Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
2018 i9 MacBook Pro throttling discussion [merged]
- Thread starter M.Rizk
- Start date
-
- Tags
- macbook performance pro
- Sort by reaction score
I don't think he is trying to compare the performance between them really. He is doing his best to figure out what is causing the throttling.
No doubt about it. I run multiple VMs: win10, ubuntu, debian on a 16GB 2017 mbp i7. Why not get the best of ALL worlds. Stepped away from bootcamp a while ago as it is (and Apple admits this) under-supported. To each his or her own. The drawback used to be Parallels or Fusion had serious issues with running bare metal - utilizing all native cores and making the best of native drivers. That's what kept heavy duty transcoders away. No longer, they've come a long way. You can't beat the appeal of shutting or suspending a virtual machine down in a second. And since its file based, moving your "machine" around in a suspended snapshot state is a synch. Priceless - something bootcamp doesn't afford easily.Using a VM gets you roughly 90-95% of the bare metal performance and you're still running macOS at the same time as the host OS. Why run just one OS at a time when you can have the best of both worlds, or even toss Linux into the mix and do even more, all at the same time, almost with the performance of running the OSes on the bare metal (within ~5-10% of it)?
Running one OS at a time on today's modern high powered multi-core RAM stuffed super duper hella high speed storage-based computers, that's so quaint, really.
Hence my sweet spot: the only high-end upgraded configuration that makes sense to me is a 32GB i7 with a screaming 1T SSD. That should have been the standard model in 2018. VMs love memory and fast solid state disk. But the entry price is out of whack, especially since (shamefully) things don't seem fully baked yet...(pun intended)
That's why I too would love to see benches of Premiere running in a modern VM (even the open source ones). Why not? Pandora's box is open again.
It produces output that looks much like x264’s superfast preset. Not good enough for a lot of purposes.
I'm new here, but I follow a lot of threads and just created an account so I thought I'd throw in my 2 cents, or at least a little info to contrast this whole fiasco with.
I've been a Mac user since ~2005, Windows before that. I've made my living as a developer for ~15 years and Mac has always felt right for work. Gaming is another story, I love Windows for that.
At the beginning of 2017, I switched back to Windows because I was fed up with the garbage hardware Apple was releasing. I got a 15" XPS for work and an Alienware R3 for gaming. Both of those machines throttled pretty bad and I'm still using the Alienware for gaming. I have to under clock the Alienware and turn off turbo boost. If I don't, the thing will just cook itself and BSOD. I'm assuming this is because of both the cpu and the gtx 1070 working while playing games creating too much heat for the system to handle. The Alienware is a beast compared to these new Macbooks in both size and cooling architecture, and yet, it throttles. The XPS also had to be under clocked or it'd throttle itself down to idle speeds and the whole machine would run like garbage.
Anyhow, earlier this year I switched back to Mac for work because I missed the development environment as it provides the greatest and most efficient workflow for myself. I picked up a 2017 13" i7 dual core, the one without the touch bar. It turns out The 13" was too small of a screen for me and the dual core was not enough power.
I've just received my 15" i9 a few days ago and I think it's an amazing machine. I don't do a ton of video encoding but I do a great amount of Photoshop and XD work and use Premiere and After Effects very little. I'm a programmer so I spend a lot of time in VS Code and the command line using Node, npm, etc... This machine hasn't let me down yet and I don't foresee it will.
I've ran my own tests on the i9 just to confirm it was throttling, and it did when I transcoded a video to hevc with handbrake. I expected as much though. Any laptop I've ever used, even if it was a beast has done the same thing.
All this to say, laptops throttle because of heat. I'm not the global laptop expert, but from my many years experience and even the recent experience with Windows laptops, they throttle, it's just the way it goes when you have heat generators inside a slim form factor. If I needed some heavy lifting, I wouldn't be using a laptop, I'd be using a desktop, but for on the go heavy lifting, I trust my new Macbook Pro i9 to get me through the day.
I've been a Mac user since ~2005, Windows before that. I've made my living as a developer for ~15 years and Mac has always felt right for work. Gaming is another story, I love Windows for that.
