2018 i9 MacBook Pro throttling discussion [merged]

[The Mercurian]

they have cTDP of 35W listed tho, and perhaps they could've have nerfed them without getting so much flak, because almost no other laptops use them.
seems like they almost wanted to appeal more to gamers than pros?


2186M (i9 equivalent) is about 50$ more according to intel, so not much, probably 100$ more than i9.

Frankly, Apple could afford to put Xeons into the MBP at the prices they ask now. At least it would give them some sort of justitication why the machines are so expensive In all seriousness, the Xeons cost only $40 more

You guys are forgetting the Apple tax. Look at existing upgrade prices. Now imagine what they want for upgrade to Xeon

 

[leman]

they have cTDP of 35W listed tho, and perhaps they could've have nerfed them without getting so much flak, because almost no other laptops use them.

Do you really think that cTDP has any practical meaning? If we learned anything here, is that TDP of these CPUs can be set to arbitrary values. The TDP and cTDP seems to just marketing specs, as in — "we guarantee certain performance if you run the CPU at this particular TDP".

 

[Ploki]

You guys are forgetting the Apple tax. Look at existing upgrade prices. Now imagine what they want for upgrade to Xeon

Intel's RRPs:
Xeon E 2168M 623$
8950HK (i9) 583$
8850H (i7 2.6) 395$
8750H (i7 2.2) 395$
[doublepost=1532426025][/doublepost]

Do you really think that cTDP has any practical meaning? If we learned anything here, is that TDP of these CPUs can be set to arbitrary values. The TDP and cTDP seems to just marketing specs, as in — "we guarantee certain performance if you run the CPU at this particular TDP".

Yeah i guess you're right, it's me being in "wishful thinking" mode.

I'm still holding off till i see some official updates, i don't want to jerry-rig a new mac for it to run normally, else i'm waiting till next year's 10nm chips.

 

[Ma2k5]

Do you really think that cTDP has any practical meaning? If we learned anything here, is that TDP of these CPUs can be set to arbitrary values. The TDP and cTDP seems to just marketing specs, as in — "we guarantee certain performance if you run the CPU at this particular TDP".

Rather than the CPU TDP - isn't the issue more to do with the VRM's overheating and thus unable to provide power to the CPU, which is causing the throttling (from the recent video)?

Would this mean, Apple could technically do a revision of the VRM solution?

 

[duervo]

You're confused a bit about the i7/i9. The 50C or lower for turbo boost is not really for turbo boost. It's for the extra turbo boost that the i9 purports to deliver over the i7. And you're right: that situation is very rare, which is why the i9 is a ******** chip to prey on ill-informed consumers.

Both of them work based on thermal headroom and available power. Thermal Velocity Boost is not going to be on a CPU that does not have Turbo Boost. So, for the i9’s, it’s all part of the same steaming pile.
[doublepost=1532426639][/doublepost]

Rather than the CPU TDP - isn't the issue more to do with the VRM's overheating and thus unable to provide power to the CPU, which is causing the throttling (from the recent video)?

Would this mean, Apple could technically do a revision of the VRM solution?

That’s what current testing results from the VRM thread suggest. I doubt Apple will do a hardware revision (although, it is possible.) It’s more likely something they’ll address with a firmware update, because that’s a cheaper solution for them.

 

[Ma2k5]

That’s what current testing results from the VRM thread suggest. I doubt Apple will do a hardware revision (although, it is possible.) It’s more likely something they’ll address with a firmware update, because that’s a cheaper solution for them.

I wouldn't put past a hardware revision happening - could be very easy to do. Firmware solutions won't address the problem but mask it with a performance hit.

 

[leman]

Rather than the CPU TDP - isn't the issue more to do with the VRM's overheating and thus unable to provide power to the CPU, which is causing the throttling (from the recent video)?

Would this mean, Apple could technically do a revision of the VRM solution?

The VRM only overheats because CPU tries to pull a lot of power for prolonged periods of time. And it only does it because its configured by Apple to pull 100Watt continuously. My guess is that they do it to give the CPU as much freedom as it can get — this worked well for all the CPUs until now, but Coffee Lake simply has longer legs

Once the CPU TDP is set to something more reasonable (e.g. winterny sets it for 49Watt), the load on VRMs goes down and they no longer overheat, the performance stabilises.

Of course, redesigning the power delivery would be another solution, and one that would give more performance, but that would require a much larger laptop. According to notebookcheck's tests, only large gaming laptops that weight 3.5 kg and more can pull this off.

