[hokiealumnus]

77W was confirmed directly from Intel.

[Bobnova]

Good to know, thanks.
I was hoping it was.

[I.M.O.G.]

You are right about the power density, I did the math on it as well after posting that - because I assumed it was higher... But really it isn't higher by much. (not enough to explain the temp problems in my opinion, but lack of solder between the IHS would explain it well)

[Bobnova]

The non-solder TIM is my bet as well. .43 to .48 is a jump, but it isn't that big.
It also seems like the 3d transistors may leak a lot at higher voltage levels. That was something I noticed with BD as well, at stock volts it ran pretty cool, raise the volts a bit and that changed rapidly. More rapidly than with previous chips I've worked with.

We may be entering an era of heat output not following voltage in the classic way.

If it wasn't a $300 puck I'd rip the IHS off mine when they get here, but as it is I can't really afford to throw $300 out the window. Clearly I can to an extent as I intend to bench the hell out of them, but high voltage and LN2 seems safer than a razor blade.

[Audioaficionado]

Well if you kill it fairly soon after you get it, you could just do an autopsy and send out the special sauce for analysis

[Bobnova]

I'll drive up to medford and borrow yours
Need to go up there and hit the kart track again someday anyway

[wingman99]

Did SB have solder between the IHS? I now see that there is higher power density by a little, however intel says IB runs cooler stock and we know at stock it uses less watts than SB.

Matt i hope intel will talk with you about the overcloking with excessive heat being generated needing excessive cooling.

EDIT my sandy bridge runs so cool and fast compared to 45nm core2 quad and IB seem like a big step backwards for air cooling.

[muddocktor]

I can think of a couple of reasons why Intel might have gone back to the traditional TIM between the IHS and slug (if they did so).

1. No competition from AMD, meaning that they didn't need any higher ambient overclocking than they already have with SB for enthusiasts that don't go cold. Let's face it, nothing AMD is producing can hang with a 2600k, much less IB.

2. More profits. Smaller cores, more good dies per wafer and money savings by using a cheaper IHS>Core bonding process, offset just a little by slightly more expensive wafers=insane profits for Intel.

[Bobnova]

TIM is far cheaper than the specialized solder they were using. Even high end TIM.
I bet it saves a decent amount per chip in time and heating too.

They split the 775 stuff, e7xxx and lower were TIM, e8xxx and higher were soldered.
P4s were all soldered, they were too hot not to be.

[JigPu]

Great article The power figures are impressive, but I'm still happy with my Sandy Bridge system. Undervolting the chip, I can run 24/7 at full load with my fans inaudible and still have reasonable temperatures.

Originally Posted by Bobnova

Regardless of the scale used you cannot accurately use any method to estimate your eventual temps without knowing the ambient.
If you know your ambient, hokies ambient, and your SB core temps you can estimate your eventual IB core temps using his percentile just fine. Alternatively you can simply look at the degrees over ambient (in C) and estimate that way. Both ways will work fine, neither will be exactly accurate unless you have the same cooler and fan of course.

This.

If you want to indicate a the difference in terms of percent, do it on deltaT, not T itself. In addition to being a more meaningful statement, it has the nice property of being applicable to *any* unit: if your deltaT doubles in Celsius then it also doubles in Fahrenheit.

JigPu

[I.M.O.G.]

Secret Sauce Revealed: Looks like non-solder TIM to me (Thanks to pt1t):



Now, let's refer back to the recent past (thanks to Muddocktor), the E6XXX series of Conroe chips were much easier to cool compared to the E4XXX series. The original Conroes used the bonded TIM, and the E4XXX series used the traditional TIM.

[rge]

Well that settles that question, thanks for posting that.

Also the E7200 (which had paste) had higher core temps and also had more issues with higher temp differences between cores than soldered counter parts.

Makes direct die cooling attractive again, well except the issues with increased risk of mobo/cpu shorting from having to alter/remove cpu socket and suboptimal contact between cpu and mobo. Good for testing, more risky for 24/7 use.

I wonder if could get ahold of any solder attach or newer die attaches, some 2x higher thermal conductance than indium solder, indium graphene 160 w/mk vs solder 80 w/mk. But my guess is none would sell 1-2 units.

Either that or new direct die waterblock sold with new socket attach

[hokiealumnus]

Originally Posted by JigPu

If you want to indicate a the difference in terms of percent, do it on deltaT, not T itself. In addition to being a more meaningful statement, it has the nice property of being applicable to *any* unit: if your deltaT doubles in Celsius then it also doubles in Fahrenheit.

JigPu

Thank you for that; and you're absolutely correct. Ambient was very close to the same between the two within two degrees Celsius. I'm kicking myself for not measuring when I took it, but that room stays very close to 23°C, ticking a degree or two one way or the other (except in the middle of summer, when it can heat up more). Kicking myself for forgetting to actually measure the temperature. I guarantee ambient temps were within two degrees of each other though.

[Audioaficionado]

Looks like you would want to mount the board so this delidded processor lays flat and the heat sink will be sitting on top rather than hanging off of it.

[Bobnova]

Pretty bad contact there too, looks like there was a gap all the way around that was filled with TIM.

[Frakk]

Bobnova... yeah.

That TIM looks pretty think between the DIE cap and the Heat Spreaders contact plate.

Is there any sort of contact between the two or is it a gap filled with TIM?

I would not like to see anywhere near that much TIM between the Heat Spreader and the CPU cooler contact plate when i occasionally pull it off to reapply.

I would wonder WTF i was thinking the last time i applied.

[Bobnova]

The reality of modern TIM is that you cannot put too much and have it cause temp issues. Modern TIM thins under heat and squishes out the side. You can put too much on and have an entertaining cleanup time. You can put too little on and have temp issues. Too much? No.
If there is a gap with TIM in it, it means that the two surfaces mechanically cannot come closer together, the TIM will not stop them from getting closer if other factors allow it.

[Janus67]

So basically ideally we wouldn't see any TIM on the die/heatspreader area then if it was making perfect/near-perfect contact?

[bmwbaxter]

I hope this is only going to be an issue with review samples and ES chips. Hopefully retail will fair much better.

The contact between core and IHS is horrible! No wonder temps are so bad.

[Bobnova]

We'd see a very thin fairly translucent layer, rather than the thick opaque one shown.