Gamers red hot with fury over Intel Core i7-7700 temperature spikes Owners of Intel's new i7-7700 processors say the chips have been randomly revving up to extremely high temperatures, and Chipzilla won't give the issue so much as a second look. Reg reader Bastard-Wizard says that he and many other i7-7700 owners are finding that the chips will occasionally kick themselves into overdrive, …
It's probably just me but I'm finding Intel's attitude of late leaning towards arrogance. The whole shroud of secrecy surrounding the C2000 SoC issues and now this.
(https://www.theregister.co.uk/2017/02/06/cisco_intel_decline_to_link_product_warning_to_faulty_chip/)
Maybe it's time to try AMD again?
Its interesting that people think that being the de facto for years is a good thing. Supremacy breeds arrogance, I have said for years that Intel plays the market, not lead the industry. Tic-toc is what they have done for years instead of leaps and bounds.
Over the last few years with all the talk of green this and green that maybe people should remind companies like Intel (and companies that exclusively use their tech) that green also means not baby stepping your way through technological possibilities ... meaning instead of making 50 generations of tech to achieve what is mass producible right now maybe doing it in half or 1/4 of the steps, thus producing less e-wast and having a much lower carbon foot print never mind the amount of poisons spewed back into the environment.
I guess I have wished for many years people would stop being fan boi's so much and buy what makes more sense to the pocket book and environment. Companies that allow you to upgrade hardware a few times are and should always be preferred than ones that force full upgrades every baby step along the way.
For those thinking AMD might be a good choice are thinking since AMD has a history of sticking to a socket for a long time and allowing multiple generations of CPU to be compatible to it. Case and point, my last AMD system I started off with a quad core, then 2 years later bought a 6 core, then 2 years later an 8 core, costing me less than 800 dollars over a 6 year period and each time it felt much snappier after the upgrade. Also to boot much less energy (aka carbon foot print) and much less poisons produced. To me its a win-win situation, unless things go very wrong in the world we are gonna have to buy this stuff for life.
One of the first machines I built for myself had an AMD XP 3200. Eventually replaced with a C2D E8400 only for the better performance, not because it stopped working.
However, lately I've seen a number of AMD based machines die inexplicably. I suspect the chipset as most other components on the boards tend to be the same.
I like what AMD have done with Ryzen 7 (5 not so). I just hope there aren't any gremlins lurking.
And I hope Jim Keller comes back to up IPC again for Zen 2 and beyond. Or the IPC lag will only get worse...
Disclaimer: I own some AMD shares.
I can halfway agree-- the integrated Radeon drivers (for 780G etc) were kinda crap and regularly produced text corruption in Win7, and they had already kicked them down to "legacy" rarely-updated status after only a few years, but at least they bring OpenCL 2.0 on *anything*-- as opposed to nVidia... who can't be bothered to help anyone except with CUDA, but who at least still release drivers for their oldest 8-CUDA-cores-512-MB junk which may in fact still work with many modern mainboards and OSes, if you just need a flippin' DVI output :D
/me wants a toilet icon because every vendor blows crap in their own special way
I've built 3 machines with the i7-7700K so far. From what I've seen, the number one reason for temperature spikes is certain BIOS settings.
The 'Intel' setting allows each core to clock up/down individually according to load and yields the lowest temperatures. The default on some boards ramps up all cores in sync, unsurprisingly leading to rapid fluctuations in temperatures. I saw the chip hit 90 Deg C easily with this setting compared to 84 Deg C on the 'Intel' one. And that was with a £11 Akasa heatsink. With a Noctua NH-D15, that dropped to 68 Deg C under full load. No delidding required and no risk of liquid leakage either. The NH15-D15 is a beast of a HS though & you wouldn't want to courier a machine with this installed...
Of course, incorrectly applied thermal paste or a heatsink with only partial contact can be another culprit.
"No delidding required... Of course, incorrectly applied thermal paste or a heatsink with only partial contact can be another culprit."