 

[duervo]

I wouldn't put past a hardware revision happening - could be very easy to do. Firmware solutions won't address the problem but mask it with a performance hit.

Maybe it will be both. Short term: firmware update. Long term: hardware revision.

 

[leman]

Firmware solutions won't address the problem but mask it with a performance hit.

Why performance hit? The performance actually increases, since VRM stops overheating. After all, the config needs to match the capabilities of the machine. A larger, bigger machine can of course offer more performance. But its a matter of picking a baseline. I don't think that maximal performance you can get out of these CPUs under ideal conditions is a reasonable baseline.

 

[Ploki]

Why performance hit? The performance actually increases, since VRM stops overheating. After all, the config needs to match the capabilities of the machine. A larger, bigger machine can of course offer more performance. But its a matter of picking a baseline. I don't think that maximal performance you can get out of these CPUs under ideal conditions is a reasonable baseline.

I actually agree. If i9 can pull reasonably more than i7 2.6GHz after they fix it, I'm going to get it. If it can't, i'm skipping this model and wait for next one.

However, if they do that (limit TDP and power consumption), they're essentially false advertising it right now. Which can in my opinion, bring them more legal issues than a few oddbal geeks that know that advertised speeds overheat it.

 

[Ma2k5]

Why performance hit? The performance actually increases, since VRM stops overheating. After all, the config needs to match the capabilities of the machine. A larger, bigger machine can of course offer more performance. But its a matter of picking a baseline. I don't think that maximal performance you can get out of these CPUs under ideal conditions is a reasonable baseline.

Maybe I can reword it to mean it would limit the performance potential - Similar to how some people are using TDP/turbo boost controls to actually make their laptops perform better, but the performance ceiling is a lot lower than comparable laptops or even previous generation MacBook's as a result.

A revisioned VRM vs firmware limiting would mean we don't have to reduce peak performance (or at least, not as much as now).

 

[Ploki]

Maybe I can reword it to mean it would limit the performance potential - Similar to how some people are using TDP/turbo boost controls to actually make their laptops perform better, but the performance ceiling is a lot lower than comparable laptops or even previous generation MacBook's as a result.

A revisioned VRM vs firmware limiting would mean we don't have to reduce peak performance (or at least, not as much as now).

just for the joke of it, i went looking for geekbench multicore results by CPU for the previous gen top of the line. (the 4980H) https://browser.geekbench.com/v4/cpu/search?dir=desc&q=4980HQ&sort=multicore_score

i was suspecting to see other laptops performing better, but MacBook is actually the highest scoring...

And that holds true for pretty much every top end macbook pro 2012-2017. You see 2017 MBP chugging 5% behind, and then 2018 i9 MBPs start to pop up on page 10, falling significantly behind.

 

[leman]

but the performance ceiling is a lot lower than comparable laptops or even previous generation MacBook's as a result.

A revisioned VRM vs firmware limiting would mean we don't have to reduce peak performance (or at least, not as much as now).

Well, yeah, but do your really think that laptops like Dell XPS or others have VRMs that can deliver more power? They perform better in tests simply because they limit the CPU. Even laptops like Razer Blade puts some strict limits on the CPU. Again, the only laptops that can unlock the potential of these CPUs are large and thick gaming machines like the MSI GT75 Titan.

Revisioned VRM would mean completely redoing the power subsystem, power delivery, cooling system and chassis. Its not something that can be done quickly or easily.

 

[The Mercurian]

just for the joke of it, i went looking for geekbench multicore results by CPU for the previous gen top of the line. (the 4980H) https://browser.geekbench.com/v4/cpu/search?dir=desc&q=4980HQ&sort=multicore_score

i was suspecting to see other laptops performing better, but MacBook is actually the highest scoring...

And that holds true for pretty much every top end macbook pro 2012-2017. You see 2017 MBP chugging 5% behind, and then 2018 i9 MBPs start to pop up on page 10, falling significantly behind.

How are they running those MacBooks at 4ghz?
[doublepost=1532428466][/doublepost]

Well, yeah, but do your really think that laptops like Dell XPS or others have VRMs that can deliver more power? They perform better in tests simply because they limit the CPU. Even laptops like Razer Blade puts some strict limits on the CPU. Again, the only laptops that can unlock the potential of these CPUs are large and thick gaming machines like the MSI GT75 Titan.

Revisioned VRM would mean completely redoing the power subsystem, power delivery, cooling system and chassis. Its not something that can be done quickly or easily.

