Calibrating CPUTemp.docky

I'm wondering what it would take to calibrate the CPUTemp.docky.

I have access to a calibrated temperature probe (Fluke) that I could use to measure the exact temperature of the heat sink or similar.

If I take a few measurements with the probe and peek at CPUTemp.docky a few times perhaps I can come up with a meaningful difference. I just hope the relationship is linear.

Would this be enough to get it at least roughly (i.e. +/- 5C) calibrated?

@ssolie

ssolie ive done mine the offset is 8 here on XEG4@1000mhz 7455 chip
I used a heat gun and is within .3 of a degree of the heat gun
now the heat gun is +-.2deg accurate so my calibration should be +-.5 of a degree now

@poweramiga

I'm interested in this heat issue.

What temp were you getting? and do you recall what Temp the air in the room was?

@ssolie

Quote:
Checking the temperature of the heatsink is problematic since it takes some time to heat up while CPUTemp displays the CPU core temperature which increases much faster when switching on the system.
For an usable calibration you'd need a sensor directly at the die of the CPU.

Quote:
Prepare your system to boot as fast as possible, no user-startup, everything but CPUTemp.docky removed from WBStartup, everything not required for the test disabled in U-Boot, etc.

Switch it off and let it cool down to room temerature.

Switch it on and as soon as CPUTemp.docky is running for about 5 seconds displaying a (nearly) constant temperature use the probe and note down the results of both the probe and what CPUTemp.docky displays until you get to the normal, idle temperature of your system and it doesn't change anymore. Then start something, or several programs, which use the CPU a lot, something which uses AltiVec if you have a 74xx CPU. For example dnetc. Note down the results of the probe and what CPUTemp.docky displays again until you reach the max. teperature of your busy system. Make sure the TEMPOFFSET and TEMPFACTOR tooltypes are not set when you do this.

Now you should have a table with at least 10 pairs of real temerature (from your probe) and what CPUTemp displays without calibration for them. Using this table and real_temperature = CPUTemp_displayed_temperature * x + y calulate x and y integers in such a way that the average error is minimal and add TEMPFACTOR = x and TEMPOFFSET = y to the CPUTemp.docky tooltypes.

Quote:
That's unlikely.


In case someone is interested in more details and how bad the TAU is check TAU_Calibration_10s.pdf. I didn't find anything like that from Freescale, except that officially it's not supported at all on 74xx CPUs (even if it works ).

@joerg

Quote:

The instructions that came with my Aerogate advised against placing the probe directly onto the CPU die.

@Adept

air temp here 28 deg c comp idle @43 deg c runnining dnetc for an hour or so comp reaches 49 deg c heat sink 49 too with gun this is with the offset of 8

this is also a 7455 cpu @1000mhz core voltage 1.79v

with a 60 X 60 X 50 heat sink with 60 X 60 X 30 fan

even on 40 deg days my A1 has never failed even while running dnetc

@joerg

I have a CoolerMaster -- Musketeer installed. The temperature gauge probe is installed on the cpu heatsink. Gauge normally is 40C something. Your CPUTemp reads 43C. That's really right on!

I startup at 28C and stay at 43C.

OK, how about a round of Quake2 and a few levels?
I get 48C with the Zalman fan in high.

edit:
cpu-7451mcp at 800

@all

Maybe a little offtopic, but what brands/types tempature meters are you recommending ?

Or is it better to buy a multifunctional meter ?

@benny

http://tech.ph.groups.yahoo.com/group ... OS4/photos/view/f929?b=11

This is my CoolerMaster Musketeer. I tried looking up the product, but it's no longer available? Some of the new ATX cases come with gauges, but I'm sure that costs more.

@Snuffy

Thanks for the suggestion.

For anyone wanting to know how to best place a temp probe on their CPU.

Whatever the default settings in this docky, it matches the readings from my temperature probe. Of course, that's the temperature of the base of the heat-sync/chip-surface, not the internal temp, but it's a good indicator.

BTW, I thought that temperature measurements were broken in the PowerPC chips used in the A1. How come it suddenly seems to work?

Hans

@Hans

Quote:

There were several different revision of PPC chips used in the A1, Maybe some of them have working temperarue monitors.