2018 i9 MacBook Pro throttling discussion [merged]

[ondert]

Well.. maybe I should feel fine passing to iMac this time instead of MacBook Pro.
This was my only concern when I saw the i9 option because Dell XPS was also facing the same problems and to me, intel is just lying about the tdp values. We’ll see how this develops for the iMacs. Any 8th gen intel cpu runs hotter than their previous equivalent. For example, noctua nh-l9i cools 65w tdp cpus very well in mini itx systems.. ..until now. It can not handle i7-8700 although it is said to be 65w by intel.

 

[leman]

Any 8th gen intel cpu runs hotter than their previous equivalent.

Yeah, that is a problem. You can't just slap 50% more cores onto a die and pretend that the heat situation will stay the same. Anyway, I'll get my i9 at some point (couldn't order early, so still weeks to wait), and I will test it out. I think the problem with all these throttling tests is that they run CPU and GPU at the same time, and that messes stuff up, since probably both of them run hotter than their advertised TDP.

 

[The Mercurian]

Yeah, that is a problem. You can't just slap 50% more cores onto a die and pretend that the heat situation will stay the same. Anyway, I'll get my i9 at some point (couldn't order early, so still weeks to wait), and I will test it out. I think the problem with all these throttling tests is that they run CPU and GPU at the same time, and that messes stuff up, since probably both of them run hotter than their advertised TDP.

I would see that as a problem with the computer not the test. Surely running CPU + GPU at same time is something users might wish to do and so a valid test ?

 

[leman]

I would see that as a problem with the computer not the test. Surely running CPU + GPU at same time is something users might wish to do and so a valid test ?

You are perfectly right Still, the CPU utilisation in the first review video confuses me a lot, its way too low. That why I would like to test it myself.

 

[ugru]

I would see that as a problem with the computer not the test. Surely running CPU + GPU at same time is something users might wish to do and so a valid test ?

Moreover the laptop used in the video for comparison, the AERO 15X, has a GPU that is a lot more power hungry and that generates a lot more heat. That laptop throttled a lot less and completed the task at hand in 7 minutes....

 

[PeterJP]

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.

As I hardly follow the Windows laptop releases, thanks for the update. If it is (and stays) that bad with the MBP, then you're right.

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. (...) 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...

There are some things they can do with a firmware update. But if they need to update the firmware, why didn't they ship the new machines with the new firmware in the first place? As testing is part of any machine's life cycle before it ships, they must have seen the throttling?

Summarising: I stand corrected & if the situation is as bad as the first reviews suggest, Apple has yet another serious problem on its hands. How to go from a wow-upgrade to a woops-disaster in a few days. Certainly not the way to restore the pro market's confidence.

 

[Schranke]

Thanks to everyone here contributing to look into the thermal limitations of the 2018 models, this is what I have been most interested in since they announce it. I do not believe in those who claims it is software related, defiantly the hardware. Some of the Cinebench result, especially from the top line, is worry some.
Ran my 2017 2.9Ghz i7 just to compare and well here is what I got

As you see it holds the turbo boost much better then the i9. It should however be noted that I have replaced the thermal paste on the machine with Thermal Grizzly Kryonaut, so anyone up for trying that?

Looking forward to see how this goes as more people share their experience

 

[RobbieTT]

Sorry if this has been mentioned and I just missed it, but there was another oddity in the 13" German article.

It seems to suggest that the CPU was pulling way more power (50W+) than listed on the Intel specification for the i5-8259 TDP. If true this could suggest a possible source for the issue. As I understand it, Intel's TDP is set at maximum utilisation on all cores and at base frequency, so it should be a good reference.

If the CPU really is taking the equivalent of a 50W halogen bulb then that would truly challenge a thermal design based on dissipating the heat from a much lower 28W TDP CPU design.

 

[maflynn]

when I saw the i9 option because Dell XPS was also facing the same problems and to me, intel is just lying about the tdp values.

Dell had throttling issues on their 2017 Kaby lake models, and then used the same design for coffee lake, so their thermal throttling is no surprise. I can't say they're i9 throttling is worse or the same as the MBPs but from the reviews I've read, the machine works well for most tasks, but when you push it hard such as gaming, then it shows up. I have to believe rendering and content creation that requires a lot of CPU may fall into that category as well but I've not delved too deep in the Dell throttling issue

 

[ondert]

btw, English article has just been published on notebookcheck;

https://www.notebookcheck.net/Apple-MacBook-Pro-13-2018-Touch-Bar-i5-Laptop-Review.316648.0.html

there is no more need to read it in German, and yes, that 28w cpu pulls 48-50w... thanks intel...

 

[maflynn]

there is no more need to read it in German, and yes, that 28w cpu pulls 48-50w... thanks intel...

I wonder if you undervolt it, will that solve the problem, i.e., voltage x amperage = watts.

 

[ugru]

AFAIK 28W is TDP at base frequency.

