Are we paying for inaccessible performance?

[bill-p]

I had a maxed out MBP order about to ship but I became concerned after looking at the max TDP of the components. I hope I'm missing something because what I learned seems rather absurd if true.

The adapter outputs 85W so this is the maximum power draw of the 15-in rMBP. The TDP of the 4960HQ is 47W and the TDP for the 750M is ~45W. This itself adds up to 92W and that number would be greatly exceeded if we include the motherboard, screen and other components. Consider a program that requires maximum CPU and GPU performance while you're blasting music with maximum screen brightness (probably another 10W or more) plus Wi-Fi and BlueTooth on. Let's say you're connected to an external display via HDMI, have a USB 3 drive and a Thunderbolt 2 device too. The amount of power required to do all this would probably be 2x as much as that 85W adapter is capable of outputting.

It's not the most realistic use scenario but the point stands. The only way I see the MBP doing this is by throttling down the CPU and GPU a ton. But then, why do I pay for the highest end CPU and a discrete GPU if they're running at 60% when I actually need them? How can they get away with an 85W PSU if only the graphics card and CPU TDP exceeds that?

TDP is just a measure of how much thermal buildup the cooling system (heatsink + fan or something else) needs to be able to dissipate for normal operation. I have personally had this misconception for a long time that it's power consumption, too... so it's not your fault. But in the end, it really is not power consumption.

In "some" cases, it's quite indicative of maximum power draw, but only over a certain period of time (about 1 hour). There are many other factors that contribute toward power consumption of a system in a time frame, so when you split that 1-hour period into different segments, you'll measure different things. Unless your load is just a linear distribution of pointless instructions that are just there for the sake of overloading the CPU and GPU... then you'll see linear load, but that's highly unrealistic (and pointless). Due to so many variables, measuring actual power consumption of a computer system is actually an unrealistic project to pursue, and that's why you don't see any manufacturer claiming TDP as power consumption these days. In fact, measuring just TDP is already hard enough.

That and the 85W rating is just an indication of typical load. It's not max load. You can still overload the power supply (trying to draw more than 85W), and it may still be able to deliver that just fine. As someone else has mentioned, the system might have been designed to tap into the battery when such a high load is demanded as well... so... in the end, you may still not run into a throttling situation. Unless you try to run the world's most intensive algorithm (that makes use of both the CPU and GPU) when the battery is empty, but... how often does that happen?

Most throttling situations I have seen thus far of the rMBP is due to the thermal system's inability to cope with the load. So chances are, you'd run into thermal limits long before you reach that power supply limit. Unless you're running your rMBP connected to an actively cooled surface, but then... if you have access to such a thing, I'd think you should also have access to a supercomputer in the same space.

Basically, there are so many factors that go into this that it's really hard to judge whether you are right or wrong regarding the power supply situation. But I'd wager that under most realistic and regular conditions (yes, including gaming), we wouldn't see this power supply limit you speak of.

 

[exigentsky]

TDP is just a measure of how much thermal buildup the cooling system (heatsink + fan or something else) needs to be able to dissipate for normal operation. I have personally had this misconception for a long time that it's power consumption, too... so it's not your fault. But in the end, it really is not power consumption.

In "some" cases, it's quite indicative of maximum power draw, but only over a certain period of time (about 1 hour). There are many other factors that contribute toward power consumption of a system in a time frame, so when you split that 1-hour period into different segments, you'll measure different things. Unless your load is just a linear distribution of pointless instructions that are just there for the sake of overloading the CPU and GPU... then you'll see linear load, but that's highly unrealistic (and pointless). Due to so many variables, measuring actual power consumption of a computer system is actually an unrealistic project to pursue, and that's why you don't see any manufacturer claiming TDP as power consumption these days. In fact, measuring just TDP is already hard enough.

That and the 85W rating is just an indication of typical load. It's not max load. You can still overload the power supply (trying to draw more than 85W), and it may still be able to deliver that just fine. As someone else has mentioned, the system might have been designed to tap into the battery when such a high load is demanded as well... so... in the end, you may still not run into a throttling situation. Unless you try to run the world's most intensive algorithm (that makes use of both the CPU and GPU) when the battery is empty, but... how often does that happen?

