Am I the only one who thinks of smartphones and tablets as silent computers?
Granted, they can't do what this person's high-spec workstation can do, but they do most of the computing tasks most people use (used) noisy fanned computers with clacking disks for and in many cases do those tasks better.
And unless I'm just losing my hearing, my smartphone is completely silent as long as I don't accidentally press the Golem Invoker, er, Siri button.
I was kind of excited about this build but then I saw your comment and remembered that was a much better option.
4-6 months ago I built a new workstation for work. I had used one of those Corsair closed loop water coolers with the prior build, so I set one up on this guy. A month or so later my workstation was running really sluggishly, and I realized it was drastically throttling the CPU because of heat. I installed some software to spin up the fans on the cooler to keep it down under 100C, but now it'll get kind of loud when I run much heavy CPU.
Now, this is a pretty heavy duty workstation, 64GB of RAM and 3 displays. But, if I were doing a new machine for home to be quiet, I think it'd be a NUC. Then I'll put the box that has the 6 drive ZFS array in a closet and call it good.
I had looked at the NUCs a few months ago, and just remembered how hard it is in their product page to find something that will support 3 monitors. Looks like only the highest end models will do. I'd be tempted to build my own again, especially for my home office where I can put it in a closet next to my desk. The NUC product page lets you select requirements and then seems to give you a list of products that don't match.
Gigabyte also has their BRIX products, which are similar to the NUC.
I was at a hotel this weekend and at check-in they had a monitor with a "ThinkVantage slotted in the back of the monitor, that might be a nice setup.
Ha, the entire team at my previous job were using NUCs. Nifty little buggers. In the summer it would heat up quite a lot (not yet to the point of damage I suppose) and I couldn't even turn it off without burning myself if the PC froze.
Eventually I just bought a cheap USB desktop fan and ran it facing the NUC.
I'm using a NUC for a quiet home PC that I do mostly browsing but also some light web development on. Although it's mostly silent, I haven't been too impressed with the performance of it. It was one of the top of the line models in 2015, an i7-5557U. I was surprised to find it's quite a bit slower than my 2013 Macbook Pro 2.4 Ghz i5. On the Geekbench profile the MBP got almost 3x the single-core score of the NUC and 1.7x for multi-core. Real world performance of the older macbook is noticably faster.
It has made me think that instead of getting a NUC, for a quiet desktop system I should have just gotten a second used MBP and ran it permanently docked in clamshell mode with the monitors and keyboard attached. (with the added benefit of being able to go portable when I want to)
Neither the NUC nor my MBP is completely silent but for my purposes I find that I seldom tax either of them enough to where the fans become audible enough to be annoying. Still, I do find the difference in performance between them to be apparent just in things like iteration time on web development and IDE responsiveness.
Is this only when it’s plugged in? Could be a bad cable or charger. I had an iPhone 5 that had a super touch digitizer when plugged in, and I finally narrowed it down to cheap third party cables. Otherwise I think I would be doing a sit-in at the nearest Apple Store until they replaced it, that would drive me nuts.
Heh - same sort of timeframe - probably a Nokia 8210 though - I could reliably have my Apple hockey puck mouse "crash" if my phone got an incoming call while it was sitting on my desk in a loop of the usb cable for the mouse. It'd just stop working, and need unplugging/replugging to get it working again.
It was clearly electromagnetically "noisy", but I do't recall ever having heard any on my phones make any unexpected audio noise... (My old-and-abused rock concert and motorcycle weary ears probably can't get up as high as inverter whine any more though...)
Pretty much all cellphones would do that to CTRs. They would also go directly into the audio circuits of cheap amplifiers, to the point where you could "hear" a text or call incoming before the phone made any kind of notification.
-I presume transmit power has been lowered significantly as coverage has improved, too; your cell always transmits at the lowest level it can get a reliable connection with to preserve battery life. This should reduce interference considerably.
Also, GSM phones used TDMA (keying the transmitter on and off to occupy one of -hm- eight, I believe - time slots on a given channel.)
This is practically asking for EMC issues.
LTE, on the other hand, transmits continously (I believe - I do not work in RF engineering anymore, but try to read up on new tech every now and then.:) - much less interference-causing than the constant on/off of TDMA.
I thought LTE worked on a timeslice schedule as well. I remember hearing that was one of the problems with carrier plans to start running LTE on unlicensed spectrum, because it doesn't play nice with listen-before-talk wifi.
I'd say a good half of the smartphones I've owned over the years have had audible something, even when not plugged in and supposedly silent.
Some were barely noticeable, while one in particular (a Droid Turbo) was so loud I could hear it getting ready to receive a call from another room. This was regardless of whether they were plugged in or not, although charger whine was its own separate issue.
Thankfully it does seem to be getting better over time- my current S8 is, as far as I can tell, genuinely silent.
This happened on TDMA systems like on T-Mobile and AT&T. (Verizon & Spring were CDMA).
There's a TDMA modulation frequency at 217Hz and this interferes with all sorts of nearby audio devices. CDMA and WCDMA phones have a much broader interference spectrum, which is why you don't hear it much anymore.
No, this is all the time. It is only quiet when the phone is idle.
I considered returning it, but I find it charming. I miss the days when you could tell exactly what your PC was doing by all the sounds it was making, and I find dead silent electronics to be elegant but a little sad.
The sound is loudest if I hold it up to my ear right behind where the SoC is. It generates a unique pattern of sound based on activity, such as running your finger over the touch screen. It's not just when sound events are about to play.
My Samsung Chromebook 3 gets a touch warm but never uncomfortably so like my 2012 Retina MacBook, which lets you really feel it when your code is inefficient. (Granted, the Chromebook is a lot less powerful)
Oh...it’s not just me then..
Every couple of months I check my 2015 MacBook Pro system info because I’m utterly convinced that it has an hard drive because of the SSD noise.
It’s quite frustrating actually..
Imagine being a kid in the 80s or 90s at school and hearing the distinctive 20k tone of a CRT Television humming and wondering if you were going to be watching TV in one of your classes that day. It was like a dog whistle for kids.
I still have a 19" CRT as 2nd monitor at work because... why not?? It still works most of the time and supports a decent resolution (1400x1050 - 1600x1200 flickers). And it's a nice nostalgic conversation piece :)
Back around 2001 I took my desktop build (Celerin 300A oveeclocked to 450MHz), installed a giant passive heat sink on the CPU and PSU, put in an 8MB IDE flash drive, and network booted off a server in my laundry room. I thought I was finally noiseless. But the end result was worse. I had coil whine out of my power supply any time CPU usage ramped up. If I was in the same room, I could tell whenever my email pinged the server or if a cron job ran. It was both instructional (why is my cpu ramping and I’m not even logged in?) and really annoying.
I’m routinely annoyed by what I am reasonably sure is the sound of my MacBook Pro’s heat pipes and other warm components expanding and contracting, making creaking noises shortly after starting it up on a winter day. Then there’s the gentle gaseous hissing I would swear is the heat pipes condensing and evaporating (based on workload and laptop temperature concurrent with the noise) if I could think of a practical reason for them to be audible.
When I go to editfight.com on my rMBP the fans go crazy and it gets super hot. When I visit it on my iPhone it stays the same temperature and is silent. Kind of an extreme example but the principle is the same. Phones are better for sites like this.
Oh I didn't know that. I thought Apple recently shifted the MacBook so that it was more "pro" and the MacBook Pro so that it was less "pro", making them a lot closer to each other, almost identical. I thought I remembered a lot of criticism over that move too, here on HN.
I bought an Asus "zenbook" ux305 for this reason. It uses an Intel core M processor, which idle around 800mHz but turbo boost to around 2gHz, which I believe have been discontinued.
I was worried about performance, but it has been very acceptable. It depends what you need it for, but I can run 2 monitors, a Linux VM and atom all while streaming HD video. Or I can do light web browsing for 10 hours on battery. I love it.
On the rare occasions I need more power, I spin up a spot instance.
 If you use Linux on a laptop, install "tlp". It optimizes battery life without a noticeable reduction of performance.
And yet, in a post referencing a fan-less "Macbook" it's almost certain that the "Macbook" in question is the only fan-less laptop Apple produces, which is coincidentally called simply "Macbook."
It's unfortunate that Apple has confusing brand names, but the fact remains that the Macbook indeed has no fans so the original comment who hears fan noise is obviously using a different model of laptop.
Of course there are Macbooks without fans, but the claim that "the macbook has no fans" is a bit too broad and generic to be true. Macbook Pros are also called Macbooks. Older Macbooks are still Macbooks (they seem to be pretty durable).
So if you say that recent Macbooks have no fans, that may well be true. But it's not true for all Macbooks.
I mean these days having a powerful computer on my desk is less of a benefit - that can be a machine in the closet or sitting on a rack. The machine on my desk just needs to drive a couple displays and run a browser.
I mean it depends on what you do - but for many people it's a realistic solution.
Mere physics says it won’t. It’s always possible to pack more performance in a larger package, even if it’s just because you can more easily dissipate heat on a larger surface.
iPhones are amazingly powerful and might just be sufficiently powerful for everyday computing soon or even right now, but they’ll never surpass anything that can accommodate a larger die.
