thermal throttling More details about the throttling issues of the 15" core i9 MacBook Pro, this time with Final Cut Pro X

[D13H4RD]

9 minutes ago, JoostinOnline said:

You need more than "I saw an MSI image once" to make a claim that big.  There is no way any modern mobile CPU is taking up 150W of power. If that number is correct, then it probably had a 1070 in it or something.

The closest actual thing I found was that Reddit article I linked above. 

 

At 3.8GHz, the CPU alone draws between 105-115W of power under load according to HWInfo. I wished I had one to test at 4.8

[Brooksie359]

6 hours ago, Trixanity said:

The U series i5 and i7 are only separated by cache and clock speeds so there you go. This proposition has existed for years.

 

And I can only repeat myself again. The low power models suffer the same. So what's the common denominator? 

The whole CFL mobile lineup is plagued by power issues. That's why. It's false packaging. That's not desperation. It's reality. If Apple was the only one with issues I wouldn't even mention it but we see pretty much every vendor have issues to some degree. But people have tunnel vision so it gets turned into "hur dur Apple" which is usually universally true but there's more to it this time. Apple still have their shenanigans but when you have to avoid all but the bottom tier SKU in each lineup to avoid problems then there's an issue. I'd also call it an issue if vendors suddenly need to redesign all their products. Not an unfixable issue but I'm not convinced Intel's partners would be particularly happy having to spend even more money just to do their yearly refresh. There's a reason laptops (or any other product for that matter) don't get redesigned every year. 

I see my point flew right over your head because you still don't seem to get it. The reason why the i7 u series was considered better was because of the higher clock speed. Nobody would buy an i7 u series if it was the same clocks as the i5 but a bit more cache especially if the price difference was 300 dollars. If you gimp the i9 cpu to lower frequencies then there is no point of it being an i9 you might as well get an i7 at that point. Funny thing is the i7 version actually performs better anyways because it didn't thermal throttle itself so hard. The i9 is a high performance cpu that is supposed to have the proper cooling for it regardless of the rest if Intel's lineup. If this was a u series part then yeah I would give Intel shit for it but it's the flagship high performance mobile cpu that really only makes sense in heavy thermal solutions like a gaming laptop. You put that chip in a small laptop with bad cooling then it's obvious what is going to happen. 

[D13H4RD]

2 minutes ago, Brooksie359 said:

I but it's the flagship high performance mobile cpu that really only makes sense in heavy thermal solutions like a gaming laptop. 

Not just any gaming laptop. 

 

A behemoth with the cooling system to handle that sort of CPU. If it can draw over so much at 3.8GHz, there's no way a system designed for a 45W CPU can even remotely handle it unless you disable TB or make it so it only does so in bursts 

[avg123]

18 hours ago, ImpulseRez said:

A bit ridiculous to blame Intel on this one. Apple must have made a conscious design choice to keep the laptop thin at the expense of cooling. Intel chips are pretty good from an efficiency perspective (better than AMD at least), and being the first 6 core laptop chips are always going to be thermally dense to start with.

Also, bringing up the ARM mention is pretty absurd (performance compared to an 8th Gen 6 core Intel CPU is not going to be in the same league!)

The current macbook laptop chassis and cooling is not designed for 6 core processors. It was designed around a 4 core i7 processor. A macbook refresh  specifically designed for 6 core processors should fix the thermal throttling issues.

[PlayStation 2]

I've been seeing that a LOT of other laptops are having thermal issues with the i9. Would I put some blame on Apple? Yes. On the other hand, I've also gotta blame Intel and the consumer for complaining about wanting new CPUs then acting shocked that they suck ass at thermal control. Many laptops that aren't straight desktop replacements with the i9 seem to have very poor temperatures.

[Trixanity]

1 hour ago, Brooksie359 said:

I see my point flew right over your head because you still don't seem to get it. The reason why the i7 u series was considered better was because of the higher clock speed. Nobody would buy an i7 u series if it was the same clocks as the i5 but a bit more cache especially if the price difference was 300 dollars. If you gimp the i9 cpu to lower frequencies then there is no point of it being an i9 you might as well get an i7 at that point. Funny thing is the i7 version actually performs better anyways because it didn't thermal throttle itself so hard. The i9 is a high performance cpu that is supposed to have the proper cooling for it regardless of the rest if Intel's lineup. If this was a u series part then yeah I would give Intel shit for it but it's the flagship high performance mobile cpu that really only makes sense in heavy thermal solutions like a gaming laptop. You put that chip in a small laptop with bad cooling then it's obvious what is going to happen. 

