Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

jgbr

macrumors 6502a
Sep 14, 2007
954
1,178
It's simple, the machine gets the job done quicker in burst mode so it can throttle back quicker.
 

jamrolu

macrumors newbie
Original poster
Oct 13, 2011
24
15
Málaga, Spain
It's simple, the machine gets the job done quicker in burst mode so it can throttle back quicker.
That's not how it looked to me, it seemed the m7 ramped up significantly higher but didn't really idle much faster, definitely not twice as fast, and its power peak was easily twice as high. That isn't an overall saving.
 

jgbr

macrumors 6502a
Sep 14, 2007
954
1,178
and that means quibbling over probably seconds or minutes of battery life here.

The m3 is spiking or going into action quite reguarly wheras the the m7 deals with a task and then reverts to stock speed for most tasks.
 
Last edited:

KPOM

macrumors P6
Oct 23, 2010
18,307
8,319
My readings from IPG are .3W, 1.2 Ghz, and 33C. Not sure why the frequency is at 1.2. Anyone else know?

My guess is that Intel Power Gadget wasn't fully updated for Skylake-M and that it is giving false readings, particularly since the m3 (which has a base of 1.1GHz) and m7 (which has a base of 1.3GHz) both show 1.2GHz when idle. The program isn't really intended for end-users and isn't updated all that often by Intel.
 
  • Like
Reactions: Trey M

izzyfanto

macrumors regular
Nov 22, 2011
233
102
My guess is that Intel Power Gadget wasn't fully updated for Skylake-M and that it is giving false readings, particularly since the m3 (which has a base of 1.1GHz) and m7 (which has a base of 1.3GHz) both show 1.2GHz when idle. The program isn't really intended for end-users and isn't updated all that often by Intel.
Makes sense to me, thanks. Any ideas to an IPG alternative?
 

curmudgeonette

macrumors 6502a
Jan 28, 2016
586
496
California
Wouldn't a higher clock pull more wattage though?

Yes, but in theory it would be for a shorter time. This should let the system enter a power down state earlier.

In a modern CMOS processor, most of the power consumption goes to overcoming capacitance. Thus, in theory, toggling a signal a million time takes the same energy whether it be at 1GHz or 1MHz. However, in the latter, other parts of the system need to be powered up for an extra 999mS.

My take on what is happening: It looks like the m7 is turbo-ing up to a higher power draw, but then running into system delays (SSD access time, memory bandwidth, GPU processing, I/O delays, and so on) that keep it from completing tasks proportionally quicker. Thus it is idling at a higher clock rate, i.e. power draw. In the end, the task wasn't completed much quicker, but more power was used. Perhaps this is why all speeds of the Broadwell core-m chips were listed at the same price - and the same for the Skylake m3 and m5.

While the OP still has both computers, it would be useful to compare the responsiveness of both. Try typing at the start of long word processor document. Or change cells in a complex spreadsheet where many other cells need recalculating. In short, try tasks that take less than a second. See which computer feels faster.
 

6749974

Cancelled
Mar 19, 2005
959
963
While the OP still has both computers, it would be useful to compare the responsiveness of both. Try typing at the start of long word processor document. Or change cells in a complex spreadsheet where many other cells need recalculating. In short, try tasks that take less than a second. See which computer feels faster.
If rMB is someone's second computer, where as the first computer does the job of the workhorse (eg. 15" MacBook Pro) then this question is what matters first, I would think; Which one feels faster when just moving around various documents (as opposed to exports, renders, and such).
[doublepost=1463333150][/doublepost]I have a series of questions for the more layman (like me):

Q1: Which one feels faster, between the m3 or m7?

  • Browsing typical website (eg. News, 5 tabs open)
  • Watching YouTube Videos
  • Watching Netflix
  • Using Apple Notes
  • Microsoft Word (eg. 30 page doc)
  • Apple Mail
  • iTunes (browsing and playing music)
  • Skype video calls

Q2: In other words: which one is better for both content consumption (to replace iPad Air) and typical productivity (Mail, Browsing, Docs)

I would imagine that there is no discernible difference between m3 and m7 when it comes to "feel" (speed) and battery life. The only difference is in cost, I could imagine, in which case m3 becomes the best featured option because it "gives you" $250 back.

