Temperature at heatsink

I have a Dell Dimension 4400 that I have upgraded the processor on (now a 2.4). Dell used a proprietary heat sink retension mechanism that forced me to reuse the Dell heatsink from the 1.6.

Dell does not provide any CPU temperature monitoring in the BIOS.

I have plugged in a spare Digital Doctor 5 that has flat temperature sensors. The sensor is taped to the side of the CPU heatsink.

I am reading 51C at full load with Prime95 set to maximum temp mode.

Does this sound like a reasonable temperature when measuring in this manner?

Sounds about right for the setup you descibed.Not to patronize you ,but did you use any thermal interface material ( T.I.M. ) ? Did you clean the HS and CPU mating surfaces? Just a couple of standard questions. Did you tape the temp. sensor to the side of the CPU or to the HS itself ? THE FANMAN

reply to archilochus

Your concerns were not taken as patronizing. Thank you for voicing them.

The old heatsink was properly de-TIM'd and cleaned with 99% alcohol. I used Arctic Silver when reinstalling the old heatsink with the new CPU.

The temperature sensor was taped to the heatsink itself; securely and on an side that does not get the majority of airflow.

The fan is the stock 92mm Dell spec'd NMB that acts as case evacuator in addition to its shroud assisted CPU cooling duties. I may consider swapping out the Dell fan for a mid-speed unit like a Sunon. I'll need to make up an adapter if I do this. Dell used a proprietary 3 wire plug for the fan (similar to an analog CDROM audio cable but with 3 wires). I need to look up the NMB's specs to see how much CFM I might gain.

I was able to complete another 15 hours of Prime95 last night without errors.

Idle temp is 33C after halting Prime.

These are work machines at a small company. Whatever is done to this machine will be duplicated on five others. The CPUs have already been purchased.

Alot of fans have three wires or more.The 3rd wire usually yellow is the rpm sensing wire.It reports rpm to the bios usualy.Some comp. will not stay running is the mobo header doesn't receive the rpm signal!Is there away you can take a case temp reading on said rig? Have the rig closed up in a normal use fashion.See what the case temp.is...Then take the side door off and shoot a fan at it angled towards the rear of the case, a couple of feet away. Record CPU or HS temp.This is a way to test where the prob. might be..THE FANMAN

I'm confused Archilochus.
In your first post you said the temps sounded normal, in the second you refer to a method of isolating "the problem".

Personally, I think that your measurement method is faulty, jlundy.
Taping a sensor on the outside of your HS doesn't accurately relect the temps at the chip. You could use this a a method of comparison, but I'll bet that if the outside of the HS reads 50C then the chip is at least 15 deg. hotter.

I don't have an intel processor. Mine is an AMD XP1900+. I slipped the heat sensor between the heatsink and the chip substrate. It is touching the die on it's side. I don't know if it is possible to do this on an Intel, but it is worth mentioning. If this is possible with an Intel chip, be careful not to slip the sensor between the die and the heatsink. If you do, it will fry the chip.

I think that sensor will probally read lower than then actual CPU temperature. The diference will depend on the overall difference between the case temperature and the actual CPU temperature. If your case temperature is high, like 45 C, I don't think you have too much to worry about, since the CPU is probally under 55C. If your case temperature is lower, like 35 C, then I would be worried that the processor is running too hot.

I agree that this is not the best method of temperature measurement. I spent several hours yesterday looking for a non mobo dependant utility that would read the CPUs on die sensor.

Due to the P4's socket retension mechanism being 'closed', there is no way to slip a 'non flat' sensor in next to the CPU's die.

Later tonight, or maybe tomorrow morning, I'll try to connect the Digital Doc 5 to one of the Dells that has not been upgraded. This should provide a good comparison.

There is not an extreme amount of airflow going past the sensor point; as in a normally OC'd setup. The shroud does not cover the outermost fin at all. I am taped to this outer fin; as close to the base as possible.

I do have another sensor reading air in the case.

Case: 28C
CPU idle: 32C
CPU Load: 51C

An update for anyone who cares

The NMB fan used by Dell is thermally controlled and a not so common size of 92x32mm.

I taped a Sunon non-thermal 92x25mm into the housing for testing purposes.

New temps:
Case: 27
Idle: 29.5
Load: 45.5

I also got my hands on one of the still stock 1.6Ghz Dells. The temperatures are as follows.

Control:
Case: 28
Idle: 34
Load: 52

The temperatures do not appear to be any higher than the stock Dell temperatures at 1.6 Ghz.

I really wish I could read the onboard CPU die, but feel confident that I am not going to cook the new 2.4 Ghz processor.

I will be ordering some Sanyo 92x32mm non-thermal fans to replace the Dell stockers. The 4-6 degree improvement that they bring to the table seems like good insurance. The 50 CFM flow rate of the Sanyo fan would be comparable to the Sunon used in my test.

Are you doing away with the stock exhaust chimney setup? Your placement of the probe can only be bettered by removing the HS.
But for test purposses you show a diff.reading,I suspect all is the same as when temps were taken previously!Keep your flow up & temps down, THE FANMAN