My proposal included more cache and clock speed than the i7 so your point was useless from the get go. You have literally no point because you don't know Intel's lineup and have shown that ignorance in multiple posts now. If you can't at least look into the products we discuss, then move along please.

 

Look at Apple's other launch: MBP with U series. Still throttles to hell too. The i9 throttles in multiple other products (not all), as does i7 but to a lesser degree but it's convenient to ignore, isn't it?

Oh and it's also throttling in gaming laptops by the way. Just to be clear.

In both an Asus and Alienware laptop I believe.

[Trixanity]

41 minutes ago, Dan Castellaneta said:

I've been seeing that a LOT of other laptops are having thermal issues with the i9. Would I put some blame on Apple? Yes. On the other hand, I've also gotta blame Intel and the consumer for complaining about wanting new CPUs then acting shocked that they suck ass at thermal control. Many laptops that aren't straight desktop replacements with the i9 seem to have very poor temperatures.

Finally some sanity in this thread.

[D13H4RD]

11 minutes ago, Trixanity said:

Oh and it's also throttling in gaming laptops by the way. Just to be clear.

In both an Asus and Alienware laptop I believe.

A Eurocom G8 and an Alienware 15 R4. The MSI GT75 and ASUS G703 are fine albeit very toasty.

 

1 hour ago, avg123 said:

The current macbook laptop chassis and cooling is not designed for 6 core processors. It was designed around a 4 core i7 processor. A macbook refresh  specifically designed for 6 core processors should fix the thermal throttling issues.

It was designed for a 45W TDP CPU and the 75W TDP Radeon GPU. It’s just adequate for that.

 

Problem is that for Skylake and Kaby, even when turbo, it’s within TDP. But for CFL, because of the aggressive stock and especially turbo clocks, the TDP is exactly as Intel describes, the amount of heat that has to be dissipated based on how much power it consumes at stock speeds. When you go into turbo, you’ll most definitely go way over the TDP. Someone with an MSI laptop tested it and at 3.8GHz, the CPU alone consumes 105W of power.

 

I thought it was a combined CPU + GPU reading at first, but when I saw “GTX 1080”, no way can a laptop with a 1080 consume just over a 100W since similar configs consume 250W at least.

[handymanshandle]

2 hours ago, Dan Castellaneta said:

I've been seeing that a LOT of other laptops are having thermal issues with the i9. Would I put some blame on Apple? Yes. On the other hand, I've also gotta blame Intel and the consumer for complaining about wanting new CPUs then acting shocked that they suck ass at thermal control. Many laptops that aren't straight desktop replacements with the i9 seem to have very poor temperatures.

That and I think, how OEMs integrate the i9 don't seem to put much thought into the fact that you're putting a 6 core monster in a laptop designed for a petite 4 core 4 thread and 2 core 4 thread CPU.

 

 

Apple should've been either willing to increase thickness just a little bit and/or add more fans and/or ramp the fans up earlier. But other designs that are struggling with the i9, need to do the same thing as well. To a lesser degree but nonetheless..

 

 

It's kinda pointless to soley blame Apple for the i9 choking in their laptops, it is really a sort of "Well what the hell did you expect"?

[mrthuvi]

3 hours ago, D13H4RD2L1V3 said:

The closest actual thing I found was that Reddit article I linked above. 

 

At 3.8GHz, the CPU alone draws between 105-115W of power under load according to HWInfo. I wished I had one to test at 4.8

At what kind of work load? Avx vs non avx are very different.

[D13H4RD]

20 minutes ago, pinksnowbirdie said:

That and I think, how OEMs integrate the i9 don't seem to put much thought into the fact that you're putting a 6 core monster in a laptop designed for a petite 4 core 4 thread and 2 core 4 thread CPU.

It's a lot to do with the TDP. 

 

The claimed TDP on the I9 8950HK is 45W, the same as the i7 8850H and the Kaby/Skylake-based predecessors. However, there is a catch. 

 

While Intel all this while claims that the TDP is measured at the CPU's base frequencies, the Skylake and Kaby Lake processors were able to maintain TDP while in their boosted state. However, with the Coffee Lake CPUs, likely due to the added cores and aggressive base/boost clocks, that has changed and actually matches Intel's explanation of their TDP numbers. 