Indeed, it seems that the sort of people who actually notice this are the sort of people who tend to play around with things, as with the person quoted in the article talking about delidding - for those unaware of what this is, it means dismantling the CPU package to remove the standard heat spreader, which can easily damage or destroy the actual CPU if not done correctly. It appears the problem is not that anything is actually wrong with these CPUs, but rather than they may simply be less friendly to DIY shenanigans than previous chips may have been. Intel say there isn't a problem because when tested under normal conditions there is not, in fact, a problem. It's only those wanting to play around that notice a problem because the normal behaviour gives them less headroom to play in. Certainly a shame for those people, but "CPU runs inside safe thermal limit" is a problem that is really hard to get worked up about.
I think I get what you're saying and I want to point out there's basically no reason to think that keeping the heatspreader would make it run cooler in general or heat up less aggressively. It's an additional thermal interface with the associated thermal resistance and with Intel's choice of heat sink compound underneath yours. Last time I checked, they're nickel-plated copper but have very little mass. The most obvious proper functions are to protect the die from your heatsink under the force of the retaining clips or springs and to present a wider area for being able to move the right amount of heat through substandard heatsink compound on top-- which Intel can't control but which an enthusiast makes a moot point. If Intel's HS compound is so much better than what an enthusiast will apply to the same surface area (the die), then they should be selling *that*. But probably they just tested a lot of market offerings and picked a good one with a decent price. It's still not as good a configuration as if the heatspreader was the same metal object as, or at least soldered onto, the thing with the mass and the exaggerated surface area. Yeah I know that's kind of impossible, I'm just saying. The best thermal conductors are the best electrical conductors for a reason. So you put your heatsink where the heatspreader goes, optimization complete.
And while DIY shenanigans certainly can send your CPU to the bad place, that will probably manifest as random crashes and temperature-related crashes because the damage will probably be limited to the bonding between the die and the package (unless you crack the die-- then you must be feeling lucky, to fire it up again). These people aren't complaining about crashes or even erratic behaviour, they're complaining about what seems to be a step backward in the "more performance for less power" sense.
I have had one for a couple of months and was really annoyed when I first got it and put the PC components together, thought my Fan was faulty at first and returned it.
In the end I had to change the fan speed profile so it is up to a reasonable level of annoyance, which basically means it only ramps up the speed at a high temp. i.e. from memory there are 5 temp / speed trigger points in the Bios and all apart from the first are set to the far right.
(and no I'm not talking about their political views, which would send anyone's temperature rising)
Indeed. Language is an ever-evolving thing, so I prefer to err on the side of clarity. For that reason, I use 'application' for a desktop program, and 'app' for a phone or tablet program. In time, the distinction will become less useful as more applications run on ARM, and more predominately ARM OSs play nice with mice and keyboards. Hey ho.
( I also use 'program' for computer software, and 'programme' for theatre and television shows.)
Would never have guessed. A while back I was pondering about why it is that I strongly prefer programs, and maybe it isn't the stupid phone craze starting 10 years ago. Maybe I just hate the wordlet 'app' (and not the word 'applet') because it was traditionally only used for 'job application' which was always a way to invite some potential employer to dutifully ignore me, which they did, and they do.
"I've called them apps since the days of DOS and 8.3 filenames because they went in the apps directory, as opposed to the tools dir, pics dir, docs dir, db dir etc."Most of my DOS-using colleagues called them programs, or proggies. Application was more a Mac-user term.
My proggies/applications lived in the BIN directory* and naturally my batch files lived in the BELFRY.
* Except for those that insisted on living in the root dir.
You have to wait for the core(s) to reach the needed temperature to prime a nuclear fusion, then surface temperature will reach the designed 7700K one. But you need to keep on opening browser *or* applications, because the CPU actually can tell the difference, just like any PR.
You think update is bad, try working with WSUS. Let's see, my server has 5 updates on 5 different architectures that have been superceeded, and should be cleaned up. The cleanup process should take, oh 30 seonds? 5 minutes? wait, what, how many hours?
From what I've read, it's possible to smooth things out with some mucking about in the underlying DB.
There must be something dodgy with the thermal design of that cpu, if it even manage to spike up to 90+ degrees with serious cooling. My venerable 5930k with a nice and toasty 140W tdp never get close to 90 even under max load, and the thermal mass should be large enough to effortlessly soak up any spike.
