Reputable test: 2.6Ghz i7-8850H MBP able to maintain max. turbo for a whopping 2 seconds

[Queen6]

The newer CPUs no longer use SpeedStep, they have switched to something called SpeedShift, which is supposed to be a lot more agile and responsive than the old method, which is probably why the graphs of the throttling look so freakishly jagged. Essentially, if I recall correctly, it offloads the decision-making to the CPU itself, which cuts out a lot of latencies.

Just looked it up:

SpeedShift is extremely rapid in action, nor are any graphing applications likely to be capable of monitoring in real-time, more a trend if you will.

That said if the notebook is well designed you should observe similar results as per below.

The charts we are seeing for the 2018 MBP are very much indicative of thermal and or power throttling under a continuous load with the CPU forced to reduce frequency.

OEM's also have the option to set the PROCHOT flag to lower the point where thermal throttling occurs. Another factor is BDPROCHOT where the temperature of another component will result in the CPU rolling back frequency as cooling mechanism.

Q-6

 

[Mockletoy]

What then, in your opinion, you think you are paying for? Do you expect that CPU to maintain max boost indefinitely? It’s not really how it works.

I kind of miss the old good times, when you knew that a cpu sold as X ghz would run X ghz and that’s it. Now you buy CPUs that can run somewhere between a and b, or maybe below that - all as part of their normal operation.

Those days are not gone. If the CPU's base clock is 2.9Ghz, then that is the minimum speed at which it should ever operate when it's under load.

The i9 has a 2.9Ghz base clock with a 4.8Ghz single-core boost.

No one expects it to run anything at 4.8Ghz except in very short bursts. But it should be able to run all cores under load at its 2.9Ghz base clock 24 hours a day.

If at any point it is not able to sustain its base clock, it's not functioning properly.

Again, I don't understand why people keep making this argument. My 2016 13" MacBook Pro has a 2.9Ghz CPU that will run under full CPU load at 3.1Ghz turbo for as long as I ask it to. It never wavers. It just sits there chugging along.

My 7700HQ machine has a 2.8Ghz base clock. It will sit there all day and all night with all 4 cores boosted to 3.4Ghz under max load.

That's how it's supposed to work.

 

[The Mercurian]

I still don't NEED 32 - but I got it anyways, just in case. You can't use RAM that isn't there, if you need it. But in the end, it is about need, isn't it? Do you NEED to go 4.8 GHZ turbo? Probably not, most don't. Should it work as Apple advertises? Yes. But if it doesn't, don't buy it and quit whining. And there is a thin line between noting an issue and being concerned about it - and what we have here about the processors: obsession. But I guess this is just a normal day in the tech world.

Ok you don't NEED it. But now that you have paid a hefty upgrade price for it - how would you feel if I told you on when the machine overheated you could only actually make use of 16Gb of it and the other 16Gb would lie dormant ?

 

[Rafterman]

Ok you don't NEED it. But now that you have paid a hefty upgrade price for it - how would you feel if I told you on when the machine overheated you could only actually make use of 16Gb of it and the other 16Gb would lie dormant ?

Actually, the vast majority of the cost of my own BTO is that 4TB SSD drive, so that paltry difference barely registers. But processor speed is something you only notice if the apps you run are fast or slow. One dude ran Premier editing on a 6 core machine and it was consistently faster than the 4 core machine, So throttling to a slower clock speed seems to have made little difference, as the 6 core machine still performed better.

 

[Ma2k5]

Actually, the vast majority of the cost of my own BTO is that 4TB SSD drive, so that paltry difference barely registers. But processor speed is something you only notice if the apps you run are fast or slow. One dude ran Premier editing on a 6 core machine and it was consistently faster than the 4 core machine, So throttling to a slower clock speed seems to have made little difference, as the 6 core machine still performed better.

It’s down to value - you don’t get value paying full price for a Lamborghini with a top speed stuck at 105mph - it doesn’t need matter if you drive that fast or not, you paid for the fact it suppose to drive up to 180mph. It’s the same with the CPU’s, you didn’t get a discount for it not performing to its capabilities - you paid full price.