Most throttling situations I have seen thus far of the rMBP is due to the thermal system's inability to cope with the load. So chances are, you'd run into thermal limits long before you reach that power supply limit. Unless you're running your rMBP connected to an actively cooled surface, but then... if you have access to such a thing, I'd think you should also have access to a supercomputer in the same space.

Basically, there are so many factors that go into this that it's really hard to judge whether you are right or wrong regarding the power supply situation. But I'd wager that under most realistic and regular conditions (yes, including gaming), we wouldn't see this power supply limit you speak of.

Perhaps you're right but heavy throttling issues aren't unheard of in laptops from all manufacturers, Apple included. I'm eagerly awaiting detailed reviews.

As for the power adapter, that sure is odd. It seems very vague to rate it for typical use instead of the max that it can output. That's certainly not how computer power supplies are rated. I don't see what advantage they'd have going for a subjective measure like this when they could easily write exactly what the maximum is so I'm not sure.

I've also read that 2013 MBPs don't tap into the battery anymore. I actually prefer it would do that rather than sacrifice performance.

 

[pshufd]

> 4960HQ's TDP includes the GPU part. It's the entire package. If you're
> using the 750m then the Iris Pro graphics on the 4960HQ will be idle,
> drastically lowering the CPU's TDP.

You could theoretically use both if you were running graphics-intensive work on the iGPU and GPU computing on the discrete GPU.

My feeling is that you should get a desktop if you need true desktop performance. The laptop will always be a compromise of power, portability, heat and cost.

One of the nice things about the MacBook * line is the compatibility of the power bricks. If you're at college and you don't have your power brick with you, the chances are that many other students will have a power brick that you might be able to borrow. These come in 45, 65, and 85 Watts - you might not be able to find one that's at the rating that your model came with but you can get a charge or get enough power to run for a while.

 

[exigentsky]

Yeah of course, I understand that. I just want to make sure the components are giving me all the specifications state.

 

[Ryan1524]

Manufacturers are allowed to understate their specs. If it performs well beyond that, then great. But if the reverse is true, they won't be on the hook for it.

 

[phsphoenix]

I had a maxed out MBP order about to ship but I became concerned after looking at the max TDP of the components. I hope I'm missing something because what I learned seems rather absurd if true.

The adapter outputs 85W so this is the maximum power draw of the 15-in rMBP. The TDP of the 4960HQ is 47W and the TDP for the 750M is ~45W. This itself adds up to 92W and that number would be greatly exceeded if we include the motherboard, screen and other components. Consider a program that requires maximum CPU and GPU performance while you're blasting music with maximum screen brightness (probably another 10W or more) plus Wi-Fi and BlueTooth on. Let's say you're connected to an external display via HDMI, have a USB 3 drive and a Thunderbolt 2 device too. The amount of power required to do all this would probably be 2x as much as that 85W adapter is capable of outputting.

It's not the most realistic use scenario but the point stands. The only way I see the MBP doing this is by throttling down the CPU and GPU a ton. But then, why do I pay for the highest end CPU and a discrete GPU if they're running at 60% when I actually need them? How can they get away with an 85W PSU if only the graphics card and CPU TDP exceeds that?

My "stressbench" for this 750m 15" rmbp consisted of compiling around 10 major pieces of software using macports. According to iStats I hit 98 C on the processor, and that's not even stressing the GPU. I wouldn't want to run all the components at full throttle without some form of external cooling.

 

[koppie644]

TDP != Power consumption
Peak power != average power consumption

I had a maxed out MBP order about to ship but I became concerned after looking at the max TDP of the components. I hope I'm missing something because what I learned seems rather absurd if true.

