So I just started thinking while watching joker's last tgw episode. How long does the TIM Intel uses between the chip and IHS last? We know thermal paste doesn't last forever. I mean sure they seal/glue the IHS to the chip, but how much does that help(if at all)?
These days unless you're doing something CPU intensive often, you don't need to upgrade much. I mean I went from having a C2D e8400 from 2008 to about 2013, to having the i5 4670k I still have in my system. And that e8400 had solder between the chip and IHS. So, anyone heard and/or experienced anything involving the TIM drying up? Or maybe monitored your temps enough to see a difference with age that might suggest?...probably not, but worth asking.
I've never heard of it being an issue. Ivy bridge had problems with their TIM, as did Haswell, which was fixed with better compound for the Devil's Canyon update. Consumer chips used solder until after Sandy Bridge, and still uses them for enthusiast and server chips. but I've never heard of the TIM under the heat spreader needing to be replaced.
First of all, it's not always safe to assume that the manufacturer has correctly applied the TIM. I had a GPU from a well known/respected company that was exhibiting random crashes to a black screen and I ultimately found that only 2/3 of the processor actually had any TIM applied.
Back to your question, I can't speak to Intel specifically, but if I'm doing a spring cleaning and blowing out the cobwebs/updating a 5+ year old PC, I'll replace the TIM while I'm in there. By that time it still tends to function more, or less OK, but it's usually noticeably dried out.
Same here, I tend to replace the TIM between CPU and cooler when the PC I'm working on is over 5 years old and the owner can't tell me if/when it's been replaced already.
Severe overheating is never good for TIM so in those cases I do also replace it.
Never had to deal with TIM between the chip and IHS.
I've made a habit out of replacing it when it's more than 5 years old as well. Or, if I get a PC from some one else, I always replace the thermal paste. Even on GPU's.
Yeah, kind of figured no one has had this issue, or at least knows they have. Just was an interesting thought I had. And it would be a tough one to know as a problem as it's not easy to get to, and the people who do it, do it to new chips.
I know my 770 from 2013 I'm no longer using might need TIM replacement. Custom cooled(by manufacturer) but still hit 80C, although my fan curve wasn't exactly aggressive either. As well I think precision changed the temp target between versions vs when I used to hit maybe 78-79C tops.
Yes this issue is more common than people like to admit. I'm about to delid my 4790K because shit is getting out of hand, I've gained 15c and reapplying thermal paste is no longer helping a damn thing. (Been running 4.7 for 1.5 years) @kewldude007 had to return a 4790K because his chip hit 99c no matter what, though I think that was due to bad application not it drying up, but regardless it is an issue.
From what I understand, this situation is rather ironic. Most people complain about Intel TIM, and rightly so. It isn't the most optimal stuff. But it IS designed for longevity, not necessarily top performance. That's the ironic compromise; the stuff is designed to last, but isn't the best goop for the job. Anyone who delids has to worry more about that stuff than stock folks. I'd be more interested in how long Coolaboratory Liquid metal TIM lasts, since that's what the delidders use. The stock crap will last for years (if not FOREVER), but will suck bum in performance terms. But when you replace that stuff with the liquid metal stuff, how long will THAT last, and will you need to re-do your delidded TIM in the future if you have the CPU for years? Most nerds don't keep CPUs long enough for this to be an issue, since we get "itchy" and need to upgrade more often. :)
My gut feeling tells me that on most CPUs (emphasis on "most") the TIM underneath the IHS will last 5 to 7 years, which is about as long as the regular user keeps his CPU. Gamers tend to upgrade sooner, so they are less likely to run into this issue unless they've been running their CPUs very hot all the time.
Of course there will always be cases where there's an issue with the TIM or where Intel didn't apply enough etc etc.
Makes me wonder if that was ever an issue with soldering. I never heard of the problem, but also never heard of failed application of TIM by them until now in this thread. If it ever was with solder, then obviously it's a lot easier to replace the TIM they use now over solder.
The TIM Intel uses between the CPU die and the IHS is not the problem. It is not bad performance TIM, and I've never heard that it would ever "dry out" etc (look at the warranty for cheap MX-4). I have Ivy Bridge chips that are about five years old in production, not much problems. Motherboards fail much more often than CPUs. Sure a TIM application can fail, but that would show up in QA as the chips are binned after the IHS is glued in place.
Careful testing by several people has shown that there are notably variations in the IHS:es. It's not cost effective for Intel to make perfect IHS:es, there will be variations. The Z-height is the important part for pressure on on die. As mentioned, the Chips are binned after the IHS is glued in place, not before. So a great chip can have a "bad" IHS. The silicon itself probably has less variation compared to the IHS after a while (yields are good etc).
When de-lid is done, remove the glue between the IHS and the chip and you have reduced the Z-height and get better pressure on the die. This is the important part. It can also be necessary to polish/sand the IHS to flatten it out if it is uneven (both inside and outside, remember z-height too). It is possible to measure the Z-height on any given CPU and from that number have good idea if de-lid will help temps or not.
So if a CPU has run okay for a couple of years, it is probably safe to assume that it will run okay for years to come. i would not worry about the TIM between the CPU die and the IHS.
I wasn't so much worried, just more curious in what people thought or experienced. Thanks for the info though, makes a lot of sense, would be foolish(if not almost impossible) not to do a quick qc test while binning processors for different SKU's.
RoHS: with lead being forbidden in solder in the EU, the temperature needed to solder it would be too high, and there would be a huge amount of breakage. It is a very very very big problem in the EU. I did hardware development for about a year, and actually sourced new old stock leaded solder for the lab setup, because there was huge component breakage with leadfree solder, and it became a major burden, not just in terms of price, but also in terms of workflow. You can use leaded solder in the lab, but you can't sell any hardware that contains leaded solder. With leadfree solder, I had a breakage of about 25% of components, so imagine Intel with robotised perfectly temperature controlled machinery having breakage of up to 10-15 %, that is an unacceptable loss. So they did away with all solder, and started using TIM.
The problem is, that Intel doesn't use high quality TIM. They use cheaper silicone-ceramic TIM most of the time. It usually - depending on the thermal stress, which is user-specific - is dry as a bone and should be changed after 4-5 years. Intel chips have great thermal safety mechanisms though, you won't fry your chip even with useless TIM, but it will throttle like crazy and you'll massively lose performance, and your chip will degenerate pretty fast. This is not an issue for Intel because they only have maximum two years of guarantee to give in the EU, one year in US/UK/rest of the world.
It's not just a problem with CPU's though, it's an even greater problem with GPU's, where the thermal stress is much higher. I usually apply new TIM on GPU's about every two years. TIM is cheap, even the really good stuff, and it takes only a coupe of minutes to clean the old TIM and apply new TIM, and it's always better if you apply it yourself than the way it comes from the factory. I always apply new TIM to GPU's the moment I open the box, before I even install them for the first time, because I just know from experience that it will make a difference lol