leman is clearly a smart guy, which makes it more depressing when you see someone intelligent being willfully deceptive.
Forgive me, but where am I being deceptive? Can you point me to some studies or evidence that shows that running CPUs at this temperature will reduce its lifetime below a useful value? Al we've had so far in this thread is "everybody knows" arguments with absolutely no references or scientific argumentation.
The fact is: I don't know either. However, I am an IT professional who had over a hundred of computers in his care, most of them laptops used in a scientific environment where people routinely run hour or even day-long simulations that will push the system to its thermal limits. I don't think that I have observed a single CPU failure in all this time. And the little I know about microelectronics tells me that for the lifespans that normal users are conserved about, temperature is an absolute non-issue. Your machine will fail or get replaced long before temperature-induced electromigration will become a factor (and yes, there is also the thermal expansion stress but its not like there is anything you can do about it).
And the only relevant study/infopaper on the topic I know is this Texas Instruments document (
http://www.ti.com/lit/an/sprabx4a/sprabx4a.pdf) where they say that their processors are expected to run 10 years when operated at 105C — we are taking about 24/7/365 operation! Sure, its much smaller devices and its a different process, so we don't really know how this translates to modern behemoths from Intel or AMD but it gives you a value to think about. At any rate, its a recommended read for anyone interested on the topic.
[doublepost=1550490744][/doublepost]
One of the MacBook Pro's I returned had an issue where it was frequently and randomly bouncing off 100c. My current MBP has never seen 100c.
ALL 15" MBPs produced in last 6+ years will reach 100C when doing intense work, because that is how they are designed to operate. It's normal and there eis nothing wrong with it. The machine is built to keep the CPU operating at 100C when it draws 45W of power. It's that simple. And if you current MBP never saw this temps it simply means you never did anything remotely demanding with it.
I agree.
Intel's safety mechanism is meant as a last resort. A computer manufacturer is supposed to design the cooling system in such a way that it (Intel's failsafe) never has to kick in.
And what makes you think that the failsafe kicks in at any time? I don't know how Intel calculates turbo boost exactly (they have formulas in their public documentation but I never looked at it in detail), but its a function of current temperature and power draw. The way MBP works is that it keeps the CPU at 100C when it reaches its thermal design power. In other terms, it's where the balance point is reached. The cooling system dissipates exactly the heat produced by the CPU and the CPU stays at the same performance level. There is no failsafe shutdowns or radical throttling occurring (you'd notice that since the frequency would radically drop — the early power management bug with the 2018 MBP is a great example for this scenario).
[doublepost=1550491154][/doublepost]
This
techguided.com thread under the TJ Max section made sense to me. But again, I don't claim to know much about this.
The article overall is a good overview for beginner overclockers and computer enthusiast but the Tjunction section is all over the place. First he says "So, ultimately, if your processor is running close to its maximum allowed operating temperature, it is fine for the time being.". Which I agree with. Then he says "However, if your processor
is consistently operating near its maximum operating temperature while under load, that could be a sign that something is wrong.". I also agree with this, but... what does this even mean? Sure, getting Tjunction Max temperatures when your CPU is idling is certainly a problem. Not when its under full load though (which is the only situation where you should see this temps).
At the same time he write "[its fine...] unless it runs at a level close to its TJ Max for extended periods of time". Why? Where did he get that from? Where does Intel documentation state this? All of this is completely subjective and not based of any kind of evidence. So is it ok or is it not? And what does prolonged periods of time mean in the first place?
[doublepost=1550491331][/doublepost]
What I'm worried is that I think Apple is removing the temperature protection for short periods of time, if you run Cinebench you'll notice the temperature and power do not correlate any more when the temps hit initially 100C and flat line at this level. Looks like temperature sensor saturation while the actual is way higher, and this may kill the chip eventually, or destroy BGA joints after some number of cycles.
The power flat lines too though. And it flat-lines exactly at or around the chips TDP, which makes perfect sense. The chip simply stops boosting further and an equilibrium is reached. I don't see any evidence for sensor saturation, but maybe I didn't completely understand what you mean.
EDIT - I’m talking about 8th gen obviously, you couldn’t run earlier generations at almost twice the TDP.
Oh you could but you had to find the way around Intel's clock limiter
Overclockers have been doing this stuff on unlocked CPUs for ages. And Coffee Lake is almost an unlocked CPU since Intel put the limits much higher then on previous generations...