The adapter outputs 85W so this is the maximum power draw of the 15-in rMBP. The TDP of the 4960HQ is 47W and the TDP for the 750M is ~45W. This itself adds up to 92W and that number would be greatly exceeded if we include the motherboard, screen and other components. Consider a program that requires maximum CPU and GPU performance while you're blasting music with maximum screen brightness (probably another 10W or more) plus Wi-Fi and BlueTooth on. Let's say you're connected to an external display via HDMI, have a USB 3 drive and a Thunderbolt 2 device too. The amount of power required to do all this would probably be 2x as much as that 85W adapter is capable of outputting.

It's not the most realistic use scenario but the point stands. The only way I see the MBP doing this is by throttling down the CPU and GPU a ton. But then, why do I pay for the highest end CPU and a discrete GPU if they're running at 60% when I actually need them? How can they get away with an 85W PSU if only the graphics card and CPU TDP exceeds that?

 

[Radiating]

Still, I question why they'd simply not use a higher wattage rated adapter instead of dipping into the battery. What would be the disadvantage?

Higher wattage adapters are very large. There is almost no realistic use scenario that would use maximum power from every component for more than a short period. The battery is nearly 100 watt hours. Meaning it can provide 33 extra watts for 3 hours, you can probably max out everything and still be fine for over 4 and a half hours. If you're doing that regularly then there is something wrong.

Apple made the choice to use a smaller power adapter so that it's 2/3rds the size.

 

[exigentsky]

Higher wattage adapters are very large. There is almost no realistic use scenario that would use maximum power from every component for more than a short period. The battery is nearly 100 watt hours. Meaning it can provide 33 extra watts for 3 hours, you can probably max out everything and still be fine for over 4 and a half hours. If you're doing that regularly then there is something wrong.

Apple made the choice to use a smaller power adapter so that it's 2/3rds the size.

Are you sure the 2013 models still tap into the battery? That is something I read may have changed.

 

[nateo200]

Was wondering about this with my Ivy Bridge early 2013 15" rMBP as well. I have an old 60watt charger for my 13" and if I start to do even modest GPU + CPU heavy stuff it stops charging and in some cases the battery is effectively a passthrough device aka it will sit their and not charge, only run on wall power and still drain the battery connected to the wall...pretty insane, I run Intels power gadget and with the dGPU on doing the math it easily exceeds 95watts let alone the 85watts the charger is rated for. Even using the regular 85watt charger charge up time comes to a crawl. So it definitely taps into the battery with the 60watt charger hooked to an 85watt machine, never seen it tap into the battery while plugged in with the 85watt and I've even tried it by doing some obsurd things like capturing video with my Intensity Shuttle thunderbolt, while exporting an After Effects project, while running 4 external hard drives (one thunderbolt HDD, one firewire HDD via thunderbolt adapter, two USB 3.0 HDD's), while rendering out H.264 from FCP X, I was at 97% battery life while doing this and the estimated full charge up time was about 28hours and 33 minutes and according to the intel power gadget the CPU was running 40watts constantly, the i7 3630QM in my machine is rated for 45watts too...dunno if 40/45watts is considered throttled but it was definitely maxed out and running at 3.3GHz...as soon as the rendering stopped and CPU wattage went back to 8-15 watts and then 5-10watts with just the capture card. Hope that answers some questions....I can do further tests easily since I run my machine hard as a video editor although I only run an Ivy Bridge early 2013 rMBP so it might be different. Haswell might have much different results as its touted as being more power efficient so YMMV. I've built PC's that exceeded their power supply units rating and can say its no good...just never figured a laptop could do so much (haven't built PC's since before 2010 but still pretty knowledgeable about all things). Anyone report issues on their 15" rMBP with tapping into the battery on the 85watt charger? Interesting stuff..

 

[Radiating]

Are you sure the 2013 models still tap into the battery? That is something I read may have changed.

Easy way to test this. Connect the 45w MacBook air charger and see if it kills benchmarks scores.

This is highly unlikely though.

 

[nateo200]

Easy way to test this. Connect the 45w MacBook air charger and see if it kills benchmarks scores.

This is highly unlikely though.

