Intel's TIM continues to create heat bottleneck, many years after Ivy

[DrMrLordX]

All but the AM3+ FX series it seems. So all FMx. Tho it seems that there is conflicting data if 7850K and 7870K is solder or TIM.

All KV-A1 Kaveri chips use TIM. All GV-A1 Kaveri chips use solder.

 

[superstition]

Except it is more heated

Heated isn't the right word. It has mostly been strawmen, red herrings, and various fallacies thrown about — with a specific strategy of repeating the same rebutted bits again and again. I would use the word noisy.

There were some useful posts earlier on but the noise increasingly drowned out the signal. I admit that the way I framed the discussion was a tad dramatic but that doesn't explain the level of nonsense in that thread. You should have seen some of the posts that were scrubbed. I was called all sorts of things, like a fraudster and so on by a poster that I think is a sockpuppet. One post alone had like five different ad hominems in it.

Attempts to rebut include:

1) People should feel very grateful that Intel lets them buy their products.

2) Enthusiasts don't matter at all to Intel so they're fortunate they can buy anything at all. (I guess Devil's Canyon was marketed to casuals and there was no PR about improved TIM.)

3) There is no evidence that the TIM causes a heat bottleneck.

4) There is no evidence that delidding is even worse to support given the thinner substrate.

5) There is no way to see that Intel has a strong incentive to use substandard TIM to encourage delidding.

6) "No one is forcing you to X." (Which avoids the fact that Intel is forcing us to have only certain choices that we can make, something the OC3D article's author pointed out clearly.)

7) Intel is a totalitarian regime so we should never expect them to make products we actually want to buy.

8) Consumers are completely powerless (despite a capitalist market economy), should remain that way, and should be pleased about it.

9) Few enthusiasts care about the heat bottleneck due to the use of low-performance polymer TIM.

10) I am a crazy/bad/naive/fraudulent/incompetent/young person.

11) If you want solder you should stop whining and get a Haswell E (which ignores the fact that that does not have a Skylake core for things like H.265 QuickSync, the efficiency improvements made to the cores such as how integer is handled, the power consumption improvements from 14nm, and the fact that six - eight core CPUs are not needed for gaming rigs but fast quads that are not heat bottlenecked by substandard TIM are in demand for that).

12) There is no evidence that gaming is driving purchases of higher-end PC components, like i7s (see point 11 about the need for six - eight cores).

13) Solder or liquid metal TIM would be so expensive per chip that enthusiasts would be very upset (riiight).

14) Solder can't be used because the process is too small (buy apparently it could be used for Haswell E but not regular Ivy).

15) Solder is too unreliable to be trustworthy (no proof for this was given, of course).

16) It would take a lot of money/effort/time to be able to use solder (as if solder is a brand new thing).

17) Businesses have only one purpose = produce profit for shareholders. As a result of there being just one purpose, they can completely ignore what potential/past/current want to buy. Businesses don't have a second purpose (don't alienate the customer so they will pass on buying your product, spend less on an inferior product of yours that has a lower margin for you, or purchase another company's product).

I may have forgotten others but I have yet to see data or logic that actually rebuts what I was saying and the data I presented to back it up. I haven't checked the thread for a while so it's possible something substantive has been added but I wouldn't hold my breath.

 

[superstition]

The TIM isn't the issue, the gap between the IHS and the core has always been the issue.

Even if that's true then why is there a gap?

With solder there would be no gap, correct? Then it is a problem with the TIM since the use of TIM creates the gap.

 

[superstition]

why is this even a poll. the TIM situation is scandalous, and all cpus should have solder-quality TIM.

I made this poll because certain folks elsewhere make claims like:

1) very few enthusiasts care about this
2) more enthusiasts want to buy chips with the polymer TIM

 

[frozentundra123456]

"Heat Bottlenecked" seems a strange term. It is not like you could overclock to 5.5 or six ghz with solder. You eventually reach other bottlenecks anyway.

In any case, I would prefer both that Intel use solder (at least on the K chips) and made a mainstream Skylake hex core. However, I dont really see it as a huge deal. I think the best you could hope for with perfect solder is what, 200 or so more mhz? Maybe 5% more overclock? Doesnt seem like something worth getting into a lather about, really.

 

[superstition]

the best you could hope for with perfect solder is what

16C less at stock and 20+ C less overclocked, according to the 6700K data in the first article, from Liquid Pro. Top-tier solder may be better.

Doesnt seem like something worth getting into a lather about, really.

20C of extra heat is a big deal to a lot of enthusiasts.

 

[beginner99]

"Heat Bottlenecked" seems a strange term. It is not like you could overclock to 5.5 or six ghz with solder. You eventually reach other bottlenecks anyway.

Yeah you don't get that much more due to process limitations. However it maters greatly for power use and noise of the CPU is 60 or 90 degrees hot.

 

[hawtdawg]

Even if that's true then why is there a gap?

With solder there would be no gap, correct? Then it is a problem with the TIM since the use of TIM creates the gap.

The glue creates the gap, that's why removing the glue fixes the problem. You could probably slap the thing back on leaving the Intel TIM on it and still drop 15C+. You could put ketchup between your CPU and heatsink, and it wouldn't be 20C hotter than good TIM. The TIM is not the problem.

 

[coercitiv]

The glue creates the gap, that's why removing the glue fixes the problem. You could probably slap the thing back on leaving the Intel TIM on it and still drop 15C+. You could put ketchup between your CPU and heatsink, and it wouldn't be 20C hotter than good TIM. The TIM is not the problem.

Doesn't the gap serve a purpose? Is thermal expansion negligible?

