The great core/heatspreader myth

[MaximumGoat]

I always see it written that larger cores (i.e. 256kb => 512kb cache tbred to barton for example) overclock better etc. because they have a larger surface area, now it is true that the AVERAGE thermal output per unit of surface area has lowered, which is what makes people think its better. But remember when you add extra cache, the size of the ACTUAL "processing part of the chip" has not got any bigger, and still has same thermal output and surface area, just because there is an extra 256kb of cache sitting beside does not mean the heat is transferred any quicker, it would do if the "processor part of the chip" was mixed in with the cache, but this isnt true. the heat doesnt magically spread evenly over the whole chip, thus extra cache doesnt help in cooling. If anything extra cache would make it worse as it would generate heat of its own, but this is cancelled out by the fact that new cores are probably optimised slightly and dont have to access memory as much if there is more cache.

another way of explaining this is that when you add cache the hot part of the chip doesnt get any bigger, thus has same thermal output per unit area. This is quite difficult to explain, and i am bad at explaining things but it is true.

p.s. this is assuming transistor size is the same of course

now, heat spreaders dont help either as they just add another barrier to the heat to get to the heatsink, the only possible help a heatspreader adds is that it is in contact with the side of the chip as well as just the top, but heatspreaders are good though i think for just protecting the core.

forgive me for being so bad at explaining things,

 

[c0d1f1ed]

Silicon has excellent heat transfer. A proof of that is that in flip-chip packaging it doesn't have much trouble conducting heat to the bottom side of the die. But I'm not sure if this would influence overclockability of processors with larger caches because they are a completely new stepping. Celerons often indicate the opposite.

A heat spreader does not make contact with the sides of the die. Besides protecting the die, the most important thing about them is that they make better contact with the die than what you get with cheap thermal compounds. Dell employers (and others) won't take the time to neatly apply Artic Silver III so it's done mechanically. That's Intel biggest buyer so in average a heat spreader does improve thermal conductivity.

 

[slvr_phoenix]

Besides protecting the die, the most important thing about them is that they make better contact with the die than what you get with cheap thermal compounds. Dell employers (and others) won't take the time to neatly apply Artic Silver III so it's done mechanically. That's Intel biggest buyer so in average a heat spreader does improve thermal conductivity.

Maybe I'm just having an off day, but I don't understand how that makes any difference whatsoever. You <i>still</i> need to apply a TIM between the heatsink and the heat spreader. So the spreader really has no effect whatsoever on improving the thermal conductivity between the core and the heatsink, no?

And also, don't major OEMs (like Dell) typically use heatsinks that have preinstalled phase-change 'pads' so that they don't even have to mess with thremal grease at all, neither mechanically applied nor human?

<font color=purple><pre><A HREF="http://www.winamp.com" target="_new">Winamp<b><font color=blue>3</font color=blue></b></A> and freeform skins, the best thing since sliced llama loaf. (Now with more beef.)</pre><p></font color=purple>