Geekbench scores are interesting, but not necessarily translate straight into real-world performance. Intel CPUs have a much richer instruction set for example. Peak performance vs sustained performance is another issue. GPU performance, disk size and speed, available RAM, battery time at a certain usage etc. are other performance factors. Apple could decide to put faster CPUs into its macbooks at the cost of less battery time.
This is all alluded to in the article and the macrumors post that the article is based upon, here are some quotes:
> "Sure, that doesn’t mean the A11 Bionic can do all the things a desktop CPU does."
> "Though the iPhone X and the iPhone 8 offer impressive Geekbench scores, how that translates to real world performance remains to be seen."
There's no question that the iPhone chips deliver amazing performance, but there's a reason people still lug Macbooks around.
It does seem like something is missing in the comparison. High end x86 CPUs draw something like 30W idle, which would drain an iPhone X's battery in minutes. Do Apple/Arm really have some magic technology that makes their CPUs orders of magnitude more power efficient?
Geekbench always seems like an odd benchmark - the variability between runs alone is kind of odd. If I could run a compiler on an iPhone, for example, would I really see similar performance to my MBP?
One thing to note, x86 is known for being spectacularly inefficient for mobile workloads. It's not magic that makes an ARM CPU more efficient, it's just different design considerations.
For an example of this, check out the Intel Atom line of processors which were mostly x86 processors but designed to be mobile and power sipping. Whether they were successful at that, or in terms of performance, I'm not sure. But they get down to single digit TDP, which is how many watts of power you can expect one to use while under load.
Same, some feedback that something is happening is great. This is what annoys me the most about macbooks, they give 0 feedback when they're off. It's always a bit of a spincter squeezing moment on monday mornings when I open the macbook, sometimes it went out of power and shut down completely, but there's no indication that it's on or off or working or whatever. Black screen, and you kinda have to push the button for an arbitrary amount of time and wait for an arbitrary amount of time until there's any feedback.
Whereas on a normal computer, a light goes on and the fans start spinning, not because it's useful but just to indicate that it's on now. It's a miracle.
I have an OTG USB hub with Ethernet. Mouse, keyboard, and external drive can plug in all at once, and if I have a place to plug the other end of the cable I get wired network too. It even has power input to drive the peripherals. Add something like GNURoot with the Debian chroot and it's nearly desktop Linux on a tiny screen. It's more of a pain to set up during a meeting than a dedicated docking station would be. It also doesn't currently charge the phone (although there are ways to modify it to do that). It's nice for novelty but a little unwieldy. Using an OTG cable with a keyboard with a built-in hub reduces some of the cable clutter.
Some of the Chromebooks and such have no moving parts. I'm probably taking my Pinebook to the next conference I attend.
The latest MacBook is essentially mobile hardware with desktop peripherals and software. Seems like the right way to go rather than bloating a smartphone with desktop software given that the amount of people who want their smartphone to be their only interface to everything is likely a very small minority.
I don't think there's a way to gather that statistic. Some people might suddenly get a project that involves software that leverages GPGPU, some might use it professionally, then there might be kids experimenting. Plus as the job market involving AIs grows, so does the GPGPU market.
0% is different from none. 0% means that it's a very small amount, not even enough to be 1%. The implication is that while it could be true, it's such a rarity that it is not in fact a counterargument to "GPU implies gaming".
correct. which makes the top comment pointless (or missing the point?)
Under-performant computers can and have been silent for a while. A phone falls into that category.
The trouble is making a good performance computer silent.
And even the case the article advertises, is pure garbage. I had the smaller ones (and the author should really have bought the black anodized one!). It works fine while underpowered. But as soon as you hit the 5h compile/rendering levels of workload, that thing cannot move heat away without airflow. period.
I can hear a high pitched whine emitted from most electronics. Granted for smartphones the display or transceiver has to be powered up and it has to be a quiet room with no white noise. It's a lot better than it used to be, I knew I was getting a call or text from my old Nokia 5160 before it did just from the whine.
Smartphones are capable of computing and are sometimes very powerful, but the analogy is totally out of wack in my opinion.
If you can live without using an actual computer, it means you don't really need computers. You can check your email and browser the web on a Kindle, on your TV, or even your car.
Everything is a computer, then.
I think a computer is a productivity tool. Smartphones (and I'd definitely say tablets, too) are to consume content, not produce it. Some companies (most notably Apple I think) believe otherwise, but I think they'll have to come to realize that smartphones and tablets are horrible to produce most kinds of content.
Certainly in the rough sense of 'Turing machine' a smartphone is clearly a computer. And, even in other senses, I can connect a keyboard to my S5, open up Termux, run ffmpeg, open up Emacs --- all on the smartphone itself, not through ssh or anything. But it's hardly as productive as a regular laptop or desktop, of course, but that's largely due to screen size, lack of keyboard (or lacklustre keyboard), and weakness of processing capacity compared to a desktop/laptop.
His old DOS machines certainly weren't servers. Not having a GUI does not make a machine a server and having a GUI does not make a computer not-a-server (especially since we have graphics chips inside CPUs now).
I don't know about iDevices, but this is very much supported on a lot of Android phones. Samsung's lineup even has it as a bullet point feature, and they sell docks which let you connect your phone to a screen, peripherals, and switch to a mouse-oriented UI without having to do _anything_ weird.
At that point, you're running a mouse-centric, multi-window OS with a wide array of software, that can run basically whatever you want.
So that's a computer, definitely, right? When does it stop being a computer? If you disconnect the display? Is it using a stylus instead of a mouse? Maybe the software keyboard instead of a hardware keyboard? (But then is the MS Surface not a computer when you detach the keyboard case?)
Even though you're technically wrong (the best kind of wrong), I can see where you're coming from. Even microcontroller ICs without enough memory to store this comment are real computers and are silent, but if you say you've built a silent computer and then unveil some Arduino contraption, expect eyes to roll.
That said, I still think smartphones qualify as computers even by your productivity definition. Newer smartphones would sit somewhere above older netbooks on a ranking of overall utility.
In the past when I tried to build a silent PC, I found that even after removing / stopping all the fans, there was still often an electronic humm or buzz left over. That is when I gave up.
Later I changed desks to one that had one of those built in computer cabinets made of thick particle board. That did as much to silence a pc as all the tens of hours of effort I had put into meticulously researching and specc'ing the build before.
This type of noise is arguably worse than that produced by a well-managed fan setup.
If you have fans that spin at a constant RPM, you can fairly easily tune out the noise.
Coil whine will vary depending on the load of the system (e.g., when you start/stop scrolling a web page), making it much more difficult to tune out.
Never heard coil whine for years and years decades infact of PC building. Until last year I got my shiny new fancy pants blast furnace, (aka GTX 1080 TI), which has near dead silent fans at idle/light loads. The minute that bad boy starts working an obnoxious screeching/whine starts.
Super annoying compared to the rest of the build being a beast of a machine and watercooled that's so quiet I'm more likely to hear the noise floor on speakers than the PC (which is on the desk, next to said speakers).
Coil whine with graphics cards is such a frustrating experience, because there is no informed consensus as to what the underlying cause truly is. Everyone has their own anecdotal reason:
a) Maybe coil whine is an intrinsic factor in the manufacture of graphics cards, similar to dead pixels on displays. "Luck of the draw" when obtaining one is the only way to win. Cycle through RMAs until you get one with little to no coil whine.
b) Or, it depends which company you buy from: each of Asus, EVGA, Gigabyte, MSI, Zotac et al are supposedly better or worse than the others.
c) Or, it's not a problem with the GPU at all; rather, it's an indication of a poor quality power supply (PSU).
I've never seen an informed analysis from an industry engineer who has a goddamn clue what they are talking about. NVIDIA could probably enlighten us all with an exact-science explanation, but that seems unlikely. My uneducated guess is that the situation is closest to option 'a' above, and that rejecting units for coil whine during quality control would drastically reduce production yield.
Alternatively, if you wanted to really dig into this, you could try to find every switch-mode power supply and replace them with (much less efficient) linear power supplies. You'd need a larger water-cooling system though.
Have you looked at the current requirements for CPUs and GPUs?
We're talking 100A or more. Even if the input was 3.3V, which it isn't, a linear regulator down to 1.1V would have to dissipate 220W or more. You'll need a bigger heatsink for the linear regulator than for the CPU...
And designing a 250W-capable linear regulator is not as simple as just hooking up a LM7805.
So with a modern CPU and GPU, your talking ~400W of power to the actual components, and nearly 500W wasted in the linear regulators. This of course also means you have to get a 1000W PSU as a bare minimum.
Here's the problem: not only do you have to dissipate 200W+ for your pass element, you need to drive it ultra fast with an extremely fast analog circuit that can withstand the massive magnetic fields (which pretty much means you need a PCB).
Yeah and good luck driving a 200W linear element (if it even exists, lol) with a few op amps--the driver which should deliver a few amps into the gate/base of the pass element, which in and of itself is a pretty difficult challenge.
