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 MBP 15 i9 Vega 20, Throttling to 1.7 ghz!!
- Thread starter IdentityCrisis
- Start date
- Sort by reaction score
I think this is a good place to discuss these thermal numbers. Running Heaven's benchmark as well as driving the CPU to 100% will stress the cooling to see if it functions as well as test the efficiency of the CPU. I'm seeing more & more posts where people don't think this is important, but it actually is on your pocketbook. Why?
Given the same GPU, the i7 & i9 can either perform the same or different in this scenario. Granted, CPUs do vary, but not by much at all. If we see a significant difference in the downclock, then you'd want to go for the higher clock. My experience says that the i7 should win out in max performance...ususally because less voltage is being pumped in due to lower cache, cores & hyperthreading. This tells you which will perform better under max load. This is also only practical if you don't replace your MacBook every couple years. I manage to stretch my machines about 11 years & they run 100% CPU & 100% GPU by the time I'm hitting year 6.
This is also the first design that I suspect the i9 may have been more efficiently designed to handle the higher frequency. Usually, the ix series denotes the number of cores & the configuration in hyper threading. That is not true in this case. It simply denotes higher frequency & difference is cache. We will find out.
Another thing brought up: both the CPU & GPU will thermal limit. That's actually not the case. Thermal control is something CPU controlled. Out of every test I have done in the past on GPUs, they don't thermal limit & if they hit their limit, they either fry, but more likely just fail until they cool. GPU Overclockers are familiar with this. Plus, we are talking the same GPU model so the difference in performance will lie with the CPU which rests on the same heatsink. You can't do a comparison using the i7 & Vega 16 vs i9 & Vega 20...there's no common variable we can use.
1.8 on the frequency across multiple cores isn't bad. Single core performance will suffer, though...& that shouldn't matter for the most part as we will likely be single tasking high performance, unlike what this test demands.
Given the same GPU, the i7 & i9 can either perform the same or different in this scenario. Granted, CPUs do vary, but not by much at all. If we see a significant difference in the downclock, then you'd want to go for the higher clock. My experience says that the i7 should win out in max performance...ususally because less voltage is being pumped in due to lower cache, cores & hyperthreading. This tells you which will perform better under max load. This is also only practical if you don't replace your MacBook every couple years. I manage to stretch my machines about 11 years & they run 100% CPU & 100% GPU by the time I'm hitting year 6.
This is also the first design that I suspect the i9 may have been more efficiently designed to handle the higher frequency. Usually, the ix series denotes the number of cores & the configuration in hyper threading. That is not true in this case. It simply denotes higher frequency & difference is cache. We will find out.
Another thing brought up: both the CPU & GPU will thermal limit. That's actually not the case. Thermal control is something CPU controlled. Out of every test I have done in the past on GPUs, they don't thermal limit & if they hit their limit, they either fry, but more likely just fail until they cool. GPU Overclockers are familiar with this. Plus, we are talking the same GPU model so the difference in performance will lie with the CPU which rests on the same heatsink. You can't do a comparison using the i7 & Vega 16 vs i9 & Vega 20...there's no common variable we can use.
1.8 on the frequency across multiple cores isn't bad. Single core performance will suffer, though...& that shouldn't matter for the most part as we will likely be single tasking high performance, unlike what this test demands.
It honestly isn't too bad. I did a geekbench on my built i9-9900k, and its not much more than that.
I believe it depends on the software. When I use handbrake to convert video it begins at 3.8 for a minute and then drops and stays at 2.9 while maxing all cores. Final Cut Pro on the other hand usually stays between 2.2-2.6 when exporting movie files. Benchmarks don't always tell the true story.
Every chip will throttle to a degree. The burst mode is that peak they try to sell you on. You need to find a good balance on what you need it for vs the cost. Mine is going to be long term ownership which means I want the best CPU to drive the background processes when this thing starts getting old. That means looking into the i7 vs i9. Eventually, the OS & whatever background service as well as the non-graphical computations games do will eventually drive the CPU up pretty high...despite some people's arguments. One of those two chips are going to win on the high demand & that's the one I want.
cooling on this model it's a total joke, few grams of copper
Nature of the beast as Apple has focused heavily on thinner chassis versus performance. Given the size of the MBP the performance isn't too bad and now with Vega 20 a much better proposition, although personally; Vega should be the default dGPU at these price points not an additional cost.
