[time-nuts] Thunderbolt temperature spikes
hmurray at megapathdsl.net
Mon Jul 14 17:51:34 EDT 2008
> Looking back through some old logs with my New and Improved Spike
> Finder (tm), it appears that spikes seem to occur, on average, around
> every 2-3 hours and their effect shows up in the data for around 20
> seconds... so figure on 1 part in 500 of the temperature data is
> corrupted by their effect. It is almost certain that the spikes are
> not caused by real temperature changes, but are the effect of some
> noise or firmware glitch. In a system where performance is measured
> in parts per trillion, something that effects .2% of a critical
> parameter's measurement is bound to have an adverse effect.
What do you mean by "not caused by real temperature changes"?
Do you mean the external temperature isn't changing so you don't expect the
internal temperature to change? Or do you think there is a bug in the
temperature measuring stuff?
It would be interesting to add a couple more sensors on nearby chips.
My guess would be either of 2 events:
The CPU is doing some extra work.
There is a bug (hardware, firmware, both...) such that there is an
occasional bus clash for a short period.
If the CPU is doing the work, then the temperature of the CPU should show the
spikes, and maybe the memory chips too.
If it's a bus clash, then the spikes should appear on the pair of chips
involved in the fight.
These are my opinions, not necessarily my employer's. I hate spam.
More information about the time-nuts