2018 i9 MacBook Pro throttling discussion [merged]

[Viper-FFM]

I did the test few posts back. The i9 actually performs worse than the i7 if you use the 45W limit:
View attachment 771992

Now THATS what I was looking for since three days.

Under the same Watt limit, the i7outperforms the i9. Question: Does the i7 perform better when the i9 is untocuhed in regards of wattage limit etc.?

Glad I downgraded to th 2.6ghz i7 but choose the 32RAm option

 

[augustya]

Just returned the i9 unopened. The i7 two point six will arrive on Monday or Tuesday.

Might have to “work” from home on that day.

Uh !! What ! You returned a product without opening it !! Just because you read reviews !! That's shocking !!

 

[kotlos]

Question: Does the i7 perform better when the i9 is untocuhed in regards of wattage limit etc.?

Yes. Both of them improve with the watt limit & Cinebench but the i7 improves even more.

Again people should be aware that setting the 45 watt limit can actually hinder the performance with tasks that don't push the CPU very hard. With many tasks the i9 can perform much better than the i7. Its only when maxing out the cores for more than few seconds that i9 gets crippled.

 

[iMacDragon]

Yes. Both of them improve with the watt limit & Cinebench but the i7 improves even more.

Again people should be aware that setting the 45 watt limit can actually hinder the performance with tasks that don't push the CPU very hard. With many tasks the i9 can perform much better than the i7.

Indeed, for brief blips the cpu should be let to max out, but if load becomes sustained it should be limited down to a level the machine can cope with.

 

[Falhófnir]

I'm certain Apple will explain this through one of the "journalists" they keep in their pocket.

‘You’re using it in a hot country wrong’?

 

[Ploki]

Grr...I’m still out and want to try on my i9 but I am not a huge techie and worry I might screw up whatever limits and kill my machine! Haha!

iirc Volta keeps a 30second window before applying undervolt settings, so you can disable it if you experience crashes. as far as i know undervolting cannot damage your hardware, it just runs unstable if too severely undervolted.

 

[kotlos]

iirc Volta keeps a 30second window before applying undervolt settings, so you can disable it if you experience crashes. as far as i know undervolting cannot damage your hardware, it just runs unstable if too severely undervolted.

Under macOS with the 2018 models, the Volta does not change the voltage of the CPU. See page 34 for my tests.

 

[Ploki]

Under macOS with the 2018 models, the Volta does not change the voltage of the CPU. See page 34 for my tests.

wait what? how did that guy get 1000 on cinebench, only by limiting TDP? Or does notthing work?

 

[Feenician]

Volta doesn't allow higher than 45W, right? I am sure we can get even better results with 50W~55W and the cooling system can still handle it just fine.

It doesn’t. If you have bootcamp you can use the Intel XTU utility to set an arbitrary TDP value. If someone can test that with higher TDP on the i9 it should be informative.

 

[Ploki]

It doesn’t. If you have bootcamp you can use the Intel XTU utility to set an arbitrary TDP value. If someone can test that with higher TDP on the i9 it should be informative.

here we are on forums fixing mess apple made. sheesh

 

[mr.anthonyramos]

here we are on forums fixing mess apple made. sheesh

Maybe Apple should hire some of us, maybe we’ll do a better job. Haha.

 

[iMacDragon]

wait what? how did that guy get 1000 on cinebench, only by limiting TDP? Or does notthing work?

Yes, only touching TDP and then get > 1000 rather than ~900

 

[kotlos]

wait what? how did that guy get 1000 on cinebench, only by limiting TDP? Or does notthing work?

Limiting the wattage gets you a 10% boost, then allowing few seconds of high wattage before setting the 45W limit gives you another 5% improvement to reach 1000 on Cinebench. As others have said, higher wattage limit could give better performance.

Changing the voltage with the current version of the Volta app has no effect in the performance as you can see in my test in page 34. Maybe the developers can update the app to support the new processors or even allow higher wattage limits. Send them emails!

 

[OC40]

Did you get the 16GB or 32GB?

32gb for sure.

Uh !! What ! You returned a product without opening it !! Just because you read reviews !! That's shocking !!

Meh. I ordered the i9 on a last second whim. I’m not one to jump to conclusions based on others reviews; however, the consistency related to the heat is what concerned me.

Unopened, yes. I didn’t want to spend time configuring a laptop, wiping it and then configuring a new one.

 

[Feenician]

I did the test few posts back. The i9 actually performs worse than the i7 if you use the 45W limit:
View attachment 771992

You also ran a test where you didn’t throttle the i9 for a few seconds before limiting it to 45w and that seemed more promising though right?


[doublepost=1532196802][/doublepost]

Limiting the wattage gets you a 10% boost, then allowing few seconds of high wattage before setting the 45W limit gives you another 5% improvement to reach 1000 on Cinebench. As others have said, higher wattage limit could give better performance.

Changing the voltage with the current version of the Volta app has no effect in the performance as you can see in my test in page 34. Maybe the developers can update the app to support the new processors or even allow higher wattage limits. Send them emails!

I emailed them last night. They have a tremendous short term opportunity to make money here!