I've had a GTX 460 and a GTX 660 Ti before, and once I upgraded to a GTX 970 i was disappointed by the coil whine. I could even see the effect of the power drain on my secondary screen when I booted up GTA 5. Guess we'll have to live with the fact that the tech cannot keep up with the demand and quality, resulting in more 'unstable' electronics.
My fiancee has a mouse(Logitech MX Master) which has a very-high pitched coil whine, but only when you move it. It's the most infuriating thing in the universe, but apparently she can't hear it. So yes, random coil whine is about 10000x worse than fans.
I've had my top-of-the-line Dell XPS serviced(motherboard fully replaced) 3 times because of the coil whine issue, and it's not going away, and Dell said they will not replace it any more. It's one of their most expensive laptops and they can't get such a basic thing right.
Last year when I was deciding which laptop to buy for personal use I ruled out the XPS15 (partially) on coil whine (I'm 37 but thanks to never going to concerts and rarely listening to music beyond half volume I still have decent high frequency hearing).
In the end I went with a Thinkpad and having seen the issues people I know have had with the XPS15 I'm pretty glad I did.
Unless you've tested OP's fiancee's system yourself and found that to be the case, I wouldn't be so sure.
I had a Logitech G500 with awful coil whine and the opposite problem: it would stop whining when moving and start when idle. I suspect it had to do with the power saving mode that lots of mice have, where the laser power supply ramps down to dim the laser illumination after a period of no detected motion.
Not necessarily. Possible, but I had a cheap HP mouse that whined briefly when it first came out of sleep mode. Couldn't have been the computer, since it would make the same noise when the PC was turned off.
Some noise, depending on the texture can be even pleasuring.
coil whine is highly unnerving while low fan sound is relaxing.
we like stimulus, the clicky keys of my old hp48 is neat, the insertion sequence of pioneer 32x slot-in cd drive was amazingly subtle; not long ago I revived an old HP tape drive, the tape rolling and the head gear was also beautiful.
Also, it was as cute as informative, it's a clear state change side channel. Often software notification about hardware are decoupled so much that you don't trust it; plus they're invasive, unlike a tiny led, a click, a tiny motor ramping up.
> Some noise, depending on the texture can be even pleasuring.
I like the sound HDD make when grinding (except when I don't know the reason for the grinding... looking at you svchost.exe).
> coil whine is highly unnerving while low fan sound is relaxing.
Which is why I have been putting off getting a new laptop for years now. Most seem to suffer from coil whines and I can't stand it (to the point I ended up using an old eeepc 1000he rather than a brand new 16 inches VAIO some years ago).
I've switched to Fractal Design cases with their heavy sound-proofing and been thoroughly impressed. I use Corsair RM750 which never powers its fan on, and I used to have a all-in-one water cooler for the CPU, but I moved to a Noctua design with a large 140mm fan. The water cooler had a 120mm fan that I replaced with a 120mm Noctua--but it was too close to the rear vent of the case, and thus noisy.
However my graphics card (RX 480) is quite loud, and one bearing is making noises.
But the Fractal Design case has really dampened the sound. For my home server, it's using a RM500 (which also never turns the fan on), and a low-profile Noctua CPU cooler. No other fans, but I do hear the 6 HDDs spinning and seeking when it's real quiet in the room.
I don't hear any coil whine, except when using headphones plugged into my desktop's speakers--probably the result of the speaker system's power supply. Klipsch ProMedia, if you're curious.
I have one of the Corsair budget quiet cases under my desk. I can attest that it has excellent bang for the buck, for being a basic black case with useful soundproofing/absorption.
As far as noise reduction for CPU cooling, I'd suggest buying more air cooling than you need for a modest TDP CPU. Between that and my fanless Seasonic power supply, and SSD, the only noiseI can ever hear from my machines is from the GPU.
I had a Corsair H50, it lasted about 5 years or so. The pump died on it. It was actually quite annoying to track down. At idle it would run for an hour or two before shutting off. But start a game it would only last a few minutes.
Seems obvious in hindsight, but I had no fan (pump) speed warning or anything.
The only time I achieved silence was when I moved the computer case outside the room and used a 2m VGA cable and USB chord extenders. That silence was weird though. No audio feedback at all from the computer, just the clicking of keyboard and the mouse.
Pretty much. I've never really shut down my main machines when i'm not using them.
I work from home, so i'm on it several hours every workday, combined with the fact that I tend to have multiple things in-progress all the time means it would be a giant pain in the ass to shut it down fully.
Hehe, you get used to it after a while. Sometimes I wonder where the heat is coming from when my PC has some higher load and my hand moves over it. Those are the moments when I remember the days when I couldn't hear the vacuum when the PC was compiling ;-)
A long time ago, about the time someone tried to coin the term invisible computing, I figured out how to hook a monitor arm to my couch, and I put the tower behind it. 3 inches of padding can absorb a lot of sound.
The case matters a lot. I've got a Fractal Design Define C, which dampens a lot of noise. Supposedly the R5 is even better.
This was my second silent PC. The first one still had moving disks, but I went for as few fans as possible, and had passive coolers on the internals. This time I did the opposite: lots of fans, but have them spin as slow as possible. This works very well.
But coil whine and electronic hums are easy to overlook when you're choosing parts. It's worth looking at not just the fans and the power use (more power needs more cooling), but also the quality of the electronics.
I've a completely silent PC too  and I was lucky, as mine doesn't have any relevant electronic buzzes.
But when I walk to the backside of my desk I can hear some electronic buzz from one of my monitors. Whats funny about it: I have that monitor since a few years now and before I built that silent PC and turned my desk to another direction, I never noticed the buzz from the monitor :D
I had a Geforce 280 that would scream like hell whenever it was at full power and its framerate went below about 10 or above about 100 FPS. I was glad when it broke some other way and got replaced under warranty.
I keep my CPU running at a confortable 50°C to 60°C.
There are 5 fans in my tower, two on the CPU cooler, only one of those two fans is running constantly, at only 200rpm. All the others aren't running most of the time.
I don't need my computer to run at a cool 30°C all the time. The hardware can run very hot without any issues. And when all the fans eventually kick in under load, it will always keep under 70°C anyway.
It does mean things run warmer, but generally moderately specc'ed pcs don't generate enough heat and have enough reliability margin that it is not a problem. If you were some game enthusiast or crypto miner running multiple flagship GPUs on an 850W power supply, then I probably would not recommend this approach.
> (Astute readers will notice they are all AMD (Socket AM4) motherboards. The whole Meltdown/Spectre debacle rendered my previous Intel system insecure and unsecurable so that was the final straw for me — no more Intel CPUs.)
That's a silly and extremist position to take. "insecure" is relative, not absolute. It's a certainty that the software he's running has far more quantity of vulnerabilities and a much longer history of them. I don't know his exact use case, but arguably his use case isn't one where Spectre is particularly more severe than even a userland, non-priv-escalation vuln. (eg ransomware doesn't require root access to hold all your files hostage.)
> Eliminate the moving parts (e.g. fans, HDDs) and you eliminate the noise — it’s not that complicated.
Ha! And yet it deserved a detailed blog post. I'm surprised he would say this even after the amount of effort he spent.
> however their GPU drivers are terrible, especially under linux.
Imo the AMD drivers are way better than Nvidia's drivers. They're included in the kernel and therefore open source. Compared to Nvidia's proprietary drivers that have horrible support with a lot of compositors and lack support for DDC/CI over DisplayPort. The Nouveau drivers are better (slower performance but better compatibility), but are unable to change the clock speed (and are set to the minimum).
The AMD drivers "just worked". Selling my RX480 for a GTX 1070 was the worst decision I made when it came to compatibility. Now I can't even get Vsync to work with this Nvidia crap.
Well it's a good thing nvidia is catching up on being horrible. The 367.xx driver's dkms module doesn't even compile for me on Ubuntu 16.04 anymore, and later drivers make some gstreamer-based apps stutter a bit.
And Windows is behaving weirdly as well since I installed the latest drivers. Black bar on top of full screen programs after waking from hibernation, The HD audio driver not letting pulseaudio start (I need it to get sound from WSL) and crashes when multiple 3D accelerated VMs are open. And restarting the GPU driver (with either the shortcut or through device manager) is what solves all the issues and they only occur with the latest driver.
And the crappiest part is that there's nobody that can help. Getting someone from nvidia to respond on their forums is basically luck, and I'm not a huge company that can get their reps to get someone to help me.
The open source drivers of slightly old GPUs are ok. Not that good, but not bad either. And they are open, so they will come well integrated with your distro.
What puts AMD GPUs in a weird situation where you can expect them not to work very well when they are new, but to improve until you can forget about them. (The inverse of the NVidia GPUs, that work ok when new, but slowly loses compatibility with time.)
NSG S0, once out, will most likely be the go-to case for such setups. Until then, an HDPLEX H5 is cool.
