which is more accurate
 

Hey all I just popped in my new XP 1900+ and put a compunurse probe right up against the die. Now my MBM temps are reading 43°c CPU 35°c system. My compunurse probes are reading 35.7°c CPU 29.7°c system so which is more accurate?

Just from the temps I get I would say the compunurse is better than your mobo.

My pm is broke here completely now If you need me email bah@webmedic.net

I think it's a problem with vbullitin because hardforum isn't working rigth either. Maybe the database though.

Also those are real close to what that equipment was getting my shop. The compunurse that is.

Thanks web. Man the temps with this setup are awsome. I knew the MOBOs thermistors were off. Now I'll just have to go do some OCing.:D

Have fun I have a couple abit boards comming and a soltec to test oc'in with.

I would be interested in knowing how that soltek does. That purple ray looks like a nice MOBO.

Thats the one I got the drv5 and for the abit I got the kx7-333 and the nvidia one with built in video. I whould really like to try an epox also.

The epox's and abits seem to get the best overclocks but dont support the athlon diode yet. Thats why I picked up the drv5. It supports the xp diode for temps.

Since the temp diff is so big its fairly easy to determine if the compunurse os ok or not. Simply put the probe and another thermometaer in a glass of icewater (alot of ice) then it should read 0C and then put them in boiling water. Then they should read 100C. then you can take water that is around 40C and measure it with them. If they show the same temp within a degree or so the you can be fairly certain that the probe is ok.

Or just do the measurements at a 5C intervall starting at 0c and work your way up. That way you get a feel on how the probe responds to different temps.

/Dix

maybe your compunurses aren't as close as they could be to the core?

Or make yourself an on die reader. I found that my probe was off by 3 degrees (hotter reading) than the die itself.

Better yet the kt400 motherboards will be hitting the market whithin the next month or so. At that time You can get an new mobo and all the amd motherboards from now on are required to be able to read the xp diode. This will make all this much easier.

But for now yes A few checks are in order to make sure your compunurse is in order . However I don't think you will find the compnurse to be to far off. The hardest part is to get it next to the cpu core without crushing the thermal reader. I crushed one of mine between the cpu and the waterblock.

The thermistor is touching the core. I have two of them and I've switched them back and forth (system/CPU, CPU/system) and the readings are the same so I'm sure they're accurate, also I picked up another LCD thermometer today at Fred Meyer in the garden dept. It's a dual indoor outdoor type so I'm using it for ambient room temp and system temp and the system temp is reading the same as the compunurse. I would like to use the compunurse probe I was using for the system temp and use it for the water temp, is there a way to waterproof the probe and will this kill flow? BTW I picked up a transfer pump for $10 there too and it works awsome for getting the water out of your system without making a mess and getting water in your case:D

Whats a transfer pump? Well I'm not sure about the thermistors that come with the compunurse but the ones that came with my digidoc I was able to use to test water temps.

Hey web. The transfer pump is a hand held pump used for transfering oil from a bottle to the fill tube on a car (or trans mission fluid...) It has a hose you can put in the res or fill tube and a hose you can put in a jug and you just pump it by hand, like a bicycle pump, and it removes all the fluid from your system real easily;) I found it in the automotive dept at Fred Meyer. I'll bring it by next weekend and show you.

A compunurse touching the core is accurate at measuring temps of the outside of the core. Insocket thermistors are a joke and should be disregarded altogether. Mobo's that read the on die senson in the cpu are accurate at reading the temps from the CENTER of the core. An on die reading will yeild about a 7-10C difference (hotter) than a compunurse touching the core. The reason is one is measuring the INSIDE of the core (on die) and one is measuring the outer surface of the core.

If either measurement is the same as the actual CPU temperature then it would be pure luck.

That's a good point, in many ways.

The original question here was about which one is more acurate: the mobo cpu temp probe, or the compunurse's probe right up against the core?

The compunurse is very easy to check for accuracy, with a glass of water.

Mobo CPU temp probes are not only located on a different area of the CPU than the core, but some mobos actually CALCULATE the core temp based on the reading of that probe, even though it's on the opposite side of the CPU, and is nowhere near the core.

The compunurses probe, as placed here, will measure the temp of th edge of the core, not the temp of the core itself.

So to answer the question, your mobo CPU probe is probably reflecting a temp that is closer to the real core temp, but is not very accurate in that, it may be off by maybe 2 or 3 degrees.

Your compunurse probe however is accurate to +/- 0.1 degrees, but it's not inside the core, and it's not right over the core, so it's probably reading something lower than the actual core temp.

As for the inside case temp, the compunurse is the clear winner here. The mobo probe is probably too close to other components to effectively give you an accurate temp.

BigBen:

You are mistaking accuracy for resolution. If Compunurses were accurate to 0.1C then noone would use thermocouples :)

The accuracy of a Compunurse is no better than +/-3C. You can't just measure the temperature of water at 1 temp; need to calibrate.

Problem with compunurse is error in remounting it each time, and the issue of where on the core to place it.

Temperature compression is an issue with any probe that is not directly measuring a temperature.

and you're right, resolution is not the same as acuracy, that's a good point. It's misleading to see a device show temps with changes by +/- 0.1 deg, when it's not even that accurate.

i don't believe you have gotten an answer to your Q about how to mount a water temp probe. take my simple idea:

slip a 1/4" piece of tubing over the center of either a brass coupler or a piece of copper piping. cut your waterline (where you want to mount the probe) & use that coupler to splice the tubing back together. slip the tubes on flush to the 1/4" piece, pry up the 1/4" piece with a screwdriver, & slip your probe underneath it. this way the metal will stay at the temp of the water & the probe will stay insulated from case air.

i used this method on my rig & get very reliable readings, a solid seal, & almost zero impedance to waterflow. looks pro too;~)