 

[MongGrel]

because some idiot is going to buy it crank down the heatsink crack it and complain on forums about how he got ripped off.

then he'll rma it and intel won't cover physical abuse and he'll complain even more.

so to skip all that they just won't not include one.

Pretty much this. You see enough people just bend/break pins here and there installing a CPU.

Would be like trying to sell a new car without air bags and seat belts these days.

It's not hard to organize people these days.

An organized Intel boycott

Intel doesn't care on way or another about anyone "shilling" for them, I'm sure.

Good luck with that one

 

[Borealis7]

IDC tested the Intel TIM (re-applied it after heating it on a spoon like a Heroin addict ) and his findings were that it's not good.

http://forums.anandtech.com/showpost.php?p=33785983&postcount=53

so even if you fix the gap, replacing the TIM would still be required, and that's not a trivial action to perform.

 

[coercitiv]

IDC tested the Intel TIM (re-applied it after heating it on a spoon like a Heroin addict ) and his findings were that it's not good.

http://forums.anandtech.com/showpost.php?p=33785983&postcount=53

so even if you fix the gap, replacing the TIM would still be required, and that's not a trivial action to perform.

From the post you linked to:

Now obviously this result is not indicative of the actual performance of the CPU TIM that Intel is putting on their Ivy Bridge CPUs, these results are botched because of the inability of the TIM to be recovered and reapplied to subsequent surfaces like my 2600k's IHS.

All well, I guess for now the mystery of "how good, or crappy, is the stock Intel CPU TIM?" will have to go unanswered.

 

[ShintaiDK]

It'll still take a couple of years before we start seeing the biggest payoffs of those investments. Looking forward to it, though.

Heck, this year is pretty damn exciting. Potential Zen launch at year's end, new graphics cards... Kaby Lake might actually not be the letdown that Skylake was for me.

Good thing I'll have my own tech website up and running in maybe a month's time

I see 2016 as a really dull year. A cooler and a new GPU is the most interesting parts of 2016 in my book. I think it was moonbogg that also just wanted 2016 to be over with

 

[ShintaiDK]

Even if that's true then why is there a gap?

With solder there would be no gap, correct? Then it is a problem with the TIM since the use of TIM creates the gap.

There is still a gap with solder. Its no different.

 

[coercitiv]

There is still a gap with solder. Its no different.

Logic dictates there must be difference. If the gap is the same yet soldered CPUs have significantly improved thermals, then the gap cannot be the main culprit for the degraded thermals to begin with. So either the TIM is the main culprit, or the gap is somewhat different.

 

[ShintaiDK]

Solder is just better than TIM. But solder got its own cons.

 

[III-V]

An organized Intel boycott

Intel doesn't care on way or another about anyone "shilling" for them, I'm sure.

Good luck with that one

The TIM issue isn't a subject that I particularly care about. There are bigger issues that are more deserving of my attention... regardless, it's not hard to get 10s of thousands of people riled up, if you dedicated yourself to a particular cause.

Also, I'll bet that Intel does in fact care. PR can make or break a company.

 

[MrTeal]

Logic dictates there must be difference. If the gap is the same yet soldered CPUs have significantly improved thermals, then the gap cannot be the main culprit for the degraded thermals to begin with. So either the TIM is the main culprit, or the gap is somewhat different.

Solder isn't really a terrible good conductor of heat, as far as metals go. Standard 63/37 is ~50W/mK, and some solders used for die attach will go from 50-70W/mK. That's still close to an order of magnitude lower than copper at 400W/mK. Still best to minimize that bond line and get the heat into the IHS ASAP.

If idc's measurements on his 3770k were correct (0.16mm gap between IHS and CPU for the 160mm² 3770k die), you'd expect a thermal conductivity of 60W/mK to give a thermal resistance of 0.0167K/W, or a temperature rise of under 2C for a 100W load through the bulk of the solder. It's not negligible, but it is still pretty small. Common thermal pastes could have a thermal conductivity an order of magnitude higher than solder though, so it doesn't take much for 1-2C to become significant.

 

[superstition]

50-70W/mK. That's still close to an order of magnitude lower than copper at 400W/mK.

Far better than polymer TIM. That's like 5, right?

 

[superstition]

Logic dictates there must be difference. If the gap is the same yet soldered CPUs have significantly improved thermals, then the gap cannot be the main culprit for the degraded thermals to begin with. So either the TIM is the main culprit, or the gap is somewhat different.

Yes, the gap hypothesis doesn't add up unless one says it's the consequence of using polymer TIM in which case it's still a problem due to the use of polymer TIM.

 

[Dufus]

because some idiot is going to buy it crank down the heatsink crack it and complain on forums about how he got ripped off.

then he'll rma it and intel won't cover physical abuse and he'll complain even more.

so to skip all that they just won't not include one.

Mobile CPU's have been without an IHS for many years and until Broadwell enthusiasts changed out CPU's.

Maybe if/when DT CPU's go BGA then perhaps no more IHS.

 

[superstition]

Mobile CPU's have been without an IHS for many years and until Broadwell enthusiasts changed out CPU's.

Maybe if/when DT CPU's go BGA then perhaps no more IHS.

Athlon XP chips were available on the desktop without an IHS but I don't think Intel is ever going to sell a bare CPU because it could be a warranty nightmare.

 

[Burpo]

Didn't Mac Pro LGA1366 Xeons come with out IHS?
http://www.ebay.com/itm/Intel-Xeon-...8&pid=100005&rk=1&rkt=6&sd=121849135476&rt=nc

 

[superstition]

Didn't Mac Pro LGA1366 Xeons come with out IHS?

Giving OEMs a bare die is one thing.

Apple also used liquid metal for its highest-end G5 models as far as I know.

 

[Burpo]

Obviously available to the public, so warranty doesn't look like a problem.. Intel does what it wants..