 

[Mockletoy]

Ok you don't NEED it. But now that you have paid a hefty upgrade price for it - how would you feel if I told you on when the machine overheated you could only actually make use of 16Gb of it and the other 16Gb would lie dormant ?

Exactly. People aren't getting what they paid for and they're making excuses for why that's just fine because it happens to be something they don't personally think they "need."

I think part of my problem is that back in the day, when I was a kid, I used to love building and tweaking my own computers. Didn't have much money, so I always agonized and chose every component so carefully to get the maximum bang for my buck. And I also spent hours and hours tweaking and fiddling to get everything just right.

A machine that is constantly overheating and throttling, even if it more or less works, it would just drive me bonkers, knowing it was going on and there was really nothing I could do about it. Apparently, that doesn't bother some folks, which is great for them, I suppose, but I don't get why they can't understand why some of us would find it bothersome.

All that aside, though, if Apple is charging people $300 for an "upgraded" CPU that is worse than useless for real work, that's just not okay as a matter of principle.

Actually, the vast majority of the cost of my own BTO is that 4TB SSD drive, so that paltry difference barely registers. But processor speed is something you only notice if the apps you run are fast or slow. One dude ran Premier editing on a 6 core machine and it was consistently faster than the 4 core machine, So throttling to a slower clock speed seems to have made little difference, as the 6 core machine still performed better.

That's really not the point. At all. It's not about whether the 6-core CPU is faster than some other CPU that came out years ago, it's about whether that 6-core CPU is functioning properly.

Here is basically what you're saying: you used to drive a car that got 18 miles per gallon, right?

Now you've bought a new car that gets 30 miles per gallon. It runs super hot all the time and it leaks gasoline like crazy, but at the end of the day, despite the heat and leaking gas, you still get slightly better than the 18 miles per gallon your old car got, so, shoot, it's all good.

 

[The Mercurian]

Actually, the vast majority of the cost of my own BTO is that 4TB SSD drive, so that paltry difference barely registers. But processor speed is something you only notice if the apps you run are fast or slow. One dude ran Premier editing on a 6 core machine and it was consistently faster than the 4 core machine, So throttling to a slower clock speed seems to have made little difference, as the 6 core machine still performed better.

Way to avoid answering the question.

 

[Rafterman]

It’s down to value - you don’t get value paying full price for a Lamborghini with a top speed stuck at 105mph - it doesn’t need matter if you drive that fast or not, you paid for the fact it suppose to drive up to 180mph. It’s the same with the CPU’s, you didn’t get a discount for it not performing to its capabilities - you paid full price.

So its spec bragging?

 

[Ma2k5]

We all have different views - what I will say is, speaking from experience, companies give notice to criticism. You can praise them night and day and you won’t get anything back. Criticism however, can lead the company to change things to benefit the consumer.

So if you are happy, great, but if people are not happy with something and want to make noise - let them. It will eventually spread and could make that product you love even better. You won’t lose out .

 

[Rafterman]

Way to avoid answering the question.

I did answer the question. The 6 core systems perform better in most cases and chasing clock speeds is just spec bragging.

 

[Ma2k5]

So its spec bragging?

Value proposition. You wouldn’t buy a 350ml can of Coke which only had 250ml of Coke for full price would you, even though the 250ml quenched your thirst?

 

[Mockletoy]

So its spec bragging?

No. It's paying extra for a 2.9Ghz CPU to batch process photos and render videos so you can earn some money to feed your family and finding out that it runs boiling hot at 2.2Ghz - 2.6Ghz instead and you wasted your money.

 

[EugW]

- They run Cinebench under Bootcamp. Is there any reason for it?

Do they do this with all their Mac Cinebench runs? If so, that may explain why their Mac Cinebench results often seem odd.

Here are Notebookcheck's 2017 MacBook Core m3-7Y32 Cinebench results:

https://www.notebookcheck.net/Apple-MacBook-12-2017-Laptop-Review.230656.0.html

Note the hard throttling for run 8, and that by run 10 the score is only 226.

Here are our MacRumors.com forum member results, all done in macOS Sierra:

https://forums.macrumors.com/thread...tained-cpu-load-10-runs-of-cinebench.2073415/

First of all, NONE of us got that hard throttling whether it was an m3, i5, or i7.
Second, after 10 runs the speed was still 246 for the m3. (That's on a wood table. On granite the score was 253 at 10 runs.)

