Sensors temperature readings.

[cizzi]

I just wanted to know if these values are normal on my quad core 2.4 intel cpu.

coretemp-isa-0000
Adapter: ISA adapter
Core 0: +37?C (high = +100?C)

coretemp-isa-0001
Adapter: ISA adapter
Core 1: +35?C (high = +100?C)

coretemp-isa-0002
Adapter: ISA adapter
Core 2: +33?C (high = +100?C)

coretemp-isa-0003
Adapter: ISA adapter
Core 3: +36?C (high = +100?C)

Thanks.

[mlau]

They should be since they come directly from the CPU.

[Swoopy]

Well, they're well above room temperature (unless you like it really hot) so they're not reporting absurdly low.
These temperatures would seem okay to me for an idle system with cpu frequency & power management working.

[Cpt-Klink]

no thoes built in sencerors are usaully crap and alot of the time way off from what it actully is

[technolope]

How good are temp readings? How hot is too hot?

How unreliable are the reported temperatures? And do the motherboard heat-safety cutoffs use the reported temperature or some other number?

I ask because I am seeing odd temperature readings from my Phenom 9500/Gigabyte MA790FX-DS5 system. Case temp is fine (never more than 47C), CPU temp is a little higher than I expected (standard heat sink, 43 idling, 55 under 4-thread load for a few minutes, 62 under 4-thread load for half an hour, then it crashes), and a third temperature which Linux/lm_sensors/GnomeSensorApplet reports as 86C, even just after the machine has booted.

Is the mobo causing this? Could Linux ber reading the data incorrectly?

[Swoopy]

How unreliable are the reported temperatures? And do the motherboard heat-safety cutoffs use the reported temperature or some other number?

I ask because I am seeing odd temperature readings from my Phenom 9500/Gigabyte MA790FX-DS5 system. Case temp is fine (never more than 47C), CPU temp is a little higher than I expected (standard heat sink, 43 idling, 55 under 4-thread load for a few minutes, 62 under 4-thread load for half an hour, then it crashes), and a third temperature which Linux/lm_sensors/GnomeSensorApplet reports as 86C, even just after the machine has booted.

Is the mobo causing this? Could Linux ber reading the data incorrectly?

If the server crashes after more than half an hour of full thread load (even if it is only at 62 degrees CPU temp) then you already have issues.

Regarding the VGA card cooling,
I've seen this with my neighbour's system which contains a GeForce 8600GT OC edition.
Turned out that when we closed the case, a cable got pushed by the lid so as to block the fan on that card from turning.
So make sure the fans on all the relevant components are turning properly once the system starts to warm up! (The Geforce card of my neighbour survived, barely)

[technolope]

If the server crashes after more than half an hour of full thread load (even if it is only at 62 degrees CPU temp) then you already have issues.

Regarding the VGA card cooling,
I've seen this with my neighbour's system which contains a GeForce 8600GT OC edition.
Turned out that when we closed the case, a cable got pushed by the lid so as to block the fan on that card from turning.
So make sure the fans on all the relevant components are turning properly once the system starts to warm up! (The Geforce card of my neighbour survived, barely)

I've got an older GT 7600 OC, but it never gets above 52C. I'm running with an open case, and all of the fans are turning. There are no major flow intrusions---just the regular cables, and the RAM sticks are somewhat close to the CPU.

Thanks for the sanity check on the CPU temperature. Does it point to the CPU or the mobo as the culprit?

[Swoopy]

I've got an older GT 7600 OC, but it never gets above 52C. I'm running with an open case, and all of the fans are turning. There are no major flow intrusions---just the regular cables, and the RAM sticks are somewhat close to the CPU.

Thanks for the sanity check on the CPU temperature. Does it point to the CPU or the mobo as the culprit?

Disclaimer: I am answering this based on limited information and without hands-on experience of the system.
While I used to be a professional hardware troubleshooter, my most exact answer to this question that I can give you is, right now the problem could still be caused by many things.

First off, the open case. Why?
Opening the case, depending on the design, can actually cause not enough air to flow from bottom-front (cold) to top-back (warm), thus limiting the amount of airflow by leaking what cooling is being provided back to the atmosphere.
Then again, opening the case can also be an improvement.
If you run the system with an open case and the ambient temp sensor on the mainboard lists 47 degrees C, then your system is in too warm an environment for it to run at the speed you're trying to run it I'm affraid.

Even with the sensor close to the CPU, 47 degrees is a lot for the ambient air temp around the mainboard temp sensor. If that's just a peak (maybe with the case closed?) and it averages around 40 degrees C, that would be more normal.