My desk has a H5 on it, housing an i7 8700 (non-K) and a GTX 1060. The TIM under the heatspreader is replaced with Thermal Grizzly Conductonaut and Thermal Grizzly Kryonaut is used as every other TIM that the case setup needs. The CPU is on stock clocks with a voltage offset of -30 mV. The GPU has the power target reduced to 90% and clocks increased by 130 MHz, so that it is effectively undervolted as well. The PSU is a Seasonic Ultra Prime Titanium 650. Prime95 with AVX throttles really, really fast, under a minute, perhaps, but is a very unrealistic load. Non-AVX stress tests and FurMark take a while to start throttling (20 minutes?), as the thermal capacity of the aluminum case is quite big. After hours of gaming, the GPU and CPU float around 80 C while providing full stock performance. I don't do 3D rendering (other than in-game) or video en/decoding, so have not had long, real-world, full loads to see how temperatures behave with those.
From the discussion I've had and forums I've read, I think that people are afraid of putting more power in passive cases and having their components at "high" temperatures, despite those being rated for them.
>Prime95 with AVX throttles really, really fast, under a minute, perhaps, but is a very unrealistic load
I suppose blender would thermal throttle the cpu as well. If you run any non-Xeon/non-Laptop Intel chip (greater than 2k series) and care about temperatures - delid the bugger. (Xeons are soldered, laptop chips don't have IHS). Intel uses something that's worse than toothpaste, plus tons of glue between the die and the IHS. If you see temperature deltas under full load more than 9-10C between the cores, the thermal paste between the die and the IHS might have missing spots or have dried out.
In your case removal of the IHS altogether would provide decent results.
You might wish to check the VRMs, they are rated at 125C but if the case is hellishly hot inside, they might not be able to dissipate the heat.
Somewhat unrelated, but maybe you can shoot me down since you seem to have some experience?
Metal is an incredibly good conductor on its own, and the properties of thermal paste (typically) are just barely better than air. So long as your cpu and heatsink are fairly flat surfaces and mashed together physically, it seems like either forgoing or having the absolute minimum amount of paste is ideal. I've used a razor to leave an absolutely minimal layer of paste (e.g. filling in sub-millimeter surface structure) on my latest build, and cpu temperatures are well within a reasonable range. But I'm also not trying to OC the cpu or anything.
>...and the properties of thermal paste (typically) are just barely better than air.
I am not certain how you have managed to come to such a conclusion. Thermal conductivity of air is around 0.03W/(m·K). Good thermal, non-conductive paste is like 12.5W/(m·K) (or 400 times better than air). Conductive ones are in the region of ~40-80 W/(m·K) and Aluminium is 237W/(m·K). Also air also expands pushing the cooler and CPU away.
Normally you if choose between "too much" and "too little" paste, you pick the former. The pressure pushes out the unneeded amounts.
Don't, you need to multiply the raw conductivity by the linear distance occupied by the thermal paste? I presume that distance will be at least two orders of magnitude larger than that occupied by air in a metal contact only setup.
I would be extremely surprised if increased pressure due to air at higher temperature played any role whatsoever unless the bolts connecting the heatsink and cpu were very loose. If anything, I'd expect the increased conductivity of air at higher temperatures to dominate.
I'd also expect there to be effects at the metal-paste and paste-metal interfaces which reduce the effective system conductivity (i.e. phonons are much more likely to reflect in this scenario than in a metal-metal interface).
It's very impractical/expensive for mass products to make the surfaces in question so flat that no thermal paste would be needed. Many tests and reviews have been done. Even if top-of-the-line coolers came with perfectly flat surfaces, Intel's heatspreader is not -- otherwise it would cost so much more. Also, heatsinks can be applied with a lot of force, which usually pushes out the "unneeded" part of the thermal paste. In a bind, even lipstick, toothpaste, chocolate and other silly compounds work better than nothing, so I'm not surprised that you're getting ok results even with a touch of thermal paste.
A fun thing to try is using a modern low-end CPU (latest i3s, Pentiums, Celerons) without its cooler. Not advised by Intel, of course, but you might get into your OS of choice even before it starts throttling. I'm somewhat comforted by the fact that a CPU automatically powers of once it reaches something above 100 C (103 maybe?) and throttles a few degrees before that. Those temperatures shouldn't leave the silicon damaged.
In practice, thermal paste is a must. If you don't like those (I personally don't, they get everywhere by accident and can be tough to remove), try getting an IC Graphite Thermal Pad which is reusable and rivals really good, if not the best thermals pastes, according to the limited number of reviews I've seen. I think that its practicality beats better results in non-highest-end applications.
Blender might throttle. Does it come with benchmarks/tests that I could use to gauge thermal performance?
The CPU is delidded! I've got another i3 4300 delidded as well running under a NoFan CR-80EH. Delid + Conductonaut + Kryonaut made the difference between throttling vs hovering around 90 C in FurMark + Prime 95. When integrated graphics aren't used, the CPU runs cooler, of course, and didn't throttle with MX4 thermal paste and no delid.
I do fear that VRMs are running too hot. When selecting components, I picked those that come with some heatsinks on VRMs at least. The motherboard is an AsRock Fatal1ty Z370 Gaming-ITX/ac (non ITX motherboard wouldn't fit in the case anyway with an ATX power supply). The graphics card is Gigabyte's cheapest offering and has a small sink across the VRMs. I'm hoping that undervolting will help keep the VRMs in check.
Clearly a work of passion and I appreciate OP sharing.
I enjoyed earlier days of "Silent PC" building, ten or fifteen years ago. For example, building a silent tower or desktop for a DAW or softsynth back then in a recording/studio environment required some ingenuity. SSD? Not on a hobbiest budget. I recall one build, not mine, fully immersed in a bucket of oil (mineral?) for passive heat dispersion.
Today, as a new reference point, any MacBook Pro within the last few years may qualify as truly silent for many people's everyday usage. It does for me. And when I do heat up the CPU/GPU with heavy tasks, the fans spin up but then they go away completely as soon as the hard work is done. Back to silent.
No more spinning platters or crappy fan bearings or poorly engineering airflow nowadays. :-)
There's no hacker pride in buying off-the-shelf, so the performance bar for DIY is higher. Progress!
I'm glad you enjoyed the post. My first silencing attempt (in the early 1990s) was to seal a hard disk in what amounted to little more than a couple of oven bags, custom-make a long IDE cable, and then — literally — dangle the drive in a bucket of water that was under the desk the computer sat on. That ended poorly, with thermal cycling inducing microfractures in the bags, which resulted in leaks, a dead drive and a blown controller on the motherboard. Ah well, we live and learn. Take it easy.
It used to be fun trying to silence a computer. Figuring out which HDDs were single-platter, undervolting processors, finding low-rpm fans, etc. used to be a challenge, and have great rewards for those of us who care. Now, check out silentpcreview.com, hasn’t been updated this calendar year. In some ways I miss the old days, but most modern laptops and even desktops are reasonably quiet, and full silence is easy to achieve.
I've had the same experience with my MBP 2017. So many people don't realise that the aluminium chassis isn't just attractive, it's actually a massive heatsink, and it's the reason the machine needs barely any internal airflow.
Silent means that it has no sound, but "sound" either means the sensation of perceiving vibrations, or the actual vibrations that cause the sensation. 0 dB SPL is the threshold of hearing, so it is "silent" in the first sense.
If the air moves at a constant speed I don't think you'd get sound. Any fluctuations probably won't be in the audible spectrum, although whether that still counts as 'silent' is more of a philosophical question.
I've been in one. Back in HS, we took a trip to the GIANT anechoic chamber at Georgia Tech. It was incredible. If someone was facing away from you and yelling as loud as they could, you could barely hear it from behind them. You're standing on basically a giant net on top of a pit of foam blocks, and every wall is wildly shaped undulating foam.
It's really surreal and a number of people actually became nauseated at the sensation.
I'm pretty sure that there's some measurable amount of coil whine or other noise coming from capacitors and other small components, making it more than -infinity dB, and probably more than 0dB. The sound might not even be within human detectable frequencies.
It may register below 0dB SPL (in an anechoic chamber, anyway) depending on the manner in which it's measured, which is a complex discipline in itself. Sound pressure dissipates over distance, for example, so the distance at which the measurement is made has bearing, as does the angle at which it's measured.
(This is unabashed pedantry. But I'm on my lunch break, so...)
So these sorts of builds are way harder than they look. You need to spend a good amount of time planning out motherboards, and willing to go back to the shop to buy another motherboard (because this motherboard might have a big honking capacitor where you expected to run a heat-pipe).
So you either need to carefully look at pictures / guestimate the location of components, or hope that someone else on the internet has figured it all out for you. Alternatively, you could just be ridiculously lucky.
But I don't really believe in luck personally. If this guy has been doing these builds for as long as he claims, he probably had a fair bit of thought go into the particulars of this build. The kinds of thoughts that don't go very well into blog posts because they're uninteresting (but very important). Like, does X combination of components fit or do I do Y instead?
With respect to the build itself: I'm surprised he was able to get 60+W CPU and a 60+W GPU in there. Most silent builds I see basically use laptop parts (30W or less) to keep the heat down.