 

[laz232]

A CPU that is bouncing its clock speed all over the place because it's continuously running into its thermal safety limits and having to down clock is just not for me,

My professional opinion is that this might indicate badly applied thermal paste - As the temperature rise Is so rapid

 

[Mockletoy]

My professional opinion is that this might indicate badly applied thermal paste - As the temperature rise Is so rapid

I saw someone on one of these many threads -- I think it was here, but I'm not certain -- who'd repasted their machine and gotten somewhat better results. Somewhat better being slightly higher clocks for slightly longer before hitting 100C and throttling.

 

[EugW]

I saw someone on one of these many threads -- I think it was here, but I'm not certain -- who'd repasted their machine and gotten somewhat better results. Somewhat better being slightly higher clocks for slightly longer before hitting 100C and throttling.

They redid the thermal paste on a brand new machine?

 

[BorderingOn]

Well, there are two issues. Raw speed and cores. Apps that use multiple cores will run faster at a lower clock speed than fewer cores at a faster clock speed. So, for many apps, 6 slower clocked cores will be faster that 4 cores at a higher speed. So it is not really a "ripoff", though the clock speeds Apple advertises should be accurately portrayed. Which is why I asked, what about ACTUAL apps, not benchmark testers. From the tests a few have done on the internet, it seems apps do tend to be faster with the 6 cores, regardless of clock speeds

Did you notice the i7s also have six cores?

 

[Mockletoy]

They redid the thermal paste on a brand new machine?

Yes, it was someone with the 2.6Ghz i7. I just tried to find it but I've read so many threads on so many sites, they all blend together. Anyway, they said it was somewhat better, which is to be expected. Stock thermal paste is usually machine-applied goopy garbage. But they were still throttling.

I really think it was someone here. Sorry I can't link to it. If I come across it again I'll let you know.

 

[EugW]

Yes, it was someone with the 2.6Ghz i7. I just tried to find it but I've read so many threads on so many sites, they all blend together. Anyway, they said it was somewhat better, which is to be expected. Stock thermal paste is usually machine-applied goopy garbage. But they were still throttling.

I really think it was someone here. Sorry I can't link to it. If I come across it again I'll let you know.

Man. That's nuts! I would have just returned it and gotten the 2.2 GHz i7-8750H instead, since it's already blistering fast.

The last two machines where I recently redid the thermal paste were a 2008 Core 2 Duo 2.4 GHz white MacBook running Linux that would ramp up the fan with just mere regular surfing, and a 2010 Windows 10 machine where I upgraded the 2.9 GHz Athlon II X2 435 tri-core CPU to a 2.8 GHz Phenom II X6 1055T hex-core CPU. For the MacBook I really needed to do it, and it turned out it had a really crappy and partially dried up thermal pad. Putting new Arctic Silver there helped greatly. Now I can surf quietly on that machine. For the Athlon I was swapping the CPU so it had to do it too. It was quiet before and it's still quiet with the same cheap stock Acer cooler but now has over twice the performance.

I would never dream of doing it on a brand new several thousand dollar machine under warranty.

 

[laz232]

I saw someone on one of these many threads -- I think it was here, but I'm not certain -- who'd repasted their machine and gotten somewhat better results. Somewhat better being slightly higher clocks for slightly longer before hitting 100C and throttling.

hmm, The rapid cycling between turbo and severely under clocked looks very strange though, I have not seen it on any other machine. I've tested both of my older MacBooks. basically reminds me of an unstable control system, where the thermal resistance of the heat Sink and fan assembly is high as it is the effect of thermal capacitance, whereas the CPU itself has a very quick response To changes in heat dissipation due to clock frequency.

The average cpu temperature is therefore the correct maximum of say 95° but in order to achieve this rather than a sustained say 2.5 GHz operation it instead Oscillates wildly whilst the cooling system cannot react as quickly.

