Replies to post #33226 on Advanced Micro Devices Inc (AMD)

[dougSF30]

You're missing the point. Read what he wrote again, more slowly.

Doug

[Pravin]

Chipguy,

" Your individual FET must be quite brilliant to
discern the operating frequency's specific contribution to its"

Now you're just being arguementative. It's a pretty simple concept. Of couse the package and heatsink are not varying. Increasing the frequency increses the power dissipation which increases the temperature, which increases the leakage. At a high enough (local, on-chip) temperature, thermal energy will be enough to generate electron-hole pairs that are available for conduction across the channel. This carrier generation should increase exponentially as the thermal energy approaches that of the bandgap energy, and result in an avalanche process.

Pravin.

[sgolds]

chipguy, I'm not sure why you wanted to persist in this point -

Your individual FET must be quite brilliant to
discern the operating frequency's specific contribution to its
junction temperature from the other design and operational
variables.

No one here claimed any such thing! Your original post #msg-2967659 seemed, in a bit of a snippy way, to apply the little man method of thinking to me: ...individual FETs are somehow aware.... That is, as soon as you get to a level of complexity that you can not explain, simply draw a little man in your diagram who is doing the actual work.

Didn't even go for it even in grade school when this was used commonly in animated science films! I certainly have not developed the 'little man' world view as an adult.

When I see an area of knowlege where I am not schooled, I attempt to understand the empirical evidence so that I can understand how something predictably works, even if I do not know why it works. That is the methodology that I applied before making my statements about clock rate and power consumption; other gentlemen here filled in the details of why processors work this way.

Empirical reasoning is how most science progresses. Often we can prove something works well before we can explain why it works. We see this especially in medicine* - how many years was it between the proof that asprin surpresses pain, and the understanding that it inhibits neural transmitters in synapses? Anwer: About 100 years (late 19th century to late 20th century).

However, for many centuries before that deer stags would rub their antlers (at developmental stages when they are painful) against the bark of the willow tree, a rich source of asprin. Since I did not conduct a systematic review of the literature on transistor heat & power usage and apply a statistical analysis, my assertion was more like that deer stag - not scientifically verified, but still correct.

*An example from physics: Gravity. Galileo proved the basic principle at the Tower of Pisa, but it took Newton to partially explain why it works, and Einstein explained it to a deeper level.