Who's temperatures should i trust?

I

illuminate

[H]ard|Gawd
Joined
Nov 24, 2007
Messages
1,895
alrighty, i've recently switched to linux, from vista home premium 32bit.

In vista, speedfan and coretemp said that my 1.8ghz dual core processor was idling at 47-50c with the stock heatsink. No overclocks or anything. I knew that was high, so i was gonna go get a new HS. Then, my convert to linux happened (and its amazing!) and, linux is saying that my processor idles at 29-31c. That, i think, should be more accurate for such a low clock processor.

So, are the linux temps right? Or the vista ones?
 
Best ones to rely on are the ones reported in BIOS setup. Those will indicate which third-party software installation is reporting accurately.
 
ok. If i got by my BIOS, the linux temps are correct. shew, thats good to know..
 
Catweazle said:
Best ones to rely on are the ones reported in BIOS setup. Those will indicate which third-party software installation is reporting accurately.
Click to expand...

I have *never* seen anyone recommend that. I've always seen (as well as used) third party tools to monitor the temperatures. The BIOS ones are usually some of the most inaccurate readings, and many applications just read the BIOS sensor.

I'd recommend giving coretemp a try. From my understanding, it reads the temp from the processor instead of the BIOS sensor.


On a side note, just because a computer runs at a "lower speed" (1.8ghz instead say 2.5ghz), it doesn't mean it's going to run cooler. There are many other factors that come into play.
 
Demon10000 said:
I have *never* seen anyone recommend that. I've always seen (as well as used) third party tools to monitor the temperatures. The BIOS ones are usually some of the most inaccurate readings, and many applications just read the BIOS sensor.
Click to expand...

Fella is asking which temps are 'right'. The ones reported by tools under vista or the ones reported by tools under Linux. I told him to go check BIOS reported temps. What's so amazing or alarming about that?

Yes, BIOS reported temps can be 'off'. Sometimes, and on some mobos, by as much as 10 degrees, actually. So what? Software tools can be (and often are) 'off' too. Witness the bloody topic question!


Two out of three tools concur with a reading, and you got a reasonable baseline to operate from. Temp monitoring on a 'pooter is a relative process rather than an absolute one anyways. Only a nong would think that there's a 'safe limit' you can push the rig to and keep it there. Idea is to get it running as cool as you can, and then monitor the thing for sharp and/or sudden unexplained increases, cause those are the signs of trouble. See a hoik in temp level which is unexplained by weather or operating load? Check what's going wrong! Simple, eh?


If the thing is being used as a 'pooter it'll get unstable before it blows up, so that risk shouldn't really be an issue. If it's being used as a toy, with heaps of volts pumped unto it for a big overclock, it'll quite likely get unstable and then blow up really soon after, but then if you're doing that sort of thing you mustn't really give a fuck if it blows up.

But most of the time all you're really doing is monitoring for rises so you get a warning to go check stuff. For that you need a baseline reading, and it doesn't really matter if that reading is a completely accurate one or not. as said, it's 'relative'.
 
Catweazle said:
Fella is asking which temps are 'right'. The ones reported by tools under vista or the ones reported by tools under Linux. I told him to go check BIOS reported temps. What's so amazing or alarming about that?

Yes, BIOS reported temps can be 'off'. Sometimes, and on some mobos, by as much as 10 degrees, actually. So what? Software tools can be (and often are) 'off' too. Witness the bloody topic question!

*snip*
Click to expand...

So much wrong information here. Geez. The BIOS reads temps from an onboard temp sensor. This is completely inaccurate and 95% of the time reports numbers that do not represent the actual state of the CPU. I'm not familiar with Linux temperature monitoring, but I would guess it's probably the same deal as with the BIOS for general compatibility (only newer Intel CPUs have DTSes). Utilities like Core Temp and Real Temp on the other hand pull the temperature readings as distance to Tjmax from the DTSes embedded into the actual CPU cores. Although the DTSes are not completely accurate as the temperature gets farther away from Tjmax, the readings are still much, much more realistic than the onboard mobo sensors, and if any values are to be relied on, these are the closest you can possibly get to the real thing short of drilling a channel into the IHS and mounting a thermal probe in it, like Intel and the [H] do in their testing.

In short, Core Temp and Real Temp give realistic readings. The BIOS and Linux monitoring tools are very inaccurate and should be ignored completely.
 
Zero82z said:
So much wrong information here.
Click to expand...

You say that as if everything I've mentioned is incorrect or unreliable information, yet the only genuine point of contention you've raised is the claim that internal chip sensor is more accurate that the sensor used for BIOS reporting.

Even if it is true that the internal sensor gives a 'more accurate' numerical reading (hell, even if you accept that the thing is somehow divinely inspired and gives 100% accuracy of numerical reading, which I don't) the fact is that your comments miss the point of what I was saying, 'Realistic reading' doesn't really mean diddley-squat, because it's not really even useful.