To me this seems like a classic case of control loop instability, does anyone know if the on CPU turbo boost settings are configurable by micro code / Manufacturer?
[doublepost=1532149830][/doublepost]

Stock thermal paste is usually machine-applied goopy garbage. But they were still throttling.

Personally I doubt that the quality of the thermal paste is that important or that much of a price driver. I think it's more likely that the poor worker is not applying it optimally, because you're better off using a bit too much then a bit too little. And we have seen from the replacing videos that doing a very good job is extremely time-consuming. factory workers do not have the time or interest when they're working long shifts in some Chinese sweatshop

 

[geromi912]

hmm, The rapid cycling between turbo and severely under clocked looks very strange though, I have not seen it on any other machine. I've tested both of my older MacBooks. basically reminds me of an unstable control system, where the thermal resistance of the heat Sink and fan assembly is high as it is the effect of thermal capacitance, whereas the CPU itself has a very quick response To changes in heat dissipation due to clock frequency.

The average cpu temperature is therefore the correct maximum of say 95° but in order to achieve this rather than a sustained say 2.5 GHz operation it instead Oscillates wildly whilst the cooling system cannot react as quickly.

To me this seems like a classic case of control loop instability, does anyone know if the on CPU turbo boost settings are configurable by micro code / Manufacturer?
[doublepost=1532149830][/doublepost]

Personally I doubt that the quality of the thermal paste is that important or that much of a price driver. I think it's more likely that the poor worker is not applying it optimally, because you're better off using a bit too much then a bit too little. And we have seen from the replacing videos that doing a very good job is extremely time-consuming. factory workers do not have the time or interest when they're working long shifts in some Chinese sweatshop

It happens with my XPS 9560 too. Out of factory it can’t maintain the GTX 1050’s boost clock of 1645Mhz, and instead of keeping a stable 1.35Ghz, it would tumble down to 900Mhz and then climb back up to 1.65Ghz.....and throttle again.

 

[leman]

No one expects it to run anything at 4.8Ghz except in very short bursts.

The person who started this thread apparently does. I mean, look at the title. I wonder how many laptops — or even desktops out there — will be able to maintain max single-core turbo on full 6 core load for more then 2 seconds?

But it should be able to run all cores under load at its 2.9Ghz base clock 24 hours a day.

Yeah, wait, but isn't it what is happening? Lets take the notebookcheck test for the 8850H. They write that in their sustained cinebencz run, "The test runs for 42 minutes and we see an average clock of about 2.7 GHz for all 6 cores, which is slightly above the base frequency". From performance standpoint, its the same as running constant 2.7Ghz.

True, the clock fluctuates wildly, which is weird. I think notebookcheck guys are correct in saying that its a drawback of Apple not restricting the CPU TDP. It tries to go as fast as possible, gets quite hot, and then needs to briefly cool down before winding up again. They could probably reach more sustained clocks by introducing a limiter — at a cost of reduced max performance though.

I would like to see more detailed analysis of these fluctuations. If they changes something in Coffee Lake and it changes its frequency more often, the observed fluctuations could be also artifacts of sampling. E.g. if it goes between 1.8 and 3.5 Ghz 50 times per second, the effective speed— for all intends and purposes — be 2.65Ghz. That is, in every small time interval, the speed will be averaging to around 2.6. If you sample only 2 points per seconds though, you will of course see wild fluctuations....

I'd like to see how it looks on macOS as well. Notebookcheck use Windows for some reason.

BTW, Dell's XS 15" quad core Coffee Lake partially shows similar fluctuating pattern: https://www.notebookcheck.net/Dell-XPS-15-2018-9570-8300H-GTX-1050-97Wh-Laptop-Review.308420.0.html

I couldn't find frequency graphs of any other Coffee Lake machine they tested though...

If the CPU's base clock is 2.9Ghz, then that is the minimum speed at which it should ever operate when it's under load.

I don't think that its how base clock is defined by Intel. They say that base clock is the clock that the CPU runs on all cores when generating its TDP of thermal power.

Again, I don't understand why people keep making this argument. My 2016 13" MacBook Pro has a 2.9Ghz CPU that will run under full CPU load at 3.1Ghz turbo for as long as I ask it to. It never wavers. It just sits there chugging along.
My 7700HQ machine has a 2.8Ghz base clock. It will sit there all day and all night with all 4 cores boosted to 3.4Ghz under max load.