Q-6
Last edited:
As I've written in the other thread, this is under a stress test, an unrealistic scenario that is designed to test the limits of the cooling and power system. The MBP is a thin and light laptop and is not designed to be operated under such circumstances. And in fact, this is a good result, since it throttles less than some other laptops in the same category (for example, Dell XPS throttles down to 1.5Ghz, Razer Blade to 1.7Ghz and Asus Zenbook Pro to a whopping 1.3Ghz).
Most importantly, you will probably never experience this kind of throttling in a practical situation, unless you try playing games while transcoding a bunch of 4K videos or do something similarly silly. This is a stress test, designed to push the machine to an unreasonable limit. Its like testing a car by loading it up to the safely limit, chaining it to a tree and then pushing the throttle as hard as you can trying to get it up a gravel road up a hill.
The only purpose of this test is testing the limits, and its doing exactly what its supposed to do — clocking down to prevent damage. If you want a machine that can do better — start looking at large and heavy gaming and workstation laptops.
[doublepost=1542971875][/doublepost]
Because Coffee Lake is clocked so weirdly compared to previous generations and it seems that it lacks a frequency limited for multicore operation, all of these CPUs should reach similar clocks in multi-core sustained operation. This was not the case with previous CPUs simply because Intel has placed a clock limit that stopped the CPU from getting higher. So yes, if what you do for living is encode videos or do raytracing on the CPU, the i9 is not worth it.
Where i9 is better though is burst performance, since its max turbo is higher.
Thermal control on CPU and GPU works very similarly. Macs are a bit special in this regard anyway, since they have dedicated hardware that monitors all this stuff and controls power levels. Its simply that the management system priorities the GPU before the CPU. The assumption is as follows: if you are actively using the GPU, than that is probably the more critical component for the user experience. In most games for example, you won't notice if the CPU frequency goes down a bit, but throttling the GPU will have a clear noticeable effect.
Single core performance won't suffer much, simply because less cores are active. In the above video, the stress test sees CPU maintaining 28 Watts while running on 1.7-1.8Ghz all 6 cores. In another words, the combined sustained TDP for CPU and GPU is around 65 watts, which makes perfect sense for the 85 watt power adapter. These 28 watts though are more then sufficient to keep two CPU cores at above 4Ghz. I can't give you exact figures, since I've sold my i9 and currently waiting for the new model with Vega
Most importantly, you will probably never experience this kind of throttling in a practical situation, unless you try playing games while transcoding a bunch of 4K videos or do something similarly silly. This is a stress test, designed to push the machine to an unreasonable limit. Its like testing a car by loading it up to the safely limit, chaining it to a tree and then pushing the throttle as hard as you can trying to get it up a gravel road up a hill.
The only purpose of this test is testing the limits, and its doing exactly what its supposed to do — clocking down to prevent damage. If you want a machine that can do better — start looking at large and heavy gaming and workstation laptops.
[doublepost=1542971875][/doublepost]
Because Coffee Lake is clocked so weirdly compared to previous generations and it seems that it lacks a frequency limited for multicore operation, all of these CPUs should reach similar clocks in multi-core sustained operation. This was not the case with previous CPUs simply because Intel has placed a clock limit that stopped the CPU from getting higher. So yes, if what you do for living is encode videos or do raytracing on the CPU, the i9 is not worth it.
Where i9 is better though is burst performance, since its max turbo is higher.
Thermal control on CPU and GPU works very similarly. Macs are a bit special in this regard anyway, since they have dedicated hardware that monitors all this stuff and controls power levels. Its simply that the management system priorities the GPU before the CPU. The assumption is as follows: if you are actively using the GPU, than that is probably the more critical component for the user experience. In most games for example, you won't notice if the CPU frequency goes down a bit, but throttling the GPU will have a clear noticeable effect.
