Void YOur Warranty Internal Diode Tests

[Mikewarrior2]

First Internal Diode Heatsink Roundup

Very interesting findings:

Millenium Heatsinks
note the 7C Diode temp difference, but the ~2C socket-thermistor temp difference

Antec Reference Heatsink
According to socket-thermistor, it easily beats the Glaciator II. Switch to DIode, voila, the winner is reversed... More evidence for those who say that backside temperature must mirror core temp. Secondary factors, airflow, etc certainly play a role in socket-thermistor readings.

Volcano 7
Fan at Lowest Speed = 49C Diode/47C Socket-thermistor. Fan at highest speed = 47C diode, ~42C Socket-thermistor???? The higher fan speed has exagerrated gains when measured with the socket-thermistor.

So there you have it... And unfortunately, due to Asus socket-thermistor temp compensation, it over compensated for the 2 dyantron units. It actually overreported temps in 2 cases. Bear in mind, with another mb without hte same level of temp compensation there will likely not be any cases of socket-thermistor's over-reported past internal diode.

These results, in addition to the void your warranty grease findings last week (7-5C internal diode change, 0C socket-thermsitor change, should eliminate any doubt as to socket-thermistor *accuracy* or use as anything other than an individual fire alarm for individual cases. Even comparisons with socket-thermistors, even on the same mb are pretty much shot due to how heatsinks effect the socket-thermistor differently.


mike

 

[Soccerman]

Mikewarrior, I don't think anyone doubts that the Diode is much better than the socket thermister..

and I don't think that Anand needed to worry (I think he had a synthetic testbed) about using a thermister.. HOWEVER, IMHO it probably is better to measure the temperature INSIDE a running computer, because you have air movement that doesn't exist in most normal synthetic tests.

 

[Mikewarrior2]

No, this wasn't directed at Anand... his reviews since they built the synthetic test platform have been fine...

however, youd' be surpised at how many people still believe that socket-thermistors are effective measurement devices... that's what this post was intended for, to rid some of those rumors.

Also, in addition, 90+% of the reviews out there are stilll socket-thermistor based.



Mike

 

[Soccerman]

true..

anywho, it's an interesting roundup! the Volcano 7 doesn't fare too well at all.

can't wait for a Thoroughbred + KT333A mobo!

 

[nortexoid]

shouldn't some environment factors play a role in temp. readings?

i.e., as environmental "constants" more so....

is the internal diode reading the temperature off the surface of the core of the CPU?...if not, I don't see how it can be said that it's any more accurate than socket-thermistor readings (simply because it's, say, closer to the core)

then again, i'm not really hardcore into reading hte temperatures of the various parts of my computer, so....

 

[Mikewarrior2]

Nortexoid,

Yes, some enviromental factors do play roles, such as humidity.... but as long as it is fairly constant, and ambient case temp is measured for each individual test (per both amd and intel recommendations: 1 inch from CPU fan inlet), then the effects are minimized to a large degree.... Some sites choose not to measure ambient case temp for each individual test, rather just maintain room temp... well, thats better than not maintaining room temp, but the fact is that air circulation, heatsink design, etc, all effect ambient case temp within a set "test" area. Changes of 2-3C aren't uncommon due to different heatsinks, and without measuring it, the results are clouded.

Internal diodes are embedded within a CPU Core. its not off the surface of the DIE, nor near the die, but it is the "die" temp. Of course, different parts of the CPU die are different temps, but an internal temp measurement is the least likely to be effected by "cheating" secondary pathway effects by clever heatsink designs.

They are by far the most accurate form of cpu temp measurement. Socket-thermistors, on the other hand, are the worst, by virtue of measuring the most easily manipulated secondary pathway and attempting ot measure CPU temp through various insulation layers, leading to the compensation of "socket-thermistor temps" we now see with internal diode tests.


Mike

 

[mschell]

Mikewarrior2
The relative humidity level of the air inside (or outside) a computer case has virtually no affect on the efficiency of the CPU heatsinks.
Because we (humans) use evaporative cooling (sweat) to control our body temperature, we can feel a temperature difference with higher or lower humidity levels as a high humidity level limits evaporation which makes us hotter.
Humid air is less dense and the reduction in density does affect the airflow properties in and around a heatsink but the affect is slight in the density range we can live in.

!

 

[cookieman]

While the findings are true I think that the main factor that make the termistor so unreliable is that different HSF make different
air turbulencies around the socket thermistor. So if the socket thermistor would be isolated from this air movement than it would report better.

I said better and not real temps....

I've isolated my thermistor with some foam (but it touches the cpu very well) to improve the readings. The full load temps got higher and switching from full load to idle made the reading to fall farr less sharply then before.

Cheers,

 

[nortexoid]

what i meant to imply is that heatsinks are meant to cool the surface of the die, not the inside of the die....and moreover, i question how representative diode readings are in light of this simply because they measure the "inside" of the die (or as you say, the die itself) which, it seems, is always going to be hotter than the surface which is being directly cooled...maybe i'm misconceiving the way the die heats up, but the way i see it, it's always going to be generating a near constant temp. before the effects of cooling can even reach it...non?

 

[Mikewarrior2]

MSChell,

Thank you for teh clarification...! . Much appreciated.


Nortexoid,

Heatsinks Cool the CPU core, not just the DIE surface. In order to understand this better, you have to understand that the die surface is thermally conductive. Its merely a sealant to protect the core from damage, and is a superior thermal conductor, therefore any cooling from the DIE will effect the core. Of course, there is some thermal "gradient" between the die surface and internal core, but it is far less signifcant than the thermal gradient between the cpu core and the backside thermistor, which must pass through several layers of either fiberglass or ceramic PCB.

Why don't you go look at the reviews before jumping to some conclusions about how the internal diode reads temps. They aren't constant temps. If anything, socket-thermsitor temps show much less movement between idle and full load than do Internal diodes. Also, internal diode temps rapidly rise and fall with load and non load on the processor, as they should. off all the temp measurements available, they show the most and quickest change in temps.


Mike

 

[Mikewarrior2]

T_bird Chip Cutaway

IN case this is why you're confused... the top silver colored section of that picture is the cpu DIE. The Gray part is the ceramic PCB.

Within the CPU "die" is sealant along with the CPU core. Heatsinks directly cool both. The thermal gradient isn't .5C/W, but more like .05C/W. It exists, but the thermal gradient of other measurements is much more, so this still remains the best possible type of CPU temp measurement.



Mike