Case designs with an air funnel over the CPU portruding into the case from the removable lid improve upon this, and can improve CPU cooling by 5 degrees C or more.

62 degrees is quite low for a heavily loaded CPU. (I realise that many people could answer this statement by truly claiming that they never see their CPU top 55 degrees, or 51, whatever. With 4 cores on a chip blasting away, 62 degrees doesn't strike myself as very hot).
My own dual-core routinely runs around 65 degrees C under full load, sometimes for days, with no issues at all.
I am not familiar with the new quad-core but is it possible the system / mainboard front-side bus is clocked to a higher rate than specified by the manufacturer ("overclocked")? When overclocking, the crashes that tell you you're 'over the edge' coincide with higher temperatures.

Everything else being normal I wouldn't expect a CPU to cause a crash at 62 degrees C already, assuming that the sensor values are being reported correctly.
If you can touch the CPU cooler for more than 2 seconds without blistering your skin, I'd say the temp is still okay and being reported correctly. If it feels hot enough to fry an egg on, chances are it's actually warmer than 62 degrees.

Improving upon the CPU cooler is one tip that comes to mind.
Setting all mainboard BIOS values (back) to defaults is another.
If that doesn't help, try disabling cool'n'quiet in the BIOS power management settings and see if the system is more stable when you do that (in Windows XP, cool'n'quiet requires a driver by the way). Without CnQ, the CPU will possibly get slightly warmer.
If the mainboard is an (older) AM2 not specifically designed to run the Phenom processor, be wary of mainboard BIOS "auto-voltage" settings. Look up what they should be on the AMD website for your CPU type and set them in your BIOS manually according to that.
Oh, and make sure the cooler is properly attached, with no dirt between it and the CPU, and if thermal paste is between it and the CPU, make sure it's evenly spread so there is contact over the whole surface.
If the thermal paste that came with the cooler (usually a grey goo sticker) is damaged, it's better to remove it and buy a tube of new paste at a PC shop and spread that evenly on CPU or cooler.
Make sure the cooler settles flush with the surface of the CPU when placing it?

P.S. Try reading in your BIOS what the 86 degree temperature is labeled as.
Some mainboards have a sensor on the power modulator (PWM). If that is running hot (and it can get above 90 degrees on my system easily, but 86 is too hot for an idle system) chances are that you have one or more BIOS voltages set too high for the component they are feeding OR not enough heat can be dissipated.
Both may trigger a crash when the system is placed under load.
Typically the PWM circuitry sits (assuming a vertically placed mainboard) between the CPU and the back of the case, above the PCI cards and below the PSU.
Placing a large fan at the back of the system near this area, if the case has a grating there, can improve the health of the system immensely, especially if the PSU only has a small air inlet inside the PC case.

[technolope]

First off, thanks for the well-thought-out reply. I understand the limited nature of the information I gave, and I don't think anyone is going to slight you for misdiagnosis.

First off, the open case. Why?
Opening the case, depending on the design, can actually cause not enough air to flow from bottom-front (cold) to top-back (warm), thus limiting the amount of airflow by leaking what cooling is being provided back to the atmosphere.
Then again, opening the case can also be an improvement.
If you run the system with an open case and the ambient temp sensor on the mainboard lists 47 degrees C, then your system is in too warm an environment for it to run at the speed you're trying to run it I'm afraid.

The case is open because I am still working onthe machine. I was testing it with several drives disconnected. In this instance, closing the case led to 2-5C higher temperatures. It's 20-21C in my apartment, so it's hard to imagine the case temperature being 47. That's why I question the accuracy of these sensors.

62 degrees is quite low for a heavily loaded CPU. (I realise that many people could answer this statement by truly claiming that they never see their CPU top 55 degrees, or 51, whatever. With 4 cores on a chip blasting away, 62 degrees doesn't strike myself as very hot).
My own dual-core routinely runs around 65 degrees C under full load, sometimes for days, with no issues at all.

The CPU is rated at 95W, which seems average for performance CPUs, 4 cores or not. My older CPUs (Athlon XP 2 GHz, Sempron 2 GHz) both appear to run much cooler: 44C idle, 47-49C loaded (and both have smaller heatsinks and fans than the Phenom).

I have read tons of forum posts (here and elsewhere) and many hardware articles and newegg reviews, and the teperature ranges that people report are wild: 32C to mid-60s. Many claim that CPUs are only rated up to 70C, but don't back those claims up. It seems to me that 65C is too hot, even for a loaded CPU. I've decided to get an aftermarket cooler for the Phenom, regardless of the outcome of these tests.

