3 ms·

I have the following approach to fanless computing: 1) I buy any desktop I like for its form factor 2) I either use the onboard graphic or a fanless graphics

by TekMol 10y ago

I have the following approach to fanless computing: 1) I buy any desktop I like for its form factor 2) I either use the onboard graphic or a fanless graphics card 3) I set the threshold for all fans that are accessible via fancontrol to a level that keeps them permanently off 4) I pull or cut the cables of any fans that remain spinning People always tell me I am "insane" to do so. But I have been doing this for years now and never had a problem.

detaro 10y ago
It works in the sense that the hardware will survive quite a while, but you are likely giving away a lot of performance to thermal throttling (at least if you have a mini-PC, it can be factor 5 - 10 or so)
djsumdog 10y ago
Modern CPUs will just downclock themselves. I mean it's not insane; you won't damage the components, but you will be running A LOT slower than you could be. Do you run tools to see how far down your CPUs frequency scaling is taking it?
- pktgen 10y ago
  I was about to write the same thing. The CPU won't be damaged, but it may underperform. The bigger concern would be other components, especially motherboard VRMs which seemingly tend not to have any thermal failsafe (on occasion, there are reports of them catching on fire).
- TekMol 10y ago
  I would be surprised if any downscaling takes place. I run some heavy computation software from time to time and compare the performance to other computers. Many of them in datacenters. I would probably notice if a machine suddenly downscales. That said, I would be happy to measure it and post the results. Running the computations now. According to "top" it saturates 2 of my 8 CPUs. So far, "lscpu" reports the CPU at 1600 MHz. "sensors" reports the CPU Temperature is slowly rising. Was 48°C at IDLE. Is 50°C now. 15 minutes later. CPU temp was raising so slow that I stopped it now. Don't think it will get to anything critical ever. It was at 52°C. CPU speed was still at 1600MHz.
  - alphapapa 10y ago
    Isn't 1.6 GHz quite a low frequency? What CPU do you have?
- oakwhiz 10y ago
  A lot of CPUs will cycle-skip which causes worse performance than just manually reducing the multiplier. Especially for real-time tasks like audio or video.
csydas 10y ago
I wouldn't call it insane (except for the cable cutting, that seems like it's just unnecessary when you can just remove the component or likely modify with software), but I also think it's ill-advised for most use cases that aren't yours, which is likely why most people are saying it's nuts. A lot of modern computing is just greedy with resources, and will gladly eat up as much as is available, be it games or web-browsers. My first computer building experience I didn't fasten the heatsink to the mobo and it would lose it's contact with the CPU when the case was upright; powering the computer on when it was upright would allow it to load for ~ 10 seconds then shut down with no warning. CPUs and other high performance components hit their thermal limit fast, meaning they get really hot and stay really hot pretty fast without some mitigating factor. To me it seems illadvise to not give the components all the help you can to keep them cool. Granted, my first home built was with an e8400 wolfdale, and the power and thermal efficiency of CPUs has come a long way since then, but they still get toasty without a bit of assistance. Well planned passive cooling is great - my first GPU was passively cooled and it served me well for a good 5 years until a decent night of blackjack in Vegas afforded me a new GPU. But I could still see how it struggled when the card started to reach it's thermal limits. The card ended up getting donated to a friend of mine, and I think she's still using it successfully to this day. But that being said, the passive cooling on the card was nearly triple the size of the card itself.

by TekMol 10y ago

detaro 10y ago

It works in the sense that the hardware will survive quite a while, but you are likely giving away a lot of performance to thermal throttling (at least if you have a mini-PC, it can be factor 5 - 10 or so)

djsumdog 10y ago

Modern CPUs will just downclock themselves. I mean it's not insane; you won't damage the components, but you will be running A LOT slower than you could be. Do you run tools to see how far down your CPUs frequency scaling is taking it?

pktgen 10y ago
I was about to write the same thing. The CPU won't be damaged, but it may underperform. The bigger concern would be other components, especially motherboard VRMs which seemingly tend not to have any thermal failsafe (on occasion, there are reports of them catching on fire).
TekMol 10y ago
I would be surprised if any downscaling takes place. I run some heavy computation software from time to time and compare the performance to other computers. Many of them in datacenters. I would probably notice if a machine suddenly downscales. That said, I would be happy to measure it and post the results. Running the computations now. According to "top" it saturates 2 of my 8 CPUs. So far, "lscpu" reports the CPU at 1600 MHz. "sensors" reports the CPU Temperature is slowly rising. Was 48°C at IDLE. Is 50°C now. 15 minutes later. CPU temp was raising so slow that I stopped it now. Don't think it will get to anything critical ever. It was at 52°C. CPU speed was still at 1600MHz.
- alphapapa 10y ago
  Isn't 1.6 GHz quite a low frequency? What CPU do you have?
oakwhiz 10y ago
A lot of CPUs will cycle-skip which causes worse performance than just manually reducing the multiplier. Especially for real-time tasks like audio or video.

csydas 10y ago

I wouldn't call it insane (except for the cable cutting, that seems like it's just unnecessary when you can just remove the component or likely modify with software), but I also think it's ill-advised for most use cases that aren't yours, which is likely why most people are saying it's nuts. A lot of modern computing is just greedy with resources, and will gladly eat up as much as is available, be it games or web-browsers. My first computer building experience I didn't fasten the heatsink to the mobo and it would lose it's contact with the CPU when the case was upright; powering the computer on when it was upright would allow it to load for ~ 10 seconds then shut down with no warning. CPUs and other high performance components hit their thermal limit fast, meaning they get really hot and stay really hot pretty fast without some mitigating factor. To me it seems illadvise to not give the components all the help you can to keep them cool. Granted, my first home built was with an e8400 wolfdale, and the power and thermal efficiency of CPUs has come a long way since then, but they still get toasty without a bit of assistance. Well planned passive cooling is great - my first GPU was passively cooled and it served me well for a good 5 years until a decent night of blackjack in Vegas afforded me a new GPU. But I could still see how it struggled when the card started to reach it's thermal limits. The card ended up getting donated to a friend of mine, and I think she's still using it successfully to this day. But that being said, the passive cooling on the card was nearly triple the size of the card itself.