Another member of the Haswell club

[inf64]

Do a simple google search and you will find results like this:
http://www.overclock.net/t/1404303/haswell-temps-seem-too-high-something-wrong
http://forums.whirlpool.net.au/archive/2109844
http://forums.whirlpool.net.au/archive/2111243
http://www.tomshardware.com/answers/id-1716005/cpu-overheating-playing-games.html
http://www.bit-tech.net/news/hardware/2013/06/06/haswell-heat/
http://www.xbitlabs.com/articles/cpu/display/core-i7-4770k_12.html (describes the issues of overshooting the vcore under load, hits 95-100C under linx with great air cooler @4.4Ghz; behaves much worse in this regard than either IB or SB )

I could probably find some from this forum too, I just didn't search it.

 

[ShintaiDK]

Do a simple google search and you will find results like this:
http://www.overclock.net/t/1404303/haswell-temps-seem-too-high-something-wrong
http://forums.whirlpool.net.au/archive/2109844
http://forums.whirlpool.net.au/archive/2111243
http://www.tomshardware.com/answers/id-1716005/cpu-overheating-playing-games.html
http://www.bit-tech.net/news/hardware/2013/06/06/haswell-heat/
http://www.xbitlabs.com/articles/cpu/display/core-i7-4770k_12.html (describes the issues of overshooting the vcore under load, hits 95-100C under linx with great air cooler @4.4Ghz; behaves much worse in this regard than either IB or SB )

I could probably find some from this forum too, I just didn't search it.

You never even checked your links did you? The first one is OC between 4.2 and 4.5Ghz for example. Then articles about OC?

Let me help you, we talk about stock cooler and stock clocks. And issues with motherboards using auto turbobin OCs.

 

[ShintaiDK]

Just a note on the motherboards that does this (Essentially most Z boards.). The CPU warranty is actually void the first time it goes beyond stock turbo.

 

[inf64]

Links #2,3,4 are not OCs. The article speaks about the issues of overshot vcore under full load in prime95/linx while using aftermarket air/WC solutions on very slight OCs.

 

[ShintaiDK]

Links #2,3,4 are not OCs. The article speaks about the issues of overshot vcore under full load in prime95/linx while using aftermarket air/WC solutions on very slight OCs.

Link 2 isnt overheating, its 90C. And most likely a Z87 board ith turbo OC as well.
Link 3 is OCing and looks more like a troll post.
Link 4 uses a mobo that got the the turbo OC on by default.

Was that the best you could find? You claim it happend on boards that didnt use the multicore boost. Yet you supplied ZERO edvidence.

 

[inf64]

How many boards actually use this multicore boost feature then? I do not care which boards are affected as users should get a functioning product if the board is specced to support the SKU (like those boards officially do). What I see here is intel and Asus scamming their customers as the end product is throttling under load. There should be a big disclaimer on intel's product sheet that states the "incompatibility" with those manufacturers that sell overpriced "feature packed" boards that actually do not work as they should with Haswell. Also there should be warranty void warning clearly stated and visible. One doesn't pay ~500$ to get broken product at stock settings.

 

[ShintaiDK]

How many boards actually use this multicore boost feature then? I do not care which boards are affected as users should get a functioning product if the board is specced to support the SKU (like those boards officially do). What I see here is intel and Asus scamming their customers as the end product is throttling under load. There should be a big disclaimer on intel's product sheet that states the "incompatibility" with those manufacturers that sell overpriced "feature packed" boards that actually do not work as they should with Haswell. Also there should be warranty void warning clearly stated and visible. One doesn't pay ~500$ to get broken product at stock settings.

Most Z series boards got it. And its on by default.

Why is Intel scamming? Its not Intel autooverclocking the CPU beyond stock speed on the stock cooler.

