VR-Zone: Ivy Bridge high temps ARE caused by thermal grease

Pyroja said:
Get back to me when this change from Intel ACTUALLY leads to "a constant influx of field incidents due to heat failure".

Also, "a single large business literally spending 10's of millions of dollars on PC hardware alone" =/= "Best Buy shoppers".

This is only a big deal to us enthusiasts and overclockers. A multi-million dollar business won't be OCing their CPUs, they'll buy the ones that perform as they want at stock. And this change isn't going to affect them because, as it turns out, somehow Ivy Bridge happens to keep on truckin' like it's supposed to despite running a few degrees warmer than SB.

And for all this fuss, there are still people out there getting BETTER overclocks on their IBs than they did on their SBs. As always, this OCing game involves a lot of luck.

Does this change suck? Yes. I'm not thrilled about it, either. Is it some big conspiracy from Intel? Uh, highly doubtful. Does this make IB a total turd? Not at all.

It's not a change that makes me happy, but I think it's being blown quite a bit out of proportion.
Click to expand...

I think you misunderstand me. I never said that IB was going to lead to an influx of field incidents due to overheating. I said that it's tested. I've posted in the past that IB will typically run cooler and draw less power than SB, until that OC threshold is passed, then it's like hitting a brick wall. Not relevant to the cash cow corporate customers, this is apparent. It would be a stupid ass move for OEMs to throw system cooling to the wayside, as that would lead to a lot of frustrated customers and brand tarnishing.

Corporations are the primary source of revenue for OEM companies and thus for a component supplier like Intel. Best Buy customers and we DIY'ers don't really have an overall large impact in total sales volume, but if the revenue capability is there and somewhat significant due to higher profit margins, then hell yes do it.

Intel is cranking out millions upon millions of IB CPU's right now and will continue to in volume for quite some time. If they can save a nickel or dime per unit by eliminating FS and using TIM, then they've just paid for a lot of annual salaries with that kind of savings. Makes perfect business sense. Sucks for us. I really wish our K series IB's has FS under the IHS, but I also understand this would likely cost Intel more money to do since the fabrication and manufacturing is tooled for TIM on every CPU being made.

I've also stated in previous posts that I love my IB. It can do anything and everything I ask of it with flying colors, and that's without OC'ing it, so heat it not an issue. Thing runs at 100% load cooler than my X2 6400+ with a Thermalright Ultimate 90 did stock clocks.
 
Bottom line Intel didn't deliver the same high quality processor they have in the past. Shame on Intel for not delivering the same high standards that we all have grow to trust Intel to deliver. Best thing moral thing for Intel to do is to step it up put on high quality TIM or go back to fluxless solder. Which ever works best. I'm sure Intel's engineers have it all figured out.
 
Since when has Intel ever worried about the moral thing to do?

We should also make Intel allow BCLK overclocking again while we're at it. It's the moral thing to do, allowing consumers the ability to overclock $50 chips so that they get more for less.
 
SonDa5 said:
Bottom line Intel didn't deliver the same high quality processor they have in the past.
Click to expand...

For you. For the millions of people that Intel actually cares about (okay, the couple of OEM customers they care about) it is a great chip that delivers the same high quality Intel is trusted to deliver.
 
Forceman said:
, the couple of OEM customers they care about) it is a great chip that delivers the same high quality Intel is trusted to deliver.
Click to expand...

Can OEM IB 3570k or 3770k even over clock in the system BIOS?
 
SonDa5 said:
Can OEM IB 3570k or 3770k even over clock in the system BIOS?
Click to expand...

Probably not - I'm assuming OEMs would be using non-K chips though. The point being, that for the vast majority of customers that use the chip at stock, it is an improvement over SB in every way. So saying it isn't the same high quality processor they've delivered in the past is overstating it a bit.
 
SonDa5 said:
Can OEM IB 3570k or 3770k even over clock in the system BIOS?
Click to expand...

Forceman said:
Probably not - I'm assuming OEMs would be using non-K chips though. The point being, that for the vast majority of customers that use the chip at stock, it is an improvement over SB in every way. So saying it isn't the same high quality processor they've delivered in the past is overstating it a bit.
Click to expand...