Am I wrong in that thinking?

Q3: If m7 is a better option, list the tasks in which m7 makes a discernible difference. (In other words, who is being rational in spending an extra $250, and what are they using the m7 for?)



 

izzyfanto

macrumors regular
Nov 22, 2011
233
102
Yes, but in theory it would be for a shorter time. This should let the system enter a power down state earlier. In a modern CMOS processor, most of the power consumption goes to overcoming capacitance. Thus, in theory, toggling a signal a million time takes the same energy whether it be at 1GHz or 1MHz. However, in the latter, other parts of the system need to be powered up for an extra 999mS.

My take on what is happening: It looks like the m7 is turbo-ing up to a higher power draw, but then running into system delays (SSD access time, memory bandwidth, GPU processing, I/O delays, and so on) that keep it from completing tasks proportionally quicker. Thus it is idling at a higher clock rate, i.e. power draw. In the end, the task wasn't completed much quicker, but more power was used. Perhaps this is why all speeds of the Broadwell core-m chips were listed at the same price - and the same for the Skylake m3 and m5.
That first part makes sense to me, yet I have a question: If the m7 takes more power to get a task completed and then rests, why wouldn't the m3 plodding behind with less power but a little longer, ESSENTIALLY equal out?

The second, are you basically saying that the m7 chips are too fast for the SSD, RAM, GPU & I/O? If that is what you're saying, how come i7 quad cores don't have that problem?

Thanks for the dialogue.
 

curmudgeonette

macrumors 6502a
Jan 28, 2016
586
496
California
That first part makes sense to me, yet I have a question: If the m7 takes more power to get a task completed and then rests, why wouldn't the m3 plodding behind with less power but a little longer, ESSENTIALLY equal out?

In theory it should!

The second, are you basically saying that the m7 chips are too fast for the SSD, RAM, GPU & I/O? If that is what you're saying, how come i7 quad cores don't have that problem?

The tests that the OP chose showed the m3 and m7 completing the tasks at about the same time. Why would that be? Perhaps other parts of the system didn't let them finish faster, causing the m7 to twiddle its idle loops at the higher power draw. We know the SSD and RAM aren't faster in the m7 version. The m7 GPU can do calculations a bit faster, but pixel shoving is limited by the same RAM speed.

The quad cores in retina laptops all have an Iris Pro GPU, i.e. they have the large L4 cache. It's obvious Apple feels this cache is necessary to make up for the memory bandwidth eaten by running a large retina display. Only the rMB and rMBP13 run a retina display without an L4 cache, but both also have only dual core CPUs.

The memory intensive benchmarks between the 2015 rMB and 2016 rMB show a huge speedup, far more than the simple ratio between 1600 and 1866 memory. One explanation is that the iGPU is using a significant portion of the bandwidth. In that case a modest memory speed improvement pays off doubly - the iGPU uses a lesser percentage, and the available portion is faster too.

So - it looks to me like the m7 is simply too fast relative to the rest of the system for its speed to be effectively harnessed.
 
  • Like
Reactions: 6749974

neo_cs193p

macrumors regular
May 17, 2016
242
291
I'd like to know if the battery life of the M7 is longer than that of the M3. If you still have both computers, could you have them do the same tasks with full charges, same screen brightness, same sound volume etc. and record how long they last?
 

mackie8

Suspended
Nov 6, 2014
898
663
i cancelled my M7/256 and settle for M3/256 ... i cant justify a slight speed bump for S$350 ...
i tried 1.1ghz 2015 and 1.3ghz 2015 before but i dont feel much speed difference.
 

izzyfanto

macrumors regular
Nov 22, 2011
233
102
i cancelled my M7/256 and settle for M3/256 ... i cant justify a slight speed bump for S$350 ...
i tried 1.1ghz 2015 and 1.3ghz 2015 before but i dont feel much speed difference.
Good choice, my reasoning as well.
 

deeddawg

macrumors G5
Jun 14, 2010
12,466
6,563
US
Surely this is still a matter of long term projection of the value of the comptuer to handle future tasks and OS? Longetivity performance wise over the period of apple care would be better with an m7?