 

[R0H1T]

In other unrelated news ~

Intel Cracks Down on Motherboard Vendors Offering Overclocking on non-Z Chipset

Over the past couple of months, motherboard vendors from across the industry offered BIOS updates for their motherboards based on Intel B85 Express and H87 Express chipsets, which enable CPU overclocking for Intel's unlocked Core processors denoted by "K" brand extension (Core i7-4770K, i5-4670K). This reportedly hasn't gone down well with Intel. The company charges anywhere between $35 and $45 per piece of its Zxx Express and Xxx series chipsets to motherboard vendors, while Bxx and Hxx series are significantly cheaper ($25 to $35). Sensing a clear threat to its revenue, from the prospect of motherboard vendors coming up with high-end or overclocking-ready (strong CPU VRM) motherboards based on cheaper chipsets in the near future, Intel cracked down on them.

Intel is giving final touches to a CPU microcode update that restricts Core "K" Haswell processors from overclocking on chipsets other than Z87 Express. A microcode update can be deployed both through BIOS updates, and surreptitiously through Windows Update. Intel's used the tried and tested "stability" bogey to justify the update. While it's true that motherboards based on B85 and H87 tend to feature weaker CPU VRM, there's nothing to say that ASUS wouldn't have gone on to design its next ROG Maximus on H87 Express, and save on manufacturing costs. While it's purely hypothetical, something like that wouldn't be in Intel's commercial interests. What next? Intel will push this new microcode update on to motherboard vendors, instructing them to issue BIOS updates with it; and future batches of Intel "K" CPUs may not support overclocking. If that isn't enough to contain the problem, Intel may give Microsoft a ring, and ask it to push the update through Windows Update. It tried that once in the past.

Source

Intel becoming the new (bad)apple ? :whiste:

 

[ShintaiDK]

The prices are pretty wrong. And the Techpowerup headline is hillarious at best. Since there is no crackdown on mobo makers.

Z87 cost 48$, H87 cost 44$. I doubt the 4$ is why Intel does it. Specially when people need to buy a CPU that cost 19$ to 27$ more.

However what Intel wont accept, is an unrealiable platform. The same reason why Intel entered the motherboard business in the first place.

But not so much drama and sensationalism in that

The original article is here btw:
http://www.hardware.fr/news/13249/intel-bloquer-oc-k-h87-b85-express.html

Intel will block K OC on H87/B85 Express
We finally got a response from Intel concerning the sustainability of overclocking processors K on B85 and H87 Express chipsets:

"Intel Plans to release a firmware update That limits processor core overclocking to Intel Z87 WW30'13 based platforms."
Bad news, Intel will next week make available a new firmware (firmware somehow processor) which will block this feature on the processor. A much more radical than the correction firmware Management Engine that could only make more sustainable solution.

Therefore the possibility to overclock on H87 or B85 Express is not sustainable, because if the updated firmware is usually via a bios update (optional so) it is not the only way. So even if the manufacturers do not integrate the new firmware microcode in their bios that could still be updated at time or another by a patch to the operating system on the occasion of the release of a firmware fixes a critical bug. Finally, it is likely that processors manufactured in a few weeks it will integrate automatically microcode.

Anyway, move along, there will unfortunately soon be nothing to see!

Update: We still asked for confirmation to our contact at Intel that the update is indeed that of the processor microcode and not the Management Engine. We are still awaiting a response.

 

[R0H1T]

The prices are pretty wrong. And the Techpowerup headline is hillarious at best. Since there is no crackdown on mobo makers.

Z87 cost 48$, H87 cost 44$. I doubt the 4$ is why Intel does it. Specially when people need to buy a CPU that cost 19$ to 27$ more.

However what Intel wont accept, is an unrealiable platform. The same reason why Intel entered the motherboard business in the first place.

But not so much drama and sensationalism in that

The original article is here btw:
http://www.hardware.fr/news/13249/intel-bloquer-oc-k-h87-b85-express.html

I'd say that forcing a microcode patch through BIOS/Windows updates is nothing less than scandalous & its not like Intel doesn't sell mobos with errata like the recent USB/sleep bug right?

 

[ShintaiDK]

I'd say that forcing a microcode patch through BIOS/Windows updates is nothing less than scandalous & its not like Intel doesn't sell mobos with errata like the recent USB/sleep bug right?

Why? Its a hack to allow it in the first place and its not validated for the platforms. Was it also scandalous when AMD disabled the ability to unlock cores? The obvious answer is no.