When the CPU goes in turbo mode power draw is more significant. That is why keep using USB-C for powersupply, in a i9 laptop, is a dumb move. In tasks using both CPU and GPU at full speed could be another limiting factor, aside from heat dissipation...

I wonder if you undervolt it, will that solve the problem, i.e., voltage x amperage = watts.

Undervolting could help but it does not make miracles....you can gain 3%-5% performance, not more...

 

[donawalt]

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.

Along this line, I am wondering if it's something simple like the temperature sensors are sending incorrect info/incorrectly calibrated/bad batch.

 

[maflynn]

When the CPU goes in turbo mode power draw is more significant. That is why keep using USB-C for powersupply, in a i9 laptop, is a dumb move. In tasks using both CPU and GPU at full speed could be another limiting factor, aside from heat dissipation...

There are utilities, at least in windows to disable turbo mode, that might help (at the expense of performance but it should run cooler)

 

[Ma2k5]

btw, English article has just been published on notebookcheck;

https://www.notebookcheck.net/Apple-MacBook-Pro-13-2018-Touch-Bar-i5-Laptop-Review.316648.0.html

there is no more need to read it in German, and yes, that 28w cpu pulls 48-50w... thanks intel...

A read about TDP: https://www.intel.com/content/dam/doc/white-paper/resources-xeon-measuring-processor-power-paper.pdf

An excerpt is as follows:

If you are a thermal engineer, the processor TDP specification is very important because your thermal solution (fans, heat sink, etc.) must be able to be able to dissipate the rated TDP value. Intel and AMD both agree on this point. If a thermal engineer designs to ACP, it is likely the design will be undersize and not keep the processor within its thermal specifications.

So is the responsibility on the laptop manufacturers rather than Intel? I am not a CPU expert or anything, was just wondering if it's Intel's fault or not?

 

[leman]

So is the responsibility on the laptop manufacturers rather than Intel? I am not a CPU expert or anything, was just wondering if it's Intel's fault or not?

It boils down to differing expectations. Strictly spoken, the CPU is not necessarily required to keep up the turbo boost frequency indefinitely. For some folks, operating below the max turbo already constitutes throttling. Overall, its very confusing because no guarantees or even guidelines are really given. And the CPU can also run below the base clock if it is under-utilised, putting it into a "power saving" mode.

But yeah, I've looked at the notebookcheck article and its really disappointing. Its obvious why Apple waited for Coffee Lake — they didn't want to downgrade the GPU to 620, but its very surprising that Kaby Lake R would perform better.

 

[maflynn]

was just wondering if it's Intel's fault or not?

My knee jerk reaction is that its a computer company's fault. Big sellers like dell and apple get the cpus very early and have a long time to test them and determine how best to use those cpus in their computers. Its funny, but Dell and Apple are the two makers that seem to have throttling issues with the coffee lake chip as well. The Razer (which only sells the i7), reaches temps into the 80s and at times 90s but doesn't really throttle, I really like the heat chamber design being used, and I love the look of the razer, but I opted for a MBP so I'll have to wait and see how my i7 works. I don't push it hard, but my kids do with some of their stuff (I hear the fans blasting on my old MBP)

 

[Eason85]

As a long-time PC user, I am happy that the majority of Mac users have finally woken up to thermal throttling in MacBooks. Just because the fan isn't loud doesn't mean thermals are good.

 

[leman]

As a long-time PC user, I am happy that the majority of Mac users have finally woken up to thermal throttling in MacBooks. Just because the fan isn't loud doesn't mean thermals are good.

What do you mean by "finally woken up"? The 2016 and 2017 models don't have any thermal problems with the CPU. The 2018 uses new CPUs, so its a different thing.

 

[Ma2k5]

My knee jerk reaction is that its a computer company's fault. Big sellers like dell and apple get the cpus very early and have a long time to test them and determine how best to use those cpus in their computers. Its funny, but Dell and Apple are the two makers that seem to have throttling issues with the coffee lake chip as well. The Razer (which only sells the i7), reaches temps into the 80s and at times 90s but doesn't really throttle, I really like the heat chamber design being used, and I love the look of the razer, but I opted for a MBP so I'll have to wait and see how my i7 works. I don't push it hard, but my kids do with some of their stuff (I hear the fans blasting on my old MBP)

It boils down to differing expectations. Strictly spoken, the CPU is not necessarily required to keep up the turbo boost frequency indefinitely. For some folks, operating below the max turbo already constitutes throttling. Overall, its very confusing because no guarantees or even guidelines are really given. And the CPU can also run below the base clock if it is under-utilised, putting it into a "power saving" mode.

But yeah, I've looked at the notebookcheck article and its really disappointing. Its obvious why Apple waited for Coffee Lake — they didn't want to downgrade the GPU to 620, but its very surprising that Kaby Lake R would perform better.

I agree, they had the chance to test it and tailor the cooling solution to the chips as they saw fit. They would have been well aware of the limitations.

Technically speaking, you could indefinitely run these CPU's even at turbo-boosted frequencies if your cooling solution allowed for it. Of course you won't have a thin laptop anymore!

