2018 i9 MacBook Pro throttling discussion [merged]

[Macgeekswe]

Tell that to owners of the 2014 (and to a less degree the followup) Retina iMac(s), those things will hit 105C under sustained load... (Although they do seem to be surprisingly failure resistant at this point.)

This is a limit put in by Intel. It's part of the specs for the 8th generation. Anything higher than 100C is considered dangerous today, across all manufacturers. I seem to recall the 3rd and 4th generation having their limit at 105 or 110 degrees Celcius.

At the moment, I will still believe that this is a new processor with software not specifically optimised to use it.

The fact that you can put it in the freezer and instantly see a huge performance improvement pretty much kills the possibility that this is a software issue.

At the moment, I would like to think that this is a new processor with software not specifically optimised to use it.

This does not mean I did not read the thread. Written words do not always equal truth, truth is something you find for yourself and believe in. I would be naive if I were to just assume these things so soon after a release.

Then why are you presenting it as a fact in your reply (quoted below) that it's a "software optimization"?

That is Apple giving users the option to rant. 😛

Do not worry just yet. This is only software optimisation that needs to be done, that is all.

 

[Aea]

The fact that you can put it in the freezer and instantly see a huge performance improvement pretty much kills the possibility that this is a software issue.

You know what.

How humid does a fridge get?

 

[Macgeekswe]

You know what.

How humid does a fridge get?

You need a freezer. Also you need to be able to somewhat seal the door fairly well otherwise you risk turning the freezer to into an oven. That being said, humidity in a freezer is fairly low but I wouldn't recommend it.

 

[Malus120]

This is a limit put in by Intel. It's part of the specs for the 8th generation. Anything higher than 100C is considered dangerous today, across all manufacturers. I seem to recall the 3rd and 4th generation having their limit at 105 or 110 degrees Celcius.



The fact that you can put it in the freezer and instantly see a huge performance improvement pretty much kills the possibility that this is a software issue.

Yeah at this point the evidence against the i9 isn't looking good (and to be fair, it's not just Apple struggling with these chips).

People trying to argue "But, but, the software's just not optimized yet," aren't thinking clearly IMHO. While there are certainly (hopefully?) likely to be things Apple can do on the OS/firmware side to mitigate these issues by adjusting the clockspeed curves to improve the all core maintainable speed (by sacrificing the boost speeds in these scenarios), if the i9 is actually as thermally limited as it appears, they can't program there way around physics...

 

[Aea]

You need a freezer. Also you need to be able to somewhat seal the door fairly well otherwise you risk turning the freezer to into an oven. That being said, humidity in a freezer is fairly low but I wouldn't recommend it.

I feel like it's an SO question, but what about the fridge's humidity?

 

[Macgeekswe]

I feel like it's an SO question, but what about the fridge's humidity?

Pretty sure it's higher than a freezers hehe.

 

[hoon2999]

I can't believe that there are people still defending apple on their poorly designed thermal solutions. Thermal issue exists. It's a fact. I mean, come on now. The macbook performs drastically well once they put in the freezer.

The thermal throttling issue is not rare in laptop. I mean, you can have so much of fans packed into the "portable" device. But many laptops (especially the expensive ones) tends to solve throttling issues somehow. And people did not expect macbook pro that costs at least $3k to have these kind of issues. I'm glad that I did not pull the trigger on i9 CPU because I knew it's too good to be true with their chassis design.

 

[verius]

I am beginning to rethink my decision to order the 2018 15' 2,6 i7 , maybe ill wait and see if the thermal throttling affects the i7s badly also.

 

[enri1357]

I can imagine running a demanding windows VM with 5 Cores (10 threads) will kill this i9 and would get the same low cpu clock throttling. Not to mention I expect I would run several of them together once I get 32GB ram....

I was planning for this usage, there is not much turbo boost strategy optimization Apple could make for my case because BELOW base clock throttling is really a big problem. I experienced similar issue with my 2011 Sony laptop.

