Why isn't Intel fixing the 'delidding' issue?

[TuxDave]

Intel dictates stock speeds. With better cooling, they could make the stock speeds higher.

With better cooling, you could potentially get higher PERFORMANCE since thermal throttling would not kick in. However, I'm not sure what you're getting at with stock clock speeds. If at 80C the silicon can work at 3.1GHz instead of 3GHz at 100C, you would still have to put in another thermal throttle point at 80C. Am I missing your point?

 

[JimmiG]

Come on, that's low.

Also these chips are guaranteed for 7 years at 110C. You want more speed? Feel free to do the legwork yourself.

Maybe, but my point is that even though the 22nm dies themselves are extremely advanced, the actual final assembly of the CPU package is very low-tech, with extremely loose margins of error and not a lot in the way of quality control. If a CPU doesn't overheat at 3.5 GHz, it's shipped out.

 

[Concillian]

Edit: I'm sure they could figure out an easy way to 'spray" an appropriately thin layer of compound on if the mass production/ automation part is the issue?

Money is the answer. The answer is always money.

If you've ever worked in this kind of industry you'd know that this is 100% about what the Investment vs. return analyses say is the way to go.

 

[Concillian]

Intel dictates stock speeds. With better cooling, they could make the stock speeds higher.

And what would that buy them? The price segments haven't changed in years now. The tiers are set up, and they aren't going to introduce a new tier. There's also little to no incentive for them to bump clock speeds an extra 100 MHz. That's less margin, less yield, more potential warantee liability, etc...

It's a business. Think like a business, not like they're a group of techies trying to push out the best thing since the wheel.

 

[Nec_V20]

They still want to sell their other processors and if Haswell were so much more in front then that would put the kibosh on that to a great extent.

The other reason could be the old IBM "Golden Screwdriver" scam in that Intel without changing anything except for the name, and getting rid of the gunk, could bring out a higher clocked version with the same production line.

The same kind of scam has been done recently by AMD with their 9590 which is just a relabelled 8350 which has been factory overclocked.

 

[kimmel]

lets make some wags (wild ass guesses):

Cost differential between solder and TIM: $0.25
Ratio of people who care about solder vs TIM: 1 out of 10,000

So, if you and everyone else that cares about solder offered to pay roughly $2500 extra per chip for solder everywhere instead of TIM Intel might listen to you. If not, you don't care enough to balance out the economics for Intel.

These are numbers I completely made up. There is no "issue" since I doubt you want to pay over 2k for a processor to get solder.

 

[Torn Mind]

High temps that do not go beyond the "danger" cliff simply do not cause any major harm to the CPU. Hence, they use cheaper materials that are good enough to dissipate energy at stock and pocket the change. Do note, that a semiconductor company must invest some of its money in R&D(engineers and those fabs) or else they will bite the dust; not even Intel is immune.

 

[Maxima1]

lets make some wags (wild ass guesses):

Cost differential between solder and TIM: $0.25
Ratio of people who care about solder vs TIM: 1 out of 10,000

So, if you and everyone else that cares about solder offered to pay roughly $2500 extra per chip for solder everywhere instead of TIM Intel might listen to you. If not, you don't care enough to balance out the economics for Intel.

These are numbers I completely made up. There is no "issue" since I doubt you want to pay over 2k for a processor to get solder.

If it's spread out on purchases, then it's nowhere near $2500. Although I'll agree it isn't really an issue.

A similar dilemma is found in laptops, but it is the opposite. They keep putting optical drives in them, even though many people never use them, and you can buy it separately if you have to have one. Yet companies like Dell want to double down and put in slimmer and slower optical drives.

 

[Rakehellion]

1.) They don't need slightly faster chips to be competitive.

Didn't stop them from releasing Haswell! :awe:

But they could get a pretty significant performance boost from applying the thermal compound the right way. How much more can they charge by going from 3GHz to 3.3GHz?

 

[jvroig]

The answer is simply because there is no "delidding issue" as far as Intel is concerned.

They don't bother to "fix" it for the same reason they don't bother to bundle a beefier stock heatsink: at stock, everything is fine and within expected parameters already.

for the minimal amount of engineering/legwork it would take to drop 10-20C it seems short sighted on there part not to fix it. Lowering your thermal envelope by 20-30% for fractions of a penny, not doing it seems silly. We're talking 10-20C at stock clock here. That is good for everyone.

Not everyone.
1. For one thing, it's not good for Intel. They don't care that you'd be 10-20C lower at stock clocks than you already are. They've already lined up the parameters and hit the targets. They don't need the customers to have 20C lower operating temps, because the current temps, voltage and clocks are all set to ensure the CPU survives long after all other components have failed. So all Intel will accomplish is increase costs for no gain on their end.

2. It's good for everyone, sure, but most customers wouldn't care either way. >99.9% of Intel's customers don't care if they would enjoy (or are enjoying) 20C lower operating temps. For practically all of Intel's customers, as long as it works at stock and the CPU will not fail before a new computer is bought and hits Intel's expected power and cooling parameters, everything is perfect already.

3. That really leaves only us, the overclocking enthusiasts, to actually consider this a good thing. It is certainly problematic for us, and would definitely benefit us if Intel were to go back to having no appreciable gap underneath the IHS. But we aren't a significant market, and for Intel to base their decisions (on where to save costs and where not to save costs) upon us instead of the >99.9% of their target market is just unthinkable.