My 130w 3930k cores sit at 35-45 C but spikes to around 50 C under load when transcoding or something like that. Go-go water cooling!
It would sit at 60C before the water cooling and I was concerned, but that could have been a poorly installed fan.
It also returns a 20% higher cpubenchmark average than the (4 generations later) i7-7700 <pointlessly-smug/>
Not having the luxury of a more up to date model, the two 2.66GHz X5355s in my home office system provide central heating through the worst of winters, and currently are hitting 80-90C continuously with their stock metal heatsinks and fans when running around half the cores. If all eight get busy, the fans go in to helicopter mode somewhere about 95C at which point I usually start reigning things in. Not great, I know, and I have reseated the sinks with proper thermal paste but they have survived years.
In theory, the Xeons are supposed to be more resilient - interesting to see how the i7's last.
"For the i7-7700K, momentary temperature changes from the idle temperature are normal while completing certain tasks like opening a browser or an application," a spokesperson said. "We've looked into the reports and have not seen any unexpected behavior or indication of performance impact."
Opening an application really shouldn't put the temps up to the red line.
"For those annoyed by the constant on-and-off frenzy of cooling fans, Intel suggests you set the speed control so that the whirring kicks up gradually, rather than all at once.
We're sure customers will find that recommendation to be wholly satisfying and adequate for their concerns."
From the amount of people still complaining I would say the whole press release is inadequate.
Even though Intel are saying the the chip is "within limits", the life of an IC is shortened the longer it is operated at it's maximums.
What is worse here is the manifestation of a temperature "spike". The sudden temperature changes adds thermal stresses to the silicon. Computers are more reliable if they operate at a steady temperature. That is one of the reasons that some of us never switch our systems off.
Starting from cold is the most likely time for a fault to occur due to the large current inrushes into the un-clocked silicon. It seems that Intel are looking to increase sales by selling us kamikaze chips!
Thank goodness for AMD coming back into the arena.
Instead of changing how my fan works to come up to speed more slowly when the chip's temperature rises, which could put the chip at risk, wouldn't it be safer to just have the fan running at maximum all the time, so when a sudden temperature spike comes, it will be taken care of?
If the temperature spikes are, as Intel says, within the operating parameters of the chip, then that should take care of the problem, except, of course, for the noise and power consumption of the fans.
Host FoilWrapperPotatos.com not found: 3(NXDOMAIN)
Hurry up and register that thing! I need to get my wrappin' on with some foil, then apply it to some potatoes and then we'll have something!
Yeah, my ancient Intel Core Duo Mac Mini runs at up to 218°F while doing the Handbrake nasty. The stupid design has the heat coming out of the bottom foot and some out the rear exhaust port, just below the main port. You'll need to use your Targeting Computers. But I digress into Episode I. Anyway, I just built a platform for it out of an upside down toaster oven tray hot glued to some discarded video game packaging plastics, and now I've got something! It drives the heat out to the metal tray, and I can monitor it there, provide extra external cooling, or shut the beastie down. It was doing that by itself for awhile, hence my homemade heat-sink. I got a new one this week to take over duties, but that little guy won't die. The new one has zero heat dissipation on the chassis, and I can't tell it it has a fan or not yet. It's been running cool as a cucumber, even when the load was over 2.5 for some hours. Core i5 that new one.
Can't help thinking this problem in the article might be the OS that is also doing idle loops or other unnecessary stuff just because the cores are there. It shouldn't be using those other cores until there is a wide-thread load or general load that some subset of the procs cannot cope with. Generally speaking.
Does not work like that.......
For every 10 deg above 25 deg junction temp you loose 50% of the life at 25 deg, this is well known for all designers.
The cooler & closer you run to 25 or below the longer the life.
The 90 deg. matters WHERE it is measured., junction core case or air.
If you are measuring 90 at the case you have massive problems, due to thermal lag the die & junction will be WAY higher.
"Time for a lie down and a cup of tea."Have some sympathy to go with it; I believe it helps ;-)
I'm only a grammar Nazi when editing, otherwise I "forgive them for they know not what they do". When English classes stopped correcting spelling mistakes and grammar they created a generation of illiterates. Getting upset doesn't help much...