Single core performance won't suffer much, simply because less cores are active. In the above video, the stress test sees CPU maintaining 28 Watts while running on 1.7-1.8Ghz all 6 cores. In another words, the combined sustained TDP for CPU and GPU is around 65 watts, which makes perfect sense for the 85 watt power adapter. These 28 watts though are more then sufficient to keep two CPU cores at above 4Ghz. I can't give you exact figures, since I've sold my i9 and currently waiting for the new model with Vega
Last edited:
thin chassis means problem with coating, from 2012 to present times problem with coating it still not solved, 6 years passed and Apple still serving same rubbish coating technology for LCD for premium price, soldered SSD is the most unreliable solution and also not upgradable, I wouldn't recommend this machine even for enemy
Last edited:
Certainly for the asking price Apple needs to do a lot more, equally Apple knows that many are tied to the OS so they don't need to try too hard.
Ultimately Apple is selling a brand to the masses and will do the bare minimum it can to maintain it's status, while maintaining & growing it's margins.
Q-6
Yeah, and you have to fix their crappy thermal paste if you want to get the best out of the even that.
It is the same with iPhones, even if they slack, people (including me) feel somewhat trapped in iOS - and while you have hundreds of Android options, there is only 1 manufacturer of iOS phones... Galaxy S10 might change that? We will see.
I broke out of Apple ecosystem some years back as I don't care to have all my eggs in one basket and frankly monopolies never serve the customer. I don't feel that I'm missing out on anything and I have a great deal more diversity. The Mac was my weapon of choice for many a year, however as Apple chooses not to serve it's professional users, I have no longer have need of Apple...
Q-6
Love the direction Microsoft are taking. They announced yesterday that they are making WinForms, WPF and WinUI open-source - can't wait to see what pro-design and pro-developer features they include next year. I think next year Microsoft/Google will make it very difficult for Apple to slack.
Last edited:
I accept MS as I need it, Google I want rid of, certainly I very carfully control it on the PC, basically YouTube only and no SW. A closed system may serve some, however those with greater needs an open system will always serve far better.
TBH if I could switch professionally to Linix I would and duelboot to a very stripped down W10 for my very limited gaming need. I digress, Microsoft is moving in a good direction and all power to them. I don't love some aspects, equally it's easily shutdown with a little knowledge.
Q-6
And at the same time, all those APIs are inseparably tied into Windows OS. What about this is pro-design and pro-developer? To benefit from this, you need to a) commit to .NET framework and b) commit to Windows ways of designing GUI. The one who benefits most out of it is Microsoft itself, since they do this in an attempt to make more software compatible with Windows platform.
And that quote you posted, no, just no. I never did driver development, so I can't comment on that. NT is certainly a good OS design and it is very much possible that their driver development model is great, but the test of the text doesn't make much sense to me. Yes, MS has good developer tools and they have some very smart people working on programming language design and stuff. But their APIs and designs are still quite crappy (I've been programming for Windows for years) and they have very little artistic vision or continuity. Thats why Windows 10 is still a terrible mess UI and functionality wise. It doesn't help that you have some "IQ 150" (lol) developers if all you do is put a bandaid on top of the bandaid. Thats why Windows apps stills struggle with HiDPI displays — because Windows core design team just doesn't think stuff through.
[doublepost=1544100406][/doublepost]
[MOD NOTE]
Lets get back on the topic, this isn't about MS, we have other locations for this sort of discussion
Hey maflynn, if you decide to delete the off topic parts, could you move them into a different thread instead? I think its a very interesting topic to discuss and it would be a shame for the time people put in writing posts to just see them gone...
I used handbrake on my new i9 MBP, and was pushing it pretty hard. It was getting around 2.8-2.9.
hm. other people seem to be getting better results (and also higher GB scores)
Well, no one has used handbrake yet that I have seen which really pushes the cpu to render the videos.