New probe shows that motherboard CPU temp is totally off.

New probe position(compared with 3 other probes for accuracy):



^ Sorry for focus. The camera seemed more concerned about the case. It is a thin probe. And has no pin lockdown problems at all.



Very hot summer weather at the moment with room temp 26C:

Northwood P4 2.8, DFI Infinity 865pe

CPU sensor:40 idle 46 load
CPU probe: 27 idle 44 load

Motherboard temperature: 27c
Case: 23C

Surprisingly the load of the motherboard cpu temperature is quite accurate. But idle is hugely off.

I have been annoyed with my idle for a long time. I can't believe how far off it is.

Newman_SCO said:

New probe position(compared with 3 other probes for accuracy):

^ Sorry for focus. The camera seemed more concerned about the case. It is a thin probe. And has no pin lockdown problems at all.

Very hot summer weather at the moment with room temp 26C:

Northwood P4 2.8, DFI Infinity 865pe

CPU sensor:40 idle 46 load
CPU probe: 27 idle 44 load

Motherboard temperature: 27c
Case: 23C

Surprisingly the load of the motherboard cpu temperature is quite accurate. But idle is hugely off.

I have been annoyed with my idle for a long time. I can't believe how far off it is.

Click to expand...

I wouldn't trust a sensor under the CPU for accurate readings... A more accurate method would be directly next to the die, but you can't do that with intel IHS'd processors (unless you somehow deIHS it).

SolidxSnake said:

I wouldn't trust a sensor under the CPU for accurate readings... A more accurate method would be directly next to the die, but you can't do that with intel IHS'd processors (unless you somehow deIHS it).

Click to expand...

I know what you mean. But with a probe directly under the cpu is maybe the closest most users will be able to achieve.

Thats cool. I wish I had done that instead of placing my probe on the side.
I cant be bothered taking my scythe ninja off though to do this.

You should test the accuracy of that probe using 50c hot water and 28c water, Dip into the 50c water and then quickly into the 28c water and make sure it stays at 28c and doesn't actually report lower than that. I've done your method but with a extremely smaller more accurate probe. I found out that the probe your using tends to go off as much as +/-7C when there is sudden temp changes, especially when it's extreme. The underpart of the CPU is probably going to be the hottest part of it too. Consider when you put that probe under the cpu your creating a small 1-2mm spacing that doesn't align right with the HS as it tries to stay properly flat against the IHS, seating won't be exactly flush even though it looks so to your eyes.

I have probes that look similar to that probe but they are from a different manufacturer. This one is thinner compared to my others and at most is 0.5mm raising the cpu. Instead of a round wire on some of the others, it has a flat edge which is roughly the width of 2 sheets of average paper going through the pins. The pressure from the heatsink is compressing it too. I trimmed the probe and wire sleeving for best fit.

I don't have ability to compare 50c water and 28c water. I could try the fridge/freezer(-18c - report 0 I presume) verse room temperature(normally 23c) for a large range of temperature difference. If I have the time tomorrow I might give it a go.

i just put my probe on the side of my IHS it reports bloddy accurate temps and the probe looks exactly the same.... ya kno i just took the IHS off my P4 2.0A today... mabey i should do it to my main rigs P4 3.0c

thats also not a good idea like ppl have said because it tilts the cpu a lil, and there is a chance you r gonna short out some pins

At the side it is not close enough. It returns slight higher than case temperatures in my situation.

thats also not a good idea like ppl have said because it tilts the cpu a lil, and there is a chance you r gonna short out some pins

Click to expand...

I guess you have to see it yourself to understand how snug it is!

Newman_SCO said:

I have probes that look similar to that probe but they are from a different manufacturer. This one is thinner compared to my others and at most is 0.5mm raising the cpu. Instead of a round wire on some of the others, it has a flat edge which is roughly the width of 2 sheets of average paper going through the pins. The pressure from the heatsink is compressing it too. I trimmed the probe and wire sleeving for best fit.

I don't have ability to compare 50c water and 28c water. I could try the fridge/freezer(-18c - report 0 I presume) verse room temperature(normally 23c) for a large range of temperature difference. If I have the time tomorrow I might give it a go.

Click to expand...

Well you must have a very unique probe if it's 0.5mm because most of those standarized probes that come with fan controllers, mb's are usually thicker than that, especially on the actual sensor portion. This is what I use:


It's very accurate and has no rebound in temp swings when subject to high and low sudden extremes. I've tested it several times in varying temps against a laboratory digital thermometer. I used to use a smaller diameter version that was only 0.47mm thick but I only bought one because it was so damn expensive: had silver leads and were so thin that the hair-fine lead broke to easily after moving stuff around in my case too much. $40 down the drain for that probe. They were both 10k ntc thermistors. If you ever need to get another the pic above is only $10. You can contact Kevin at http://mcshaneinc.com, he has treated me well and really taken time to explain the problems of temp sensors to me. Turned out his product is so accurate that you'd have to have a thermometer that has a resolution better than 0.1c to notice any inconsistency. I'm using this one: EXCELLENT thermometer for the price: http://cgi.ebay.com/NEW-HANNA-CHECK...ryZ50974QQssPageNameZWD1VQQrdZ1QQcmdZViewItem
and I've even check the accuracy of the above thermometer via boiling point and my restuarant's state regulated RTC probe.
Mcshane's 10k thermistors work on both mb's and fan controllers but they are not suited for being under the cpu as I soon found out because the ends are polyamide doesn't compress at all (unless your able to get one of those micro ntc's from Betatherm as a "sample" ~0.37mm) . I found out my Dual core Opty 170 was getting 54-56c when it was placed under/between the pins at full load 2900mhz 1.55vcore.
I had it under my cpu for about 2 months because I knew that the leads and sensor are coated with non-conductive polyamide (very resiliant plastic). But it's just not good for watercooling if you want to make sure everything is a 0 degress North.

