Ivy Bridge's thermal density "problem" is a myth which has already been busted

 

Yes, but my point is the 'myth' part is that the temps involved may not really be problematic. What kills components is heat content, not temperatures! Heat content is a measure of total energy whereas temp is a measure of speed of random particle motion. You are suggesting there is no myth, but that the solution to the alleged problem is delidding, no?

You didn't state in your original post, but you did test the chip in your mobo before beginning the delidding exercise, right?

 

Yes, but my point is the 'myth' part is that the temps involved may not really be problematic. What kills components is heat content, not temperatures! Heat content is a measure of total energy whereas temp is a measure of speed of random particle motion. You are suggesting there is no myth, but that the solution to the alleged problem is delidding, no?

 

Sure the temps are problematic, if you desire to overclock, which is the only reason people perform this procedure.

Sure the temps are problematic, if you desire to overclock, which is the only reason people perform this procedure.

 

Sure, that's why they perform the procedure. But...is this because they believe the high temps are problematic when they really aren't, or because they actually are? I ask the question because the higher temps in IB processors are higher than w/ SB processors, so this could easily be assumed to be problematic but may not be the case.

 

What is the empirical evidence to warrant this risky procedure, over for example applying an equivalent cooling solution you would use for a SB platform for OC purposes? I am arguing it may be less relevant to go by temp, but instead consider the physics. The argument that Intel engineers inadvertently didn't consider the physics when they chose a different TIM or whatever the alleged culprit is seems on its surface exceedingly unlikely. People clamored over high temps in IB processors and likely this reaction was based on seeing temps they weren't used to instead of, for example, not being able to OC their new IB processors with a cooling solution on a par with what is needed for OCing a SB or Core2 Quad or whatever processor. I think the temps might well be a red herring...they're real, but they aren't as relevant as assumed. Therein lies the myth...

 

Pure conjecture of course. Does removing the TIM lead to lower temps? Apparently it does. How much does this matter? If the theory I am proposing is sound, then IB is actually a superior processor over SB, because it would have been understood/designed to be high temp tolerant, and doing something physical (i.e., delidding) to provide more heat loss than external cooling alone is something you can't do w/ SB.

Sure, that's why they perform the procedure. But...is this because they believe the high temps are problematic when they really aren't, or because they actually are? I ask the question because the higher temps in IB processors are higher than w/ SB processors, so this could easily be assumed to be problematic but may not be the case.

 

What is the empirical evidence to warrant this risky procedure, over for example applying an equivalent cooling solution you would use for a SB platform for OC purposes? I am arguing it may be less relevant to go by temp, but instead consider the physics. The argument that Intel engineers inadvertently didn't consider the physics when they chose a different TIM or whatever the alleged culprit is seems on its surface exceedingly unlikely. People clamored over high temps in IB processors and likely this reaction was based on seeing temps they weren't used to instead of, for example, not being able to OC their new IB processors with a cooling solution on a par with what is needed for OCing a SB or Core2 Quad or whatever processor. I think the temps might well be a red herring...they're real, but they aren't as relevant as assumed. Therein lies the myth...

 

Pure conjecture of course. Does removing the TIM lead to lower temps? Apparently it does. How much does this matter? If the theory I am proposing is sound, then IB is actually a superior processor over SB, because it would have been understood/designed to be high temp tolerant, and doing something physical (i.e., delidding) to provide more heat loss than external cooling alone is something you can't do w/ SB.

 

The high temperatures inhibit one's ability to scale clockspeed as desired via overclocking, due to the thermal threshold value of the Ivy Bridge processors established by Intel as "TJMax" of 105 degrees C on any given CPU core. The empirical evidence has been obtained by anyone wishing to overclock their Ivy Bridge past ~ 4.5GHz without de-lidding, and observing temperatures in excess of 100 degrees C which causes thermal throttling to kick in to protect the CPU by lowering clocks and voltage, thus reducing performance. This is why I state the stock thermal interface material is problematic.

 

The argument is not that Intel engineered the CPU incorrectly, but that the manufacturing choices they have made are limiting users' potential to overclock the chips.

The argument is not that Intel engineered the CPU incorrectly, but that the manufacturing choices they have made are limiting users' potential to overclock the chips.

 

What do you think Intel's intention was in this regard? My only argument is that this was by design. Nice folks have discovered an effective work around. I wonder if there isn't a better way to de-lid them, not that I ever will! Hanging on for Haswell which of course remains to be seen. TJ max for SB was 95? So again, the chip can handle 105 w/o downthrottling.

 

What do you think Intel's intention was in this regard? My only argument is that this was by design. Nice folks have discovered an effective work around. I wonder if there isn't a better way to de-lid them, not that I ever will! Hanging on for Haswell which of course remains to be seen. TJ max for SB was 95? So again, the chip can handle 105 w/o downthrottling.

 

Without sounding too much like a conspiracy theorist (I hope), I think they did it to lit top-end performance in an attempt to sell far more expensive Extreme Edition chips which otherwise have little reason for being, IMHO.

Without sounding too much like a conspiracy theorist (I hope), I think they did it to lit top-end performance in an attempt to sell far more expensive Extreme Edition chips which otherwise have little reason for being, IMHO.

