RAM and CPU options for 2017 Macbook 12 in

[Coolio1234]

I've got a pretty specific question about the new 12 in MB with Kaby Lake processors. I've been sifting through all possible reviews and keep running into the same thing:

Does the 8gb or 16gb config make sense for these type of laptops? As someone who plans to browse the web, stream video, listen to music, use MS Office and perhaps light Photoshop or video editing in the future, I am starting to think that the 16gb option is overkill for a laptop like this. I don't plan to run any VM's, game or anything crazy like that. I'm set on getting 512gb, and the i7/m7 makes sense as well since there isn't a big difference between m5/m3. What situation would you all recommend I get the 16gb config?

 

[EugW]

For your usage, 8 GB is probably fine, but if you have a second user who doesn't log out consistently (which may add another 2 GB RAM's worth of usage) or if you do more heavy multitasking occasionally, or eventually do decide to add a VM, then 16 GB can make sense. Plus, it depends on how long you plan on keeping your machine.

I am someone with mostly similar usage in a MacBook to you. However, I am in a 4 person household, and my wife is one who annoying never logs out of anything. I haven't decided if I need to use a VM or not either. Right now I don't use a VM but due to some changes coming to my workplace's IT rules, I'm not sure if I might have to use a VM in the near future. Plus I tend to keep my laptops for much longer than most people, and as you know, memory usage creeps up with time just due to OS and software updates. My last Mac laptop was a 2009 MacBook Pro. I started with 2 GB RAM on that thing, and now I have 8 GB in it, but I had originally guessed 4 GB would have been enough for the life of the machine. Thus, I just went with 16 GB for the new 2017 MacBook, as you can't upgrade RAM in these things after the fact.

But again for your usage as described 8 GB is probably fine, at least for now.

As for the CPU, it sounds like the m3 is also all you need, but you can't get the 512 GB SSD with the m3, so it makes sense to upgrade. Personally in your shoes I'd get the 512 GB i5. Sure, the i5 isn't that much faster than the m3, but it should be noted that the 2016 m7 throttles like mad so that in some testing the 2016 m5 actually benched faster than the m7. I don't know if the same in 2017 or not, but I'd be concerned about that possibility since these are fanless machines after all.

tl;dr:

I'd recommend a 512 GB m5 with 8 GB RAM for your current described needs, but if you think your memory needs could change and you plan to keep this machine for a very long time, then you could consider a 512 GB m5 with 16 GB RAM.

 

[Coolio1234]

For your usage, 8 GB is probably fine, but if you have a second user who doesn't log out consistently (which may add another 2 GB RAM's worth of usage) or if you do more heavy multitasking occasionally, or eventually do decide to add a VM, then 16 GB can make sense. Plus, it depends on how long you plan on keeping your machine.

I am someone with mostly similar usage in a MacBook to you. However, I am in a 4 person household, and my wife is one who annoying never logs out of anything. I haven't decided if I need to use a VM or not either. Right now I don't use a VM but due to some changes coming to my workplace's IT rules, I'm not sure if I might have to use a VM in the near future. Plus I tend to keep my laptops for much longer than most people, and as you know, memory usage creeps up with time just due to OS and software updates. My last Mac laptop was a 2009 MacBook Pro. I started with 2 GB RAM on that thing, and now I have 8 GB in it, but I had originally guessed 4 GB would have been enough for the life of the machine. Thus, I just went with 16 GB for the new 2017 MacBook, as you can't upgrade RAM in these things after the fact.

But again for your usage as described 8 GB is probably fine, at least for now.

As for the CPU, it sounds like the m3 is also all you need, but you can't get the 512 GB SSD with the m3, so it makes sense to upgrade. Personally in your shoes I'd get the 512 GB i5. Sure, the i5 isn't that much faster than the m3, but it should be noted that the 2016 m7 throttles like mad so that in some testing the 2016 m5 actually benched faster than the m7. I don't know if the same in 2017 or not, but I'd be concerned about that possibility since these are fanless machines after all.

tl;dr:

I'd recommend a 512 GB m5 with 8 GB RAM for your current described needs, but if you think your memory needs could change and you plan to keep this machine for a very long time, then you could consider a 512 GB m5 with 16 GB RAM.

Thanks for the detailed response! So I plan for this to be my primary laptop for at least 4-5 years, so the 16gb option seems like it will be worth it. Now are you saying that from more of a future proof standpoint or a performance one? My concern with the 16gb of ram is that it would be wasted due to the similar reason the i7 would be; once thermal throttling kicks in the MacBook would get really slow. Am I correct on this assessment, or does having more ram not get affected by this thermal throttling aspect? Ultimately I'd like this to last a while.

And going off the i7, here's my situation; B&H has the i7/16gb/512gb ssd for sale at $1799, and the i5/16gb/512gb ssd for sale at $1729. The difference is only $70, and the savings is bigger on the i7. I've been scouring the internet to see if the same i7 throttling issues come up for the 2017 model, but I can't find anything. The only thing I can find is that the m-series chips overall weren't made for reaching those higher speeds, but can't confirm whether the i7 in 2017 will have those problems. In what situation would an i5 be better than i7?

