• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Ivy Bridge's poor TIM, what about future generations?

Turbonium

Turbonium

Platinum Member
I've noticed my i5 3570 temps at stock, with a stock cooler, are still rather high (even with the BIOS cooling curve set to maximize cooling at the cost of noise).

If the ambient/room temperature is really warm (say 26-27°C), and I'm running the chip at full load, it will hit as high as around 96°C. With proper ambient temperature (~20°C), the chip will sometimes go as high as 83°C.

I'm assuming this is normal though for Ivy Bridge due to the TIM used. Did future chips go back to soldering? If not, will they? This is actually quite bad, all things considered.
 
Full load with AVX?

I wouldn't necessarily base entire opinions on personal experience alone.

However, using the stock cooler with synthetic stress programs does bring Ivy temps up, but even at such unrealistic conditions, they are still well within the range that Intel and OEMs are comfortable with warranting.

Your temps do seem high compared with:

load-temp.png
 
My i5-3570 is 58c max with a CoolerMaster 212 EVO @ 4GHz with 700rpm silent fan speed on Prime (AVX). There may well be some issue relating to your TIM, or needing heatsink reseated, etc, but the first thing to do is test it with a better cooler (nearly all stock coolers run far hotter than they should).

Edit: What speeds are you running at? If you're not intending to overclock, you could always try undervolting it, which can easily knock 10c off.
 
Last edited:
Turbonium said:
I'm assuming this is normal though for Ivy Bridge due to the TIM used. Did future chips go back to soldering? If not, will they? This is actually quite bad, all things considered.
Click to expand...
If you've read Idontcare's delidding experiments, you'd remember that the TIM's actually good stuff, and is comparable to the best thermal pastes available on the market. If you were to use the stuff on a heatsink, it'd perform very well.

The primary issues are that the thermal conductivity is low, and there's a gap between the IHS and the die. Back when Intel used solder, the gap wasn't an issue. This is where thermal conductivity comes into play.

Mn1nr31.png

Since solder has over 10x the thermal conductivity of the TIM used in IVB/HSW, there's no heat dissipation issue with Sandy Bridge and its predecessors.

Haswell, so far, uses the same TIM (which is probably Dow Corning TC-5688). The upcoming Devil's Canyon SKUs will use an improved TIM, which might be solder (probably), or might be something else. I'd imagine those improvements will be passed on to Broadwell-K as well.

So... enough with the physics/history lesson. As others have commented, your temps are higher than they're supposed to be with a stock cooler. I'd definitely look into re-pasting and re-seating the heatsink. Hopefully you've got some thermal paste and some IPA on hand.
 
I almost wonder if Devils Canyon is an entirely different mask that Intel stiched together to try to seperate out the heat better (at the expense of die space?) to eek out a few more Mhz, like it was an R&D experiment.
 
I'm not praising Intel for the TIM vs solder choice, still it's curios how widespread is the misconception about the poor paste when it's actually so good instead...
It's really all a node problem that occured at 22nm process, of course it's more evident in the overclocker community because if at stock you're close to T-max then ther's no more margin for them. Blame the mobile crowd and the fact that Intel isn't engineering for top desktop performance anymore instead.
 
Homeles said:
So... enough with the physics/history lesson. As others have commented, your temps are higher than they're supposed to be with a stock cooler. I'd definitely look into re-pasting and re-seating the heatsink. Hopefully you've got some thermal paste and some IPA on hand.
Click to expand...
I'm really paranoid about damaging something or simply "wearing out" the socket, even though it's probably an either "you break it or you don't" sort of situation.

Are the temps I'm at now (mid 80s) really an issue in terms of anything tangible? I don't care much about lowering the life of my CPU by a couple years.
 
Turbonium said:
....
I'm assuming this is normal though for Ivy Bridge due to the TIM used. Did future chips go back to soldering? If not, will they? This is actually quite bad, all things considered.
Click to expand...

I think Ivy/Haswell extreme edition Xeons are soldered but the others are not.
And your hot sample might be caused by a slightly higher than normal gap (which the solder would automatically alleviate) and might be a good candidate for delidding and 'lowering'.
 
A decent cooler should make a big difference. Stock coolers just about get the job done, usually.
 
Are you sure the stock cooler is attached correctly? Those plastic things can be a pain.

