensive materials, and it is hard enough, and have high melting temperature, so manufacturing also more expensive than from plastics.
When micro computers was inside 15W TDP this was not a problem, as heatsink was tiny, but for modern commodity designs, 25W+, need to use all surface of notebook as heatsink (at least one side of "shell"), this would be very expensive.
It's not perfect (headphone jack doesn't work), but been really happy so far. Probably the best $ per performance laptop I've ever purchased.
Edit: Just did a search of junction temps for 10th, 11th, and 12th gen i5s... all 100C.
Tablets mostly spend their energy budget (under normal operation) on illuminating the screen and most of the action content they see would either classify as webpages or video encoded with codecs they have hardware support for decoding efficiently.
A generic PC meanwhile runs a heavy OS and is often asked to do large volumes of local processing tasks. E.G. that bloody 16GB Outlook mail database that has to be stored locally because your power user that demanded a laptop has to be able to read all of their past email while they're in the field with no reliable net connection.
Oh, they also demand that it has to be light. Fans are cheaper and lighter than putting enough metal inside to do something passively; and also run cooler since even a pathetic volume of airflow is many times better than waiting for convection alone to move the heat.
You can blame the noise of the fan on the push for crazy flat laptops though. There's no reason a still light but thicker laptop couldn't use a larger slower moving fan. Well, other than the perception of thick as heavy and thus bad.
FWIW I've taken apart many laptops over many generations and the general trend seems to be that OEMs always use a cooling solution that's just-so good enough, regardless of available space and laptop cost. When you give an OEM a 40 % reduction in power through a new CPU generation, you're giving them a choice: a) spend the same amount of money on a cooling solution that's as capable as before, so now you'll get a quiet and cool laptop or b) make the cooling solution 40 % smaller, making it cheaper, and keeping noise and heat roughly similar.
It's the same with workstations. Although they have plenty of room, but OEMs often install barely-capable CPU coolers that are noisy (rather than something like a large Noctua cooler that can keep most CPUs cool with barely any noise). Or another big annoyance: many OEMs (even reputable brands like Dell) install a PSU that is just enough for the load. If you want to replace the GPU by something that requires a bit more power, you'll end up replacing the PSU as well (assuming that you can get one that is compatible with their custom mainboards).
I'm not sure what the current technology is like as I haven't built a PC in a long time, but it used to be the case that efficiency would drop off quite rapidly if you underpower it. e.g. your fancy 600W 95% efficiency PSU would only be 60% efficiency at 200W power draw.
> OEMs often install barely-capable CPU coolers that are noisy (rather than something like a large Noctua cooler that can keep most CPUs cool with barely any noise).
My old AMD Athlon mainboard came with a very rattly and noise cooler for the chipset; I ended up replacing it with an after-market heat-sink that I got for €10. Pretty bad design IMO, but it saved ~€3 on the mainboard price.
A lot of the old Atom CPUs (Atom 330 etc.) had fanless CPUs, but because they were almost always paired with an old inefficient chipset they all came with a little cooler on the chipset. Ended up replacing a few of those, too.
This. I have got an old HP EliteDesk from work. It uses a standard SFF (not sure if it's the right term) PSU, but the motherboard connector is custom. Of course the fan makes a terrible noise. And even if I'm able to find a replacement PSU that would fit in the case (even fanless! - I'm happy with the integrated graphics), I also have to look for an MB connector adaptor.
[0] https://github.com/n4ru/FakeSlice
With a deep-travel backlit keyboard, FreeDOS, WordPerfect 5.1, DataEase 4.53, SuperCalc 5, and JPI TopSpeed Modula-2. (Oh, and Elite+, Prince of Persia, and the Infocom back catalogue for down-time).
Seriously. My dream machine.
So we don't know, how much this costs for them, and even, is M1 project profitable.
I bought my Asus UX305C about five years ago now. It's fanless, and the CPU is definitely a little slow, because it's an Intel atom. However, I wasn't buying it for the CPU - I was buying it for the fact that it was fanless, had a decent keyboard, hidpi screen, enough RAM, and sturdy metal case. I don't need a fast CPU, because everything I do is run on a chunky server somewhere else. It feels like an expensive laptop because of the metal case and nice screen.
I was so pleased with it that I recommended an Asus laptop to someone else, but unfortunately that one does have a fan, and the case is made of plastic, and it feels really cheap. The touchpad stops working properly when you hold the laptop case in a particular way, which bends the plastic and stops it being able to detect fingers.
[1] https://www.bestbuy.com/site/asus-17-3-chromebook-intel-cele...