32gb of RAM will be plenty, and 8 Zen-cores is plenty strong. The GPU is a bit weak for gaming, but he should be able to play plenty of the lower-end games without much issue, even at 1080p / 60fps (probably Overwatch for example). He probably can't run Witcher3 at 4k on Ultra, but such a GPU would blow out his thermal-design completely.
Yep, it took weeks to finalise the parts. I had to contact multiple vendors and get them to check component heights, track down motherboard layouts, compare review photos from different angles, plan heatpipe runs, resolve conflicts with VRM heatsinks and power connectors, watch assembly/review videos over and over again, scrutinise manuals, determine hardware/OS incompatibilities... probably well over 80 hours of work before I could even order the parts.
But, like you say, even though I had done all that, there was still the possibility that I might have missed something and would need to return/replace some parts and rework the plan. My mouse hovered over the "Order" button for quite a while before I finally committed. An anxious moment.
Virtually none of that methodical and boring research/prep work made it into the post — it's just not that interesting. Necessary, yes, but not something that the vast majority of readers would want to read.
The setup should be able to handle a 105W CPU and 75W GPU. At this point 65 + 75W is confirmed. When I OC the 1600 I should be able to verify up to ~95 + 75W.
The strategy was to start with a really well engineered passive case, then select components that could be pushed right up to the thermal limits. It's worked well so far. I'll keep publishing all of my test results for those that may be interested.
Personally, I just don't like computing at a distance. I actually like hearing things like fan noise or hard drive sounds, as long as they're not incessant, because it can serve as a subtle notification that the computer might be doing something in the background that I might not have realized otherwise.
On a slightly different but related subject, I live on a boat with a diesel engine. Being able to hear the engine run and hear the change of tone is something I couldn't live without.
I often run the engine to heat my water and the very slightest change of tone has my ears pricking up, panicking with "I've not heard that sound before! What's up with it? Is it broken? Am I going to sink??" running through my head.
Fans don't tell the whole story. Network traffic is super useful to see as well. Also, if your computer never gets hot enough, it can be doing plenty of unwanted background work without setting off the fans.
I did the same but the tower is in my basement. I ran an HDMI and USB cable up thru the wall and into a keystone panel behind my TV. Other than occasionally having to manually flip the tower back on after a power outage it worked pretty flawlessly. Totally silent and no tower to hide in my living room.
At the risk of me too! I have a Dell T420 with eight spindles in the loft above our bedroom, plus assorted UPS, APU2 based router, switches and a mini ATX based thing for MythTV.
I screwed four lengths of 4" x 2" between two roofing joists to make two trapeziums. I then screwed a 3/4" thick sheet of chipboard to the bottom and then wrapped the top and sides in roofing felt (some protection in case of roof failure).
The loft is lagged 200mm deep with 3/4" chipboard on top which deadens the noise nicely. The Dell was running the fans quite a lot today when I went up to check on something - 25C outside in Somerset, UK today.
On a more serious note, would it be possible to achieve said silence without too much compromise by some variant of water cooling, with a good sized reservoir? And use the heat to drive circulation rather than a pump?
Because there is no pump the water will flow much slower, therefore you need much bigger pipes. Think at least 3x the diameter you would think you need.
Also because you are depending on thermo siphon you have to ensure you have to ensure the water flow works with gravity. The heat source must be at bottom, and the radiator must be at the top. You don't get any flexibility go around something that might be in the way, the pipes must always be sloped in the correct direction.
If you are using anything other than water you need to pay attention to specific gravity (and how it changes with temperature), and viscosity. Probably something else that I'm not thinking of.
It probably can work. I have an antique tractor that uses such a system. A 2.2L, 16hp engine, has 24L of coolant, and the pipes between the engine and radiator are noticeably bigger than my truck with a 7.3L 250HP engine. Remember that this was designed to run at just under the boiling point of water (antifreeze is too modern). You probably want your computer considerably cooler, which implies even larger systems than a similar active one.
Yes, just use a flat (i.e., no high fins) passive cooler on the CPU (this won't fit with stock parts), lay the motherboard flat and submerge about 1cm deep in a refrigerant boiling in the range (max room temp 5K) and (max healthy cpu temp -20K). Flourinert and Pentane are suitable, the former is really expensive and the latter as flammable as gasoline.
If you replace the cooler with a 1mm copper plate and use a pump to archive a flowspeed across the cooler between 5 and 20m/s, you can get crazy overclocking. You might get subzero clocks without any active refrigeration.
The benefit is that all your ancillary components are cooled too, and you have superior performance with a pumnp, and even performance without one.
Ensure some radiator with gravity-backflow gets the vapor, e.g. just use a central heating one above and ensure liquid can't pool in the piping, and that that is not too narrow.
From what I recall mineral oil has long term effects on the HW, or in other words, I think I've seen some pictures of mineral oil cooled MBs with degraded/melted plastics. And the oil it's a logistic mess.
Instead the other day on Youtube I don't recall how I landed to a 3M infomercial about their Novec dielectric liquid used for immersion cooling. That stuff looked interesting, but they didn't spilled (pun ^__^;) details about cost or health issues tho.
The difference between Novec 649/1230 (they seem to be the same) and Perflourohexane/FC72/'Flourinert (well, 'the' classical kind)' is that the former is somewhat less inert and apparently slightly toxic to the central nervous system (high doses only), and the latter is sufficiently non-toxic to be used for allowing burn victims to breathe, as it shields their lungs from damage, and is able to provide enough oxygen by dissolving it to prevent suffocation. Rats don't drown in it if submerged and if it is sufficiently aerated to contain enough oxygen.
The other difference is that Novec has a Halon-style effect on fire, whereas Flourinert is at least not sold with that intention. Also FC72 decomposes slowly into a toxic, but afaik not dangerous to the elctonics, product, which has to be scrubbed.
Both of those have a vastly higher global warming effect than 'just' using Pentane, but that is rather flammable so one would probably want to ensure the oxygen concentration in the room is too low to result in a flammable mixture. This would probably mean that a human couldn't breath without external oxygen supply, but that should not be more expensive if it is a suitably low-physical-maintenance location, i.e. the systems are installed once and at most swapped out when obsolete or in case of component failure. A human could probably make due with an oxygen tube in his nose and some way to prevent the exhaled oxygen from sticking around in the (probably low-airflow) room.
The cost for FC72 is about 300$/kg, keep in mind the density is about 2.3 times that of water. Novec should be somewhat below, I think (otherwise there is little reason to offer it).
The difference is that mineral oil cools by convection and my solution cools by nucleate boiling. With my solution, and a flow or iirc about 20m/s, you could cool a delidded AMD EPYC 32 core at 600W. I.e. you are not limited by cooling, you are limited by how much current the LGA pins can handle before they loose stiffness (due to resistive heating in the pin and at the contact area) and then spiral out of control due to the weak contact pressure increasing resistance and leading to the ping disconnecting. This easily goes over to the others, there are documents from iirc Intel about how this happens/works, and how to prevent it.
I considered soldering the LGA pins to the CPU, but due to the craziness of this idea I haven't researched it further.
The thing is, that a dual-socket 32 core/socket EPYC clocked/overvolted for full stability and an overvolting-induced lifespan of 5 years, which does not have problems with power delivery on the way between the interposer (the thing the dies are mounted to and that connects them to each other and provides LGA pads) and the wall socket, would probably be the most affordable (TCO, i.e. including electricity and maintenance) system for non-distributed compiling of reasonably parallelizable software.
It would be 'the' ultimate workstation in the sub-$15k range, this would be the full system including the single-quantity amount of, in this case, Flourinert (which is not toxic; there was even research to fill the lungs with it for high-G environments, as it prevents lung collapse. There were issues circulating the liquid fast enough due to the much higher viscosity compared to air, so they dropped it.) and a custom-manufactured (from a local metal shop) containment case for the system and work from a plumber to provide suitable piping. The only non-included thing is the manufacturing and mounting of the nozzle that provides the required flow rate across the CPU to keep up with the heat.
Short answer: mineral oil does not provide suitable cooling for high-TDP processors.
It's possible using a low temperature phase-change system, but you need some absolutely immense radiators. The Calyos NSG-S0 chassis will passively cool a high-end gaming PC, but it weighs 22kg and costs €1000.
Probably not - you need to push water over the radiator blocks at a decent clip and I'm pretty sure you need pump. You could try a heatpipe into a big radiator that sits in a big tank of water, though.
Peltiers are only really worth it if you need sub-ambient temperatures (i.e. extreme overclocking). You then need to carry away the heat of the Peltier in addition to that of the processor, making expelling the heat even more difficult than it was before.
If you are interested in this topic then http://www.silentpcreview.com/ is your friend, they have been doing this since 2002 although certainly activity has been lees the last few years. The forums are also good.
IME components need to be selected for low coil whine (and the related brethren, namely electrostriction and piezoelectrics), because it easily dwarfs (both in loudness and annoyance) the air flow and fan noise of a well-built air-cooled machine.