 

[coffeejunkee]

http://www.anandtech.com/show/6989/...aswell-gigabyte-msi-asrock-and-asus-at-200/24

"In terms of performance the Extreme6/AC comes with MultiCore Turbo as standard, giving that extra boost over motherboards that do not have it."

.. Sure hope it does, its the reason I picked it over Asus when going non-K

Well, Asus got hammered pretty badly by some reviewers for having multicore turbo enabled by default on their Z77 boards, 'unfair advantage' blablabla. They disabled it by default on their Z87 mobo and now they don't sell :hmm:

The question is, why is shooting +0.1v on AVX loads? Would it be unstable if it wasnt for that extra pump in voltage on those kind of loads?

I wonder about this too, Prime/Linx already cause the most heat of any program, no need to increase it even more. Sandy/Ivy are plenty stable without a voltage boost.

Prime95 is not a power-virus program though. Prime95 is an actual program that people use to find Mersenne primes.

Yes, but for stress-testing we choose 'torture test'. No idea how hot Prime gets when you use it for what it is actually designed to do.

 

[cytg111]

Just a note on the motherboards that does this (Essentially most Z boards.). The CPU warranty is actually void the first time it goes beyond stock turbo.

So, put your brand new Intel CPU in 9 out of 10 Z-boards will void its warranty? Good luck with that one in court.

edit : also, looks like im going into undervolting territory until I decide(or not) other than not-stock cooling.

 

[PPB]

I wonder about this too, Prime/Linx already cause the most heat of any program, no need to increase it even more. Sandy/Ivy are plenty stable without a voltage boost.

Probably they found out that the arch can run stable at lower volts on any code path other than AVX. So considering that this one is more of an exception than the rule (the adoption rate is slow), they decided to go for lower volts with the exception of when AVX code is run.

Anyone currently owning a haswell rig should test a real world program that uses AVX, and check if it ups the voltage by 0.1v too, and how it affects temps.

 

[coffeejunkee]

Could be, my Ivy uses around 1.050V under Linx load. Would be nice if Haswell owner could confirm 0.950V under non-avx load.

 

[cytg111]

Most Z series boards got it. And its on by default.

Why is Intel scamming? Its not Intel autooverclocking the CPU beyond stock speed on the stock cooler.

Allright, then riddle me this, how much of temp rise can we expect from this "MCT" ? 5 degrees? 10 degrees ? Right now, where I live, it is about 25 degrees celcius ambient. I guess that means that stock Haswells - with no MCT are not certified for running in africa, middle east or anywhere else in the world where ambient temps go above 35C.

 

[Kenmitch]

Could be, my Ivy uses around 1.050V under Linx load. Would be nice if Haswell owner could confirm 0.950V under non-avx load.

At what clock speed is your Ivy stable at 1.050v in LinX without avx support?

LinX without the AVX support does not up the vcore .100v on my motherboard when running stock settings.

It's either a uEFI infancy issue or it's just the way Intel programmed the VID on Haswell for some reason. Leaning towards the VID currently.

Quick and dirty results on my 4670k running stock settings. I had to use a massive -0.140v offset to get it this low tho. Using a fixed vcore eliminates the xtra 0.100v the VID calls for when using AVX extensions on my motherboard.

Not sure if it would be feasible or 100% stable at these settings as the chip idles at 0.568v's with this offset.

 

[coffeejunkee]

Ivy (or at least my cpu) uses the same vcore for both avx and non-avx Linx, at stock settings it's slightly fluctuating between 1.048 and 1.056 according to cpu-z.

You say non avx Linx doesn't give vcore bump but what about the avx version at stock settings?

Afaik I know it's indeed intended behaviour, but the question still is: why?

 

[Kenmitch]

Ivy (or at least my cpu) uses the same vcore for both avx and non-avx Linx, at stock settings it's slightly fluctuating between 1.048 and 1.056 according to cpu-z.

You say non avx Linx doesn't give vcore bump but what about the avx version at stock settings?

Afaik I know it's indeed intended behaviour, but the question still is: why?

