Need help determining if temp sensor is bad

[CitizenInsomniac]

I recently built a system with an AMD64 Venice 3500+ CPU on an Asus A8V-E Deluxe mobo with Zalman CNPS7000B-AlCu cooling (stock paste).
The temperatures reported have been nothing short of sensational. At 2640 MHz and 1.425V the CPU temp maxed at 59C, while at 1.5V the temp reached an amazing 63C. Through all of this the system remained stable and rock solid. This lead me to believe that something might be wrong with the temperature sensor on this motherboard (or the BIOS). So I did a very simple test:

I turned off the machine and let it cool down for about 1 hr. I don't have a termometer in my apartment but I would guess the room temp to be about 20C or less. I turned on my computer and went straight into the BIOS. The temperatures read:

CPU: 37C
MB: 32C

Is it possible that after only 15 secs of activity these would be the temperatures? I doubt it, and I need your help in confirming my doubts.

Can you please try doing the same thing? Turn off your machine, let it cool for 1hr or more, then turn it on and read the CPU/MB temp values straight from the BIOS. If my assumption is correct, I'm guessing for most people these will be in the low 20s.

 

[CitizenInsomniac]

This is the only post in the last 2 days that nobody has replied to. Doesn't anybody have the time to help out with some info?

 

[Borgschulze]

Try reseating the Zalman and use less thermalpaste, too much will make it hotter.

 

[CitizenInsomniac]

Thanks, I've already tried that - I even changed to a different paste - it made no difference.

Right now I'm chiefly interested in numbers other people are seeing upon cold bootup. If it turns out that everyone else is seeing numbers way lower than me when their machine is cold, that would be a sure indicator that my temperature sensors are bad.

 

[CitizenInsomniac]

(bump)

Anyone? Bueller?

 

[dr_dirtnap]

I'm not sure, but I think the BIOS places your CPU on constant load on your CPU or something like that. It's not like being idle in windows, where I get my lowest temperatures.

My venice 3800+ doesn't go over 10 degrees ambient temp with my XP-90C under full load, so something is seriously wrong. What does the bottom of the Zalman look like when you take it off? Is it making good contact?

I have a winnie 3500+ at 1.5v @ 2.6 with a XP-120, and full load temps reach 54c. Which I think is high, but I haven't had much luck with the XP-120's. I went from a SLK948u to the XP-120 and my temps went up. I would go back to the SLK, but I'm too lazy to my MB out again to change the backplate, and the 120mm fan is alot quieter pushing more air than the vantec tornado that I had on the SLK.

 

[CitizenInsomniac]

I'm not sure, but I think the BIOS places your CPU on constant load on your CPU or something like that. It's not like being idle in windows, where I get my lowest temperatures.

Right, but you're talking about the difference between BIOS readings and Windows readings after the machine has already been running for a while. In my test, I let the machine cool off for at least 1 hr. It only takes me about 15 seconds between turning on the machine and getting into the BIOS hardware monitor section. I don't believe that's enough time for the temperatures to rise 10 degrees Celsius or more.

My venice 3800+ doesn't go over 10 degrees ambient temp with my XP-90C under full load, so something is seriously wrong. What does the bottom of the Zalman look like when you take it off? Is it making good contact?

Yeah, the entire heatsink is making contact and it's firmly in place. Before I run off and buy a new heatsink, I'd like to make sure that my MB is not just reporting incorrect tempeatures. If you have time, can you check out your "cold" temperatures next time you turn on the machine after it's been off for a while? Thanks!

 

[zer0signal667]

Right, but you're talking about the difference between BIOS readings and Windows readings after the machine has already been running for a while. In my test, I let the machine cool off for at least 1 hr. It only takes me about 15 seconds between turning on the machine and getting into the BIOS hardware monitor section. I don't believe that's enough time for the temperatures to rise 10 degrees Celsius or more.

!

How long does it take a light bulb filament to start glowing once you give it power? There's a huge power density in that chip, it's not unreasonable to think that it can heat up thta quickly. However, I would still be skeptical of any readings given by the onboard sensors, there's no gaurantee on their accuracy.

 

[CitizenInsomniac]

Basic physics would seem to suggest that 15 seconds is not enough to raise the temperature by 10+ degrees by contact. Even a car engine takes minutes to heat up - or at least that's how long it takes for the sensors to register it.

But my point is exactly that the onboard sensors can be faulty. If they are faulty, it's likely they will be offset by a fixed value. We can discuss this endlessly, but what I'd really like to know is: what temperature values are people seeing upon cold boot-up before getting into Windows?

 

[RawsonDR]

Sorry, I don't have BIOS cold boot info.

But unfortunately I don't think your tests are conclusive to a faulty sensor. The fact is that it can raise that quickly. Idle your CPU in Windows and then load it. Mine, going from idle 1.3ghz to load and 2.6ghz (and raising voltage) can change almost 20 deg Celsius within 1 refresh of the CPU temp sensor. Yours should act the same. All A64's seem to act this way. Of course other chips, like AXP's, never did. It's a function of the chip and how the temp sensor relates to it.