One is the good old P1 125Mhz router we had like 10 - 15 years ago. The only active cooling in that system was in the PSU and that thing throttled it's fan off during normal operation. It had zero moving parts outside of the PSU. You couldn't even get fans for those CPUs. These old systems were fun :)
And beyond that, this makes me think of the computer requirements of coal mines, and I think mining in general. In a mine, a piece of compute has a defined maximum energy emission per square inch surface. If you exceed that, you risk coal dust explosions, which are rather inconvenient, loud and adverse to throughput. The stuff in the article is very similar to industrial computers in such a setup. Quite interesting.
At my college, we had a bar with a custom account management system (it managed the credit of each students).
The thing was all text/ncurses.
The terminal was an under-clocked PII, with all the fans removed and connected to the server using a serial cable and it was kind of necessary since the environment was quite hostile and sticky (spilled beer, some fat deposit, etc).
If I recall correctly before the PII, it was a VT-100 console, so even more basic than the PII.
A higher wattage than 400W is utterly pointless for a fanless PSU; if you need to dissipate constant loads of >400W of heat, you will need a fan somewhere (exotic options like mineral oil cooling aside).
Well you can always add absurd amounts of surface area and mass to the heat sinks to take advantage of thermal inertia, or possibly change components to simply raise your safe operating temperatures to absurd levels. But that's pretty involved. Spacecraft take care of huge temperatures with no fans or even conductive or convective cooling. But that requires enormous, sail-sized surfaces to take advantage of extremely slow radiative cooling.
It would be an exotic option for sure. And not something you'll find pre-built.
Wow. That thing has more surface area than a motorcycle engine. I guess it really needs it with such a relatively low temperature gradient, not to mention the absence of airflow from movement. I'd love to see how that really performs.
I considered a fanless PSU when I bought my current set-up, but at least back then I didn't find them cost-effective: for a cheaper price I could buy a higher-rated, at least as efficient model from the same manufacturer, and even though it has a fan, it is equally not making any noise in most use cases because it's not running. (I've never bothered to find out when it turns on; later on I added an external GPU to my case, and when its fans start -- not in desktop use -- you can't hear the PSU fan anyway, although I certainly wouldn't mind it running when the case temperature goes up, which it certainly does in certain use.)
Speaking of silent computing; unfortunately my monitor is "louder" than my PC in desktop use for most brightness levels. Anyway, personally I've given up the goal of doing a solid-state PC; a single fan chosen correctly in a case is virtually silent and makes the cooling design much easier; in my experience most PSUs and GPUs produce similar amounts of noise even without fans (and with bad luck, they can be worse).
I think a good route would be to get as efficient a fan-cooled psu you could find and modify it yourself. You can always take the PCBs out of the case and add bigger heat sinks, change the TIM under those heat sinks, add water blocks if needed, desolder components and swap with higher wattage components, etc.
Or you could just submerge the whole machine in mineral oil. That technique really sucks, though. As cool as it is to have a fully submerged computer, it's hell to clean literal liquid laxative out of a PCIe slot when it's time to upgrade the graphics card.
For what it's worth, I've been using a Corsair AX760 (80+ Platinum) with a i5-6600 and a GTX 1070 in a Thermltake V31 case (stock fans) for a couple of years now and I don't think I've ever heard/seen the fan spin up, even under load while gaming in the Southern California summer heat.
Completely silent and zero time boot are the two features I still miss from my C64. The only thing that came close in the meantime was my Palm III, but it wasn't a desktop computer. To be fair I rarely boot a device (be it mobile or notebook) nowadays, but still..
About two years ago I wanted to make my pc completely silent as well. It turned out that the cheapest solution for me was to put the pc in another room that I don't sit in and put long hdmi + USB cables through the wall. I also soldered on a longer cable with a power button so I can start the pc from my room. The only downside for me is that I had to run cables along my walls through a conduit but I think in total I paid less than 40€.
If you're going to count all possible side channels as output mechanisms, then I'd imagine that it is indeed impossible to build a computer without an output mechanism.
Unless you limit your scope to say, Earth, and somehow build a computer outside of Earth's light cone. It is thus causally disconnected from Earth and has no output mechanisms that can be read from anyone on Earth.
Anything that can be used can be misused, if you wanna get into the philosophy of it. And any information that is intelligible to one party can be made intelligible to another party, pending whether quantum weirdness™ pans out or not. Even if entanglement becomes a consumer-level thing, I bet eavesdropping techniques will eventually be developed, even if they only operate on the conventional side of the quantum computer.
But this is more about security engineering philosophy. The Intel management engine relies on a lot of security-by-obscurity, which is not the best way. But I think that Intel chips are passably secure as long as you watch what you're doing.
If I were to try to setup a system with as much security as possible, I'd go with OpenBSD, in a Tempest Room, without a direct connection to the Internet. I don't recall the exact procedure, but there is a way to securely get information into and out of such a system, but it involves CDROMs, not USB memory sticks.
For these cards the giant heatsink is connected to the VRMs so cooling is a non issue. In OP's pictures, he seems to have replaced the heatsink of his card with one from the case without cooling the VRMs.
From personal experience, I tried to silent cool an RX 550 (50 watt vs 75 watt of 1050 ti) without VRM heatsink and the VRM temnperatures were up to 100-103C during a torture test. Adding a fan lowered the temperatures by 20C.
Both the hi-side and lo-side mosfets (they might be 2in1 on a card like this) are rated for their amperage at 125c and are typically capable of dissipating that through their contact with the pcb. Airflow is nice, but not required. Neither is direct contact with a heatsink. Those are for higher end cards (who also rate their mosfets at 125c for good reasons. Max tCase for mosfets is usually 150c. Beyond that they just die.
If I remember correctly that card still needs some air flow to work, normally provided from case fans. That would be problematic in a completely fanless system. A Ryzen APU might be better, as the case would cool it.
The design of the DB4 is such that the graphics card is mounted vertically. Fresh, cool air flows in through the grille at the bottom, over the VRM area, then up and out the top of the case.
I have more small heat sinks I could put on the mosfets, but they are relatively low profile so I'm not sure they would improve things much.
A single, large heat sink, with long fins (that would protrude above and beyond the nearby capacitors) would seem to make more sense — but given there are no mounting points I could use to clamp it down I'd be a bit concerned about the long-term effect of torque.
Quick note for everyone who might want to build a mostly-quiet home theatre PC. There are several versions of the geforce1030 that use passive cooling/no fans. With H265 acceleration in the GPU drivers, more than fast enough for 2160p, 60fps movies in VLC. They're right around $85.
Same basic chip architecture as the much more expensive 1070/1080. Not something you'd want to run newer 3D games on, but perfect for media center use.
There are several passively cooled geforce1050 (not ti), but you really don't need it for any form of 4k video. My output setup is to a 70" 60Hz TV over HDMI 2.0, with a ryzen 1500X that is slightly overclocked, this setup handled 2160p60 H265 up to around 300 Mbps, at which point it CPU bottlenecks. Which is a vastly greater bitrate than any bluray or torrented 4K content.
If you wanted a setup slightly more capable of 3D games then the 1050 would definitely be a consideration.
Well, 300 Mb/s it's a lot of headspace, it's nearly nine times the bandwidth of a Bluray. ^__^
I'm itching to stuff one of the recently release Raven Ridge APUs in a 3 liters case with a PicoPSU and have a portable, reasonably powerful desktop that I can stuff in a backpack, I'm waiting for a little more mature Linux support, at least a stable 2.4.17 kernel...
No. A vacuum doesn't transmit heat at all. The chips will get very hot. Liquids (generally) transfer heat better that air. You can completely immerse your computer in 3M Fluorinert. That will transfer the heat. Then, cooling fins to transfer the heat into the environment.
The challenge is removing heat. Putting things in a vacuum chamber doesn't make them colder - vacuum is actually a pretty good insulator, so you'd need some mechanism to get the heat out of there with9out using air.
If you go the other direction, some people have built computers they submerge in mineral oil or other similar media. That works pretty well for removing heat from components more effectively than air convection.
You could, but it would be thermally quite bad: The only way for parts to cool would be through radiation, which is way less effective than convection (be it forced by fans or passive, as in OP's design.)
As someone who has used the fanless 12" Retina MacBook, I can't put in words how nice of a feeling it is to have a fully silent computer. It's hard to understand why without actually experiencing it yourself, but it was a really big deal. If not for the awful keyboard ruining the experience, I wouldn't ever give it up.
It may sound (no pun intended) like an exaggeration, but I think more powerful computers going completely silent (fanless) will be the next most noticeable breakthrough since we got Retina displays on computers in 2012.
I doubt that. A while back I removed all of the moving parts from an old 486 and netbooted it, only to find out I could actually hear the network card making very faint screeching noises whenever there was network activity. You've probably heard the same effect on a much larger scale, at much lower frequency, when you hear the 60Hz hum of a "silent" transformer.
Kinda unfortunate timing for him that AMD released Ryzen 5
2400GE about the time as the article was written, which seems like pretty much perfect chip for silent reasonably powerful computer. Sure, it might not be quite as beefy as his 1600+1050Ti combo, but the energy efficiency should be much better. 2400GE is rated at 35W, while 1600+1050Ti is at 65W+75W=140W, or full four times more.
