4 ms·
`sudo powermetrics` will tell you die temperatures. It's possible the author has a bad sensor, in which case the fans will ramp to max whenever the machine is n
by ridiculous_fish 7y ago
`sudo powermetrics` will tell you die temperatures. It's possible the author has a bad sensor, in which case the fans will ramp to max whenever the machine is not asleep.
FWIW I run the same machine (except 32 GB) and a 4k external display. It's silent when idle (~45C) and quiet even when loaded (~97C).
- SavageBeast 7y agoTHANKS! Didn't know that existed!
- O5vYtytb 7y ago> quiet even when loaded (~97C). Yikes, that seems like an issue on its own. It's generally not good for parts to get that hot regularly.
- ztjio 7y agoModern intel parts run that hot no problem, what is normal for temps changes wildly over time across product lines. The current gen CPUs will self throttle at 100C.
- bscphil 7y agoOn the other hand, the fact that the user upthread's computer is quiet even at 97C strongly implies that throttling is happening.
- hnuser123456 7y agoApparently, intel accepted that laptops will just never have decent cooling, so they let the high power chips get right up to 100C and throttle based on how much cooling is available. If your laptop CPU is at 100C and the fans are quiet, you're wasting a lot of performance potential.
- eptcyka 7y agoThe chips themselves will probably outlast their users or at least their own useful lives quite comfortably at 100°C. If sat on a desk, not in something mobile. Heat cycling the motherboards will definitely shorten their effective lifetime significantly. There's no reason not to have proper cooling. Having said that, modern Thinkpads are not great in this regard either.
- BossingAround 7y agoYea no, Apple would rather run hot things quietly... That has been the case ever since retina MBPs came out.
- drunken-serval 7y agoMy 2008 MacBook Pro ran at those temperatures for many years just fine.
- bayindirh 7y agoRunning close to TMax values for extended periods of time is not a big problem. Servers in our HPC cluster runs with similar temperatures even with enough cooling and we change CPUs very very rarely.
- heavenlyblue 7y agoWhy not?
- t34m90fmei90w 7y agoComputers are not humans. 100C is just fine, lots of people are running their bitcoin miners for years on that temp with no issue.
- woadwarrior01 7y agoThanks for the tip. I hadn't known about powermetrics before. I've had a 16" MBP with the same specs as OP for almost 2 months now, and I use it with a 4K external monitor, and haven't noticed the any issues with fan yet. **** SMC sensors **** CPU Thermal level: 47 GPU Thermal level: 0 IO Thermal level: 0 Fan: 1828.89 rpm CPU die temperature: 64.80 C GPU die temperature: 68.00 C CPU Plimit: 0.00 GPU Plimit (Int): 0.00 Number of prochots: 0 Incidentally, the `Number of prochots` field in command's output made me chuckle. At one of my previous jobs, I spent some time on a team that owned the fleet wide server auto-remediation service and the prochot alarm and the resulting CPU throttling would end up reducing the available CPU capacity in clusters and sometimes knocking out entire services. I have bittersweet memories spending sleepless nights trying to debug it.
- jimsmart 7y agoIntel Power Gadget might also prove handy here (assuming it runs on the newer MBPs?) - it charts some useful power/performance related stats, including temp/power/speed, making it clear to see when/if the system is throttling/boosting/etc. https://software.intel.com/en-us/articles/intel-power-gadget https://software.intel.com/en-us/articles/intel-power-gadget