Question for Cathar

[UnloadeD]

Originally posted by Since87
To the best of my knowledge, Bill varies the input voltage to his heating elements to keep the power at the right level.

I don't recall him ever mentioning that. They way I understood it, he keeps a constant voltage to the die for all tests. I'm sure he'll jump in here and clear it up for us.

peace.
unloaded

[BillA]

'most' wb tests are run at 70W, but the actual value does vary very slightly
- do note that it is the actual value, as recorded at each measurement point, that is used in all calculations related to those specific combinations

here is a breakdown:
- heater voltage is measured ~1/4" from the heater's coil from the voltage sensing leads used in the 4-wire setup from the regulated power supply
- heater current is measured and corrected for the meter's 'burden' (fro a Keithley 195A this is 0.0148A with a measured current of 1.4224A, with this heater 49.462VDC applied measured with a Fluke 8842A)
- the resultant Watts are then calculated and then reduced by 1.6% to 'correct' for the apparent secondary path losses (heat applied but not going through the wb)

this value, to 5 sig figs, is the one used for all calcs relating to temps taken at that same moment
note that my temps have only 4 sig figs, and this is further reduced by me to 3 sig figs for the "C/W" (or T/W if I'm looking at the bp temps)

the absolute accuracy/uncertainty of the measurements depends on the stability of the measurand and the capability of the instrument,
and an advantage to the continuing reduction in sig figs as the calculations progress is the effective elimination of propagation errors

as I suspect Since87 can describe, this whole topic becomes complex very quickly; fortunately my measurand values are fairly large (ignoring the RTD resistance changes)

[Since87]

Right, it doesn't matter whether the power into the die simulator is exactly at 70 Watts. It does matter that you know the power into the die simulator accurately at each data point.

(I'd skimmed through Bill's description of the die simulator setup and seen that he was doing it right to the nth degree. I hadn't remembered the details though.)

[8-Ball]

As I suspected.

BillA, have you any desire to put together a heat source based on the Hammer dimensions. I have a few ideas for Hammer waterblocks, and if I did build any of them, I'd definitely want to get them tested properly.

8-ball

[BillA]

am now doing the initial parts of this
2 distinct elements, well - 3 perhaps

the mounting geometry:
no real biggie, put the studs where ever they need to be

the clamping force:
AMD is now 24lbf
P4 is 100lbf
Hammer is to be ~90 or 100lbf (?) this may be speced, have not looked

the heat source size:
this is a pissing contest, though to me (and others) it is quite clear
see Joe Citarella's thoughts here, with which I, and others, do NOT agree at all

my 'issue' is with the IHS heat spreading
one can measure its thickness, go to Waterloo and plug in the values
and that is it, ain't gonna get no more
- in fact rather less due to the TIM joints

so the upshot, per me, is that the actual/effective heat source is NOT going to be substantially larger than the CPU
- the lateral heat dispersion will occur in the bp of the wb or hsf

I'm thinking strongly of staying with my present 10x10mm die
- also enables the direct comparability of data

and why would a 'good' wb not be so on any CPU ? eh ?
Hammer 'special' = marketing 'special'

[Les]

Originally posted by unregistered

my 'issue' is with the IHS heat spreading
one can measure its thickness, go to Waterloo and plug in the values
and that is it, ain't gonna get no more
- in fact rather less due to the TIM joints

It is a can of worms but:-
For a Cu Heatsink/Waterblock/IHS.
Used Waterloo for thermal Resistance when change in Flux Channel . 1D sums for other resistance . Added up the resistances.
Differences between "Purple Dotted" and "Blue Dotted" is what would expect to show in a "Joe Citarelli type" experiments* with Cu coolers and reducing Cooler Dimensions to fit IHS**.


* http://www.overclockers.com/articles708/index02.asp
** Calculating for the reduced contact area without reduction in heasink base area involves an another "Spreading Resistance" calculation - cannot be bothered(confused enough as it is)

[UnloadeD]

I didn't understand Joe's experiment or results. To start with the pages were linked backwards. Then on one of the temp tables, for one HSF the on die temp was lower than the IHS temp. I assume its from some kinda error (maybe not?). Anyway it confused me. And finally to eleminate/isolate cooling properties of the IHS, he hacked up the HSF rather than removing the IHS! WHatever he was trying to do, I sure couldn't follow him. Glad you guys mentioned it, I thought maybe it was just me 8)

peace.
unloaded