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

[TerryMathews]

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.

See, I would be a dick and intentionally warranty exchange them over and over until it was perfectly within thermal specifications just to punish intel for being dicks with the IHS

 

[Ketchup]

OP, I think you are asking the wrong question. The question shouldn't be "why isn't Intel fixing" it. It should be "when are dies going to become so small they stop lidding."

 

[rgallant]

The IHS issue started with Ivy. Sandy's IHS was soldered.

But to answer your question. Nobody here knows for sure but it could be engineered that way to address a thermal expansion problem Intel identified during development. Or maybe it is cost. Whatever the case, Intel has its reasons and we will only be speculating.

Also, the thermal improvements are really only seen when overclocking. At stock there isn't much benefit.

they did but no one likes it
-the biggest back lash are peeps buying a $600.00-1000 system but have high temps with their $19.99 hyper cooler.if those peeps along with peeps that have rooms and case temps near 30c ,then whine about high temps running some 3rd parity stress program for 5 days ,get a clue things might look better.

 

[Homeles]

See, I would be a dick and intentionally warranty exchange them over and over until it was perfectly within thermal specifications just to punish intel for being dicks with the IHS

Boy, you are quite the rebel there. Least effective way of sticking it to the man ever.

 

[blackened23]

I don't think it's a huge issue anyway. I mean I get 4.7ghz on my 3770k fairly effortlessly at 1.275V and my real world temps never exceed around 65C (with a H100 cooler). Sure, they creep up to 88C during prime95, but my real world temps never, ever match prime95.

I think some folks read into stress testing temperatures too much. I've never even approached those temps in real world use, even at 100% CPU load. The main purpose of stress testing is to ensure CPU overclocked stability in the long term, beyond that I don't ever pay attention to what temps prime95 gives. There's really no need to.

 

[calc2]

What I ended up doing with my 4770k was delid, lap the bottom of the ihs, and use painted on mx-4, I might switch to CLP later. Then I found the proper settings for my cpu using linpack 10.3 and linpack 11.0. I used the linpack 11.0 settings as default multiplier/voltage (40x 1.20v) and set linpack 10.3 multiplier for turbo mode (43x). I also set the max tdp to be slightly under what linpack 11.0 used at 43x 1.20v, which would eventually fail, which turned out to be around 170w.

 

[JimmiG]

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.

We don't know yet how other apps will that use AVX2 will behave. Linpack may be an outliner.

Does the Aida64 FPU test use AVX2 or just AVX? Its temps are much lower than Linpack 11. The benchmarks included with Aida64 are however AVX2 optimized:

AVX2 and FMA optimized 64-bit benchmarks for Intel “Haswell” processors

I couldn't get my CPU over 75C running those. They're very short though, so it's possible the temperature would have climbed if they were longer. But I don't think they'd be anywhere near the Linpack 11 temps. With Linpack 11, the temperature shoots straight up into the high 90's within 1 second of starting the test.

Remember Furmark? It caused several GPUs to overheat and crash, yet AMD and Nvidia dismissed it as a "power virus". Of course, it would be harder for Intel to dismiss their own application...

 

[ShintaiDK]

I havent experienced a single case of throttle. Not even close. Even with AVX2 load mine still runs cooler than the previous 3570K. Stock cooler that is.

In terms of GPUs, its different. GPUs are designed never to be able to handle full load. Unless you buy Tesla cards. The reason is that game loads are (almost) always light loads.

So you can pick between driver modified speed on GPUs. Or simply remove 20-30% performance.

 

[seitur]

Because making some of your overall product offering worse for certain percentage of your clients in certain situations can lead to earn more money either at present, perspectively or both.

Intel know it is holding you by your balls and is simply using this to it's advantage. History as old as world, nothing new, business as usual. You get the point.

 

[ShintaiDK]

Because making some of your overall product offering worse for certain percentage of your clients in certain situations can lead to earn more money either at present, perspectively or both.

Intel know it is holding you by your balls and is simply using this to it's advantage. History as old as world, nothing new, business as usual. You get the point.

I doubt the other 99.x% part of all buyers would think its a good idea to pay more for their CPUs, just so you can OC higher.

 

[Chiropteran]

In terms of GPUs, its different. GPUs are designed never to be able to handle full load. Unless you buy Tesla cards. The reason is that game loads are (almost) always light loads.

So you can pick between driver modified speed on GPUs. Or simply remove 20-30% performance.

Half dozen machines running with multiple video cards each mining bitcoins 24/7 for years on end disagree with you.

 

[seitur]

I doubt the other 99.x% part of all buyers would think its a good idea to pay more for their CPUs, just so you can OC higher.

You must have mistakenly assumed my post contained condemnation of those practices, while in reality it carried no emotional load.

I was merely answering OP question.

 

[tweakboy]

I think Intel knows what its doing. If you wait for the Store to get it in oct 16th

Why is there a big delay between RTM and RTM comes out to public ?

 

[Anarchist420]

They do it because of IP. intel sucks because of IP and they've held back progress forever. We wouldn't even have hardware 3d accelerators if they hadn't made such lousy CPUs back in '98.

The TIM is good enough at stock speeds.

That's what you think.

Because it's not an issue.

That's what you think. I don't want my damn CPU heating up the room especially when it doesn't have to.

 

[videogames101]

someone link me an example of a haswell chip throttling at stock

 

[Torn Mind]

That's what you think.That's what you think. I don't want my damn CPU heating up the room especially when it doesn't have to.