If that's a non-K processor, I'd get an aftermarket cooler, before voiding the warranty by delidding.
 
jpiniero said:
I almost wonder if Devils Canyon is an entirely different mask that Intel stiched together to try to seperate out the heat better (at the expense of die space?) to eek out a few more Mhz, like it was an R&D experiment.
Click to expand...

Doubtful. Intel couldn't possibly sell enough single-K SKU "Devil Canyon" chips to recover the investment required to make such a specialized and unique CPU layout.

It would be possible had such a re-designed Haswell layout been the basis for the entire Haswell Refresh lineup such that Intel could amortize the layout expenses across all the chips they will sell as Haswell Refresh chips. But that isn't what happened with Haswell Refresh.

I suspect Devil Canyon will be soldiered and specially binned for higher clockspeed capability on the basis of lower Vcc per GHz, but other than that it will be your standard run-of-the-mill quadcore haswell chip.
 
Turbonium said:
I'm really paranoid about damaging something or simply "wearing out" the socket, even though it's probably an either "you break it or you don't" sort of situation.

Are the temps I'm at now (mid 80s) really an issue in terms of anything tangible? I don't care much about lowering the life of my CPU by a couple years.
Click to expand...
Well, those stock coolers just use plastic clips. They don't really exert much strain on anything. Not much risk involved at all, especially with those coolers.

Aside from running a slightly higher power bill, and the fan getting loud at times, it's not going to make a difference.

I've personally ran into issues with them not seating well. The first system I built was idling at 60c the first time I put the heatsink on. I like to press down on the metal part of the heatsink, in an effort to get it to spread out the paste and make better contact.
 
It's definitely not the clips being improperly seated, I triple checked it, and have had prior experience with this exact (and mediocre) cooler before.

I wish I knew the cooler was this bad though (I knew stock ones aren't great, but this one is bad even for stock). At this point, I would get an aftermarket cooler with the rear bracket thing, if it wasn't for the fact that I have to take apart half my system to do that, and go through all the risks that involves. :/
 
SAAA said:
I'm not praising Intel for the TIM vs solder choice, still it's curios how widespread is the misconception about the poor paste when it's actually so good instead...
Click to expand...
The TIM is changing in Devil's Canyon; that alone tells me the old one was no good.

It's really all a node problem that occured at 22nm process, of course it's more evident in the overclocker community because if at stock you're close to T-max then ther's no more margin for them.
Click to expand...
It's actually nothing to do with the 22nm process. Did you not see the numerous delidding experiments online where changing the TIM made a huge drop in temperature?

Also as far as transistor/mm2 goes, AMD/nVidia have massively higher densities on their GPUs, yet their die shrinks consistently reduce temperatures.
 
BFG10K said:
It's actually nothing to do with the 22nm process. Did you not see the numerous delidding experiments online where changing the TIM made a huge drop in temperature?

Also as far as transistor/mm2 goes, AMD/nVidia have massively higher densities on their GPUs, yet their die shrinks consistently reduce temperatures.
Click to expand...
It actually does have a bit to do with the process. FinFETs perform worse than planar FETs at high voltage (> 1.3 V)

http://download.intel.com/pressroom/pdf/kkuhn/Kuhn_22nm_Device.pdf

Slide 13.
 
Homeles said:
It actually does have a bit to do with the process. FinFETs perform worse than planar FETs at high voltage (> 1.3 V)
Click to expand...
Again, numerous delidding experiments put temperatures roughly back in line with SB, including when overclocking. That alone tells me the process has little (if any) effect on the issue.
 
OCGuy said:
I'm still wondering what is causing discomfort temp wise?
Click to expand...
Is this directed at me?

If so, I was wondering if it's actually going to slow anything down. Afaik, it isn't in throttling territory, so it won't.

Really though all I was asking was if the TIM is going to be changed to soldering in the future, and people said it will be for Devil's Canyon. I was just wondering, really.
 
Intel haven't mentioned Soldier yet, from what i've read from press releases it's worded so it gives the impression it is improved TIM. I wonder what made them change over in the first place?
 
BFG10K said:
Again, numerous delidding experiments put temperatures roughly back in line with SB, including when overclocking. That alone tells me the process has little (if any) effect on the issue.
Click to expand...
Regardless, new nodes are supposed to increase performance, not stand pat. The process is certainly part of the issue.
 
I pulled the lid off my 3570K immediately, applied better TIM, and put it back on. Temps dropped 15C under load. YMMV, but seems worth it to me.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top