Take the bulb example. If you turned on a cold light, you wouldn't feel significant heat on the glass for quite awhile. And likewise you wouldn't feel the bulb cool down for another few minutes after you turn it off. But if you have your fingers on the filament, you instantly feel heat when you turn it on and the pain almost instantly subsides when its off. I believe this relation is similar to the way A64 and AXP (and other) temp sensor methods relate to each other.

What's more is that your cold boot temp data was actually opposite I was looking/hoping for. I was half expecting to see temps like 50 or so, which would then indicate that maybe your temp sensors were just inaccurate. But your numbers did really make sense (your CPU at least). From 20 deg to 37 is definitely possible. As has been mentioned, your hardware is beyond idling when in BIOS. So those numbers do seem to fall in line where I would expect between 20 deg and 60 deg max.

Unfortunately at this point its hard to tell. Your numbers do still make sense. Its possible to be running at 60 deg celsius and still be stable. Nothing yet has called out to me that having bad CPU contact just can't be the case. So at this point I would assume that, although you did say you remounted multiple times. Start thinking harder about it... the base off the heatsink is flat, its mounted with enough pressure, there is not too much thermal grease (in fact I would suggest you check that one, try using very little if you haven't yet), etc etc. Can't think of any other good ones right now.

The only other thing I'd check is if other people with your motherboard have similar or normal temps... but I assume they do (having normal temps) or there would be more posts about it.

edit: Just wanted to come back and add that I'm not dismissing the faulty sensor. Could definitely be that, its just that if it is its hard to troubleshoot and I don't know what you could do to find out definitively.

 

[peterhoang2002]

on my motherboard, dfi lp ultra-d, temperatures can jump up really fast! monitoring it in windows with the included ite guardian program, it can be 35c idle in windows and right when prime95 torture test is started, the temperature can be 52c in a few seconds.

 

[CitizenInsomniac]

But unfortunately I don't think your tests are conclusive to a faulty sensor. The fact is that it can raise that quickly. Idle your CPU in Windows and then load it. Mine, going from idle 1.3ghz to load and 2.6ghz (and raising voltage) can change almost 20 deg Celsius within 1 refresh of the CPU temp sensor.

Doesn't that seem a bit extreme? On my system when I start Prime95 tests, the CPU temperature only goes up about 6C during the first 30 seconds. My voltage is fixed though.

Take the bulb example. If you turned on a cold light, you wouldn't feel significant heat on the glass for quite awhile. And likewise you wouldn't feel the bulb cool down for another few minutes after you turn it off. But if you have your fingers on the filament, you instantly feel heat when you turn it on and the pain almost instantly subsides when its off.

I was always under the impression that the sensor is on the motherboard, underneath or close to the socket. Are you saying that it's actually in the chip itself?

What's more is that your cold boot temp data was actually opposite I was looking/hoping for. I was half expecting to see temps like 50 or so, which would then indicate that maybe your temp sensors were just inaccurate. But your numbers did really make sense (your CPU at least). From 20 deg to 37 is definitely possible. As has been mentioned, your hardware is beyond idling when in BIOS. So those numbers do seem to fall in line where I would expect between 20 deg and 60 deg max.

Hmm, that made me think of something so I did another experiment. This time I used a stopwatch to measure the time between hitting the ON button and reading the CPU temperature from the BIOS. I measured roughly 20 seconds. After I recorded the first temperature, I recorded 8 more temperatures over the course of the next 120 seconds - from the BIOS. In other words, I took a CPU temp reading every 15 seconds.

Here's what I saw (0sec marks the first entry I read from the BIOS):

0sec ... 32C
15sec ... 35C
30sec ... 35C
45sec ... 36C
60sec ... 37C
75sec ... 38C
90sec ... 38C
105sec .. 39C
120sec .. 39C

So aside from the initial 3C jump over 15 seconds, the rate that followed was pretty constant - roughly 1C every half a minute.

BTW, I should note that my temperatures were much lower this time than at the time I wrote the original post because I did this one at night and the original in the middle of the day.

Unfortunately at this point its hard to tell. Your numbers do still make sense. Its possible to be running at 60 deg celsius and still be stable. Nothing yet has called out to me that having bad CPU contact just can't be the case. So at this point I would assume that, although you did say you remounted multiple times. Start thinking harder about it... the base off the heatsink is flat, its mounted with enough pressure, there is not too much thermal grease (in fact I would suggest you check that one, try using very little if you haven't yet), etc etc. Can't think of any other good ones right now.

I think even if there is an error offset, it's probably no more than 7-8C. It would be interesting to see people report boot-up temps more often, not just Prime95 max temps. The spread is perhaps even more interesting than the max values, I'd think.

 

[DamienThorn]

On my board, with a Zalman AlCu7000 I hit around 42-45 idle, and 56+ when under load. In my case, there are several people with my particular chip/mobo combo that seem to have such high temps. Even when at the highest I've ever seen it (61 degrees) all of the fins of the heat sink were cool to the touch, meaning that either there are some wicked cool cooling properties associated with zalman aluminum causing heat to flow off the fins faster than normal, which is about as likely as someone reinventing the periodic table in the next minute or so, or more likely the ASUS boards have some kind of an issue with the thermal sensor. Various people have done tests using external probes, and the chips were usually detected as being about 10-15 degrees cooler when touched by the probe than what the bios/windows/linux sensors detected.

As a note, I have a clawhammer, not a venice core, so take that into consideration.