I understand your frustration OP, and I feel what you feel as a fellow enthusiast. In real life though, for Intel, this is not a problem and there is no delidding issue, no matter how beneficial it would be for us.

 

[cytg111]

Because it's not an issue. The standard TIM is absolutely fine for stock, and even moderate to high OC...

But it is not.. These chips throttle at stock running the right/wrong payload .. I was going to go h100i partly to get out of throttle territory - part to reduce noise, but I read somewhere that this dude is throttling (stock) even with the h100i (which in itself poses a great number of questions) .. untill i get that debunked im staying on the stock cooler and throttle at stock when running prime95 like payloads ..

edit : running rig in sig

 

[TerryMathews]

But it is not.. These chips throttle at stock running the right/wrong payload .. I was going to go h100i partly to get out of throttle territory - part to reduce noise, but I read somewhere that this dude is throttling (stock) even with the h100i (which in itself poses a great number of questions) .. untill i get that debunked im staying on the stock cooler and throttle at stock when running prime95 like payloads ..

edit : running rig in sig

If your chip throttles at stock, it's defective and intel will replace it.

 

[calc2]

If your chip throttles at stock, it's defective and intel will replace it.

There have been several articles around the net about the fact that Haswell engineering samples performed much better than mass production Haswell. Not that many people run Linpack 11 so that may be why this issue isn't as well known as it should be.

 

[cytg111]

The other reason could be the old IBM "Golden Screwdriver" scam in that Intel without changing anything except for the name, and getting rid of the gunk, could bring out a higher clocked version with the same production line.

- I dont think so, evidence so far suggests that delidding and gaining 10-20 degrees south might give you another 200-300Mhz but it might also give you a big fat 0. Thermals in these ranges doesnt seem to impact Hz scaling very much (haswell, ivey)

 

[cytg111]

If your chip throttles at stock, it's defective and intel will replace it.

Then there's alot of defective haswells outthere ..
And I cannot be without a cpu for the replacement period and that is IF the RMA gets approved ..

 

[TerryMathews]

Then there's alot of defective haswells outthere ..
And I cannot be without a cpu for the replacement period and that is IF the RMA gets approved ..

That doesn't change the fact that intel offers a warranty for issues such as this.

 

[cytg111]

That doesn't change the fact that intel offers a warranty for issues such as this.

alot of facts dont change given whatever unrelated information gets reiterated. let me put it another way : I dont get it.

 

[TerryMathews]

alot of facts dont change given whatever unrelated information gets reiterated. let me put it another way : I dont get it.

If intel over applied the TIM on your processor, it's defective. No different than any other product.

You're treating a processor different than a microwave. To you, it is. To intel, it isn't.

 

[cytg111]

I get that? Not playing the blamegame at all .. even if Intel was charged with a massive lawsuit over this (or not), i wouldnt change my stand or perspective at all .. dont really give *'s * about it... It is a hell of a processor, best mainstream x86 on the market, even with this little catch.
But I suppose there would be value in RMA'ing it, if enough did .. Maybe it would get fixed? Sadly im not a CPU-patriot .. I just need to find a way to get my microwave fixed(h100i or whatever). (cannot live without my microwave for 14-21 days ..)

 

[Kenmitch]

But it is not.. These chips throttle at stock running the right/wrong payload .. I was going to go h100i partly to get out of throttle territory - part to reduce noise, but I read somewhere that this dude is throttling (stock) even with the h100i (which in itself poses a great number of questions) .. untill i get that debunked im staying on the stock cooler and throttle at stock when running prime95 like payloads ..

edit : running rig in sig

I run a H100i in my rig in sig. It takes around 4.5ghz and Linpack 11 to get in the throttle zone.

Best to run fixed vcore at whatever your comfortable with. Running stock settings feeds too much juice under avx loads.

 

[solidsnake1607]

Intel wont fix that, with that issue users try to dellid it and many failed then they buy again intel chips, they keep winning money and we all lose lol in my opinion with a corsair h100 and a good aftermarket thermal paste and high airflow coolers you cant get very high temps, Not over 90 degrees.

 

[2is]

Intel should have certainly done a better job with Haswell, be it a soldered IHS or have included a better cooler, seeing that even at stock speeds, it can start to throttle with the stock HSF when AVX2 instructions are thrown at it.

 

[Kenmitch]

Intel should have certainly done a better job with Haswell, be it a soldered IHS or have included a better cooler, seeing that even at stock speeds, it can start to throttle with the stock HSF when AVX2 instructions are thrown at it.

What uses avx2 extensions currently?

Linpack 11 torture test seems like a unrealistic use of it to me.

 

[2is]

What uses avx2 extensions currently?

Linpack 11 torture test seems like a unrealistic use of it to me.

Might be unrealistic to you. That doesn't change the fact that it causes Haswell to throttle under 100% stock conditions (clocks, voltage, cooling) Previous processors easily managed to stay within their thermal threshold under unrealistic tests @ stock conditions. Haswell should be able to do the same.

 

[VirtualLarry]

What uses avx2 extensions currently?

Linpack 11 torture test seems like a unrealistic use of it to me.

Not at all. Distributed-computing apps, once they start to use AVX2, will likely put a similar load on the CPU. And I agree, if a stock setup throttles under such load, the the CPU is defective. If they all do it, then there should be a class action. The fact that retail samples of Haswell overclocked much worse than ES chips, is kind of a smoking gun, IMHO.