which thermistor is the correct one to monitor?
 

ok, i've been pondering this for a while now, and it kinda concerns me a bit.

i run mbm on my rig, and it seems to be consistent with bios and winbond temp monitors, so i assume it's fairly accurate. i don't know if all systems do this, but my system monitors case temp, cpu diode and cpu core.

for the cpu, which is the temp i need to pay attention the most? i assume that most people are more concerned with the cpu diode, as is amd, but i want to make sure. for example, if amd says a processor's "oh shit" temp is 75deg, does that mean a reading off the diode, or the core temp? i was wondering because my core temp is consitently about 15deg warmer than the diode.

the system idles now around 38-40degC on the diode. i assume when most people are talking about cpu temps, that's what they're referring to (this is what i always thought). i'm also concerned about the accuracy of the core temp - is there a temp sensor in the cpu the mb is reading? i have the bios set at the moment to shut down the system if cpu temp reaches 63deg, i'm also assuming that reading is taken from the diode.

anyways, just wondering which was the more important number, to make sure i'm not roasting my kit.

It depends on the Software. MBM-5, and CPUCool use different diodes, or Thermistors to monitor software, and most times you have to assign these diode/thermistor addresses. Especially with MBM-5 there's a specific diode it monitors which will set al the alarms off, because it's insanely high. I've seen it read either 157C or -84C and it's simply re-setting it to another on the list. It's usually the second diode fo the CPU. I'll bet my solder-dropping collection, it's just reading the wrong device. The CPU should be about 20C above the Motherboard diode reading, depending on cooling. In my case I use some extreme watercooling placement, and usually my CPU temps are 5C to 11C below my mobo temp.

http://www.ksbrainstorms.com/RBXnozz...ETI3749MHz.jpg

This is employing the DD RBX with Nozzle-5 installed (it truly replicates the Cascade's impingment into cup design) the rubber gasket, and the multiple holes cut into the lucite on the number-5 directs mutiple streams into the RBX's heat-voids. I know many eschew the RBX design in favor of "real" H20 blocks. Lest we forget at one time Danger Den was the PolarFLO and Cascade innovator of it's time. They've done a lot bringging real water-cooling to the market, and dare I say Koolance as well. Please don't flame me for mentioning the latter.

The reason my temps are so low is location, location, and location, of components, and the case.

http://www.ksbrainstorms.com/2004_0128Image0039.JPG

http://www.ksbrainstorms.com/2004_0130Image048.JPG

Anyway I'll get back to ya with specific thermistor assigment in MBM

heh... welcome to the board :). that's good and all, but didn't really answer my question.

mbm5 with my board monitors the on-die temp sensor and the cpu socket sensor.

which one is the one that i need to worry about? which one when it hits 70 does the cpu go BOOM?

right now i'm shoting 39-56-45 (cpu socket, cpu diode, case, respectively).

Dear God, please tell me that isn't the SETI screen saver you are running Liquid3D... :cry:

The CPU diode temperature is the important one. The thermal diode is a single transistor on the CPU die, identical the the million of others on there, wired up such that YOU can use it to measure the temperature inside the CPU.

And it won't do BOOM, it'll lock-up or give lots of errors. A year or so back on a dare, I ran Prime95 on a Palomino (stock voltage and speed) while throttling down the water flow. It was still stable at 96 C and 1.5 liters/hour.

groth, thanks for the answer. guess i have more cooling to do, the on-die diode is uncomfortably high (just when i thought i was doing fairly decent :) ). i'm surprised i haven't had more problems then, considering. when it's running at 55-57 idle, i wouldn't be surprised if it hits 70, or at least pretty damn close to it. argh!

odd, i dont really trust the Diode at all,

Room = 30'C
Case = 31'C
Water temp = 30'C
CPU Socket = 32'C
Waterblock (right above/next to CPU Core) = 31'C
Diode = 17'C!
(all temps from MBM5/Digital Thermister readout)

... somehow i dont belive that a Diode in the CPU is cooler than the case.

:shrug:

interesting... i'm curious tho. when people talk about running their cpu's on aircooling, they're always talking about having their cpu's running in the 40's or whatever, i assume that's all got to be read from the socket (excuse my bad grammar). it doesn't make sense otherwise for me to be running so much higher. i mean, my aircooler isn't exaclty the best, but it's in the top 10 or so, so it's acceptable, and i have decent airflow through the system. it also seems that winbond (at least by default) doesn't monitor the cpu diode - only the socket. which is kindof disturbing if the diode is the one needed to be watched more carefully.

althought with your reports guandi, that kinda throws a wrench in the works. what mb are you running?

DFI Lanparty NFII Ultra (nForce2 SPP nb/MCP-T sb Chipset) with latest driver revs, pheonix 6PG Bios + Genie mod Bios set to Turbo/Aggressive (will wait till i get Corsair XMS4000 before tweaking that)

got a AMD XP2800 Barton (stock 2080), 1.65v/12.5x185 = 2314mhz stable @ room temp.

funny thing is, i was running a little over room temp before the little overclock, i'm afraid to go further incase i have to put a room heater in front of my heatercore to stop ice formation :D

damn, i must have the wrong gear. i know i'm running a tbred, but i've sure got to dump a lot of juice in it to get it to run faster :(

Most of the OEM hardware monitor programs won't show the thermal-diode temperature for the simple reason that it IS higher (and frightens people). Then there's boards like the Abit NF-7 that use a dedicated over-temp protection chip that doesn't even report the temperature. And let's not forget that manufacturers like to 'adjust' the data (Asus is famous for 10 C changes with every BIOS revision).

I suggest a rough calibration of your thermal-diode. Notwithstanding the recent fun of C/W values varying with heatload and temperature, you can make a plot of multiplier vs. delta-T (diode - water). At a theoretical zero-multiplier, the delta-T should also be zero. If it's not, it gives a rough estimate of the thermal-diode's error. See also pHaestus's first thermal-diode article.

groth : that sounds a little too much like maths to me, i think i will steer well clear. i have to take a calculator with me to buy penny sweets.

i prefer compunurse type thermisters. i put a couple on the same heat source right next to each other (usually in a star shape) and i take the average reading, the one thats closest to the average is usually the one that is the nearest to the actuall temp when mesured with my infra-red thermometer. although they are never more than 1'c out,s they all get used eventually, i just make a note of how much its out by.

i am a hardware fanatic, i dont trust software as far as i can throw a server rack, if you cant see it, i'm pretty certain it doesnt exist!

It's not that hard, a few BIOS changes, at bit of time with your favorite burn-in program, a piece of graph paper. But, yea, you have to really want to know that internal temperature.....

Hey, at least I didn't suggest the hardcore add-an-external-diode-reader-and-calibrate-in-a-stirred-water-bath that pH has been doing recently. :)

yeah, thats just crazy ;)

Being new to watercooling I'd like to ask a related question to this. A lot of folks distrust the quality of on mobo sensors and their relations such as on dye P4 temp probes. I have Asus, and begin to wonder if its wiser to go with a Digidoc array of sensors vs MBM. I own neither atm, but use but currently use the Asusprob.

Is it wiser to go with Digidoc, calibrate if necessary but have an independent temperature method, or does MBM do just fine?

Thanks for letting me draw on your collective experience

this explains the very low p4 temps some people are getting on asus motherboards.

http://www.legitreviews.com/Reviews/coretemp_1.shtml