In my above images the LinX is the one without AVX extensions. IBT on the other hand has AVX extensions. Guess I could try the LinX with AVX extensions if you want. Would most likely give the same .100v bump that IBT does.

I just tried the LinX with the AVX extension support....Same results as with IBT the .100v bump is active.

 

[ShintaiDK]

Allright, then riddle me this, how much of temp rise can we expect from this "MCT" ? 5 degrees? 10 degrees ? Right now, where I live, it is about 25 degrees celcius ambient. I guess that means that stock Haswells - with no MCT are not certified for running in africa, middle east or anywhere else in the world where ambient temps go above 35C.

I cant say since I havent tried or seen the difference with Haswell. But if we look on IB. You could go from 80-85C to throttle at 105C by changing 4/3/2/2 to 4/4/4/4 and using stock cooler. I assume the VIDs are the main cause of it.

 

[coffeejunkee]

In my above images the LinX is the one without AVX extensions. IBT on the other hand has AVX extensions. Guess I could try the LinX with AVX extensions if you want. Would most likely give the same .100v bump that IBT does.

I just tried the LinX with the AVX extension support....Same results as with IBT the .100v bump is active.

Thanks for those, but if you undervolt you're not using stock vcore. I'm wondering if the bump happens when you use stock speed and stock vcore, ie 'auto'.

I haven't used the stock cooler, but for me the difference in temps between mct and non-mct was very small. I'm able to run Linx at 4.0GHz all cores using same vcore as stock though, and difference in temps vs stock is just 1 degree so it's hard for me to believe mct could cause 15-20 rise. Then again, the stock cooler is really really crappy (or maybe op needs to check the push-pins again).

 

[Kenmitch]

Thanks for those, but if you undervolt you're not using stock vcore. I'm wondering if the bump happens when you use stock speed and stock vcore, ie 'auto'.

I haven't used the stock cooler, but for me the difference in temps between mct and non-mct was very small. I'm able to run Linx at 4.0GHz all cores using same vcore as stock though, and difference in temps vs stock is just 1 degree so it's hard for me to believe mct could cause 15-20 rise. Then again, the stock cooler is really really crappy.

The same thing happens when using the stock vcore. Using auto or normal gives you the .100v bump with anything AVX it looks like. The only way to eliminate it is going to a fixed vcore value.

I was trying your " Would be nice if Haswell owner could confirm 0.950V under non-avx load " and the only way is to use a massive - offset value.

 

[coffeejunkee]

The same thing happens when using the stock vcore. Using auto or normal gives you the .100v bump with anything AVX it looks like. The only way to eliminate it is going to a fixed vcore value.

I was trying your " Would be nice if Haswell owner could confirm 0.950V under non-avx load " and the only way is to use a massive - offset value.

Ok thanks, so then I guess it's safe to assume the vcore bump is the cause of the stock cooler failing, even at stock settings.

PPB suggested that Haswell runs at lower vcore than Ivy under non-avx loads and that the bump merely puts it at 'normal' levels, so that's what I was asking. Considering your quite large negative offset though I'd say Haswell uses the same vcore, but I'd like to know for sure what vcore your cpu uses at stock non-avx Linx.

 

[Kenmitch]

Ok thanks, so then I guess it's safe to assume the vcore bump is the cause of the stock cooler failing, even at stock settings.

PPB suggested that Haswell runs at lower vcore than Ivy under non-avx loads and that the bump merely puts it at 'normal' levels, so that's what I was asking. Considering your quite large negative offset though I'd say Haswell uses the same vcore, but I'd like to know for sure what vcore your cpu uses at stock non-avx Linx.

At work now.

I can confirm it uses way too much vcore at stock settings when AVX hits tho. Same vcore fixed value gets me 4.2ghz stable even under AVX loads.

 

[cytg111]

I cant say since I havent tried or seen the difference with Haswell. But if we look on IB. You could go from 80-85C to throttle at 105C by changing 4/3/2/2 to 4/4/4/4 and using stock cooler. I assume the VIDs are the main cause of it.

- Oh well, guess it is incentive to go for a better and perhaps more silent cooling solution ala megahelm.