At the beginning of 2017, I switched back to Windows because I was fed up with the garbage hardware Apple was releasing. I got a 15" XPS for work and an Alienware R3 for gaming. Both of those machines throttled pretty bad and I'm still using the Alienware for gaming. I have to under clock the Alienware and turn off turbo boost. If I don't, the thing will just cook itself and BSOD. I'm assuming this is because of both the cpu and the gtx 1070 working while playing games creating too much heat for the system to handle. The Alienware is a beast compared to these new Macbooks in both size and cooling architecture, and yet, it throttles. The XPS also had to be under clocked or it'd throttle itself down to idle speeds and the whole machine would run like garbage.
Anyhow, earlier this year I switched back to Mac for work because I missed the development environment as it provides the greatest and most efficient workflow for myself. I picked up a 2017 13" i7 dual core, the one without the touch bar. It turns out The 13" was too small of a screen for me and the dual core was not enough power.
I've just received my 15" i9 a few days ago and I think it's an amazing machine. I don't do a ton of video encoding but I do a great amount of Photoshop and XD work and use Premiere and After Effects very little. I'm a programmer so I spend a lot of time in VS Code and the command line using Node, npm, etc... This machine hasn't let me down yet and I don't foresee it will.
I've ran my own tests on the i9 just to confirm it was throttling, and it did when I transcoded a video to hevc with handbrake. I expected as much though. Any laptop I've ever used, even if it was a beast has done the same thing.
All this to say, laptops throttle because of heat. I'm not the global laptop expert, but from my many years experience and even the recent experience with Windows laptops, they throttle, it's just the way it goes when you have heat generators inside a slim form factor. If I needed some heavy lifting, I wouldn't be using a laptop, I'd be using a desktop, but for on the go heavy lifting, I trust my new Macbook Pro i9 to get me through the day.
If you download Volta, set TDP to 45 and run that encode again it shouldn’t throttle. It won’t run at turbo frequencies, but it should reach and maintain base frequencies.
I did this as well and I had the same results as others in this thread. Hoping Apple will resolve this in a future (hopefully not too far future) update that prevents us from having to mess with any 3rd party software. If I was a video editor, I'd probably be a little more concerned with all of this, but I'm not constantly stressing my machine, so using Volta whenever necessary doesn't seem like a big issue for myself, but others, perhaps.
It seems like a question if and when they’ll apply this type of method (and also why the heck they didn’t do it in the first place). Right now though, and as always with Apple, the silence is deafening.
I might have missed the sarcasm but at the end of the video after multiple runs of cinebench you can see the extreme throttling of the CPU in which wattage/temperature/frequency oscillate:
Not sure if this has been posted yet here, but intel removed Mac Version power gadget today. It was gone for a few hours and replaced with an updated version. It's quite possible that 3.5.2 was not compatible with mobile i9. I have not tested the new version 3.5.3, but maybe it will be more accurate as to what the heck is going on.
https://9to5mac.com/2018/07/21/intel-power-gadget-mac-removed/
https://9to5mac.com/2018/07/21/intel-power-gadget-mac-removed/
I was referring to the constant 2.6-3.4 ghz clock speeds at around 7:40 minutes there about. Just another test to analyze. Im not an expert but I haven't seen any tests producing these results so wanted to know the community's opinion.
[doublepost=1532219890][/doublepost]
You forgot to post this link buddy https://software.intel.com/en-us/articles/intel-power-gadget-20
I just did a few tests transcoding a video in Handbrake to HEVC and used stock settings, Volta undervolt, Volta power limit and Macs Fan Control. I'm on the 15" Macbook Pro 2018 i9 with 32GB RAM and 512GB SSD. Below are some of my results with seconds it took to transcode the video.
Stock Settings: 44 seconds.
Volta undervolt (-55mV) with full fan speed: 43 seconds.
Volta Power Limit to 45w: 41 seconds.
Volta Power Limit to 45w with full fan speed: 40 seconds.
edit:
I tested with the latest intel power gadget as well as 3.5.2 (the version that was removed) and they both showed the same info, so I don't think there was a conspiracy lol.
