Quote:
|
Originally Posted by snowwie
so the TTV setup is flawed? how is the TTV test setup less representative of our application than the "other" (copper die) setup? or is there a discussion on these forums that I missed?
|
The TTV setup does not monitor the actual die temp, just the IHS surface temperature.
If the wb base-plate flexes, along with the IHS, and I can assure you that CPU silicon is flatter than a very flat thing, then the heat-flux pattern between the die and the IHS is changing with respect to that flexing. Naturally, if the block is flexing and therefore so is the IHS, then the IHS is not being pushed flat against the CPU die, and so the heat distrubution between the CPU/heat die and the IHS is now different. Since the thermal probe is merely measuring the temperature of the IHS surface, its readings are therefore going to be affected by any flexing.
IF both the heat-die center temp, AND the IHS surface temp were known, then I'm almost 100% positive that we would be able to witness and quantify these effects.
As it stands, the temperature of the CPU die / heat die is not known, yet it seems that some people are willing to dismiss this as being irrelevant.
I have been privy to data generated on the TTV that defies logic in terms of the TTV's accuracy when base-flex is present, however the data is not available for public consumption. If only that data were available for public consumption, then the true picture of what I'm describing would be laid bare for all to see.
Market forces being what they are, however, prevents such disclosure.
If the heat-die temperature were being monitored as well, then the TTV would be representative of CPU temperature, and would therefore be more suitable.
Quote:
|
if the TIM joint (inc IHS) is poor, then should it be ignored? if the TIM joints (still including IHSes) are inconsistent from the factory (as is widely reported on forums for AMD's processors), then it should be tested and stastical variances quantified. no?
|
Exactly my argument. If such is not measured and is not known, then how can one confidently state that there is no systemic error present?