Heat energy is measured in Joules. Temperature is measure of hot or cold at a specific point. The fact is that Ivy Bridge and Haswell might run hotter, but consume less electricity and hence heat the room up less because they use less Joules per second than their previous counterparts. The reported higher temperatures due to having a worse thermal conductor in contact with the silicon die.

 

[piasabird]

Intel is a monopoly and they dont care.

 

[dmens]

They do it because of IP. intel sucks because of IP and they've held back progress forever. We wouldn't even have hardware 3d accelerators if they hadn't made such lousy CPUs back in '98.

Implying releasing Haswell source code would somehow allow people to make better TIM/lid thermal solutions. Also implying the x86 spec hasn't been open to user suggestion for decades.

You make the rest of us anti-copyright abuse people look bad.

That's what you think.

It is fine for stock use. Show me any stock CPU in the past decade that has exceeded 100C with its stock cooling configuration properly mounted.

inb4 100C is too much, I've seen chips last months in far worse conditions and higher voltages.

 

[JimmiG]

Heat energy is measured in Joules. Temperature is measure of hot or cold at a specific point. The fact is that Ivy Bridge and Haswell might run hotter, but consume less electricity and hence heat the room up less because they use less Joules per second than their previous counterparts. The reported higher temperatures due to having a worse thermal conductor in contact with the silicon die.

If anything, de-lidding my 4770K caused it to heat up the room more.

Before delid, I was running 1.218V / 4.3 GHz. Heatsink was cool to the touch and the air being exhausted from the case felt the same as the air in the room. After delid, I can push 1.323V / 4.5 GHz with similar temps, but the heatsink is now warm to the touch, and the air being exhausted from the case is warmer than the room.

So not delidding is a great way to trap that heat under the IHS, preventing it from getting out and heating up your room

 

[Ketchup]

If anything, de-lidding my 4770K caused it to heat up the room more.

Before delid, I was running 1.218V / 4.3 GHz. Heatsink was cool to the touch and the air being exhausted from the case felt the same as the air in the room. After delid, I can push 1.323V / 4.5 GHz with similar temps, but the heatsink is now warm to the touch, and the air being exhausted from the case is warmer than the room.

So not delidding is a great way to trap that heat under the IHS, preventing it from getting out and heating up your room

That isn't exactly a fair comparison as you are running more voltage now (and faster) so putting out a decent amount more heat.

 

[Torn Mind]

That isn't exactly a fair comparison as you are running more voltage now (and faster) so putting out a decent amount more heat.

I think his point is that his temps are very not different (practically constant) at a higher voltage and clocks yet the heat output is greater.

Had he not bumped up his voltage and clocks, the temp readings would have been lower. Pretty much, the ultimate point is that the temperature reading on the CPU die gives no information on heat output at all.

As another example, my family's Atom laptop can shoot up to around 60 degrees Celsius even though it pulls much less power than the old Pentium 4 desktop we have.

 

[Anarchist420]

Implying releasing Haswell source code would somehow allow people to make better TIM/lid thermal solutions. Also implying the x86 spec hasn't been open to user suggestion for decades. You make the rest of us anti-copyright abuse people look bad.

Sorry about making others look bad. And while I realized that the x86 spec has been open to user suggestion (AMD even had an input in the 64 extenstion), it costs a hefty fee to make an x86 processor if you're not intel.

As for haswell souce code being released that wasn't what I meant... I meant that intellectual property gives intel an advantage and that they don't have to care as much about their products as they would without IP.

 

[frozentundra123456]

I think his point is that his temps are very not different (practically constant) at a higher voltage and clocks yet the heat output is greater.

Had he not bumped up his voltage and clocks, the temp readings would have been lower. Pretty much, the ultimate point is that the temperature reading on the CPU die gives no information on heat output at all.

As another example, my family's Atom laptop can shoot up to around 60 degrees Celsius even though it pulls much less power than the old Pentium 4 desktop we have.

Exactly. Temperature of the chip and heat output are two separate issues. Temperatures are a problem of their own, because they can cause throttling or damage the chip. But that does not necessarily mean more heat output. That is related to the total power consumed, as the energy the chip uses is converted to heat. I don't see heat output as an issue at all with ivy/haswell, but due to the small die size and poor heat transfer, temperatures are definitely an issue, especially with overclocking and perhaps with haswell at stock under avx2.

 

[TerryMathews]

If anything, de-lidding my 4770K caused it to heat up the room more.

Before delid, I was running 1.218V / 4.3 GHz. Heatsink was cool to the touch and the air being exhausted from the case felt the same as the air in the room. After delid, I can push 1.323V / 4.5 GHz with similar temps, but the heatsink is now warm to the touch, and the air being exhausted from the case is warmer than the room.

So not delidding is a great way to trap that heat under the IHS, preventing it from getting out and heating up your room

If your post is serious, I don't think you understand how thermal systems work.

You increased voltage and frequency, which means you did increase thermal output. You stated that measured core temp remained consistent.

If the heat isn't being retained in the processor, it has to be discharged: into your room.

Imagine a sink, a faucet with a clog, and a dial to control flow. You had the dial on 8, you fixed the clog and turned it up to 11. What is really making you use more water, fixing the clog or turning up the flow?

 

[Nintendesert]

I personally think Intel should just continue selling their chips as is and then sell a $1,500-$2,000 dollar chip, the -ESE. This stands for Extreme Soldered Edition, the ESE will have been properly soldered and the enthusiasts will eat it up. $$$ for stock holders.