What do you mean? Like if the computer would throttle the GPU/CPU based on the power connection thus not performing as well or if the battery is tapped into while plugged into the wall like a passthrough? Just want to clarify that...I know a guy who ran a 17" MBP off a 45watt charger and he kept wondering why his battery was dieing while editing with the wall charger plugged it :O eBay seller claimed it was a 85watt charger... :O

 

[exigentsky]

Easy way to test this. Connect the 45w MacBook air charger and see if it kills benchmarks scores.

This is highly unlikely though.

I have a Macbook Air only so I can't test how its charger would affect MBPs under load if that's what you mean. Besides, the test would just show that you're using the wrong charger. Of course, it would tap into the battery then. A better test is to use a MBP's own charger and run some insane stress test benchmarking tools similar to Kombustor or Prima95 while plugged in and watching the battery.

 

[bill-p]

Perhaps you're right but heavy throttling issues aren't unheard of in laptops from all manufacturers, Apple included. I'm eagerly awaiting detailed reviews.

As for the power adapter, that sure is odd. It seems very vague to rate it for typical use instead of the max that it can output. That's certainly not how computer power supplies are rated. I don't see what advantage they'd have going for a subjective measure like this when they could easily write exactly what the maximum is so I'm not sure.

I've also read that 2013 MBPs don't tap into the battery anymore. I actually prefer it would do that rather than sacrifice performance.

Most heavy throttling on any laptop, Apple's included, are due to thermal constraints and much less due to power supply constraints.

And I think you are misinformed. Even desktop power supplies are rated for typical wattage scenarios. Max power consumption is rarely (if ever) stated on any power supply because of variations during the manufacturing process, and also because of variations in components. We want to look at electrical circuits as perfect absolute values, but they not. You always have that slight variation that actually gives a different maximum rating for each power supply.

And even disregarding that, you can "overload" any power supply. No exception. How the power supply handles the extra load above its specifications is highly dependent on how it was designed. In Apple's case, they simply tell the computer to go look for more juice from the battery.

If 2013 MacBooks don't tap into the battery, then they have to either throttle down to reduce power consumption, or overload the power supply adapter even more, which then brings the risk of an explosion.

You seem to be under the impression that "tapping into the battery" somehow causes performance to be throttled, but that can't be any further from the truth.

 

[HurryKayne]

I fear we are paying ,eventually,for inacessible Machine.
I mean Ifixit gave 1 out of 10,glue everywhere,ram soldered,display completely sealed.
This is disturbing me more than the lack of real discrete.
I hope they stepback ,at least with Pro machines.
Please.

 

[exigentsky]

Most heavy throttling on any laptop, Apple's included, are due to thermal constraints and much less due to power supply constraints.

And I think you are misinformed. Even desktop power supplies are rated for typical wattage scenarios. Max power consumption is rarely (if ever) stated on any power supply because of variations during the manufacturing process, and also because of variations in components. We want to look at electrical circuits as perfect absolute values, but they not. You always have that slight variation that actually gives a different maximum rating for each power supply.

And even disregarding that, you can "overload" any power supply. No exception. How the power supply handles the extra load above its specifications is highly dependent on how it was designed. In Apple's case, they simply tell the computer to go look for more juice from the battery.

If 2013 MacBooks don't tap into the battery, then they have to either throttle down to reduce power consumption, or overload the power supply adapter even more, which then brings the risk of an explosion.

You seem to be under the impression that "tapping into the battery" somehow causes performance to be throttled, but that can't be any further from the truth.

I don't think tapping into the battery would cause performance to be throttled unless you run out of battery or the computer gets too hot. I'd prefer it to tap into the battery rather than reduce performance (although new models may not). However, I do think it's not the best idea for a mobile workstation because it means it's not designed with intense use scenarios in mind. I don't like the idea of losing battery cycles when I'm trying to actually get work done. If all I needed it for was light tasks, then there is no need to get a MBP and I could simply get a MBA or some other much cheaper laptop.

There are other computers, like the to be released m3800 that are designed to be capable of sustaining max loads while still charging the battery and even maintaining a 2.6 turbo on all 4 cores. It can do this indefinitely with no heat issues and good temps. for the CPU. This is true for professional mobile workstations in general but most of them can't match the MBP's form factor, weight and style.