 

Outside of Extreme Ed chips having a different TIM to obviate this issue, is there another interpretation? You aren't saying Extreme Ed chips employ different TIM are you?

 

Outside of Extreme Ed chips having a different TIM to obviate this issue, is there another interpretation? You aren't saying Extreme Ed chips employ different TIM are you?

 

I am because they do. With the current Extreme Edition chips and the previous run of the mill standard Sandy Bridge parts, Intel actually solders the CPU core to the IHS with flux-less solder. This has a much higher thermal transfer rating than traditional thermal pastes. However, the liquid metal products mentioned in this thread close the gap with a thermal transfer rating of 82 w/(m*K). The average thermal paste is somewhere in the 5-10 range and from what I have heard the flux-less solder Intel uses has approximately the same thermal conductivity as the Liquid Metal products. However their paste used in Ivy Bridge parts is on the low end of the range of thermal transfer ratings as far as pastes go (i.e. cheap). Explains the massive temperature drops we see from de-lidding, wouldn't you say?

However their paste used in Ivy Bridge parts is on the low end of the range of thermal transfer ratings as far as pastes go (i.e. cheap). Explains the massive temperature drops we see from de-lidding, wouldn't you say?

 

No. :blink: http://forums.anandt...2261855&page=20 #493

 

The empirical evidence has been obtained by anyone wishing to overclock their Ivy Bridge past ~ 4.5GHz without de-lidding, and observing temperatures in excess of 100 degrees C

 

Who is the "anyone" btw?

 

No. :blink: http://forums.anandt...2261855&page=20 #493

 

 

 

Who is the "anyone" btw?

 

Anyone who has done the procedure and chronicled the before and after results. On the 4 I've done the average temp drop when overclocked is between 23 and 31 degrees C. The variability is due to the cooling capacity of the cooler used in each instance, ranging from a meager Coolermaster 212 heat sink to a custom water cooling solution, with a couple Corsair H100s in between.

 

If your own results don't fall in this range , the choice of TIM and it's application are absolutely crucial. As explained in my previous post, standard pastes just don't cut it, but they do offer from 5-10 C improvement over the stock paste, in my experience.

With the current Extreme Edition chips and the previous run of the mill standard Sandy Bridge parts, Intel actually solders the CPU core to the IHS with flux-less solder.

 

OK, thanks for that. I would agree, that was a decision they made intentionally very likely for the reason you ascribe, i.e. to separate their EE chips from ROTM IB chips, and moreover to make it less of a no-brainer to remedy the situation, hence the resistance to taking on the risk involved of damaging an IB. What is the cost difference between EE and ROTM IB now? Do we know the part number for EE Haswell, and when is it to debut now, Q2 or later? Perhaps they will return to fluxless solder on Haswell parts...

 

Addendum: I see since IB-EE is not even out yet, is projected to be out Q3 2013, so one can safely guess non-EE Haswell desktop variants WILL NOT sport fluxless solder, else there would be little point in releasing IB-EE from all of the above considerations.

 

That one can get 4.5-4-5Ghz out of ROTM IB is pretty important though. Same same for ROTM Haswell. If good water cooling can get either of these chips to those overclocking levels, that's pretty darn good on its own especially when paired w/ fastest RAM albeit its small positive effect on total performance, a decent SATA3 SSD, and makes getting the extra 5-10% clockspeed out of de-lidding (or pricey EE chips) very much moot, especially w/ the hazards involved. FSX's nature is that an additional 5-10% is, I'm afraid, more about bragging rights than anything else. Never will buy into the idea that that extra few % matters more than what it is, 5-10%. When you load up the NGX at FSDT's CYVR in FTX PNW scenery, full cloud cover, rain, autogen and other sliders full right, 10% improvement over 22FPS (or whatever you can get in this scenario w/ stock water cooled IB) will yield a whopping 24FPS, which is hardly significant. Therefore, de-lidding is what it is, mostly about bragging rights I'm afraid, as are the EE chips. Now, if ROTM IB chips couldn't get to those OC levels, then the issue is very much changed. I see you are on a 4.9Ghz IB chip, and so you are saying you are getting a whopping 6% high clockspeed for the risk involved. I rest my case...

Anyone who has done the procedure and chronicled the before and after results. On the 4 I've done the average temp drop when overclocked is between 23 and 31 degrees C.

 

I did not ask about temp drop.

 

I asked who has tried to oc above 4.5GHz and had temps above 100°C.

I did not ask about temp drop.

 

I asked who has tried to oc above 4.5GHz and had temps above 100°C.

 

It's all over the web Rob. IB gets far too hot with overclocks in excess of 4.4 / 4.5GHz. And like TechGuy said, it has nothing to do with the new 22nm process, and all to do with the thermal interface

 

I did not ask about temp drop.

 

I asked who has tried to oc above 4.5GHz and had temps above 100°C.

 

Run a stress testing app that actually uses the CPU fully, not FSX, and you'll see drastically higher temps on Ivy Bridge when overclocking with additional voltage. Go far enough and you'll hit throttling limits at 105 C. De-lid and you can push your chip another 200-300MHz with much lower temps as well. You'll hit the voltage wall before you hit the thermal wall with Ivy of you de-lid, vice versa if not.

 

This has all been said before in several threads on the matter.