Strange Temperature Readings - Is This Even Possible?

OmegaZero · Post by **OmegaZero** » Mon Jun 06, 2005 10:05 am

To start with, my system is:

Intel P4 Northwood 3.2Ghz
Thermalright XP-120 w/Nexus 120mm
2x 1GB Crucial Ballistix PC4000 (2.5-3-3-6, 1T)
Asus P4C800-E Deluxe (Intel 875P) w/stock passive NB cooler
Gigabyte GeForce FX5200 128MB AGP w/stock passive cooler
M-Audio Audiophile 192 24-bit/192kHz 4-in/4-out
Western Digital Raptor WD740GD 74GB 10k RPM
GUP SmartDrive 2002 Copper
Plextor PX-712SA DVD-RW
Seasonic S12-430
Antec SLK3000B w/Nexus 120mm exhaust
Arctic Silver 5

Modifications to the case include cutting out all stock fan grills with my trusty Dremel, removing the TAC duct (since it doesn't line up with my processor on this board), and covering the side panel vents with thin black plastic and aluminum ducting tape. I also covered all openings in the front and rear of the case with the aluminum tape, with the exception of the filtered front input. All wires inside are tucked away and sleeved, and all drives are SATA so there are no substantial airflow obstructions. A quick incense test shows a strong airflow current in the front and out the exhaust & power supply vents (even without an intake fan). All fans are powered using motherboard headers, so they are controlled by Asus' Q-Fan system. All speeds and voltages are stock - no overclocking or undervolting.

The ambient temperature in my room is generally 22C (central air).

After a two day burn-in period, including periodic powering off to allow the AS5 to "settle in", my idle temps are: CPU 35C, System 30C, HDD 34C with fan speeds of: CPU ~800RPM, Exhaust ~700RPM, and PSU ~650RPM. I should note here that all temperatures and speeds are recorded and verified using SpeedFan, MBM 5, and the Asus AiBooster Utility. Time for stress testing...

I ran Memtest86+ through 10 complete passes without error.

Prime95 Blend Stress-Test with CPUBurn (just to be sure it stayed at 100%CPU usage) ran for 30 hours and 18 minutes, at which time I stopped it. Highest temperatures recorded for each component were CPU 47C, System 35C, HDD 38C. Fan speeds seem to ramp up as expected as temperatures rise - very impressed with the behavior of the fan controller.

Now for the weirdness:

Just for kicks, I decided to test using two instances of SP2004. This is a utility that, as I understand it, uses a variation of the Prime95 stress test and adds some error checking/reporting functionality. I used this because I can run two copies of it simultaneously (the stock version of Prime95 doesn't seem to work well doing this - see my post on that topic here). Everything seems to be going fine; showing the same behavior as the Prime & CPUBurn after 6 hours. Then at 7 hours I hear the alarm go off - CPU Temp 88C!! Wouldn't it have rebooted LONG before it got that hot?! NOTE: All fans showed to have been operating normally at that point. 88C temp was shown by the ASUS utility.

I immediately shut the tests off and did a soft restart, booting into BIOS. Temperature readings are CPU 41C and System 32C (no HDD temp available in BIOS). Could it have cooled down that fast?! I opened up the case and found that nothing seems to be too hot. XP-120 is very warm to the touch, but not hot really at all.

Ok, time to go extreme. Ran CPU Burn x4 on the maximum heat generation setting with Prime95 on Blend with all 2GB of memory. Verified 100% CPU usage and ~1.93GB memory usage for 10 hours. Maximum temperatures: CPU 48C, System 36C, HDD 37C. <shrug>. Lets try SP2004 again...

Ran SP2004 x2, verified 100% CPU usage and ~1.91GB memory usage for 8 hours with no problem. Then at 8 hours 43 minutes I hear the alarm again - System Temp 121C!!!!!!!!!!!!!

I opened up the case at this point. All fans running normally. The only thing that seems to be overly warm to the touch is the memory, but again nothing is really hot. SO strange! Reboot into BIOS again and temps show: CPU 39C and System 32C.

Anyone have any clue what is going on here? What should I do next?

[EDIT: Fixed SpeedFan typo]

lenny · Post by **lenny** » Mon Jun 06, 2005 10:14 am

I'd suspect sensor (or software) malfunction. Did you graph the temperature (or can you do it within the Asus utility)? Try Speedfan. Or a newer version of the Asus utility. I suspect you'll see a sudden spike in the temperature (e.g. steady at 47C for CPU, then next reading is 88C).

If the CPU really is 88 C, touching the HSF would have blistered your finger at least.

OmegaZero · Post by **OmegaZero** » Mon Jun 06, 2005 10:21 am

The "SmartFan" application was actually SpeedFan. That was a typo...

The readings were verified using three different utilities, though I did not think to look at/save the graphs.

I thought that those readings were ridiculous. It's even more strange that I only seem to get them when running that test application - nothing like that happens with Prime95, CPU Burn, or Sandra 2005's stress test.

Could that test being screwing up the sensors somehow?

lenny · Post by **lenny** » Mon Jun 06, 2005 11:16 am

Are you running Speedfan and MBM5 (and Asus AiBooster?) at the same time? Won't they interfere with each other if you do (contention of SMBus)?