I doubt that OEMs are going to be primarily utilizing the chipset that allow for overclocking. If they are, then they'll have a custom BIOS/UEFI that won't allow it.
 
I am going to rip my 3570k apart just for the hell of it. I am going to try applying different TIM as well as no heat spreader as well. Why not. Let's see what happens.
 
SonDa5 said:
Bottom line Intel didn't deliver the same high quality processor they have in the past. Shame on Intel for not delivering the same high standards that we all have grow to trust Intel to deliver. Best thing moral thing for Intel to do is to step it up put on high quality TIM or go back to fluxless solder. Which ever works best. I'm sure Intel's engineers have it all figured out.
Click to expand...

I disagree, they made it the way they want. They don't care about overclocking much past getting attention to buy the chip.
 
Does anybody know if unpackaged CPUs are available. I thought I read somewhere that laptops use unpackaged CPUs. Maybe so they can squeeze it in smaller spaces with custom designed heat sinks attached directly to the chip. If somebody could buy a bunch of non-lidded IBs there would be no warranty issues. On the other hand they probably wouldnt be "k" models. :(
 
youghmama said:
Does anybody know if unpackaged CPUs are available. I thought I read somewhere that laptops use unpackaged CPUs. Maybe so they can squeeze it in smaller spaces with custom designed heat sinks attached directly to the chip. If somebody could buy a bunch of non-lidded IBs there would be no warranty issues. On the other hand they probably wouldnt be "k" models. :(
Click to expand...

They also don't fit in the desktop socket.
 
InternationalHat said:
Dear AMD, plz send help.
Click to expand...

This made me laugh...
It definitely would be nice, however, I would be surprised if AMD's next gen platform even comes close to the performance and low power consumption :( Maybe with enough High quality TIM or fluxless solder !!!
 
When a system suffers poor performance because of of a technical limit or because of some sub-optimal choice I have made I am ok with that.

But we're talking about enthusiast performance products here.
You can get GPUs with premium compound, high quality coolers and expensive PCBs and people pay the premium when they're looking for optimal performance.

But people should expect the components themselves to be optimized already... then the manufacturers just leave the synergy to the builder.
I would venture a guess that people ponying up for one of these top-of-the-line Intel CPUs would pay the slight premium for a high-end TIM.... hell, I bet most would pay the premium for a purified copper top-plate.
 
Zangmonkey said:
When a system suffers poor performance because of of a technical limit or because of some sub-optimal choice I have made I am ok with that.

But we're talking about enthusiast performance products here.
You can get GPUs with premium compound, high quality coolers and expensive PCBs and people pay the premium when they're looking for optimal performance.

But people should expect the components themselves to be optimized already... then the manufacturers just leave the synergy to the builder.
I would venture a guess that people ponying up for one of these top-of-the-line Intel CPUs would pay the slight premium for a high-end TIM.... hell, I bet most would pay the premium for a purified copper top-plate.
Click to expand...

This is not in the enthusiast performance line. The 1155 socket is mainstream, the 2011 is enthusiast. In their eyes, you gotta pay to play, and with SB you all got a freebie. they're pulling that freebie now. It sucks, but from a buisness standpoint, its the right thing to do so your markets don't overlap too much.
 
Jorona said:
This is not in the enthusiast performance line. The 1155 socket is mainstream, the 2011 is enthusiast. In their eyes, you gotta pay to play, and with SB you all got a freebie. they're pulling that freebie now. It sucks, but from a buisness standpoint, its the right thing to do so your markets don't overlap too much.
Click to expand...

You're talking about platforms and I'm talking about cpus.
Everything tagged 37xx is in the "performance" series for IB.
 
Zangmonkey said:
You're talking about platforms and I'm talking about cpus.
Everything tagged 37xx is in the "performance" series for IB.
Click to expand...

???

Platform includes the CPU, it's just the 37xx is the "performance" of the mainstream, which still makes it mainstream.
 
Jorona said:
This is not in the enthusiast performance line. The 1155 socket is mainstream, the 2011 is enthusiast. In their eyes, you gotta pay to play, and with SB you all got a freebie. they're pulling that freebie now. It sucks, but from a buisness standpoint, its the right thing to do so your markets don't overlap too much.
Click to expand...

Planned obsolescence and gimping your own products is not the right thing to do.
 