 

[PeterJP]

I'm not proficient in German, but, it's more or less decipherable (and google translate does an excellent job albeit mangling the graphs) and the key graph (Cinebench multicore over time) can be found at about the 4/5 mark.
(...)

2. Given the chassis constraints, and what we know about Intel 8th gen, this behavior isn't unexpected. It's not a defect (on Apple's part) its just a fact of life that when you stuff more cores onto the same die without significantly redesigning the architecture (Intel's knee-jerk response to AMD), something's gotta give, and in this case that something is sustained multicore performance (vs a desktop where TDP/heat don't really matter, or a design that was meant to operate with four cores in a notebook).

Exactly. The article clearly states that this is a problem with Intel's Coffee Lake processors whose temperatures go crazy after short periods of use. I'm certain that we'll be seeing the same behaviour in Windows laptops. So wait a little before starting your class action lawsuits to be sure you're targetting them at the right company.

Intel is in really deep trouble. 10nm won't be here for another few years and now their alternatives aren't any good anymore. The x299 processors have the same heat problems, so they're having trouble in the HEDT market as well. AMD could not have timed their comeback better.

 

[Adamantoise]

At the moment, I would like to think that this is a new processor with software not specifically optimised to use it.

This does not mean I did not read the thread. Written words do not always equal truth, truth is something you find for yourself and believe in. I would be naive if I were to just assume these things so soon after a release.

You can believe what you want, but all symptoms point to thermal throttling.

 

[clangers23]

Prime95 ONLY puts a load on the CPU, so what you are saying is entirely false.




90-100C is going to murder your CPU long-term. At 100C the CPU will be forced to underclock even more and if for some reason hits 105C the entire machine will shut down to protect the CPU.

I wasn't referring to Prime95 I was outlining possible stresses placed on a thermal cooling solution.

 

[Macgeekswe]

Exactly. The article clearly states that this is a problem with Intel's Coffee Lake processors whose temperatures go crazy after short periods of use. I'm certain that we'll be seeing the same behaviour in Windows laptops. So wait a little before starting your class action lawsuits to be sure you're targetting them at the right company.

Intel is in really deep trouble. 10nm won't be here for another few years and now their alternatives aren't any good anymore. The x299 processors have the same heat problems, so they're having trouble in the HEDT market as well. AMD could not have timed their comeback better.

The i9 has already been in several Windows laptops for months and roughly half of them suffers from thermal throttling, however, none of them suffers from thermal throttling so bad that it actually goes BELOW the base clock - That's exclusive to the MBP.

I wasn't referring to Prime95 I was outlining possible stresses placed on a thermal cooling solution.

The only stress Aea placed on the MBP was on the CPU, basically voiding what you wrote as a reply to his post. I'm not sure what you're trying to say here cause it honestly doesn't make a lot of sense from a technical standpoint.

 

[duervo]

The fact that you can put it in the freezer and instantly see a huge performance improvement pretty much kills the possibility that this is a software issue.

That video was disappointing to me for two reasons:

1. The 2017 i7 was not put into the same freezer and tested with the same workload, to see if it also noticed similar gains in performance.

2. The presenter failed to rule out something else that could be causing the issue: They didn’t validate if the thermal paste application was messed up, suggesting a manufacturing issue, or if there is a defective heat sink. I’m not saying that any of those are actually the cause for the issue one way or the other, just that without confirming, he’s only jumping to conclusions.

His editing chops are pretty decent though.

Note: I’m addressing that video specifically. The number of people elsewhere reporting the same thing seems to suggest a valid problem, but for somebody that’s going to do a “review” and post it on YouTube, with that quality of editing, I’d expect a bit better testing. Not as hard core as Gamers Nexus, but something better than what he did there. Seems like he spent more time making the video look nice than he did on actually planning an adequate testing methodology.

 

[Adamantoise]

That video was disappointing to me for two reasons:

1. The 2017 i7 was not put into the same freezer and tested with the same workload, to see if it also noticed similar gains in performance.