I think as others have said and as you have also touched on, it is dissapointing, especially when it is unable to hold base clock frequency (if it couldn't indefinitely hold the turbo that would have been more palatable).

Perhaps they all hoped that most users will not have prolonged loads, with most users only needing bursts of power - which would be handled fine. This is why the 12" MacBook and similar is actually very snappy for many, a lot of people's use just doesn't require prolonged CPU loads.

Do you think manufacturers should maybe look at more advanced cooling solutions, perhaps even some kind of liquid cooling solution?

 

[Eason85]

What do you mean by "finally woken up"? The 2016 and 2017 models don't have any thermal problems with the CPU. The 2018 uses new CPUs, so its a different thing.

I couldn't get my 2017 i7-7920HQ to stay at max boost for more than a minute of testing. I was seeing 99-100C temps from nearly any benchmark. The macbook fan control app wasn't able to help, so I returned it.

Also, for reference scores if you have a new hexa-core MBP: My XPS 15 9570 i7-8750H running CB 15 looped for 30 minutes averages 1220 pts +/- 20. This is after undervolting and repasting. Stock I was seeing scores between 1090-1130.

The wide and silly range of defenses from the misinformed defenders in this thread have been entertaining to read.

 

[maflynn]

Technically speaking, you could indefinitely run these CPU's even at turbo-boosted frequencies if your cooling solution allowed for it. Of course you won't have a thin laptop anymore!

I really wished we had the intel XTU app available to us in macOS. I used that on my Razer (before returning it), to undervolt and turn off turbo mode. Actually with the Razer, now that I think about it, I was able to turn off turbo mode in the bios. Again, it would be nice to have the level of control on my mac, but that's un-apple like.

 

[Eason85]

You are perfectly right Still, the CPU utilisation in the first review video confuses me a lot, its way too low. That why I would like to test it myself.

There are likely two (or even more) types of throttling that are going to affect the 2018 MBP:

The first is thermal throttling, which seems to just be the TJMax of the manufacturer to prevent damage to itself (100C). Apple's problem is that they like to have their machines run as silently as possible at the expense of heat build-up. Personally I will choose noise over a shortened lifespan of my machine (especially one as un-repairable as the post-2015 macbooks).

The second is PL throttling, which refers to the power limits that the chips are allowed to draw. 6-core coffee lake (i7-8750H) draws double the power that kaby lake did, and very few laptops can sustain 77W of power draw from the CPU alone. The MBP is limited to 100W, and then you have speakers, SSD, memory, screen, not to mention the OTHER processor, the GPU. PL throttling is going to be a major issue with even the i7, but the i9 will be hit by this the most.

The third type is one I'm not sure about yet except for the one guy who said his CPU dropped to 800Mhz, and this is VRM-induced throttling. Essentially, the voltage converters that feed the CPU, if not cooled properly, become too inefficient and are unable to supply the CPU with enough current, which often leads to very low PL1 limits (5-15W) being enforced upon the CPU, resulting in 800 Mhz clocks. This was the kind seen on the XPS 15 9560.

 

[leman]

Do you think manufacturers should maybe look at more advanced cooling solutions, perhaps even some kind of liquid cooling solution?

I think that CPUs that are advertised to output 28Watt under load should also output 28Watt under load

Lets look at it critically. Producing hotter and hotter chips and requiring more cooling is not the way to go. What we need is architectural improvements, true performance-to-watt increase. Intel was supposed to bring out new CPU architectures, what, two years ago? Instead we are getting the same old Skylake core, just pushed more and more to the limit and getting fancier and fancier name (Kaby Lake, Coffee Lake, Whiskey Lake, Amber Lake, all for the variations of essentially the same design). Since they can't really make new chips, what they do is sell factory overlocked chips, now with more of the same cores, using clear tricks like short bursts of very high boost frequency etc. etc.

It was clear from the start that Coffee Lake is a compromise, an attempt to get more performance out of old architecture. I would have hoped though that the performance-per-watt would have been improved. Instead, it seems like they simply overclocked the CPUs with little to no efficiency updates. If so, the result is of course predictable...

Anyway, I'll wait for my i9 model, test it our rigorously, and if I see significant throttling, it goes back until truly new CPUs are available.

 

[aevan]

My knee jerk reaction is that its a computer company's fault. Big sellers like dell and apple get the cpus very early and have a long time to test them and determine how best to use those cpus in their computers. Its funny, but Dell and Apple are the two makers that seem to have throttling issues with the coffee lake chip as well. The Razer (which only sells the i7), reaches temps into the 80s and at times 90s but doesn't really throttle, I really like the heat chamber design being used, and I love the look of the razer, but I opted for a MBP so I'll have to wait and see how my i7 works. I don't push it hard, but my kids do with some of their stuff (I hear the fans blasting on my old MBP)

Unless you plan on running benchmarks, compare some hypothetical scores and measuring %&$^ sizes with other computers, your MBP will be an extremely fast computer, capable of most demanding tasks.