 

At boost during extremely demanding loads like benchmarking, a Core i7 7700HQ consumes around 42-45W of power, below its TDP and well within the expected range. However, the i9 8950HK consumes around the same at its base frequency. At 3.8GHz, the CPU can consume well over 100W, which is extremely high for any mobile CPU. 

 

The problem with the MacBook Pro and other laptops with the issue is that their cooling systems were likely designed with a 45W TDP CPU in mind. As such, because a boosted 8950HK can greatly exceed that figure, the system is easily overwhelmed and thus, throttling occurs. 

 

Only thing I can say is that in the case of the MacBook Pro, the best it can hope for with this cooling solution is a max frequency of 2.9GHz constantly with very brief turbo. 

[D13H4RD]

9 minutes ago, mrthuvi said:

At what kind of work load? Avx vs non avx are very different.

Prime95 + FurMark

 

A 7700HQ in the same test consumed 42-45W. A Ryzen 7 1700 in the ASUS GL702ZC consumed around 62W

[mr moose]

9 minutes ago, mrthuvi said:

At what kind of work load? Avx vs non avx are very different.

So are normal and power virus loads according to Intel.  TDP is only rated for workloads that are encountered in actual software. Intel do not use programs like prime95 when calculating the TDP rating.

 

[handymanshandle]

4 minutes ago, D13H4RD2L1V3 said:

It's a lot to do with the TDP. 

 

The claimed TDP on the I9 8950HK is 45W, the same as the i7 8850H and the Kaby/Skylake-based predecessors. However, there is a catch. 

 

While Intel all this while claims that the TDP is measured at the CPU's base frequencies, the Skylake and Kaby Lake processors were able to maintain TDP while in their boosted state. However, with the Coffee Lake CPUs, likely due to the added cores and aggressive base/boost clocks, that has changed and actually matches Intel's explanation of their TDP numbers. 

 

At boost, a Core i7 7700HQ consumes around 42-45W of power, below its TDP and well within the expected range. However, the i9 8950HK consumes around the same at its base frequency. At 3.8GHz, the CPU can consume well over 100W, which is extremely high for any mobile CPU. 

 

The problem with the MacBook Pro and other laptops with the issue is that their cooling systems were likely designed with a 45W TDP CPU in mind. As such, because a boosted 8950HK can greatly exceed that figure, the system is easily overwhelmed and thus, throttling occurs. 

 

Only thing I can say is that in the case of the MacBook Pro, the best it can hope for with this cooling solution is a max frequency of 2.9GHz constantly with very brief turbo. 

meh just live in Minnesota or something, then most of the year it's like having an outdoor freezer.

Or turn your house/office into a large scale walk in freezer.

[D13H4RD]

1 minute ago, mr moose said:

Intel do not use programs like prime95 when calculating the TDP rating.

 

And that's understandable because Prime95 is a very unrealistic workload. 

 

But it does push power consumption to its limits. Funnily enough, I managed to find a Lenovo Y920 with the i7 7820HK and that maintained under its 45W TDP, although it ran super hot 

[mr moose]

2 minutes ago, D13H4RD2L1V3 said:

And that's understandable because Prime95 is a very unrealistic workload. 

 

But it does push power consumption to its limits. Funnily enough, I managed to find a Lenovo Y920 with the i7 7820HK and that maintained under its 45W TDP, although it ran super hot 

And that is the problem with all these threads,  people do not understand TDP and why it exists.  Intel know that if you run prime95 the CPU will draw more power than the TDP rating.   It is not the point of TDP to push a CPU to it's limit but to set a datum for engineers who have to design a cooling solutuon.  And to be honest I am sick of people claiming Intel TDP ratings are rubbish because if anyone replicates the conditions with the Intel spec for TDP they would see the CPU draw around the same as TDP maybe a few spikes over.

 

 

Long story short, you pull one of these CPU's out of a mac book and put it under a 45W cooler it would maintain base clock running any non artificial load.  Because if it doesn't you have a faulty chip or have successfully caught Intel lying and may start a class action that they will lose.

[D13H4RD]

4 minutes ago, mr moose said:

Long story short, you pull one of these CPU's out of a mac book and put it under a 45W cooler it would maintain base clock running any non artificial load.  Because if it doesn't you have a faulty chip or have successfully caught Intel lying and may start a class action that they will lose.