Reason being that the CPU a person has isn't necessarily gonna run at precisely the same temps as somebody else's. So if a person's baseline temp reading is ballpark accurate that's all a person needs. Temp readings are damn near meaningless in comparison with the reading somebody else gets. Temp readings are on genuinely meaningful in comparison with the reading you saw, last time you looked, on that same machine and using the same monitoring tool. And it doesn't really matter a fuck if that tool provides 100% accurate baseline temps or not, because all you're really looking for, if you want the exercise to remain meaningful, is an unexplained rise in temperature!

  • Not all CPUs of the same model run at the same temperature.
  • Not all people using the same model of CPU have them operating in the same environment.
  • Not all people using the same model of CPU have them stressed to the same extent.
  • Not all CPUs of the same model will become unstable at the same operating temperature.
  • etc etc etc....

All of that is given 'fact'. Always has been and always will be. Same principle applies, by the way, for GPUs, PPUs and whatever-else-PUs you could mention.
 
Demon1000- I dont know if you missed it, but i said that i use coretemp, and speedfan. They both gave me the same temperatures. But, my BIOS temps, and my temps in linux are cooler then what speedfan and coretemp say (in vista).

also, i know it depends on the athmospher, but, my PC is sitting in the same spot where i took the linux temps, same day, same... sky outside. i made sure that both CPU's were on ideling. and, Linux matches the BIOS, and windows and speedfan/coretemp were the same.

I dont know if it matters, but when i first built my PC, my temps were around...24-25 but that was also the winter, and my PC is right by a glass door. I dont really know if that matters at all, but, theres the info.
 
Under Linux, I use gkrellm to show temps with sensors as the reporting software. On Intel CPUs I have found that this is usually pretty close to the temps as reported by numerous programs in Windows. In my case, I'm getting temps for 3 overclocked Q6600s.

Athlon X2s (AM2) and temperature readings are a different story. I have an X2 4000+ running 3.0Ghz with an Arctic Cooling Freezer 64 Pro and almost all temp reporting tools like to say it runs around 40C under full load. It's obvious that is not correct whatsoever.

I'd say that if you are using sensors to detect the temps, you're probably close to the real temps. It could be reporting a little low, but that's a gamble you take with any temperature reporting software.

That said, I would still ditch the stock heatsink and fan and get a decent aftermarket one. You might look into getting a stock heatsink from a Q6600 if you know someone that has one. That should cool your processor just fine and will be good enough to allow a moderate overclock.

 
illuminate said:
Demon1000- I dont know if you missed it, but i said that i use coretemp, and speedfan. They both gave me the same temperatures. But, my BIOS temps, and my temps in linux are cooler then what speedfan and coretemp say (in vista).
Click to expand...

I did miss that -- my apologies. But my point still stands -- don't solely rely on the BIOS temp. Use some other form to validate it. According to the above responses, a 10 degree variation in temperature is acceptable. If you're OK with that (I'm not), then the BIOS is fine.

I dont know if it matters, but when i first built my PC, my temps were around...24-25 but that was also the winter, and my PC is right by a glass door. I dont really know if that matters at all, but, theres the info.
Click to expand...

It's very possible that some driver or other program slowed down you fans which caused the temperature to rise to a higher, yet acceptable rate.



e-peens aside, I was just trying to advise that the BIOS is not an acceptable tool for measuring the temperature of your processor. I wasn't trying to show anyone up or anything like that. You seem to have your answer, and lots of advise to go along with it.
 
alrighty, thanks for the replies.

I still think that im gonna pick up a new HS. Just, so even if the vista temps are right (which, when its on load with vista temps, hits around 60c) it'll stay cool. and, it also adds to the inner PC coolness scale :p

Demon1000- im not ok with 10 degree varying temps... so, ill try what i can to figure out the right temps
 
Catweazle said:
You say that as if everything I've mentioned is incorrect or unreliable information, yet the only genuine point of contention you've raised is the claim that internal chip sensor is more accurate that the sensor used for BIOS reporting.

Even if it is true that the internal sensor gives a 'more accurate' numerical reading (hell, even if you accept that the thing is somehow divinely inspired and gives 100% accuracy of numerical reading, which I don't) the fact is that your comments miss the point of what I was saying, 'Realistic reading' doesn't really mean diddley-squat, because it's not really even useful.
Click to expand...

Actually, I did read the rest of your post, and I was referring to the entire thing with my last comment. The fact that you say absolute temperatures are insignificant doesn't make much sense. Though it's true that one person's CPU likely will have some variation in terms of temperature compared to a different chip, there are still certain temperature levels that people should not exceed when trying to achieve a fully stable system. For example, many modern Intel chips experience stability issues when exceeding ~70C, which is why it's frequently recommended that C2Ds and C2Qs be kept under ~60-65C to be on the safe side. Maybe for people who run their computers at stock all the time temperatures aren't much of a concern; however, this is an enthusiast site populated with many overclockers, and for an overclocker, it is very important to monitor temps in order to obtain the best possible level of stability. That's why programs like Core Temp and Real Temp are extremely important for obtaining the most accurate readings possible.
 
You must log in or register to reply here.
Back
Top