Safe Extended-Hours PC Operating Temperatures?

[jiyeon]

Hi, so I recently got into Folding@home and the biggest requirement that folding entails is that you need to leave your PC on to run its automated work units, meaning you could be leaving your PC on to do its thing for hours.

 

My Folding@home setup is my i5-8600K clocked at 4.6GHz 1.15V and my GTX 1060 3GB, my CPU is cooled by my Arctic Freezer 33 eSports One, and the whole case is cooled by four Corsair SP120 RGB case fans flanking the perimeter.

• My 8600K will reach at a max of 60°C after five hours of constant folding, pegged at 4.6GHz at 100% utilisation.
• My GTX 1060 will reach at a max of 70°C after five hours of constant folding at the same time as my CPU.

 

I was wondering if there's any side effects that I should know about when I'm leaving my PC on for hours on end? Such as if my case fans are being worn faster than they would be if I didn't leave my PC on specfically to fold, or my cooler being slowly worn down at the heatpipes due to the constant barrage of heat dissipation.

 

An average automated, independent, unsupervised folding session is around 5 hours. I usually leave the folding client to run before I leave home for work, and finish whatever work unit my PC is working on when I arrive back home. if I'm feeliing good about my PC, I'll run the program while I'm downstairs for dinner sometimes.

 

So, I have some final questions:

1. Ultimately, are there any side effects that would negatively affect my PC from these long operating hours?
2. For the Folding folk: Would lowering my 8600K clock down to 4.4GHz do much in terms of how efficiently I get points and complete work units?
3. Would you say my temperatures are safe?

[Firewrath9]

anything under 80c for cpu and GPU is fine. lowering your clock might reduce temps, but minimal perf as GPUs can compute much more. I'm Thinking of throwing together a 400$ F@H build/server. If you have good case fans (Ball bearing or FDB) you should be fine. My friend has 15 arctic f14 fans for his mining rigs and they all still work after hundreds of hours non-stop.

[WoodenMarker]

Heatpipes are inherently robust and don't wear down over time. Fans wear down with use and more so with higher rpm but manufacturers list MTTF and MTBF accounting for fans at 100% rpm. The fans you've mentioned aren't bad and should last many years of constant use. 

60c is low for the cpu and higher temps up to 80c shouldn't be any issue. Thermal cycling is usually the bigger problem.

70c for the gpu isn't high but it isn't low either. Lower would be better still but it should be fine as is.

Your temps are safe.

 

Moved to Folding@home, Boinc, and Coin Mining.

[Gorgon]

At 60C Pascal will start to slowly drop the clock down by 13MHz steps as you approach 70C and faster between 70C and 80C due to Turbo Boost 3. You can add a Overclock to offset this at the expense of more heat but you are then in a situation of diminishing returns. More overclock means more heat and the GPUs are much less efficient at folding looking purely at PPD/W.

 

This is offset somewhat by the Quick Return Bonus (QRB) where Stanford rewards you more non linearly for faster completion.

 

I would suggest that if your 1060 is running at 70C you might need more airflow. Miners have found that 70C is safe but 60C to 70C is safer. The GPU fan speed should also be monitored. If you have to run the GPU fan at 100% 24x7 it will likely fail faster than running it at 70% which will, in turn likely fail faster than 60% etc.

 

I’m of the opinion that running a folding rig 24x7 ends up being purely a thermal balancing act. You want a chassis with good airflow to remove the heat from the GPU(s). The GPU(s) likely should be kept under 70C if possible, definetly under 80C and the GPU fans in the 40-70% range.

 

The peak and quiescent system power draws should be kept in the 30-70% range of the Power Supply and moving from a Bronze (80%) to Gold (90-92%) power supply can pay for itself quickly.

 

Folding can be done anywhere between short, “Full-Tilt” spurts or a sustained “Slow and Steady” fashion. Like life this often comes down to personal preference at the time at hand.

 

The first approach runs the GPUs at maximum TDP and the second at or near peak efficiency for PPD/W.

 

System Noise can be an issue. I prefer many larger fans running more slowly than fewer smaller fans running faster. Tweak your fan curve by finding the best sensor to drive it and then the optimal speeds for the range of airflow needed between the “Slow and Steady” and “Full-Tilt” states.

 

I’m quite literally a Fan Boy and unabashedly fond of, perphaps overly so, of 140mm Noctua iPPC 2000 and 3000 PWM models. I mean there’s sexy German engineering but Austrians seem to just take it to the next level.

 

YMMV. Keep a journal and document what you tried and how it worked.

[wONKEyeYEs]

On 10/24/2018 at 7:13 PM, seoz said:

GTX 1060 will reach at a max of 70°C

I underclocked my GPU so it stays between 60C and 65C.

[jiyeon]

2 hours ago, wONKEyeYEs said:

I underclocked my GPU so it stays between 60C and 65C.

That would surely affect the speed of progress that my GPU would make when Folding, so I'm not up for that, but I assume it's just done with MSI Afterburner?

[GOTSpectrum]

4 hours ago, seoz said:

That would surely affect the speed of progress that my GPU would make when Folding, so I'm not up for that, but I assume it's just done with MSI Afterburner?

You could try reducing the power limit and voltage (if you can reduce voltage on your GPU) to reduce temps. Also making a custom fan curve will help.