I think it doesn't matter much, as you can take a standard laptop and restrict the clock speed, and the fan will never turn on.
And laptop marketing focuses primarily on CPU speed, which is an error. It drives running CPUs at speeds that create non-linear heat/power graphs without much consideration to total system performance.
Or do you mean restricting clock speed based on temperature?
I've been buying dozens of those for people at work since 5 years now (first ones with n5000) and I still have to hear a single complaint about it.
The only point of failure of that thing is the charging port, proprietary, not usb c.
Also to be noted : one of the lightest 14 inches on the market with an excellent battery life.
I don't have any first-hand experience with this (yet), mind you, and am unable to comment on other features :)
Now Apple’s M-series chips combine power with low power usage and they are able to build fast and fanless devices like the M1 MacBook Air.
Intel and AMD have been focused on building for power and have been, mostly, ignoring efficiency for too long. Hopefully, Apple’s newfound success in this area has waken them to the need to address power consumption.
My work laptop is one of those macbook air style super light Windows ones, I thought it was fanless until I watched a teardown, the fan hardly turns on and I have to put my ear to hit to hear when it does.
> Is it impossible to passively cool i5, i7 (integrated graphics)?
Obviously not, but same as with the M1 in the air and in the pro: fan=more power and heat available. Add the fact that in many devices you have to try really, really hard to be even able to hear the fans adding them is a no brainer.
The MacBook Air M1 does not have a fan, it's passively cooled. The fan in the MacBook Pro M1 (non-Pro/Max) and Mac Mini M1 spins up very rarely.
Add the fact that in many devices you have to try really, really hard to be even able to hear
Then you have a lot of ambient noise. Everything from Intel MacBook Airs, Intel MacBook Pro 16" to Lenovo laptops make quite a bit of noise when under load (where 'load' can also mean something simple like a video call). Once you are used to a quiet machine (like an M1 Mac), it can be very annoying.
Yeah, that is what I said. Not sure why you are repeating that.
> Once you are used to a quiet machine (like an M1 Mac),
One with a fan or without?
> it can be very annoying
Yup, but you can simply turn that off by adjusting the fan and power. Allowing less fan usage on the M1 Pro just pushes performance down to the M1 air and makes the fan inaudible or even just off. However, you cannot go the other way: the Air is not going to magically grow a fan.
So most companies do exactly what apple does with the pro: include a fan.
That statement can be interpreted to say that the Air and Pro do have fans. It’s ambiguous.
I think most surface pro lineups had the i5 being fanless and the i7 with a fan. And that was a problem, because although the performance difference between the i5 and the i7 wasn't that great (most would not pay for the difference) the cooling differences made it night and day. And as someone who loves passive cooling I was forced to recommend the i7 just because it had a fan. (note: in latest lineup of the surface pro the i5 also has a fan, making the i5 version much more attractive for most (though now it cost as much as the i7 used to do ...))
There is a huge difference in cooling between passive and a very very slight breeze, intel/amd CPUs are a bit too hungry for it to make much sense in the medium-to-high performance CPUs. Though I feel not enough attention is being put into cooling of laptops (passive nor active), isn't a strong selling point I guess (though it should!).
The letter indicate the target application, which then impacts the TDP Intel is targeting (slightly dependent on i3/i5/i7, but maybe not at all). U and Y are the big ones to look at for mobile fan vs fanless. Their definition is:
Here's the ARK link for 4 10th gen i7 with the above 3 letter (plus one desktop CPU): https://ark.intel.com/content/www/us/en/ark/compare.html?pro...TDP (max power draw) based on letter:
No it is not.
Why do they not sell one? Because passively cooled x86 CPU doesn't run every fast enough (yet) which leads to poor sales number.
May be a few more years. When both IPC and node improvement catch up.
other than that, everything else I find has weak Celerons in it.
I have a more powerful desktop for heavier tasks, and this fills my portable needs very well.
(Haven’t put linux on it yet, but maybe soon. WSL is filling that void for now).
For, I think, $799 with pen and keyboard its a great mobile machine and remote terminal. If I could go back in time I'd maybe spring for the 16/256gb upgrade, though that was more than double the price.
You could go "fanless" on many laptops, though. You'll have to check beforehand, but all you need is TDP adjustment (and ideally, undervolting) support and the ability to turn off the fan or adjust the speed table with something like NotebookFanControl.
For TDP/TPL/voltage adjustments, use Throttlestop on Windows; Linux also has some tools or you can write to MSRs directly after a long night of figuring them out. AMD mobile chips can not be undervolted as of right now, btw, a real shame. But neither can Intel with Plundervolt mitigations, so that's a double real shame.