Another possibility could be that very high CPU usage could lead to a very long time between issuing a read and receiving the data, and this led to corrupted data. If this is the case, you'd see strange spikes and dips in your graph.

If you like, get an independent temperature probe and attach it to the CPU HSF and/or NB heatsink. That way you'll know without a doubt if the temperature is really out of whack.

The Greek · Post by **The Greek** » Mon Jun 06, 2005 11:42 am

That sounds like a temp sensor malfunction, at 88*C (or 121*C for that matter!) your CPU would cook, I can almost guarantee it!*

And if the CPU went from 121* to 39* in the few seconds it takes to soft reset and get into the BIOS it would probably have been damaged by the rapid cooling because of difference in cooling rates of the different materials in the CPU.

*Those temps are 190.4*F and 249.8*F for those who don't do metric!

thetoad30 · Post by **thetoad30** » Mon Jun 06, 2005 11:53 am

Don't be so sure that the CPU temp can't reach that high.

I was doing an experiment with one of my servers in a Sonata case: I was trying to see if I could forgo the 80mm fan on my Thermalright heatsink...

It was running great, but the temp on the heatsink and the CPU temp kept rising... so I decided to reinstall the fan.

Well, I'm lazy, so I wanted to install it while the system was on.

Took me a while, but when I looked back up after a few of my failed attempts... the CPU had hit a reported 82C... and yes the heatsink was scorching hot. I managed to get the fan on very quickly after that, but there was no malfunction, no data corruption, and the CPU did not shut down (the motherboard has a thermal shutdown enabled, but I don't know at what temp).

So, it could get that hot... but if you jumped from 47 to 88, I would suspect that perhaps the heavy loading of the machine caused something to go awry. Perhaps try the above recommended advice with an external probe, and if that shows everything as ok, then consider RMAing the board for a new one... I'm a perfectionist, I like things to work!

But that's just me.

OmegaZero · Post by **OmegaZero** » Mon Jun 06, 2005 11:57 am

Just to clarify (though not that it matters much), the 88C measurement was on the CPU and the 121C was on the System. They were not both CPU measurements.

I agree that there is no way that those were accurate readings, but how is it possible that I only got them when running that specific test utility? Even now I am running Prime95 with CDBurn, the processor(s) are 100% and the memory is 1.94GB/2.00GB, while the temperature is stable at CPU 46C and System 33C. I am not doing anything differently when I run SP2004 - could that utility be changing the readings on all three monitor programs?

Not to be redundant, but since this only seems to happen when running that one test program I would be inclined to think it was a software issue rather than a hardware malfuncion. No?

NOTE: All three monitor programs have been running throughout all of my testing.

thetoad30: Was that an AMD processor? It has been my experience that any Intel CPU will reboot automatically shortly after crossing the 70C mark...

thetoad30 · Post by **thetoad30** » Mon Jun 06, 2005 12:06 pm

No. This was an Intel Northwood 2.8 Ghz. Hyperthreading was enabled... so I too was surprised to see such a high heat level with the CPU... in a way it sort of scares me... as the whole reason I was trying to see about removing the fan is because its failing, the bearings are driving me nuts with their high squeal, and I don't have another to replace it right now.

I'm waiting for the day to wake up and its smoking...

drliamski · Post by **drliamski** » Mon Jun 06, 2005 3:01 pm

Dont worry about your temps, this is a speedfan error,
when i reinstalled windows and tried to run speedfan without "reinstalling" it, the cpu temp in system tray would read as 88.
If you run the tests and use a different program to report temp, I am sure you would have seen a temp of around 41.
Just reinstall speedfan and run the same tests again.

Mar. · Post by **Mar.** » Mon Jun 06, 2005 4:00 pm

I would have suggested resating your heatsink and re-applying thermal paste just to be sure, but since your system temp was also very high, I seriously doubt anything could be doing that other than some software problem. It also seems unlikely to me that two sensors would go bad at the same time.

sthayashi · Post by **sthayashi** » Mon Jun 06, 2005 4:07 pm

I had heard that funny things can happen if you run 2 monitoring programs at once. Perhaps this is the situation.

OmegaZero · Post by **OmegaZero** » Mon Jun 06, 2005 6:04 pm

Just figured this out: One major difference between SP2004 and Prime95 is that SP2004 has temperature monitoring built in. Perhaps that's what is causing these erroneous readings when using this app when it does not happen otherwise?

Makes some sense - it is very likely that the actual CPU temperature was 44C, so 88C is an understandable error. Still doesn't explain the reading of 121C on the System sensor though...

drliamski · Post by **drliamski** » Mon Jun 06, 2005 7:17 pm

I think 88 is what it reads when there is an error in the program, think about a digital clock when its off, you can see 88 if all the bars are lit.
Also as to your system temps, were they along these lines
http://img200.echo.cx/img200/9767/desktop5dr.jpg
lol

OmegaZero · Post by **OmegaZero** » Tue Jun 07, 2005 2:25 am

drliamski wrote:Also as to your system temps, were they along these lines

SpeedFan has always done that. I have one of those too - it reads a constant 126C, just like in the picture. The problem is with the top three sensor readings. They read normally at all times(in this, and two other monitoring apps), except when I am running one particular software package. See my original post for details...