Yep, you can't always wait for the perfect product to come along — sometimes circumstances force your hand.
Not only did that processor not exist at the time, but even if it did there were going to be a bunch of compatibility issues with it on Ubuntu (like there always are when new hardware is released). I simply didn't want to have to deal with those sorts of issues for — potentially — months.
Maybe the Ryzen 5 3400G(E) will be a compelling upgrade? We shall see.
> "(Astute readers will notice they are all AMD (Socket AM4) motherboards. The whole Meltdown/Spectre debacle rendered my previous Intel system insecure and unsecurable so that was the final straw for me — no more Intel CPUs.)"
Silent Pcs are hard, however the simplest way to make your PC silent is to put it elsewhere....
I worked at a VFX company, where we did interactive client sessions (think: zoom! enhance!) which required a totally silent setup. In somecases that meant trying to silence an entire rack of disks (baselite 8,http://erwanlecloirec.typepad.fr/digital_flipbook/2000/10/ba... big glowing thing at the bottom of the page)
Congratulations tp69! The article details how miners made it difficult to procure components for the build, but the irony is that the miners might be specifically interested in this build for its power efficiency.
A decade ago, I was obsessed with fans and at one point I had more than 10 in my workstation. The noise it created kind of resonated with me, like I would be able to tell when the job was CPU intensive, HDD or GPU; I guess it felt more lively!
But after the smartphone boom, I got more adhered to power efficient, noise less systems (Smartphone/Tablet/SBC's/Chromebook) & after the meltdown spectre (pun intended) my adherence got reinforced.
Anyone looking for low-cost completely silent yet portable system in laptop form-factor can take a look at Chromebooks, now that Google has announced Linux apps support it's usefulness would tend to grow.
Compulab Airtop makes a VR ready tiny gaming rig i7 gtx1060 that is similar. There are very few fanless rigs with discrete graphics for gamer and workstation use like this that I can find:
I wanted complete silence too, but realized it's way easier to just move the PC elsewhere. I moved it into an adjacent room, made a small hole in the wall for cables (10M hdmi+usb+audio) and that was it, it worked perfectly. The price was trivial - about $40 total for cables. I already had a usb extension hub.
Sadly, that's only useful for 60Hz, as high refresh rates require a short displayport cable to work. For this reason I no longer use this solution.
If you're ok with 60Hz and have compatible room layout I recommend this. The biggest advantage is that it's completely PC agnostic, which makes upgrades way easier, as standard cooling solutions are fine.
Not quite the same thing but... I'm in the middle of building a totally silent Retro DOS machine. Using an old low-spec passively cooled VIA Mini-ITX motherboard in a small case with a passively cooled internal PSU. The storage is an SD card on the main IDE channel. The motherboard also has hardware Soundblaster/ADLIB support, so all the DOS games sound really great! 
I've just written a front end launcher for the apps/games, and I found some working USB drivers for DOS that makes transferring software over nice and easy.
EPIA 5000 Motherboard. The chipset on the Motherboard directly includes Soundblaster, ADLib, and MPU-401 support. You can enable/disable each one in the BIOS.
And yes, you need the EPIA Soundblaster drivers for DOS, but that's not that different from normal Soundblaster cards. It was a bit of an uphill struggle getting it working, but it's quite simple once you get the right version of the drivers.
It's not limited to the power supply, ceramic capacitors are on basically all complex circuit boards now, and most of them will readily sing and whine. It can be eliminated through careful design (reducing ripple) or paying nanocents more for the caps, but as far as I can tell Lenovo doesn't give a fuck.
I made a similar build but with the Streacom FC9 Alpha case. I put a slim 1050Ti in there and replaced the GPU fans with Nuctua 1cm thick 9,2cm fan with Zalman Fan Mate 2 that I turn up when I play games. I have to run it without lid though, as any computer I have really, convection is your friend.
I also went with 32GB RAM because I make multiplayer games and I need to be able to run multiple clients at the same time which takes alot of memory.
This build is my last computer, peak moores law will make it so.
I've built a fanless desktop too last year. Because I only use it for browsing and as a media player, I've chosen a 35w i3 CPU with integrated GPU. This way I could use a cheap semi-fanless PSU - which never switches on its fan - because the system just never exceeds its consumption limit.
Because of the 35W CPU, the system was cheap and easy to build.
What I have learned from it, is that my screen has a silent buzzing sound... :)
I am both impressed and left wondering "Why go through such trouble?" Not in a dismissive way but an inquisitive one since it has been such a long term goal. I can see some power usage reasons given how much fans use and that noise like heat is a proxy for inefficiency whenever it isln't the direct goal but that isn't his main goal.
Ratiofarmings is on the right track. The noise coming from the computer drowns out the subtle sounds of life all around me.
Someone suggested in the comments that I should just wear earplugs instead. This was my response to him: "But the rest of the world doesn’t make annoying noises that I want to block out — only the computer does. I still want to hear the birds tweeting in the trees and the wind blowing past the awnings and the rain hitting the roof and the ice-cream truck driving by and my wife chuckling at something she’s watching and the cats running along the floor and the microwave oven heating a meal and the kettle finishing its boil. I’m not trying to disconnect from the rest of the world — I’m trying to reconnect to it."
A lot of the most interesting and enjoyable sounds in life are very, very faint. It doesn't take much noise to drown them out. Even a single fan will do it. Now I can hear those noises and get 'work' done at the same time. It's really hard to describe, but 'magical' comes close.
Because it feels nice if things just happen like magic. Even "big" things like compiling, rendering or gaming. It just happens and you hear nothing but the birds outside and the cat breathing next to you.
Of interest, "Jay's Two Cents" did this clickbaity video where they daisy chain 4 radiators together. This would be one of those entertainment value only stunts, except that they discovered that they could shut off all the fans. Having only a pump running, you can passively cool both the top-end CPU and a high-end GPU. If someone can put the pump inside a nice machined aluminum case, this could be a DIY recipe for a fanless silent PC.
This is a wonderful project. It looks very sharp and has pretty good components.
However, it's usually significantly better, cheaper and easier to build an "almost" silent system, just by using good heatsinks and very slow fans. A small amount of airflow is usually significantly better than none at all.
Why not just put the computer in the adjacent different room and install a DisplayPort and USB patch panel on each side of the wall? Monitor and peripherals in one room and the computer on the other side. It’s what I would do if I owned a house.
I've had much success with fanless industrial computers for certain applications in certain environments. Seeing someone do it manually though is super cool (especially the vram chip heatsinks... why have I never thought of that before?!)
Speaking of noisy computers and electronics, I designed a PCB with an ESP32 bluetooth/WiFi chip on it. I'm far from an expert in board design, and my board unsurprisingly emits a buzz when the Wifi/BT radio is transmitting or receiving. It's interesting though, when it's using BLE you can hear the frequency hopping as it changes channels, and then you get entirely different sounds when it stops BLE and uses just WiFi. But yeah I should probably fix that...
Possibly caused by the piezoelectric effect on an SMD ceramic capacitor, making it vibrate and hit the PCB surface at an audible frequency.
Interesting project. I was playing with the idea of running Linux on one of those little Chinese Qotom PCs without fan for a server, but I live in Portugal where it can get very hot in the summer and decided that the potential fire hazard is not worth it.
Overclocking, fanless PC, etc. are all fine if you can make sure that there is no problem if the ambient temperature raises to 40 C - or even higher, if the machine is accidentally exposed to sunlight. I don't trust most existing fanless solutions to work reliably under such conditions.
If you're removing the GPU heatsink and fans, buying a dual-fan card isn't necessarily a dealbreaker depending on the design. For example, this 1050Ti from EVGA has two fans but the board itself is pretty small. https://imgur.com/a/AFvUlZs
Real late to this thread, but for the last ~10 years I've run my desktop in the adjacent room's closet. My desk has the monitors, speakers, USB hub for peripherals, with cables running through the wall to the actual tower/UPS/etc. It's a great setup if your living space can accommodate it (the adjacent room is my bedroom, but I shut this computer down at night).
I started doing this because quiet computers just weren't available at decent horsepower. Rad to see a build like this!
Apple’s iMac is an excellent budget alternative to this approach, as long as fan noise under high load is acceptable. 99% of my usage is not under high load, so it works out to be silent effectively any time I’m not wearing headphones. Using the CPU and/or GPU and/or writing gigabytes to the SSD will wind up the fans, so this is not even remotely “always silent for power users” – but it’s always silent for general non-coder usage!
I won a silent computer 5+ years ago as a prize at a lan party. SSD drives, two of the best passively coolable gpu at the time (ATI I Believe?), passive CPU cooler, and the whole case was lined with musicians sound absorbing foam. I think it did have an optional 240mm fan but I never heard it even when I turned it on. It was so nice that I have built my newest computer in the same way (though I did put some big storage drives in it).