Planned obsolescence and gimping your own products is not the right thing to do.
Click to expand...

Both AMD and Intel do both of these all the time. Do you think Intel has separate dies for all the variations of cores and cache with its xeons on the same node? Do you think the i5s and i7s are different dies?
 
Everyone does it, it's bullshit. It's monopolist fascism.

Anyway, sure haven't seen a gimp like this before.
 
Fixall said:
This is so spot on. Strong competition makes for great products.
Click to expand...

Absolutely it does. Although I doubt Intel would have steered away from TIM, regardless. Besides, Intel probably had the groundwork on this decision laid out even before AMD released BD.
 
Samwiches said:
Everyone does it, it's bullshit. It's monopolist fascism.

Anyway, sure haven't seen a gimp like this before.
Click to expand...
We have, read this article http://www.overclockers.com/ivy-bridge-temperatures

And the others are correct. Intel, AMD, Nvidia gimp chips all the time by disabling a core and selling them cheaper. Think about it, the chips are the same, it probably costs Intel more to make a cheaper chip because they have to take a chip and damage part of it somehow. Look at the difference between the Nvidia 680 and 670 -- they just disabled a part of it and took $100 off. This is their product, they can do what they want and they do. I suppose there is some limitation because they try to be honest about telling you what you are buying, but that is about it.

This chip was gimped from the start by the TIM and anybody with the slightest bit of common sense could see it (precisely because they do it all the time). However, the question remains as to how much we can ungimp it and how expensive it will be. I dont think we will ever be able to match the FS for efficient heat transfer and thereby keep a pristine CPU package merely by replacing the TIM between the chip and the IHS, but who knows. But we should be able to bring the chip up to its full potential by running lidless with a high-quality TIM and a block designed (or modified) to make good contact with the naked chip. It would be nice to see the after market cooling mfgs come to the rescue here. :D
 
youghmama said:
We have, read this article http://www.overclockers.com/ivy-bridge-temperatures

And the others are correct. Intel, AMD, Nvidia gimp chips all the time by disabling a core and selling them cheaper. Think about it, the chips are the same, it probably costs Intel more to make a cheaper chip because they have to take a chip and damage part of it somehow. Look at the difference between the Nvidia 680 and 670 -- they just disabled a part of it and took $100 off. This is their product, they can do what they want and they do. I suppose there is some limitation because they try to be honest about telling you what you are buying, but that is about it.

This chip was gimped from the start by the TIM and anybody with the slightest bit of common sense could see it (precisely because they do it all the time). However, the question remains as to how much we can ungimp it and how expensive it will be. I dont think we will ever be able to match the FS for efficient heat transfer and thereby keep a pristine CPU package merely by replacing the TIM between the chip and the IHS, but who knows. But we should be able to bring the chip up to its full potential by running lidless with a high-quality TIM and a block designed (or modified) to make good contact with the naked chip. It would be nice to see the after market cooling mfgs come to the rescue here. :D
Click to expand...

You're assuming that all silicon comes out perfect, which it doesn't. Some don't come out perfect, like maybe some cache not working right, or not being able to hit a target frequency with a target voltage and power consumption. These, instead of being scrapped, are sold as lower end chips. Silicon yields are out of their control, they can improve it, but there will always be some bad ones.

The TIM thing is a bit different because it is something they directly control.
 
Dantrax said:
These guys say otherwise- http://www.eteknix.com/news/ivy-bridge-heat-problems-remain-even-after-ihs-removal/ - I don't understand how that could be unless the D-14 wasn't put on tight enough, or doesn't tighten up enough to make up for the IHS thickness. I would think after removing the IHS, that direct cooler contact is the way to go.
Click to expand...

Could be a lot of things. Bad contact, the fact that modern heatsinks are designed to cool the IHS and not the die itself (the heat is more concentrated in the direct die contact, which reduces the effectiveness of the outer heatpipes), the IHS works better than people think, or maybe the problem isn't the TIM as much as it is the thermal density/some other issue with the chip. Maybe Intel tested some with solder and found they still had high temps (doubtful, but I guess possible) and so decided not to bother. It's been a while since people have been de-lidding chips, and die sizes and densities are a lot different now.
 