From a compute standpoint, my mid-2011 core i5 1.7 MBA13 was still working fine for me when I bought my rMB a couple months ago. That was the base CPU for that model. I don't see where the Core i7 1.8 would have made it last any longer for me than the five years the core i5 1.7 did.

My upgrade was prompted by a need for more internal storage as well as the better screen, lighter weight, and smaller size. System performance was never a concern (in my usage); any money I'd have spent on the core i7 would have been effectively wasted.
 
Last edited:
  • Like
Reactions: 6749974

Queen6

macrumors G4
From a compute standpoint, my mid-2011 core i5 1.7 MBA13 was still working fine for me when I bought my rMB a couple months ago. That was the base CPU for that model. I don't see where the Core i7 1.8 would have made it last any longer for me than the five years the core i5 1.7 did.

My upgrade was prompted by a need for more internal storage as well as the better screen, lighter weight, and smaller size. System performance was never a concern (in my usage); any money I'd have spent on the core i7 would have been effectively wasted.

Exactly, unless your usage/workflow is pegging the CPU 100% (then rMB is the wrong machine) your wasting your money on such upgrades, unless you can monetarize the performance gain. The one that really benefits is Apple as they notoriously mark up in-house upgrades to significantly boost product margins.

Intel & Apple play the same game, bigger numbers are better for bar room computing :)

Q-6
 

Trey M

macrumors 6502a
Jul 25, 2011
964
331
USA
I generally agree with the responses on this page. You're not buying the rMB in the first place to be any kind of workhorse. Of course it does the essential tasks and performs smoothly (MS Office, all browsers, more basic video editing, etc.), but I can definitely tell there's a significant gap between my rMB and rMBP, which is now 4 years old. I certainly feel the throttling when working with lots of applications. That's not going to change between an m3 or an m7. Of course it's nice to know the m7 is helping out in some situations, but the real world difference doesn't really justify the cost. There's the appeal of having a 'maxed out' model, but that really doesn't mean squat at the end of the day.

Frankly I got mine for a steal on eBay, otherwise I would've bought the m5 for the slight speed bump and SSD upgrade (I need 512). I could justify the m5 upgrade because it upgrades both the CPU and SSD, however I would personally never recommend the m7/256 model. You can really spend that extra ~$300 on something much more worthwhile for your computer (USB-C adapter, wireless mouse, Samsung T3, etc.)
 
  • Like
Reactions: 6749974

touchstoned

macrumors regular
Apr 4, 2015
173
141
Hermosa Beach, CA
I had the 2015 1.1/2.4 Ghz base model, and upgraded to the 1.3Ghz/2.9Ghz 512GB 2015 fully equipped model. My reasoning in not purchasing the newer 2016 model was: The 2016 1.3Ghz Model priced out to 1990$ with tax, while the 1.3Ghz 2015 model was $500 less. So for $500, I would've gotten an extra 2-4 PCIe Lanes, slightly faster ram, and 200mhz in boost clock. Oh, and the extra hour of battery life. Not worth the money IMO, but who knows... maybe scoring 3000/6500 in Geekbench vs 2800/5500 is worth it if you encode a ton of video files or something.
 

6749974

Cancelled
Mar 19, 2005
959
963
Does anybody know what the GPU differences are between
  1. 2015 vs 2016 models
  2. 2016 m3 vs 2016 m7
In other words, should I get the 2016 because the GPU is significantly snappier, and should I get the m7 because it has a snappier GPU than m3?
 

touchstoned

macrumors regular
Apr 4, 2015
173
141
Hermosa Beach, CA
If you think 20-30% increase in GPU performance (with a reduction in # of supported games, because skylake drivers are less mature than broadwell) is important, then by all means go 2016. I actually think the only real difference you'll see is in battery life, especially if you go with the 2016 M3.

I have the 2015 "M7" version and it's plenty quick for what I use it for, which is web browsing and work. If I played games, I probably would've gone 2016 M7

You should also consider than an xbox one or PS4 is like $300, and the upgrade from a 2015 M7 to a 2016 M7 is nearly $500
 
  • Like
Reactions: 6749974

6749974

Cancelled
Mar 19, 2005
959
963
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.