I guess if you disable turbo boost and run demanding applications that aren’t benchmarks or stress test tools, the processor would largely maintain its 2.9GHz base.

 

Itll run hot but it wouldn’t seriously throttle

[mr moose]

1 minute ago, D13H4RD2L1V3 said:

I guess if you disable turbo boost and run demanding applications that aren’t benchmarks or stress test tools, the processor would largely maintain its 2.9GHz base.

 

Itll run hot but it wouldn’t seriously throttle

That's the whole idea.   But even with turbo on, with a realistic but hard workload and a 45Watt cooler it should not not throttle below the base freq. 

 

It think the other problem we have on the internet when it comes to discussion like this,  is that people (for whatever reason) seem to think that TDP is a hard number.  It is a definite number that has a measurable hard point for sure, but it is not a limit, like a  minimum or maximum.   It is one defined point in a scale.

[D13H4RD]

One thing I find interesting 

 

On my 7700HQ, going Prime95 (a very extreme and unrealistic workload) would have it brush right against the 45W TDP during a 3.4GHz turbo but sometimes dropping to 3.2 GHz. ThrottleStop shows that my CPU isn’t overheating but is instead TDP throttling, possibly hitting a power limit.

 

4 minutes ago, mr moose said:

That's the whole idea.   But even with turbo on, with a realistic but hard workload and a 45Watt cooler it should not not throttle below the base freq. 

 

It think the other problem we have on the internet when it comes to discussion like this,  is that people (for whatever reason) seem to think that TDP is a hard number.  It is a definite number that has a measurable hard point for sure, but it is not a limit, like a  minimum or maximum.   It is one defined point in a scale.

That’s what’s in my mind. If the 45W is calculated at the base frequency, brief periods of turbo should be possible. But being unable to maintain a base frequency (at which the 45W TDP is calculated based on heavy-but-realistic workloads) is a head scratcher.

[avg123]

8 minutes ago, mr moose said:

That's the whole idea.   But even with turbo on, with a realistic but hard workload and a 45Watt cooler it should not not throttle below the base freq. 

 

It think the other problem we have on the internet when it comes to discussion like this,  is that people (for whatever reason) seem to think that TDP is a hard number.  It is a definite number that has a measurable hard point for sure, but it is not a limit, like a  minimum or maximum.   It is one defined point in a scale.

Whats the point of making a laptop cpu that generates so much heat only a laptop weighing 10 lbs can run at max turbo speed?

 

 The market size for such a laptop is so small, intel could not have made a new cpu just for that market. 

 

I think intel developed the cpu knowing most laptops would not be able to run it at its full potential.

[Mr.Dingle]

18 minutes ago, mr moose said:

And that is the problem with all these threads,  people do not understand TDP and why it exists.  Intel know that if you run prime95 the CPU will draw more power than the TDP rating.   It is not the point of TDP to push a CPU to it's limit but to set a datum for engineers who have to design a cooling solutuon.  And to be honest I am sick of people claiming Intel TDP ratings are rubbish because if anyone replicates the conditions with the Intel spec for TDP they would see the CPU draw around the same as TDP maybe a few spikes over.

 

 

Long story short, you pull one of these CPU's out of a mac book and put it under a 45W cooler it would maintain base clock running any non artificial load.  Because if it doesn't you have a faulty chip or have successfully caught Intel lying and may start a class action that they will lose.

Just a question ... shouldnt power draw be quite a bit higher than the rated TDP of the chip ? Depending on workload of course.

[mr moose]

8 minutes ago, D13H4RD2L1V3 said:

One thing I find interesting 

 

On my 7700HQ, going Prime95 (a very extreme and unrealistic workload) would have it brush right against the 45W TDP during a 3.4GHz turbo but sometimes dropping to 3.2 GHz. ThrottleStop shows that my CPU isn’t overheating but is instead TDP throttling, possibly hitting a power limit.

 

That’s what’s in my mind. If the 45W is calculated at the base frequency, brief periods of turbo should be possible. But being unable to maintain a base frequency (at which the 45W TDP is calculated based on heavy-but-realistic workloads) is a head scratcher.

Unless apples cooler has been running close to the wire already and throwing 6 cores at it was the final straw.

4 minutes ago, avg123 said:

Whats the point of making a laptop cpu that generates so much heat only a laptop weighing 10 lbs can run at max turbo speed?

It is what it is, if OEMs don't want to make large laptops then they can still use the 4core CPUs.