2. The presenter failed to rule out something else that could be causing the issue: They didn’t validate if the thermal paste application was messed up, suggesting a manufacturing issue, and they also just immediately jumped to the conclusion that the cooling system isn’t good enough for the i9. I’m not specifically saying that it’s the cause for the issue, just that without confirming, he’s just jumping to conclusions.

His editing chops are pretty decent though.

???

The objective of putting it in the freezer was to isolate if thermals were the reason for the throttling. Mission accomplished as far as I'm concerned.

With regards to the second point, application of thermal paste is considered part and parcel to be a component of "the cooling system" onus is on Apple's QA/QC to get it right.

 

[macintoshmac]

You can believe what you want, but all symptoms point to thermal throttling.

I am not believing anything at the moment. Like you yourself said, symptoms point to throttling, in my opinion is one perspective. Mine is software optimisation. To each his own. No saltiness needed.

 

[Adamantoise]

I am not believing anything at the moment. Like you yourself said, symptoms point to throttling, in my opinion is one perspective. Mine is software optimisation. To each his own. No saltiness needed.

No one is salty here; there's just no evidence whatsoever to even guess software optimization, while there is plenty to suggest thermal throttling.

It's just wishful thinking on your part is what I'm getting at.

 

[macintoshmac]

This is a limit put in by Intel. It's part of the specs for the 8th generation. Anything higher than 100C is considered dangerous today, across all manufacturers. I seem to recall the 3rd and 4th generation having their limit at 105 or 110 degrees Celcius.



The fact that you can put it in the freezer and instantly see a huge performance improvement pretty much kills the possibility that this is a software issue.





Then why are you presenting it as a fact in your reply (quoted below) that it's a "software optimization"?

It was said to a person who usually worries exponentially if he feels there is something wrong with the software, and if there may be something with the hardware then all bets are usually off, so I thought. It was meant for him directly. if my comment made no sense to you, or if you feel I was wrong, should probably just ignore my lack of knowledge over you.

I may be wrong, I may be right, but I hope I have answered your curiosity as to why I "presented it as a fact" which after your nitpicking I am of the opinion that I have worded incorrectly, and should have been more into the probables and possibles, rather than sound like it is a fact. Thank you for pointing out.
[doublepost=1531898371][/doublepost]

No one is salty here; there's just no evidence whatsoever to even guess software optimization, while there is plenty to suggest thermal throttling.

It's just wishful thinking on your part is what I'm getting at.

No evidence to even guess software optimisation needed? Really? This is not a new concept now, right? I agree that the chip running hotter and Apple's history with thermals makes it far easier to think on the lines of thermal throttling and there is evidence of it, but is there really no reason to think that that evidence may itself be arising out of running software not optimised for 6 cores - meaning - when software is not able to utilise the cores properly, there is a chance of processor heating up due to undue load on it.

I am only making some impressions based on comments here and there that suggest that some software are hardcoded to use 4 cores only, and when run on 6 cores, are not able to effectively utilise the processor, which I know will cause overheating. Just voicing out. I could most definitely be wrong.
[doublepost=1531898428][/doublepost]

You know what.

How humid does a fridge get?

I did not say what you are quoting me as having said. 😛

 

[duervo]

???

The objective of putting it in the freezer was to isolate if thermals were the reason for the throttling. Mission accomplished as far as I'm concerned.

With regards to the second point, application of thermal paste is considered part and parcel to be a component of "the cooling system" onus is on Apple's QA/QC to get it right.

And he never verified, is my point. Just jumped to conclusions.

 

[jaredm1]

Look on the bright side / this laptop can keep you warm over Winter!

Given Apple refer to thermals of the i9 on the MBP ordering page I’m pretty sure they knew about this before release. I’m sure they did what they could to minimise throttling.

 

[Malus120]

Exactly. The article clearly states that this is a problem with Intel's Coffee Lake processors whose temperatures go crazy after short periods of use. I'm certain that we'll be seeing the same behaviour in Windows laptops. So wait a little before starting your class action lawsuits to be sure you're targetting them at the right company.