Thanks for sharing that information. I might need to import the Hanna Checktemp 1 Electronic Digital Thermometer. Exchange rate is good though.

Well you must have a very unique probe if it's 0.5mm because most of those standarized probes that come with fan controllers, mb's are usually thicker than that, especially on the actual sensor portion.

Click to expand...

I tried taking a picture of the probe but my phone is having focusing problems. Needs a clean I think.

Your method must of been much more difficult. With a 939 socket you would have to thread your sensor through 15-16 pins to get to the centre and would certainly not have the room for a normal sensor bulge. While the 478 socket only needs threaded though 6 pins and has plenty of space for the bulge of the sensor. My probe would not work at all with a 939 and I understand why you took it an extra step of getting a tiny sensor.

I know you probably know what the sockets look like, but I though I would just add them to compare the differences.









You seem very knowledgeable about thermometers. Slightly off topic here but could you please recommend me an outdoor thermometer ideally with a pc connection such as usb. (range -10c to 30c+ and weatherproof) If not could you recommend me a make or even a website? I was also hoping for dev kit.


.

Newman_SCO said:

Your method must of been much more difficult. With a 939 socket you would have to thread your sensor through 15-16 pins to get to the centre and would certainly not have the room for a normal sensor bulge. While the 478 socket only needs threaded though 6 pins and has plenty of space for the bulge of the sensor. My probe would not work at all with a 939 and I understand why you took it an extra step of getting a tiny sensor.
I know you probably know what the sockets look like, but I though I would just add them to compare the differences.
Slightly off topic here but could you please recommend me an outdoor thermometer ideally with a pc connection such as usb. (range -10c to 30c+ and weatherproof) If not could you recommend me a make or even a website? I was also hoping for dev kit.

Click to expand...

The sensor I'm using exactly fits between 939 pins without bending them. I've never seen any sensor with a pc connection of any kind. Except for possilbe super expensive RTD probes which you pay $100-$300 just for the probe and another few hundred for the module in connects to (maybe those modules have usb ports??). I did a lot of research on various sensors/digital thermometers and their accuracy. I don't think I ever saw one used specifially for pc connection. Pretty much any probe you get is going to have an accuracy of +/- 2.0C at the worst between -20 and 50c. Even the junky $14 deals will do that. When I did some chemistry work I used a couple of those in my reactions that required strick temp controls and they worked fine for my needs (Radio shack sometimes has decent ones). You can Google "weatherproof digital thermometers." Your going to find a lot of thermometers. Most will be junk, finding decent thermometers with everything you want and really good accuracy is going to be difficult and time-consuming. How much do you want to spend, they can easily go into the hundreds of dollars. Lab thermometers are usually standarized and more reliable for accuracy.
I have no idea how your going to get readings from a digital thermometer if you plan on cutting the leads off and attaching USB cables. You'll probably have resistance problems once you apply USB cables and get inaccurate readings. You have to have an instrument like a fan controller or mb "temp" connect which uses the same resistance readings and temperature curves (I don't understand the curve thing but their is Fenwell Curve and some other standarized curves which the instrument calculates the resistance against another calculation). If you call that guy Kevin at Mcshane he would be able to advise you how you could do your project.
What is a "dev kit"?

This sight is a little expensive but their websight is easy to use. Not really what you need though.
http://www.daigger.com/catalog/department?deptId=Digital+with+Probe

I don't think this place will sell to you unless you are an institution:
http://vwrlabshop.com/category.asp?...cpc-_-labs-_-labthermometers&bhcd2=1152541506

Thanks for the links. I have searched around more myself.

It seems I will have to sacrifice your recommendations of good thermometers unfortunately for the weatherproof thermometer with pc interface. First thing is price and second is I would like a all in one ready built unit. But I will certainly check out the others you recommended for my other projects.

The dev kit I mentioned contains software tools to help you control/read the hardware. So programmers can write their own software for the hardware with ease.

But I've found something better than a usb interface which I was originally interested in.

Ethernet IP temperature sensor on a Czechlovakian website. $136.297

http://www.papouch.com/en/products.asp?dir=thermometers <3rd down

I've sent an email to them hoping to negotiate a price(Worth a shot). I'm not sure on the accuracy of the device. But I have only found 2 manufacturers in existence making these devices.
Quite rare, so I don't have much of a choice. With this it doesn't need a computer connected to it. But I can read it from any computer on our house network or via the internet. And even log the temperatures to an online PHP+SQL database if I'm feeling that geeky.