Stock Settings: 44 seconds.
Volta undervolt (-55mV) with full fan speed: 43 seconds.
Volta Power Limit to 45w: 41 seconds.
Volta Power Limit to 45w with full fan speed: 40 seconds.
edit:
I tested with the latest intel power gadget as well as 3.5.2 (the version that was removed) and they both showed the same info, so I don't think there was a conspiracy lol.
You act like they haven't been making computers for over 40 years?
I have a MacBookPro15,1 (i9 with 32GB) and, amid so many facts and/or speculations (including the Intel Power Gadget utility download link removed), I looked into an alternate test.
Basically, I created a bash script that builds a large project (over 1000 C++ source files) using incrementally from 1 to 12 threads and measuring the build time of each case. Then, I compared the results with a MacBookPro14,3 (2017 i7 3.1GHz), with threads from 1 to 8.
Results are quite interesting:
As expected, the i9 is faster with only with one thread (it seems that the CPU frequency was over 4GHz). But immediately with only 2 treads, the old i7 is faster.
By comparing the two absolute minimum build time (564s and 614s), there is a gain of only 8%.
Basically, I created a bash script that builds a large project (over 1000 C++ source files) using incrementally from 1 to 12 threads and measuring the build time of each case. Then, I compared the results with a MacBookPro14,3 (2017 i7 3.1GHz), with threads from 1 to 8.
Results are quite interesting:
As expected, the i9 is faster with only with one thread (it seems that the CPU frequency was over 4GHz). But immediately with only 2 treads, the old i7 is faster.
By comparing the two absolute minimum build time (564s and 614s), there is a gain of only 8%.
Here is a very good review. Compact but very informative.I have a MacBookPro15,1 (i9 with 32GB) and, amid so many facts and/or speculations (including the Intel Power Gadget utility download link removed), I looked into an alternate test.
Basically, I created a bash script that builds a large project (over 1000 C++ source files) using incrementally from 1 to 12 threads and measuring the build time of each case. Then, I compared the results with a MacBookPro14,3 (2017 i7 3.1GHz), with threads from 1 to 8.
Results are quite interesting:
As expected, the i9 is faster with only with one thread (it seems that the CPU frequency was over 4GHz). But immediately with only 2 treads, the old i7 is faster.
By comparing the two absolute minimum build time (564s and 614s), there is a gain of only 8%.
In that video above the reviewer gets into discussing thermals at 3:30 if you're interested in just jumping to it. His testing shows completely erratic behavior with the clock speeds showing that now typical jagged line jumping up and down and not a flat line level of performance which we should be seeing if things were working even close to stable.
I can and I do appreciate that many people have discovered the Volta app and it's well known now that using that app can alleviate some of the throttling issues by bringing the temps down and properly limiting the i9 power consumption to 45 watts. Using Volta to also disable the Turbo Boost apparently does provide benefits as well and getting a nice flat line we'd expect to see. Unfortunately that flat line is at 2.6 GHz and it's an i9 with a stock speed of 2.9 so you're losing almost 10% performance from what it should be able to provide adequately for extended periods of time.
What I find disturbing (as some others have noted but not that many oddly) is that the i9-8950HK in the 2018 MBP has a TDP of 45 watts and yet we're seeing instances and examples of the power consumption reaching almost 70 watts (at about the 4:40 mark in the video he mentions it and you can see that spike occur). Yes I realize the Intel Power Gadget is measuring some other aspects as well but it's patently obvious the i9 is pulling a lot more power than designed for, and that's the curious thing as it shouldn't be happening. The i9 should not be pulling that kind of power, obviously, so yeah, there's something definitely wrong happening somewhere, most likely inside the CPU itself since it's requesting that kind of power in the first place. Yes, it can pull more than the rated TDP since that's just an average Intel calculates, but even so, close to 70 watts is a rather significant jump indeed.
