The Haswell-E Heat-spreader and Die-Size -- Implications for Cooling

[BonzaiDuck]

Of course, I can now watch CNN in the small, single-seat room during my morning routine, but I still keep my pile of Maximum PC magazines in the small magazine rack near the roll of Cottonelle. So I was reviewing the latest issue on the Haswell-E, the preceding IB-E and SB-E chips.

A photo displayed the size of the heatspreader for the "E" processors next to a photo of the IHS for the i7-4790K. the E heatspreaders are enormous.

Further, for purpose of the Max PC article, someone had delidded an i7-5xx0K processor. The die is square -- not rectangular -- apparently with proportionately more area (and IHS contact) consistent with the two (or four) extra cores. This seems roughly equivalent to the extra TDP wattage -- from ~90W for Devils Canyon to 140W for the E cores.

So it occurs to me that the E's won't require too much more cooling -- water or air -- than the Haswell refresh.

Any further observations about this?

 

[mistersprinkles]

That's about right. Greater surface area and the fact that the IHS is soldered are factors in chips like the 5820K showing similar temps at similar clocks to the 4790K on the same cooler.

You can take a 4790K to 4.6Ghz+ on an H110 and the same goes for a 5820K.

BTW the size of the IHS/heatspreader is irrelevant. The heat is transferred mainly at the part that is covering the CPU die. The heatspreader is so large as a consequence of the much higher number of pads/contacts on the bottom of the CPU package, rather than a consequence of a larger CPU die.

 

[BonzaiDuck]

That's about right. Greater surface area and the fact that the IHS is soldered are factors in chips like the 5820K showing similar temps at similar clocks to the 4790K on the same cooler.

You can take a 4790K to 4.6Ghz+ on an H110 and the same goes for a 5820K.

BTW the size of the IHS/heatspreader is irrelevant. The heat is transferred mainly at the part that is covering the CPU die. The heatspreader is so large as a consequence of the much higher number of pads/contacts on the bottom of the CPU package, rather than a consequence of a larger CPU die.

Yeah . . . no matter how much better the heatsink-base can absorb heat from the IHS, the transfer of heat to the latter is the only relevant determinant -- because it is first interface with the source and has the smallest surface-area.

 

[VSG]

Yup. I was also surprised how big the 5960x was compared to the other 2 Haswell chips (G3258 and 4770k) I have:

 

[mistersprinkles]

Yup. I was also surprised how big the 5960x was compared to the other 2 Haswell chips (G3258 and 4770k) I have:

Lock your doors. An AMD Kabini with blond hair is going to come to your house, eat your porridge and sleep in your CPUs's sockets.

"This one has too many pins"
"This one has a bent pin"
"This one is juuuuust right"

 

[BonzaiDuck]

Lock your doors. An AMD Kabini with blond hair is going to come to your house, eat your porridge and sleep in your CPUs's sockets.

"This one has too many pins"
"This one has a bent pin"
"This one is juuuuust right"

I still cannot get rid of the intuitive appeal presented by a larger heatspreader. On the one hand, only so much thermal energy per increment of time can be transmitted to the IHS by a silicon die of a given size: it is size-dependent.

Suppose, however, that a water-block or heatsink base is larger than the IHS? Heat transfer with a larger IHS would be more efficient.