PS I also noticed I omitted an apostrophe in a post I made in the small hours. Oh, the shame of it!
Sadly, this article was published one week too late, for me. I bought mine already and instantly hit these problems. Sure, I could send it back but how would that help? I'd have to send the motherboard back, too, and basically start again from scratch!
The strange and annoying thing, to me, is that the temperature spikes almost instantly and yet the outside of the chip remains so cool you can touch it. The back of the socket on the motherboard and the board itself are also barely warm. This means that you can waste all the cash you have on the best cooling in the world and it will not do any good at all - heat dissipates according to the heat equation and you can cool the cold surface of the IHS all you like, if the heat isn't getting to it, it isn't going to help.
Mine spikes from 28 degrees to 100 degrees, where the chip is throttled, within under two seconds when I fire up Prime95. I first thought it was a bug with the temperature reporting, it was so bad.
Also, if Windows Update or starting Firefox roasts the CPU, exactly how do Intel expect it to perform under real load, such as execution of my machine learning and image processing algorithms?
Personally, I think the Intel bean-counters caused this by "cutting costs" and I hope and pray that AMD school them for us. Unfortunately, Ryzen isn't going to do it unless regular generational updates keep up the pressure.
"For the i7-7700K, momentary temperature changes from the idle temperature are normal while completing certain tasks like opening a browser or an application," a spokesperson said.
"It's normal for the temperature to rise while the CPU is logging the task completion for later transmission to the proper authorities."
with the previous generations there was a Tcase value that essentially ran the fan speed and damped the temp value feed to the BIOS. gen 7 uses the hottest core ( TJuntion) and hasn't got a Tcase value - could it be that Intel CPUs have always spiked (though not to 90C ) on single cores just that nobody noticed 'cos their fans didn't spin up ?
The Delid modification I mentioned in my tip is somewhat commonly done on recent Intels. Since Ivy bridge, Intel has not soldered the IHS (integrated heat spreader) to the CPU die even on their high-end chips. Instead they use regular thermal paste, the same goop you would normally use between your cooler and your CPU's IHS. (I note that nearly all AMD chips, including Ryzen, use solder.)
Dilidding a soldered CPU is only done by mad folks, imho. But these aren't soldered, so the modification is fairly safe if you know what you are doing. The IHS is held to the chip by a form of black silicone glue.
The modification on post-Ivy Intels basically involves removing the IHS (there are specific delidding tools that can be made, some use the hammer-and-vise method, others with more money than sense use razor blades.) The crappy TIM is removed and either better paste or liquid metal TIM is applied. The IHS is then replaced.
I have an i7-7700K. The K skus have an unlocked multiplier. This is a feature that's there *explicitly* for overclocking (or in my case, underclocking.) Intel even sells an enhanced warranty for K chips covering overclocking.
Even the Pentium 4 didn't behave this badly.
It's my understanding that the issue affects some non-K chips of this generation too.
Launching a web browser should not cause a 30-50C temperature spike. I probably have a particularly bad chip, but there's enough grumbling about this to suggest that there is an actual issue. The fan and pump ramping is just an added annoyance.
It's possible that this is just an antsy thermistor. I sorted the problem for now by downclocking and undervolting the Vcore and PLL and setting my cooling system to run at a fixed, tolerable speed until 60C is hit. It no longer spikes into the scary zone. I now have an antsy 6700K, basically.
Yes, gaming boards do tend to overvolt the CPU Vcore a bit out of the box. Correcting to Intel's specs did not remedy the issue for me. Indeed, this seemed to have no effect on the spikes until I also tweaked the PLL and downclocked.
To sum, here's what it took to make the spikes (or buggy thermistor) manageable:
-Delidding
-Using a fancy-pants water cooling system
-Setting voltages and clock speeds to last generation's specs
It works now and I can now fire up a web browser without fear of a meltdown.
Intel's response is quite poor given just how many people seem to have some form of this problem. I'll definitely give Ryzen a look once the kinks are ironed out.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