I did a similar (but much more ugly) project a few years ago, and found that the result was that one could then hear various electrical crackling sounds from inductors transformers and capacitors, especially from the power supply and motherboard. It actually sounded worse than a macbook pro which didn't make as much electrical noise.
I've been interested in passively cooled PCs for a while now. Particularly, I've been considering building a computer which outputs its heat into a wall using heatpipes and a metal plate heat spreader. Although I can't find any case of this being done by anyone else, so I'm not sure if it's a crazy idea or not.
Since he described how his cpu cooling is capable of dissipating up to 105W and he uses a 65W cpu, he can overclock a little. He is also using fast memory and a 1050ti is quite capable of running pubg well. Based on that pubg should run well.
>[it] doesn’t have a single fan. Indeed, it doesn’t have any moving parts at all. It’s totally silent — 0dB.
Things without moving parts emit sound. LED lights are a simple example. Anytime you have a wave -- be it mechanical or electrical -- you'll have some of the harmonics will 'bleed' into the acoustic spectrum.
Built a for normal desktop use a nearly silent gaming MiniITX build, it hums from the graphics card when gaming. Fractal Design Node 304, ASUS Z170I Pro Gaming, Intel i7 6700k, Samsung M2 SSD, ASUS Strix Gaming OC Nvidia GeForce GTX 1070, Noctua NH-U12S CPU cooler, Corsair PowerSupply (PSU) CS550M. Runs Linux well.
I don't have a desktop PC right now but if I needed one I'd just place it in the basement or a closet (with enough ventilation), and wire the screen, HIDs, and a few USB connectors to my desk. Bam, no noise. However, it's cool (indeed) to see these silent constructions, though.
My Mac mini does not have a hard drive in it. It is completely silent most of the time. (A fan is running, but I cannot hear it even in the middle of the night.) When building software or playing certain videos, the fan speed increases with the result that I hear a quiet white noise.
I'm actually on the lookout for a wired, backlit, white/aluminium, chiclet, TKL keyboard. Basically something like the Apple Keyboard but sans NumKeypad, backlit and keyed for Linux/PC. Bonus points if it has a column of macro keys on the left.
This is a somewhat deep rabbit hole, but what I will tell you is that you'll probably have a hard time finding what you want in a chiclet keyboard because there's not much of a market for chiclet keyboards among keyboard enthusiasts. Chiclet is both a switching mechanism and a keycap; most of the keyboard-enthusiast community is working with setups that involve separate switches and keycaps that are interchangeable.
What you probably want in practice is the following characteristics: a specific key layout, all of the keys should be the same height, the key travel should be low (i.e., the amount of distance you have to depress the keys before registering a keystroke). No idea if you want the keys to make noise or not, or how stiff you want the springs to be, but that's customizable too. If you want chiclet keys because you want low key travel, you have a lot of options. If you want something really thin, there are way, way fewer options.
A single left column of macro keys is not something I've ever seen, but the one keyboard I can think off of hand that has a similar layout is the Red Scarf II, which has two columns of macro keys on the left side; unfortunately it's not currently in production. Some people use an external num pad and put the num pad on the left for this purpose. https://www.massdrop.com/buy/red-scarf-ii-ver-b-custom-mecha...
I personally use a KBD75, which is an aluminum body, tenkeyless layout with a right-hand column, and it's all programmable with QMK so you can make any key do whatever you want. For example, I have a key that, when I press it, reverses the position of the Alt and Win keys, so that I can switch between layouts for either PC or Mac and have the Alt, Command, and Windows keys always in their correct location. Images and build information here: https://imgur.com/a/5pSva2A
Agreed, it's a very deep rabbit hole. The 12" PowerBook G4 that I bought back in 2004 had the best keyboard I've ever used — I still have it and occasionally use it. But I've been trying to hunt down a desktop equivalent ever since (but with backlighting and TKL ANSI layout). Haven't succeeded so far. Still trying though.
Although my highest priority (by far) is low noise, I do prefer short travel/registration and a soft landing as well.
Meanwhile, now that Corsair's exclusive rights to the Cherry MX Silent (Red) have expired, I'm hoping to see a lot more keyboards with that switch make their way into the market. The Corsair Strafe RGB was tempting, but the light spill from under the caps was far too gaudy and distracting for my liking.
Cherry MX Silent (Red) switches with landing pads and o-rings seems like a combination that would work well (for me) for gaming.
For the DB4 and daily driver usage, though, the hunt goes on. I'll keep using the wired Apple Keyboard until someone else comes up with a TKL clone.
3M make a non conductive solution for exactly this application. I remember reading about an overclocking rig using something like it pumped through a radiator submerged in liquid nitrogen. Was pretty cool but the solution ended up turning into a gel at low temps
There are several. But the liquid will heat up and eventually require cooling again. Pumping it somewhere generates noise. Evaporating it creates costs and need to frequent maintenance. And a large enough tank for that not to be an issue is impractical.
The Australian state I live in has the highest electricity prices in the country, and amongst the highest in the world — so there's a simple financial reason to defect from the grid. Then there are the self-sufficiency angles (produce/control your own power) and the preparedness angles (grid power always fails in bushfires) to consider as well. They are all the reasons I need.
Interesting, I've been looking for a small computer that would make a good HTPC, and I thought the Skull Canyon NUC might be a good choice. I sort of assumed that that form factor would always be fanless.
Sometimes I can "hear" my processor working through my headphones. I'm not sure how, but, for example, when I scroll down a page I can hear high frequency chirping in my headphones which stops as soon as I stop scrolling.
It is caused by the increase/decrease in power consuption, as in general the audio chip isn't properly isolated from the power source of your computer. So, as more power is used (by slightly increasing the CPU speed, for example, to render the page as you scrooled it), the audio chip amplifies it.
It is a problem of some motherboards, and one way to get ride of it is to use another audio chip (for example, a small USB device that acts like a sound card).
...which is still a "USB sound card", no matter how you choose to look at it. It still needs to be a combined DAC and headphone amplifier in addition to being a set of 'phones.
Personally, I use an Audioquest Dragonfly USB DAC/headphone amp, which gives me the option to use either headphones (of my choosing) or speakers. (And unlike a larger "breakout box" USB sound card, it's quite portable, being about the same size and shape as USB thumb drives used to be.) It's a good fit for me, but if you need mic input, there are other devices - or you can still use the mic in of your regular sound card/integrated audio.
I can also "hear" something when I scroll on some computers, but I hear them with my ears, no headphones required.
On one particular susceptible computer I did a bit of testing, and scrolling white background pages made the noise, but black background ones didn't! Scrolling a completely white page also made the noise. And it depended on app, some apps were immune, it seemed to only happen on GPU-accelerated apps, like browsers.
I used to be able to hear data going over the serial port back in the modem days. Some systems used to make a buzz when scrolling too. These days I have not heard that, but I'm not sure if it's because the frequencies have gone up (outside of my hearing range) or the parts have better isolation or I'm just going deaf as I get old.
I don't think I've heard scrolling noise since I stopped using analog video signals.
These phenomena are more typically caused by ground loops and shield currents in the computer itself. Reordering plug-in cards (e.g. moving a sound card to a different slot) or isolating/removing the slot bracket can help.
Back when I was making electronic music, I noticed that I could hear pixels. When I did things that changed a region of the screen — pulling down a menu, switching windows, etc. — an audible whine would change on my audio monitors.
This was back in the CRT days, so I'm not sure if it's still an issue today, but it took me forever to figure out what was going on and drove me crazy when I noticed it.
I do find that generally USB headphones do better and minimizing interference/noise. Which makes sense, having a sound card on your motherboard is pretty much a worst case scenario. With USB it's not till outside the case that you are susceptible to RF.
What I remember finding most annoying was when you can hear the same (IIRC, actually the inverse) through your monitor.. as in it would whine ever so slightly unless you were scrolling. (NB no speakers involved)
I built a fanless (technically had fans, but only turned on if needed), semi-high-performance open bench-style system several years ago.
It's really pretty nice in theory. The reality is that some components still emit a good bit of noise, and unlike fans, the noise they tend to emit isn't very pleasant. Power-hungry graphics cards are by far some of the worst offenders; without being masked by fan noise, they can be incredibly loud and irritating. An open bench PC is definitely cool, but was kind of a mistake in my case.
Fan tech has improved to the point that I think, more often than not, you really want to have them. Otherwise it's time spent trying to eliminate other noise sources that would otherwise be conveniently and usually pleasantly masked by a little wooshing.
Completely agreed. When GPUs do this, I just declare them defective and send them back. It comes down to a lack of quality of the electrical components and simply shouldn't be a problem, especially with the current GPU prices.
I am Andrew Alan a hacker, i can help gain acccess to any social media. Generating all login informations for you. This can be useful in divorce case, court case and many more. I also make alterations to credit score, have a lot of funding programs ranging from accounts and credit cards. This and many more services i have to offer. Anyone interested in this service can text us on 3238351741, andrewalanhacks@gmailcom.
The obsessive pursuit of anything is usually interesting, but I haven't felt like noise was a real problem with any computer I've had in the last 5 or so years. They're silent nearly all the time, and they're still quiet the rest of the time.