As for the power supply, it's possible that you can get a bit more juice than it's rated for but that shouldn't be counted on. For every PC I've built or seen professionally built, the power supply is rated ABOVE what the components require (measured in TDP even if it's not 100% accurate).

I think Apple simply didn't design the MBP as a desktop replacement for people who demand a lot of performance. I will wait for reviews with stress tests but my impression from previous MBPs is that they are meant to be very fast for daily tasks and with spikes of intense use but not past 20-30 minutes here and there.

 

[magbarn]

....
There are other computers, like the to be released m3800 that are designed to be capable of sustaining max loads while still charging the battery and even maintaining a 2.6 turbo on all 4 cores. It can do this indefinitely with no heat issues and good temps. for the CPU. This is true for professional mobile workstations in general but most of them can't match the MBP's form factor, weight and style.....

Has a reviewer checked out this claim? Given Dell's previous XPS throttling issues, I'd wait for an actual reviewer to take the M3800 to task. It has some pretty hot components in a small lightweight chassis so we'll see if it either throttles or sounds like a jet engine when working hard.

 

[Mathias Denichi]

So did I ^^
Guess it went like this:
Apple: Intel! Your graphics don't deliver, how are we supposed to go iGPU only?
Intel: that's your problem. You asked for this chips, now use them!
Apple: *****! They are not powerfull enough. We have to put in a dGPU too. No way we gonna pay you for this?
Intel: how about we meet in the middle - we give you a great deal and you take all of them because noone else wants em
Apple: ok. And let's try again next year

:>

^This exactly.

 

[exigentsky]

Has a reviewer checked out this claim? Given Dell's previous XPS throttling issues, I'd wait for an actual reviewer to take the M3800 to task. It has some pretty hot components in a small lightweight chassis so we'll see if it either throttles or sounds like a jet engine when working hard.

Yeah, that's the only reason I mentioned it. As far as noise, I think it will be hard for the m3800 to be noisier under load. I've only owned white Macbooks and the Macbook Air but they all got very noisy when I did chess analysis.

If you install the Intel XTU software, the max turbo clock rates are unlocked. One active core can go to 3.4Ghz, Two cores go to 3.3Ghz, Three and Four go to 3.1Ghz. You can't adjust the amount of power that the processor gets, so the higher turbo modes have to stay within Dell's 46.25 watt spec for short term turbo (around 40 seconds) and 37 watts for long term turbo. Turbo clock speeds will be adjusted down until they fall within the power limit of 37 watts on the CPU.

You won't go past 37 watts in turbo mode until you get to 4 (of 8) threads running in Prime95. Single threaded applications will stay at 3.4Ghz forever since they are no where near hitting 37 watts You will see some quick dips to 3.3Ghz as the thread is passed back and forth between processors.

WIth 8 threads running, all 4 cores are fully loaded, they run between 3.0 - 3.1Ghz for around 40 seconds, then fall back to 2.6 - 2.7Ghz. They will stay between 2.6 - 2.7Ghz with a temp of 78C until you remove the CPU load. The M3800 never allows the CPU to fall back to the base 2.2Ghz as long as a the processor is fully loaded. The fans stay quiet.

If you then add MSI Kombustor to fully load the GPU in addition to the fullly loaded CPU, several things happen. Temps on the CPU will jump from 78C to 84C followed by both fans kicking in. At this point you will hear the fans for the first time. The temp on the CPU will fall back to 81C and the temp on the GPU will climb to 69C. The CPU will lock in at 2.6Ghz and stay there. Battery charging rates will fall from 20Whr to 16Whr. The system will stay in this state for as long as you want it to.

If you stop MSI Kombuster, the temp on the CPU falls from 81C to the mid 70s. The fans also slow down. Once the CPU hits 69C, the turbo mode starts going back up to switching between 2.6 and 2.7Ghz while the fans continue to slow down. Eventually, the system settles back to the same CPU behavior as before, but the temp will stay around 70C instead of 78C. The CPU stays cooler because the fans don't completely slow down to low speed until the system is below 50C or so. While the CPU is staying right at 70C with a full load, the fans stay at medium speed.