A Core i7 can go up to ~20W without turning the fan on, and higher with the fan at minimum (I doubt you'll hear it). The critical temperature (=fan at full speed, overriding settings) is usually over 80 degrees Celsius, which is high enough that it may never get there.
Using better thermal paste would be highly desirable, too.
The major disadvantage is obviously performance - you'd get half of whatever the chip offers, at best.
More laptops should have fanless modes, where the processor is power limited & only passive cooling is used. I believe this would be possible on many laptops. Perhaps especially gaming laptops could potentially do very well only on their passive mode while not gaming, since they have such a large focus on cooling, but given the extreme flexibility in specifying power limits for chips that's possible these days, I feel like quite a wide range could have a pretty ok passive cooling performance.
It's be wonderful if there was a good standard review process to better explore laptop noise. That seemingly is the objection to fans, yes? Rather than totalize the ask, allowing different routes to success would imo be preferred.
I think that's it, x86 isn't even close at the moment.
You said "Hot and noisy", and making it fanless would only be even hotter or, have significant performance penalty (to a point that even light users won't tolerate.)
My (Windows) ultrabook bought a few years ago could be hot and noisy (and slow) when I'm merely using browsers nowadays, can't imagine what a fanless one would do.
A fan is a big and fragile physical component with a lot of associated cost – physical size, constraints on heatsink dimensions, noise, and power.
And indeed, customers' response to Apple's fanless laptops has shown a reasonably strong economic preference for them, or it seems that way right?
With the fan off it sits at 65°C vs 55°C fan on when browsing the web and programming. It only approaches 80°C (threshold at which the fans engage) when viewing Google Maps or gaming.
I suppose I could make it run even colder if I repasted it.
In any case now that Apple isn't using TSMC's entire 5nm capacity anymore we should expect nice, fanless laptops to become more common. Lenovo recently released the X13s, which has the Snapdragon 8cx and glorious 20h+ battery life. The M1 Air is still faster, but some of us refuse to use their hardware and this device fills this niche.
So what I'm hearing is that you don't want the devices that are already on the market, probably because you think they're underpowered.
If nobody is buying the existing devices, why would they make more varieties?
(I got one and Linuxed it. The wifi was so crappy I had to compile a driver from source, for an authentic early-2000s Linux experience. Need to test 22.04. https://reddragdiva.dreamwidth.org/608064.html Love it for writing on, not so great as an actual netbook.)
If you're not running a web browser with serious intent, it should be a perfectly capable little box.
My i9 mac just makes me angry, every time docker spins up. Stuck with it for a few more years as we start going into the office.
I do run Linux, and have a fairly light-weight environment (dwm etc.) I don't know how much that matters.
My opinion is intel/amd are not incentivised to make silent / fanless designs as they are hard to market as they are not going to be the fastest. Apple are incentivised as a silent cool running computer is something people will buy.
On the other end, it generally takes a couple of iterations to get a fan-less design right, because you can't simply slap an oversized heatsink into a mobile device.
See: https://www.ofzenandcomputing.com/best-fanless-laptops/
part of the date is rechecking assumptions as sometimes we do not re-check them ourselves before posting.
Looks pretty untrustworthy in general.
People preferring fanless/quiet devices are a minority and devices made for them won't sell enough to justify the design costs.
Also, latest Intel CPUs can eat up to 300W for some reason.
I was really disappointed with the fan behavior of my Thinkpad X1 Nano -- constantly spinning, and loudly. I use third-party software to control it now. Having the fan spin for writing an email is just ridiculous.
What program are you using for writing said emails? That matters more than what user action you're performing.
As a reference, using gmail in firefox makes my X1 very loud too, but I've moved to Thunderbird as of late and it's much more careful about resource usage.
This ignores how much more wasteful with resources developers are on the desktop, though.
Of course it matters what you do on a laptop when it comes to how the fans are reacting. Are you really expecting the fan to stay silent and the computer to stay cooled when compiling the Linux kernel for example?
Different workloads have impact of the thermals, it's as simple as that. Using heavy applications will make the laptop behave differently than if you use vim.
Because, to cool passively, they need huge heatsinks.
Real passive machines are in 15W limit, but commodity x86 are 25W and more.
Military notebooks are passive, but they used all their case as huge heatsink.
Why are you still wondering is the real question here.
Apple simply invested a huge amount of resources and had the benefit of optimising the SOC specifically for the OS.
The sheer amount of legacy in x86 is why those CPUs are so inefficient and require so much cooling. ARM would not have emerged if not for its superior efficiency.
The more power any electronic device uses, the more heat it produces.