Are you sure about that? Do you know how the frequencies are sampled? How do you know its not bouncing between 2 and 4Ghz 50 times per second?

And please don't misunderstand me, I am not trying to be polemic, I just want to challenge people to look at the things in a more complex way. It is obvious that there is something weird going on with these machines. But I don't think that its as trivial as some make it sound.
[doublepost=1532164860][/doublepost]

That said if the notebook is well designed you should observe similar results as per below.
View attachment 771745The charts we are seeing for the 2018 MBP are very much indicative of thermal and or power throttling under a continuous load with the CPU forced to reduce frequency.

Queen, that is the bus frequency. I would be very surprised if it would change (as it will probably destroy your mainboard or at least make your machine crash). You have a Coffee Lake laptop, right? Could you maybe do some Cinebench runs, while capturing the CPU frequency and see whether it shows a similar fluctuating pattern?

 

[Queen6]

The person who started this thread apparently does. I mean, look at the title. I wonder how many laptops — or even desktops out there — will be able to maintain max single-core turbo on full 6 core load for more then 2 seconds?



Yeah, wait, but isn't it what is happening? Lets take the notebookcheck test for the 8850H. They write that in their sustained cinebencz run, "The test runs for 42 minutes and we see an average clock of about 2.7 GHz for all 6 cores, which is slightly above the base frequency". From performance standpoint, its the same as running constant 2.7Ghz.

True, the clock fluctuates wildly, which is weird. I think notebookcheck guys are correct in saying that its a drawback of Apple not restricting the CPU TDP. It tries to go as fast as possible, gets quite hot, and then needs to briefly cool down before winding up again. They could probably reach more sustained clocks by introducing a limiter — at a cost of reduced max performance though.

I would like to see more detailed analysis of these fluctuations. If they changes something in Coffee Lake and it changes its frequency more often, the observed fluctuations could be also artifacts of sampling. E.g. if it goes between 1.8 and 3.5 Ghz 50 times per second, the effective speed— for all intends and purposes — be 2.65Ghz. That is, in every small time interval, the speed will be averaging to around 2.6. If you sample only 2 points per seconds though, you will of course see wild fluctuations....

I'd like to see how it looks on macOS as well. Notebookcheck use Windows for some reason.

BTW, Dell's XS 15" quad core Coffee Lake partially shows similar fluctuating pattern: https://www.notebookcheck.net/Dell-XPS-15-2018-9570-8300H-GTX-1050-97Wh-Laptop-Review.308420.0.html

I couldn't find frequency graphs of any other Coffee Lake machine they tested though...



I don't think that its how base clock is defined by Intel. They say that base clock is the clock that the CPU runs on all cores when generating its TDP of thermal power.



Are you sure about that? Do you know how the frequencies are sampled? How do you know its not bouncing between 2 and 4Ghz 50 times per second?

And please don't misunderstand me, I am not trying to be polemic, I just want to challenge people to look at the things in a more complex way. It is obvious that there is something weird going on with these machines. But I don't think that its as trivial as some make it sound.
[doublepost=1532164860][/doublepost]

Queen, that is the bus frequency. I would be very surprised if it would change (as it will probably destroy your mainboard or at least make your machine crash). You have a Coffee Lake laptop, right? Could you maybe do some Cinebench runs, while capturing the CPU frequency and see whether it shows a similar fluctuating pattern?

The 8750H holds a stable 3.9GHz for the duration of the Cinebench R15 CPU test hence the score of 1273CB.

Massive synthetic stress tests such as Prime95 will result in the CPU rolling back frequency due to the PL-1 limit of 45W being exceeded. The thermals remain to be excellent if memory serves approx 70C, with an ambient temperatures of 25C.

Currently both PL-1 & PL-2 are at default as is short term Turbo (PL-2 @90W) at 28 seconds.

I'll run Cinebench with task manager on top, later today. From memory the CPU holds the line. I'm generally not into benchmarking etc. however they can serve purpose to validate system performance & stability initially.

I may also still have more of the OCCT results that may be pertinent.

Q-6