Here is the problem with everything I just typed - this is still a pre-release machine. Dell has plenty of time to tweak the BIOS and performance curves. All of this could change by the time the M3800 ships. Maybe they could add some time to that short term turbo boost.

 

[katorga]

Problem across the board

Throttling due to heat and/or power source is very common across the board now, and most users don't realize how often intel cpu's are downgrading themselves, especially on battery power.

 

[nateo200]

Throttling due to heat and/or power source is very common across the board now, and most users don't realize how often intel cpu's are downgrading themselves, especially on battery power.

Funny you mention that, I admit I don't run on battery power that often but this thing stays cooler than any other computer I've had (that includes very large spaceous ones with plenty of cooling systems in place) and yet it seams to run faster than its rated clock speed VERY often...even on battery power...I keep intels power and processor monitor widget thingy up constantly too. Never goes over 40watts (the CPU alone) but it does hover around 2.8 to 3.2GHz allot...then again I am a power user for sure. For most peoples needs I suppose throttling is good...Ivy Bridge and Haswell definitely used better power effeciency as a marketing scheme and it definitely is a legitimate claim. I mean you don't need a quad core i7 to browse the web, hell you don't need a quad core i7 to do basic picture editing (which some people consider "intense").

 

[exigentsky]

Throttling isn't the same as a CPU underclocking itself because what you're doing doesn't require much processing power. What I'm talking about is when a CPU can't handle the load and steps down a notch due to heat or insufficient voltage.

 

[jont-fu]

Throttling isn't the same as a CPU underclocking itself because what you're doing doesn't require much processing power. What I'm talking about is when a CPU can't handle the load and steps down a notch due to heat or insufficient voltage.

AFAIK the max TDP of the processor is for situations when the turbo clocks are activated; and the 2.0/2.3/2.6 clock speed is the minimum that can be sustained over long periods. The current draw would then be smaller than the maximum.

We should wait for the Anandtech review before going too far with our speculation; I would love to do stress tests myself but lack the time and knowledge to go so far.

I can see your point of the power supply capacity being too low for sustained taxing of both the CPU and the dGPU. But it's also hard to imagine workloads that could drive both chips at max, except for a few games. And the rMBP really is not the perfect gaming laptop.

 

[koppie644]

First, TDP is not power consumption (related but very different)
Second, there is no way you run the CPU to the max, the iGPU to the max, and the dGPU to the max at the same time! (47 + 45 = 92W).
Third, the battery can kick in to help if it really happens

I had a maxed out MBP order about to ship but I became concerned after looking at the max TDP of the components. I hope I'm missing something because what I learned seems rather absurd if true.

The adapter outputs 85W so this is the maximum power draw of the 15-in rMBP. The TDP of the 4960HQ is 47W and the TDP for the 750M is ~45W. This itself adds up to 92W and that number would be greatly exceeded if we include the motherboard, screen and other components. Consider a program that requires maximum CPU and GPU performance while you're blasting music with maximum screen brightness (probably another 10W or more) plus Wi-Fi and BlueTooth on. Let's say you're connected to an external display via HDMI, have a USB 3 drive and a Thunderbolt 2 device too. The amount of power required to do all this would probably be 2x as much as that 85W adapter is capable of outputting.

It's not the most realistic use scenario but the point stands. The only way I see the MBP doing this is by throttling down the CPU and GPU a ton. But then, why do I pay for the highest end CPU and a discrete GPU if they're running at 60% when I actually need them? How can they get away with an 85W PSU if only the graphics card and CPU TDP exceeds that?

 

[ha1o2surfer]

I have done some tests on this when I got my 15 inch MBp without retina (and returned it in 5 days) If you are charging the battery and running the CPU full load it will throttle then. I find it very surprising these chose to do it this way. Whether it be the size of the adapter they wanted or what. The CPU in the macbook pros throttles anyways under a medium to heavy load scenario so it doesn't require much power anyways