The best dell scores are obtained after a reflow and tinkering with cpu settings. I would not reflow a new machine personally but the folks doing it are scoring 1200 with the i9 on cinebench

 

[Eason85]

Well, yeah, but do your really think that laptops like Dell XPS or others have VRMs that can deliver more power? They perform better in tests simply because they limit the CPU. Even laptops like Razer Blade puts some strict limits on the CPU. Again, the only laptops that can unlock the potential of these CPUs are large and thick gaming machines like the MSI GT75 Titan.

Revisioned VRM would mean completely redoing the power subsystem, power delivery, cooling system and chassis. Its not something that can be done quickly or easily.

It's impossible to predict how the MacBooks will run with the same TDP as windows laptops like the XPS 15 9570 until they actually do it, because there are 2 concurrent throttling problems working against the new MacBooks. The XPS 15, small cooler as it has, has at least a better cooling solution for the CPU; I am positive that the MBP will still see thermal throttling limiting its clocks from boosting, as tests have already shown temps hitting 100C and throttling in a few seconds.

 

[leman]

The XPS 15, small cooler as it has, has at least a better cooling solution for the CPU

Then why does it throttle more on Skylake and Kaby Lake CPUs?

I am positive that the MBP will still see thermal throttling limiting its clocks from boosting, as tests have already shown temps hitting 100C and throttling in a few seconds.

Of course, since that is how its designed to work. The CPU gets to the max speed it can, which is limited by the cooling solution. And obviously that means that it will run on its max safe temperature. There is nothing wrong with it reaching 100C or not getting maximal boost. What matters are the sustained clocks can can be reached this way.

After the TDP fix, winterny reports that his i9 MBP can reach sustained 3.2Ghz on 6 cores on consecutive Cinebench runs. This is a very solid result. And its consistent to what Dell XPS 15" 2018 can deliver (more info: https://www.notebookcheck.net/Dell-XPS-15-9570-15-more-performance-by-undervolting.317738.0.html).

 

[Eason85]

Then why does it throttle more on Skylake and Kaby Lake CPUs?

Easy: those models had difference VRM configurations. As reviews (including my own) show, the 9570 has significantly more stable performance than the last 2 generations under heavy load.

 

[augustya]

He ment i7 2.6.

Since base is i7 2,2 and upgrade is i9 2,9.

And yes, all three are six cores.

There is no option for i7 2.2 GHz on the 15" rMBP which Jon is talking about. The 15" rMBP starts with i7 2.6 GHz. That is the base version.

Dude what are you talking about ? Don't confuse others and everyone else here !!

 

[Ploki]

There is no option for i7 2.2 GHz on the 15" rMBP which Jon is talking about. The 15" rMBP starts with i7 2.6 GHz. That is the base version.

Dude what are you talking about ? Don't confuse others and everyone else here !!

No, there are 3 six-core 15" models.
i7 8750H 2,2 GHz
i7 8850H 2,6 GHz
i9 8950HK 2,9 GHz

 

[M.Rizk]

There is no option for i7 2.2 GHz on the 15" rMBP which Jon is talking about. The 15" rMBP starts with i7 2.6 GHz. That is the base version.

Dude what are you talking about ? Don't confuse others and everyone else here !!

?!?!?!?!?!

 

[augustya]

Oh Ok !! My Bad....

But he hasn't mentioned anything specifically, He just says Mid-Tier 6 Core model, He does not say 2.2GHz i7 or anything higher ? So What does Mid-Tier 6 Core Model mean ?

 

[OC40]

Oh Ok !! My Bad....

But he hasn't mentioned anything specifically, He just says Mid-Tier 6 Core model, He does not say 2.2GHz i7 or anything higher ? What does Mid-Tier 6 Core Model mean ?

Posting is revoked for 2.2 days.

Low tier - 2.2
Mid - 2.6
High - 2.9

 

[AndyMacAndMic]

There is no option for i7 2.2 GHz on the 15" rMBP which Jon is talking about. The 15" rMBP starts with i7 2.6 GHz. That is the base version.

Dude what are you talking about ? Don't confuse others and everyone else here !!

Yes, there is an option for an i7 2.2 GHz (low tier). The 2.6 GHz is the mid tier version!
Edit: I am a bit to late with this post. I see you already noticed. Consider this as not posted.

 

[augustya]

And does the Mid-Tier i.e. the 2.6GHz i7 Processor 6 Core Model not Throttle ?

 

[M.Rizk]

And does the Mid-Tier i.e. the 2.6GHz i7 Processor 6 Core Model not Throttle ?

They all do. 2.6 throttles just a little less than the i9.