 

[Ploki]

I did the test few posts back. The i9 actually performs worse than the i7 if you use the 45W limit:
View attachment 771992

well this is depressing.

 

[Uhu]

The fix for apple should be easy then. CPU intensive tasks should be easily identifiable by the scheduler. Then the power target should be lowered to 45 or 55 watts. Easy as Pie.
You can even sell it like a software speed upgrade

 

[SRLMJ23]

Did you not see that after he put it in a freezer and redid the test it was 30% faster and no aggressive throttling??????

TLDR: its not the software

I agree, I do not think it is the software. How is macOS now optimized for 6 cores when the iMac Pro has 8 core, 10 core, 14 core, and an 18 core CPU's?

Also, the now ANCIENT trashcan Mac Pro had 6 core, and 8 core CPU's?

I love Apple, and hope these new MacBook Pro's get the throttling issue figured out because I really want to buy one and upgrade from my 2015 MBP, even though this machine is still in mint condition and runs amazing.

However, I just cannot blindly defend Apple on everything like a lot of people on this forum do, and that is why I am saying that I do not think it has anything to do with macOS. I can see certain software and software suites not being optimized, but not macOS. If it turns out it is macOS, it is pretty sad Apple had all this time to optimize macOS and did not, and then released machines that cannot even run right!

 

[mr.anthonyramos]

Here is a new video. Not as comprehensive as Jonathan’s but still maybe some information we could work with.

 

[M.Rizk]

The fix for apple should be easy then. CPU intensive tasks should be easily identifiable by the scheduler. Then the power target should be lowered to 45 or 55 watts. Easy as Pie.
You can even sell it like a software speed upgrade

Stop giving them cash grab ideas. Thanks

 

[Feenician]

Stop giving them cash grab ideas. Thanks

If they charged for a fix, I’m not coming back to the forum for a week!

 

[Elektrofone]

32gb for sure.

Meh. I ordered the i9 on a last second whim. I’m not one to jump to conclusions based on others reviews; however, the consistency related to the heat is what concerned me.

Unopened, yes. I didn’t want to spend time configuring a laptop, wiping it and then configuring a new one.

When you do get it would you mind running some benchmarks? Considering returning my i9 for the 2.6 but need to see some comparisons soon.
[doublepost=1532198313][/doublepost]

Here is a new video. Not as comprehensive as Jonathan’s but still maybe some information we could work with.

Yeah. At this point I want to see a comparison between the i9 and 2.6 i7 most of all. If I return my i9 that will probably be the one I go with.

 

[Ploki]

I agree, I do not think it is the software. How is macOS now optimized for 6 cores when the iMac Pro has 8 core, 10 core, 14 core, and an 18 core CPU's?

Also, the now ANCIENT trashcan Mac Pro had 6 core, and 8 core CPU's?

forgot about the cheese grater, 8 cores since 2008.

if you want hyperthreading, then the 2009 mac pro (8 cores, 16 threads)

 

[content]

The fact that the 2017 model can get an almost equal score to the 2018 i7 with 2 cores less / DDR3 and the dell 6 core i7 with a lower spec CPU can rolf stomp the MBP 2018 i7 by a 20-30% score lead under sustained load should show you how bad the situation is.

https://www.notebookcheck.net/The-n...nd-clearly-beaten-by-the-XPS-15.317264.0.html

This is precisely the point I was trying to make in this thread.

The 2018 2.6GHz 6-core is only slightly (13%) faster than the 2017 2.8 GHz 4-core base model.

And not only that, but they had this to say:

“The performance drops further after the initial run and the average result after 36 runs is just 832 points. This means the advantage over the MacBook Pro 15 2017 with the quad-core Core i7-7700HQ is just 13%, and the difference should be even smaller the more runs we perform.”

In other words, if they ran the test longer the gap between the 2018 6-core compared to the 2017 4-core would get even smaller. Even less than 13%.

https://www.notebookcheck.net/The-n...nd-clearly-beaten-by-the-XPS-15.317264.0.html

So in this sense, I feel any test comparing the 6-core 2018 models is moot. The real test needs to occur against the 2017 4-core, which is completely stable and not thermally constrained. Just as Dave Lee did...

 

[M.Rizk]

This is precisely the point I was trying to make in this thread.

The 2018 2.6GHz 6-core is only slightly (13%) faster than the 2017 2.8 GHz 4-core base model.

And not only that, but they had this to say:

“The performance drops further after the initial run and the average result after 36 runs is just 832 points. This means the advantage over the MacBook Pro 15 2017 with the quad-core Core i7-7700HQ is just 13%, and the difference should be even smaller the more runs we perform.”

In other words, if they ran the test longer the gap between the 2018 6-core compared to the 2017 4-core would get even smaller. Even less than 13%.

https://www.notebookcheck.net/The-n...nd-clearly-beaten-by-the-XPS-15.317264.0.html

So in this sense, I feel any test comparing the 6-core 2018 models is mute. The real test needs to occur against the 2017 4-core, which is completely stable and not thermally constrained. Just as Dave Lee did...

Is this what you are asking for?
#946

The Notebookcheck review was done before they themselves discovered the power limit fix.