Dantrax said:
These guys say otherwise- http://www.eteknix.com/news/ivy-bridge-heat-problems-remain-even-after-ihs-removal/ - I don't understand how that could be unless the D-14 wasn't put on tight enough, or doesn't tighten up enough to make up for the IHS thickness. I would think after removing the IHS, that direct cooler contact is the way to go.
Click to expand...
If we are trying to prove that the CPU package heat transfer system is bad at cooling, and we had ten tests on ten proven-good IB chips where we removed the IHS w/Intel TIM and replaced it with direct contact heat sinks with different TIM (say a different heat sink and different TIM in each test) and ONE OF THEM had a significant, say 5% to 10% improvement in lower core temps, and ALL THE REST either had worse or same heat levels WE HAVE STILL PROVEN THAT THE INTERNAL HEAT TRANSFER SYSTEM IS BAD!!!!!! This isnt an election, this is testing a hypothesis, and the hypothesis is that Intel did a bad job transferring heat from the bare chip to the CPU surface for dissipation. It takes just one trustworthy positive test to prove it, even if a hundred other tests fail. And once you get that one successful test, you replicate it for everybody else who wishes to improve the performance of their IB.

You cant prove that the Intel did a good job by looking at one test which didnt show any improvement, or even looking at nine tests which showed no improvement. There are so many factors involved (and so many incompetents and Intel fanboys out there) that you dont know which test was done well and there really is no way to know which combination of cooler type, pressure, TIM amount, TIM type, mounting system, etc... works the best.

If Intel did its job correctly there should be no significantly improved results at all, not one!!! AND if there are better results should be only minimally better.
 
I wouldn't say 5-10% is all that significant - certainly not enough to justify cutting the chip open. Figuring for most chips that run at 80C that would only be 4-8C.
 
nissanztt90 said:
Because all silicon is created equal :rolleyes:
Click to expand...

No but my point is both manufacturers intentionally disable features (cache, limit performance, disable fully working cores) just to have different products that they can charge different prices for. I could have gone into the rumor that all lga2011 processors actually contain 8 cores with 2 or 4 cores disabled on most. That is a bigger gimp than the example I made although I am not sure that this is confirmed.
 
Forceman said:
I wouldn't say 5-10% is all that significant - certainly not enough to justify cutting the chip open. Figuring for most chips that run at 80C that would only be 4-8C.
Click to expand...

I agree. Now if the results were a consistent 15-30% load temp drop, then it would not require a lot of thought by a lot of enthusiast IB owners willing to alter their CPU's, imo.
 
youghmama said:
If we are trying to prove that the CPU package heat transfer system is bad at cooling, and we had ten tests on ten proven-good IB chips where we removed the IHS w/Intel TIM and replaced it with direct contact heat sinks with different TIM (say a different heat sink and different TIM in each test) and ONE OF THEM had a significant, say 5% to 10% improvement in lower core temps, and ALL THE REST either had worse or same heat levels WE HAVE STILL PROVEN THAT THE INTERNAL HEAT TRANSFER SYSTEM IS BAD!!!!!! This isnt an election, this is testing a hypothesis, and the hypothesis is that Intel did a bad job transferring heat from the bare chip to the CPU surface for dissipation. It takes just one trustworthy positive test to prove it, even if a hundred other tests fail. And once you get that one successful test, you replicate it for everybody else who wishes to improve the performance of their IB.

You cant prove that the Intel did a good job by looking at one test which didnt show any improvement, or even looking at nine tests which showed no improvement. There are so many factors involved (and so many incompetents and Intel fanboys out there) that you dont know which test was done well and there really is no way to know which combination of cooler type, pressure, TIM amount, TIM type, mounting system, etc... works the best.

If Intel did its job correctly there should be no significantly improved results at all, not one!!! AND if there are better results should be only minimally better.
Click to expand...