I was with you til the last sentence. I want to clarify that my remarks, while generally applicable, were aimed at the throttling situation on the (base) 13" 2018 MBP. At the end of the day Apple has responsibility for the components it chooses to put in its machines, and the level of throttling we're seeing in early reports of the i9 are just unacceptable (far worse than what we're seeing on comparable PC laptops). If the chassis can't handle it, Apple shouldn't be offering it.

Intel is in really deep trouble. 10nm won't be here for another few years and now their alternatives aren't any good anymore. The x299 processors have the same heat problems, so they're having trouble in the HEDT market as well. AMD could not have timed their comeback better.

Couldn't agree more. Really wish (hope) Apple would (will) start dual sourcing CPUs from both Intel and AMD (and not switch to ARM for gods sake).

The i9 has already been in several Windows laptops for months and roughly half of them suffers from thermal throttling, however, none of them suffers from thermal throttling so bad that it actually goes BELOW the base clock - That's exclusive to the MBP.

THIS. Right here, is the key point. Throttling below base clock is something that really just shouldn't be happening. HOPEFULLY, Apple just massively ****ed up with the way the firmware handles thermal load, but if not then Apple really just shouldn't be offering the i9.

Given Apple refer to thermals of the i9 on the MBP ordering page I’m pretty sure they knew about this before release. I’m sure they did what they could to minimise throttling.

And therefore if this isn't something they can fix with a quick firmware update, then they knew about how severe it was in advance and sold it for a premium anyway. Really hoping there's more to this story than meets the eye...

 

[The Mercurian]

If this issue is as bad as it appears to be there will for sure be lawsuits over this

 

[Queen6]

The benchmark starts when the CPU Utilization (Bottom Graph) hits 100%, and ends when it drops to near zero. It starts off strong, and then as soon as it hits the 100C ceiling (which I suspect is a no-throttle, full-bore) zone it starts going all over the place. Jumping between 2.0 Ghz and 3.0 Ghz regularly.
[doublepost=1531889196][/doublepost]

I don't upgrade my machines all that often (3-4 years) so it's a definite improvement, my beef is with selling a top-end processor that is unable to really perform at it's full potential for more than 15-20 seconds before being passed by the mid-end. The $400 cheaper mid-end.

Straight up that's terrible results - My W10 notebook with 8750H

3 hours at full load CPU temp in the region of 70C

Q-6

 

[leman]

I will assume that you are joking. [...] The problem here is heat, not software optimization.

No, I'm not. Have you watched the video? You can see that CPU utilisation is surprisingly low. Which is a clear indicator that the software is not utilising all of the CPU's capabilities. Its very much possible that the software is detecting the computer type it runs on and chooses one of the preconfigured profiles.

As to heat: the CPU doesn't even get too hot. Throttling occurs at 100C, in the test he is not even getting close.

For clarification: I'm not saying that i9 models do not throttle, I am saying that this particular video the test is probably flawed since the CPU stats don't make much sense to me.

Nevertheless, i9 chips with 6 cores can run 12 threads and any modern quad core i7 chip can run 8 threads.

Hyperthreading is situational. It works by trying to run two threads using resources of one core. This can only work well if you have code that is stalled a lot by waits and thus has hardware resources to spare. In dense numeric code hyperthreading often leads to excessive heat generation with very little benefit to performance (it can be counter-productive actually, since it can lead to throttling). So far, my experience with latest gen CPU is to run 6 threads on a quad-core hyper threading CPU. Anything beyond that only adds overhead and actually slows things down.
[doublepost=1531905799][/doublepost]

Tell that to owners of the 2014 (and to a less degree the followup) Retina iMac(s), those things will hit 105C under sustained load... (Although they do seem to be surprisingly failure resistant at this point.)

Of course they will. Its the max safe operating temperature for these CPUs. I did quite a lot of testing of CPUs on 2016 models and the result is always the same — it seems that the cooling system is carefully designed to keep the CPU at 100C when it runs at max clock. This way, you can prevent throttling and still have compact cooling.

 

[Mathematig]

So this means that all 8th gen CPU have this problem no matter which models you buy?

That’s even worse. Why even buy the 2018 models then.

Do you have any idea what you are saying?