I am not familiar with the new quad-core but is it possible the system / mainboard front-side bus is clocked to a higher rate than specified by the manufacturer ("overclocked")? When overclocking, the crashes that tell you you're 'over the edge' coincide with higher temperatures.

At first, I just used the motherboard BIOS defaults. The system was "unstable." Then, I specifically set them to what the Phenom specifications called for, and the behavior was the same. Then, I specifically downrated the speeds (bus and CPU) and it ran a little cooler---cool enough to not crash at all.

Everything else being normal I wouldn't expect a CPU to cause a crash at 62 degrees C already, assuming that the sensor values are being reported correctly.
If you can touch the CPU cooler for more than 2 seconds without blistering your skin, I'd say the temp is still okay and being reported correctly. If it feels hot enough to fry an egg on, chances are it's actually warmer than 62 degrees.

I could probably hold my fingers to it (the hottest part of it---the side at the base) for 20-30 seconds at full-load. If that part of the sink is 5-10C cooler than the CPU itself, then I could believe the 62C reading was correct.

Improving upon the CPU cooler is one tip that comes to mind.
Setting all mainboard BIOS values (back) to defaults is another.
If that doesn't help, try disabling cool'n'quiet in the BIOS power management settings and see if the system is more stable when you do that (in Windows XP, cool'n'quiet requires a driver by the way). Without CnQ, the CPU will possibly get slightly warmer.

I turned off the CnQ, and the CPU fan ran at 3000 rpm constantly, and he machine crashed just as before. With CnQ enabled, the fan sould start off at 2300 and increase to 4500, but the machine would still crash and restart.

If the mainboard is an (older) AM2 not specifically designed to run the Phenom processor, be wary of mainboard BIOS "auto-voltage" settings. Look up what they should be on the AMD website for your CPU type and set them in your BIOS manually according to that.

It's the Gigabyte AM2+ 790FX board, designed for the Phenom FX chips. It seemed to be a future-proof board, but at also appears to be so advanced that it has numerous small problems.

Oh, and make sure the cooler is properly attached, with no dirt between it and the CPU, and if thermal paste is between it and the CPU, make sure it's evenly spread so there is contact over the whole surface.
If the thermal paste that came with the cooler (usually a grey goo sticker) is damaged, it's better to remove it and buy a tube of new paste at a PC shop and spread that evenly on CPU or cooler.
Make sure the cooler settles flush with the surface of the CPU when placing it?

The Phenoms come with a heat sink that already has thermal compound on it---you just mount it in place. The installation was quite easy compared to my previous experience. Nevertheless, when I disassembled the machine to return it (I returned the mobo and CPU), the heat sink was very tightly attached to the CPU. I had to wait 10-20 minutes after loosening the atachments before I could muscle the heat sink off the CPU. The thermal paste was spread very evenly (and thin).

P.S. Try reading in your BIOS what the 86 degree temperature is labeled as.
Some mainboards have a sensor on the power modulator (PWM). If that is running hot (and it can get above 90 degrees on my system easily, but 86 is too hot for an idle system) chances are that you have one or more BIOS voltages set too high for the component they are feeding OR not enough heat can be dissipated.
Both may trigger a crash when the system is placed under load.
Typically the PWM circuitry sits (assuming a vertically placed mainboard) between the CPU and the back of the case, above the PCI cards and below the PSU.
Placing a large fan at the back of the system near this area, if the case has a grating there, can improve the health of the system immensely, especially if the PSU only has a small air inlet inside the PC case.

The BIOS gives no indication of what the 86C might be. I only get that reading through the Gnome Hardware Sensors applet. At first I thought that the 86 was an error, because it never changed. Then, I saw it move form 86 to 87 during one test. Then, when I pointed the 8" desk fan at the motherboard (case open), it dropped to 85. When I boot the machine cold and log in as fast as possible, I see it at 82C before it heats up to 85/86C after a few minutes.

On the gigabyte, the northbridge and the power circuits on the mobo are covered with a heat pipe and a series of fins. One set of fins is "south" of the CPU, one set of find is "west," but the CPU fan only blows air directly over the "south" set of fins. I felt those once, and they were hotter than the CPU heat sink. I suspect that you are right, and that the 86C is the reading on those components. In that case, I'd guess that having a CPU fan that funneled *more* air over *both* fins is a necessity. And the crashes are then probably due to that component overheating and triggering self-protection circuitry.

I returned the mobo for a replacement, and I returned the CPU for a refund (I will get the 9550 instead). When the new components arrive, I will test them with the same hardware setup and see what temperatures I get. Maybe the heat pipes on the Northbridge weren't installed properly. I will also get a better CPU fan.

Thanks for sharing your experience!