4 minutes ago, avg123 said:

 The market size for such a laptop is so small, intel could not have made a new cpu just for that market. 

 

I think intel developed the cpu knowing most laptops would not be able to run it at its full potential.

Maybe, but then maybe there are good cooling solutions out there as other laptops don't do as bad.

3 minutes ago, Mr.Dingle said:

Just a question ... shouldnt power draw be quite a bit higher than the rated TDP of the chip ? Depending on workload of course.

Sort of, but not really,  We had a really long thread a while back on this and think the issue was that I am not very good at communicating some concepts. 

 

TDP is a set figure for a set condition. Once we know what the set condition is we can calculate the required cooler size.  What this means is that for a set frequency and workload the TDP might be 45W, but if you increase the frequency, or the voltage or the workload you alas increase the heat being dissipated.  So yes you can exceed the TDP rating in power draw and often this happens when you use an artificial workload like prime95, however the power draw and the heat dissipation typically scales together (meaning if you use a bigger cooler it doesn't matter). 

 

 

[Mr.Dingle]

7 minutes ago, mr moose said:

Unless apples cooler has been running close to the wire already and throwing 6 cores at it was the final straw.

It is what it is, if OEMs don't want to make large laptops then they can still use the 4core CPUs.

Maybe, but then maybe there are good cooling solutions out there as other laptops don't do as bad.

Sort of, but not really,  We had a really long thread a while back on this and think the issue was that I am not very good at communicating some concepts. 

 

TDP is a set figure for a set condition. Once we know what the set condition is we can calculate the required cooler size.  What this means is that for a set frequency and workload the TDP might be 45W, but if you increase the frequency, or the voltage or the workload you alas increase the heat being dissipated.  So yes you can exceed the TDP rating in power draw and often this happens when you use an artificial workload like prime95, however the power draw and the heat dissipation typically scales together (meaning if you use a bigger cooler it doesn't matter). 

 

 

Im going by the TDP < power draw since you cannot create energy out of nowhere , and if TDP was equal to the power draw , that would mean that the CPU didnt perform any useful work ? (If this applies to CPUs and the like)

[mr moose]

2 minutes ago, Mr.Dingle said:

Im going by the TDP < power draw since you cannot create energy out of nowhere , and if TDP was equal to the power draw , that would mean that the CPU didnt perform any useful work ? (If this applies to CPUs and the like)

I think you are thinking of TDP the wrong way.    The amount of heat a chip dissipates per watt of power consumption is different to the TDP.  As far as the laws of physics go, power in will be the equivalent of power out as heat + power out as sourcing + plus power remaining in the chip as heat.  So yes, in that regard the power in will always be higher than the heat power dissipated, however the heat power being dissipated is not the TDP (that is the rating for a set condition) but the result of the current condition the chip is operating under.

[asus killer]

 

 

43 minutes ago, Mr.Dingle said:

Im going by the TDP < power draw since you cannot create energy out of nowhere , and if TDP was equal to the power draw , that would mean that the CPU didnt perform any useful work ? (If this applies to CPUs and the like)

 

On 9/10/2017 at 12:29 AM, tomaatvk said:

TDP is used to indicate the amount of heat that the component in question could produce in the worst condition. Different companies do use TDP in different ways. For example:

intel's TDP is the amount of heat the worst binned chip could produce when put through the worst kind of torture (at stock settings ofcourse), while most intel chips could not even come close to the TDP they are rated at.

Nvidia GPU chips are a whole different story, they use TDP in such a way that the indicated TDP is the maximum amount of power it is allowed to pull. If the chip wants to pull more, it will get throttled to the point where it will maintain its TDP value.

I have no idea how it works with AMD CPU's, but I guess it works the same as with intel CPU's.

AMD GPU's are the same as Nvidia GPU's, but they key difference is that Nvidia's TDP is for the whole card while AMD only indicates the amount of power the silicon uses.

 

TDP is always just the amount of power, measured it watts (which is the same as joules per second), the part could consume. When we talk about computer parts, basicly all the energy goes into either movement of air/fans/HDD platters, light, sound and heat output. As you may guess, a CPU does not move a fan or produce light, etc. so pretty much all the power a CPU will consume will get transfered into heat.

 

To conclude, for your example of the 3930k. If it would consume and produce 130 watts of heat if it would work at 100% and be one of the worst chips sold. This is at the condition that it is not overclocked.