No you havent. 1 in 10 getting 5-10% better is an outlier if 9/10 trend towards getting worse. Even knowing that theres probably large systemic errors due to different testing setups, you would absolutely expect more improvement in more samples. I agree there needs to be more rigorous testing and that its still possible that there are improvements to be made, but the heat transfer system being "BAD" would IMO mean that its a significant, performance altering deficiency. That 1 in 10 chip barely even qualifies for that threshold so that seems like a stretch, especially if its true that 9/10 have stayed the same or gotten worse

I would prefer if Intel had used FS on their IHS, but in the scheme of things its looking like the more data we have, the more justifiable this decision is. I'm not cracking the lid on a CPU for 3% and I haven't seen any new data showing more than 10% which is just about approaching "worth-it" territory
 
Activate: AMD said:
No you havent. 1 in 10 getting 5-10% better is an outlier if 9/10 trend towards getting worse. Even knowing that theres probably large systemic errors due to different testing setups, you would absolutely expect more improvement in more samples. I agree there needs to be more rigorous testing and that its still possible that there are improvements to be made, but the heat transfer system being "BAD" would IMO mean that its a significant, performance altering deficiency. That 1 in 10 chip barely even qualifies for that threshold so that seems like a stretch, especially if its true that 9/10 have stayed the same or gotten worse

I would prefer if Intel had used FS on their IHS, but in the scheme of things its looking like the more data we have, the more justifiable this decision is. I'm not cracking the lid on a CPU for 3% and I haven't seen any new data showing more than 10% which is just about approaching "worth-it" territory
Click to expand...
We are engineering a solution here, it is a matter of trial and error. The statistic of 9 out of 10 failures is only applicable to the testing phase, to when we are trying different combinations of heatsink, pressure, TIM, lapping, etc.... Once you find the successful procedure, you record it and use it and your success rate will go up substantially. Pretend you are a defense department official trying to get a new hand grenade for the Army. You have ten different competing contractors who come up with ten different hand grenades. You go to trials and nine of them are out right failures, the tenth works perfectly. Do you give up and assume nine out of ten grenades will fail? No, you go to the builder to the tenth grenade and ask for more samples so you can finally test the successful design. This is not and exercise in statistics.
 
youghmama said:
We are engineering a solution here, it is a matter of trial and error. The statistic of 9 out of 10 failures is only applicable to the testing phase, to when we are trying different combinations of heatsink, pressure, TIM, lapping, etc.... Once you find the successful procedure, you record it and use it and your success rate will go up substantially. Pretend you are a defense department official trying to get a new hand grenade for the Army. You have ten different competing contractors who come up with ten different hand grenades. You go to trials and nine of them are out right failures, the tenth works perfectly. Do you give up and assume nine out of ten grenades will fail? No, you go to the builder to the tenth grenade and ask for more samples so you can finally test the successful design. This is not and exercise in statistics.
Click to expand...
Thats a pretty crappy analogy. We have one grenade, people are just throwing it 10 slightly different ways. If the results were definitively awesome (say 30% temp reduction) but we were still in the testing phase (and I agree that we are) then you'd expect maybe one or two to be at 30% where someone got a really good result, probably a few in the 10-20 range and maybe 1 or 2 the same or worse. Right now the differences are in the noise but they're all centered around <5% change.
 
Samwiches said:
Planned obsolescence and gimping your own products is not the right thing to do.
Click to expand...

This isn't planned obsolescence, that means they were designed to fail. These chips, run at their stock settings will likely run til they're legitimately obsolete.

"Gimped" is a poor way of putting it as well. These chips perform very well thermally when used at stock frequencies. They didn't design the chip for faster then that. They sell you unlocked so you're free to try to lean on it, but don't whine when it doesn't work well. It wasn't designed for it. Also, alot of people at getting 4.4-4.5 OCs out of IB. A 1+ ghz increase is still pretty epic. Thats over a 25% increase. that doesn't sound gimped to me.

Before SB, a 1+ ghz OC ment you had a pretty cheery chip, or really good cooling. SB was a jem, no bones about it, but not the norm.

Once again, this is the mainstream platform guys. If you really want top of the line performance you have to go to the enthusiast line.
 
I'm not willing to believe this is the actual solution to the "problem" until the results are repeated by multiple websites / enthusiasts.

Out of all the stories I've seen so far with sites popping-off the IHS, none have shown any improvement until this one. In addition, one of our own forum members has violated his warranty to discover that there is no load temperature decrease with a directly-seated heat sink

http://hardforum.com/showpost.php?p=1038731976&postcount=73

If this were the real solution to the problem of higher Ivy Bridge temperatures, we would have seen this result repeated many times.
 
You must log in or register to reply here.
Back
Top