lm_sensors cpu temp v.slightly different to MOBO temp on Intel D865GLC
 

Hi,

I am using lm_sensors on F10 and everything seems bang on accurate except for the CPU temperature. Where in the BIOS it reads as 54 - 56 C, lm_sensors is telling me that it is 46 - 47 C. I am surprised though as sensors gets everything else right, even the CPU heatsink fan speed! I hope someone can explain this and its a simple config issue.

I am running and Intel P4 2.8GHz 533MHz Bus with and Intel D865GLC Mobo and F10.

p.s. I looked high and low for a similar thread before posting this but I couldn't find one.

The CPU temp will show higher in the BIOS display if the BIOS is programmed to loop waiting for input. This information from the lmsensors documentation (sorry -- can't find the page).

Which is more accurate in that case? I'm inclined to believe my BIOS over sensors but it would be nice to monitor the temp from the OS. I just worry that its not telling me how hot it actually is.

The reason im so worried about it is I am trying to build a quiet box and I am tweaking the fan speeds so they run at optimum. It would be nice to see the true temp in OS so I know how my system is running. I have my conky liked in to sensors so I can monitor performance and temp. Is there a better way of doing this?

Both are accurate (assuming you have a good sensors3.conf file and the CPU is looping while displaying the BIOS menus); the CPU runs hotter when looping than in normal operation.

Do you have a sensors3.conf file for your specific motherboard?

I also have a quiet box, since January. The fans run about 50% of the time at 500 RPM or a little higher. 500 RPM is relatively fast (but almost inaudible) but they twitched and did not start reliably at lower minimum speeds. Ambient temperatures here peaked around 35 deg C. Max CPU casing temperature seen is 43 deg C.

Beware the motherboard chips are normally cooled by CPU cooling air spill; after controlling the CPU cooler's (ASUS Triton 75's) fan, the Northbridge was too hot to touch so I glued a tiny (Rs.9 ~= GBP 0.13) fan to it which was the noisiest thing in the box -- despite running at lowered voltage via a FanMate -- until replaced by a passive cooler (Nexus 2200).

I use GKrellM, not conky, and haven't used conky so cannot say which is better; AFAIK they do pretty much the same job. It is prudent (and re-assuring!) to monitor temperatures when experimenting with lowered fan speeds.

I dont know if I have a sensors3.conf for my specific mobo. Everything seems to be detecting fine for my board. Voltages all look roughly the same and the CPU Fan speed looks spot on so I am assuming it is all set up correctly, although I am fairly new to the sensors package.

I set it up from various tutorials on the web. It was pre-installed so to configure I ran a sensors-detect followed by a pwmconfig. This seems to have done the trick as, like I say, all of the values look ok. Is there anything u think I might be missing?

I have managed to get the system running pretty quiet while still retaining a decent amount of cooling (if sensors are accurate running less than 50C while in OS). I am thinking of trying for a bit more hush by mounting the two case fans (1x120mm and 1x140mm) with silicone mounts, installing 2x Nexus D12SL-12 (recommended by silent PC review) in the front using silicone mounts, then run all chassis fans at a slightly lower set of revs and sit back and admire the lack of noise lol.

I am hoping this will work ok but thats why I need to ensure that I am getting an accurate readout on the temp. I wana try and keep her below 55C if at all possible.

The trickiest parts of sensors3.conf are usually voltages rather than temperatures so your auto-generated sensors3.conf may well be OK for the CPU casing temperature. See this LQ post for a list of places to look for sensors3.conf files for various motherboards.

You may be able to get correlating information from the CPU core temperatures. What is your CPU? At least some AMD dual core models give unreliable data so you may need to be specific about the actual CPU model, including its "stepping" number.

Sounds as if you have gone a long way down the road of quietening fans. How much of the residual noise is caused by the PSU fan? If that is the main source of residual noise then your unusually elaborate case fan ideas will not solve the problem.

Regards case fans, I removed one leaving only one (in the top chamber -- I use an Antec P182 case) and drive this at the same speed as the CPU fan reasoning that the CPU produces most of the heat. If you use this technique in a single chamber case you may like to monitor the HDD temperature carefully during heavy disk I/O and low CPU load.

Silent PC Review is a great source of quietening information :)