Originally Posted by renethx
Simply there is no information on what TIM is used in Haswell-E (Core i7 5960X, 5930K, 5820K, all unlocked of course). Considering relatively low base clock of these processors (3.00-3.50GHz base, TB is unknown), NGPTM may not be used. Who knows.
Sandy-E and Ivy-E were soldered, it is a safe bet Haswell-E will be as well. Soldered IHS beats stock TIM, though having the option to de-lid does give you a chance to surpass it. (I'll pass)
Their "TIM experiment" to save a few pennies per unit got quite a bit of unexpected backlash, hence the whole DC PR
counter-spin coming into existence at all.
I would not expect them to repeat it with models that are steeped in enterprise and often in thermally limited environments such as blades or cloud nodes smaller than 1U, they are very tight fisted with clocks and TDP, have to pay out the nose for both. The volume is a lot lower and the margin a lot higher on "real" xeons, too risky. (E3s are nothing more than feature-complete consumer cpus and/or perhaps consumer cpus are bin-failed E3s)
Buyer beware when dealing with "socket 2011" in the future, intel in its usual giant corporation-style stupidity kept the pin count the same while changing the layout, no compatibility. None that knows better expected any moving from ddr3 to ddr4, but adding or subtracting a pin to avoid confusion is so blindingly obvious only a massive organization with such inertia could miss it.
Haswell-E X99 will be "2011-3"
By the way, according to all documentation so far (mostly by omission) the 20th anniversary pentium G3258 does NOT have the improved TIM, only the specifically noted DC i5 and i7
. Delid that badboy if you are going for the golden overclock...
Which reminds me, anyone have a reference as to why they chose the number 3258? I'm sure it means something but I was nerdy on other things back then.