Then he explains how doing a Cinebench run with the TDP locked at 45 watts got him the best overall score with a near flat line stable run at 3.1 GHz so slightly above stock speed which is very promising BUT again, the fact that one needs to use some third-party app (not free, either, but with that 7-day trial, there could be other such apps like Volta, I don't personally know at this moment) to get that flat line level stable performance one should be able to expect is what's important here. He also notes that the temps when the TDP is locked at 45 watts get into the 80C range, a big difference from ~100C obviously for a variety of reasons, the most prevalent one being you get better performance with moderate expected power consumption.
Could be a glitch with the i9 in terms of the TDP rating and how much power it's pulling, and obviously the actual thermal setup inside the 2018 MBP isn't that great even considering the TDP should be the same at 45 watts as last year's top of the line hardware.
It's a nice little review, however.
I can and I do appreciate that many people have discovered the Volta app and it's well known now that using that app can alleviate some of the throttling issues by bringing the temps down and properly limiting the i9 power consumption to 45 watts. Using Volta to also disable the Turbo Boost apparently does provide benefits as well and getting a nice flat line we'd expect to see. Unfortunately that flat line is at 2.6 GHz and it's an i9 with a stock speed of 2.9 so you're losing almost 10% performance from what it should be able to provide adequately for extended periods of time.
What I find disturbing (as some others have noted but not that many oddly) is that the i9-8950HK in the 2018 MBP has a TDP of 45 watts and yet we're seeing instances and examples of the power consumption reaching almost 70 watts (at about the 4:40 mark in the video he mentions it and you can see that spike occur). Yes I realize the Intel Power Gadget is measuring some other aspects as well but it's patently obvious the i9 is pulling a lot more power than designed for, and that's the curious thing as it shouldn't be happening. The i9 should not be pulling that kind of power, obviously, so yeah, there's something definitely wrong happening somewhere, most likely inside the CPU itself since it's requesting that kind of power in the first place. Yes, it can pull more than the rated TDP since that's just an average Intel calculates, but even so, close to 70 watts is a rather significant jump indeed.
Then he explains how doing a Cinebench run with the TDP locked at 45 watts got him the best overall score with a near flat line stable run at 3.1 GHz so slightly above stock speed which is very promising BUT again, the fact that one needs to use some third-party app (not free, either, but with that 7-day trial, there could be other such apps like Volta, I don't personally know at this moment) to get that flat line level stable performance one should be able to expect is what's important here. He also notes that the temps when the TDP is locked at 45 watts get into the 80C range, a big difference from ~100C obviously for a variety of reasons, the most prevalent one being you get better performance with moderate expected power consumption.
Could be a glitch with the i9 in terms of the TDP rating and how much power it's pulling, and obviously the actual thermal setup inside the 2018 MBP isn't that great even considering the TDP should be the same at 45 watts as last year's top of the line hardware.
It's a nice little review, however.
Please correct me if I am wrong, but according to notebookcheck, there is a 13% gap between the mid model 6-core 2018, and the base model 4-core 2017. And by limiting power consumption to the processor on the 2018, this gap widens. The 2018 now sees 20% more performance. However, at the cost of closing the gap between the two when doing every day work that does not tax the processor. Further it appears, they would need to limit power consumption to the 2017 as well, to make the test completely fair. I mean, the gap may get close to 13% all over again, if they did this. If they applied the same tweak to the 2017 as well.
To add insult to injury, forum member Aea has just shown here:
https://forums.macrumors.com/threads/2018-mbp-i9-testing-w-volta-fan-control.2128797/
That even when limiting the power consumption to the i9, it still has a worse score than the mid model 6-core 2018 with the same tweak. Ultimately making the i9 even closer in performance to the base model 4-core 2017. All without the same tweak applied to the base model 4-core 2017... Which may now make all of these tweaks meaningless to an extent?
I am referring to:
First article:
https://www.notebookcheck.net/The-n...nd-clearly-beaten-by-the-XPS-15.317264.0.html
Second article:
https://www.notebookcheck.net/Apple...e-performance-with-a-few-clicks.317552.0.html
To add insult to injury, forum member Aea has just shown here:
https://forums.macrumors.com/threads/2018-mbp-i9-testing-w-volta-fan-control.2128797/
That even when limiting the power consumption to the i9, it still has a worse score than the mid model 6-core 2018 with the same tweak. Ultimately making the i9 even closer in performance to the base model 4-core 2017. All without the same tweak applied to the base model 4-core 2017... Which may now make all of these tweaks meaningless to an extent?
I am referring to:
First article:
https://www.notebookcheck.net/The-n...nd-clearly-beaten-by-the-XPS-15.317264.0.html
Second article:
https://www.notebookcheck.net/Apple...e-performance-with-a-few-clicks.317552.0.html
for those who are interested in undervolting, there is a terminal program calles VoltageShift, it works in my KabyLake hackintosh laptop and may work in Coffee Lake too. It would be interesting to see the results of TDP limiting in VOlta and Undervolting with VoltageShift, also you can undervolt the iGpu.
Download and Instructions:
https://sitechprog.blogspot.com/2017/06/voltageshift.html?m=1
Download and Instructions:
https://sitechprog.blogspot.com/2017/06/voltageshift.html?m=1
https://twitter.com/tldtoday/status/1020887972780969985?s=21
New update from Jonathan Morrison. i9 real world tests.
New update from Jonathan Morrison. i9 real world tests.
While this thread has been entertaining, those that do not think MacOS and/or the firmware of the machine cannot be possibly be updated because they had 6 core support in 2003 or whenever need to stop for a moment and think.
New CPU's introduce new timings, hyper-threading, and a host of other features today that the OS must support. Supporting a 2003 CPU is nowhere near the same as supporting a 2018 CPU.
This is all beside the point.
I'm sure Apple does know their user base and designed it for what it sees as the usage of a majority of those users (who will be very happy, as am I, with their i9 and 32gig of ram).
Last message on my 15" mid-2015 i7 2.8 with AMD Radeon R9 M370X as my Time Machine restore to the new one is finally done. It'll be listed soon on Craigslist.
If I were rendering I might think again, though probably not since it's the fastest MacBook available and it's unlikely I would be using it full time for that purpose.
Edit - One other thing is that it might just be a bug in their firmware for that usage. Apple is almost never going to admit that. It would just be fixed in the next security patch or other update.
Edit2 - Other vendors do this all the time to new releases. Hardware is frequently ready before all the optimizations are done. Engineers get asked yeah, but does it work? Then, how long do you need? If it's within a month or two then marketing will override and it's released.
Personally I think they just rushed it a bit because they were, rightfully, getting beat up for not having a release at WWDC.
New CPU's introduce new timings, hyper-threading, and a host of other features today that the OS must support. Supporting a 2003 CPU is nowhere near the same as supporting a 2018 CPU.
This is all beside the point.
I'm sure Apple does know their user base and designed it for what it sees as the usage of a majority of those users (who will be very happy, as am I, with their i9 and 32gig of ram).
Last message on my 15" mid-2015 i7 2.8 with AMD Radeon R9 M370X as my Time Machine restore to the new one is finally done. It'll be listed soon on Craigslist.
If I were rendering I might think again, though probably not since it's the fastest MacBook available and it's unlikely I would be using it full time for that purpose.
Edit - One other thing is that it might just be a bug in their firmware for that usage. Apple is almost never going to admit that. It would just be fixed in the next security patch or other update.
Edit2 - Other vendors do this all the time to new releases. Hardware is frequently ready before all the optimizations are done. Engineers get asked yeah, but does it work? Then, how long do you need? If it's within a month or two then marketing will override and it's released.
Personally I think they just rushed it a bit because they were, rightfully, getting beat up for not having a release at WWDC.
Last edited:
This thread reminds me very much of back when Apple was in the G3/G4 ditch and was putting out benchmarks showing how the G4 with AltiVec could outperform x86 chips in certain workloads. I was a teenager, complete mac fanboy, and totally ignorant. I lapped it up.
Yeah no thanks.
I dont remember iphone release that overheats and underperforms. Its a huge company with huge budget not some boutique company. They had enough time to prepare.