 

[EugW]

Thanks for the detailed response! So I plan for this to be my primary laptop for at least 4-5 years, so the 16gb option seems like it will be worth it. Now are you saying that from more of a future proof standpoint or a performance one? My concern with the 16gb of ram is that it would be wasted due to the similar reason the i7 would be; once thermal throttling kicks in the MacBook would get really slow. Am I correct on this assessment, or does having more ram not get affected by this thermal throttling aspect? Ultimately I'd like this to last a while.

More RAM won't cause thermal throttling. And you can use a lot of RAM without taxing the CPU too much.

16 GB will be more future proof, but could also be very useful if you multitask a lot. However, if you don't multitask much and don't see your needs changing much, then maybe 8 GB is fine. 5 years is a long time though.

And going off the i7, here's my situation; B&H has the i7/16gb/512gb ssd for sale at $1799, and the i5/16gb/512gb ssd for sale at $1729. The difference is only $70, and the savings is bigger on the i7. I've been scouring the internet to see if the same i7 throttling issues come up for the 2017 model, but I can't find anything. The only thing I can find is that the m-series chips overall weren't made for reaching those higher speeds, but can't confirm whether the i7 in 2017 will have those problems. In what situation would an i5 be better than i7?

I just point to this 2016 test:

https://www.notebookcheck.net/Face-Off-Apple-MacBook-12-Core-m3-Core-m5-and-Core-m7.172046.0.html

 

[Coolio1234]

More RAM won't cause thermal throttling. And you can use a lot of RAM without taxing the CPU too much.

16 GB will be more future proof, but could also be very useful if you multitask a lot. However, if you don't multitask much and don't see your needs changing much, then maybe 8 GB is fine. 5 years is a long time though.


I just point to this 2016 test:

https://www.notebookcheck.net/Face-Off-Apple-MacBook-12-Core-m3-Core-m5-and-Core-m7.172046.0.html

View attachment 720462

Gotcha. Looks like 16gb it is!

And that test.. wow that is telling. I remember seeing that be the case in the 2015 model as well.. I wonder if apple fixed it in this 2017 version. Logically speaking, this seems like more of a problem with the m series chips than it does with the MacBook. I wish I could see the same comparison for the 2017 version, if history tells the future than it looks like the i7 isn't the best choice. The only evidence that really sticks out to me that is for the i7 is that the Kaby Lake processors are a lot more efficient, but I don't know whether that is enough. Just a feeling.

 

[EugW]

Gotcha. Looks like 16gb it is!

And that test.. wow that is telling. I remember seeing that be the case in the 2015 model as well.. I wonder if apple fixed it in this 2017 version. Logically speaking, this seems like more of a problem with the m series chips than it does with the MacBook. I wish I could see the same comparison for the 2017 version, if history tells the future than it looks like the i7 isn't the best choice. The only evidence that really sticks out to me that is for the i7 is that the Kaby Lake processors are a lot more efficient, but I don't know whether that is enough. Just a feeling.

If it turns out to be a consistent problem with the i7, then perhaps more blame should go to the MacBook design, as in its design isn't appropriate for the power characteristics of the i7.

BTW, not that the chips are in any way representative of each other, but I do note I returned my Kaby Lake i7 7700K iMac because it was very, very quick to heat up, meaning that any very heavy usage would get the fans going. It still performed well and didn't usually throttle, but it was loud. I replaced the machine with an i5 7600 instead, and it's very rare to hear the fan spin up. So, despite the lower performance, I'm much happier with the machine and saved a few bucks in the process.

 

[Coolio1234]

If it turns out to be a consistent problem with the i7, then perhaps more blame should go to the MacBook design, as in its design isn't appropriate for the power characteristics of the i7.

BTW, not that the chips are in any way representative of each other, but I do note I returned my Kaby Lake i7 7700K iMac because it was very, very quick to heat up, meaning that any very heavy usage would get the fans going. It still performed well and didn't usually throttle, but it was loud. I replaced the machine with an i5 7600 instead, and it's very rare to hear the fan spin up. So, despite the lower performance, I'm much happier with the machine and saved a few bucks in the process.

I guess the only other way I can fully decide on these processors is like this:

Would the i7 be superior to the m3 in terms of overall snappiness and battery life? Note that I'm not referring to things like encoding videos and the like (aka the multi core scores on GB4), but just everyday use. The charts you posted were showing performance over a sustained load, correct?

 

[EugW]

I guess the only other way I can fully decide on these processors is like this:

Would the i7 be superior to the m3 in terms of overall snappiness and battery life? Note that I'm not referring to things like encoding videos and the like (aka the multi core scores on GB4), but just everyday use.

Dunno for sure but probably not. But you can't get the 512 GB SSD you seem to want, if you get the m3. Minimum i5.

The charts you posted were showing performance over a sustained load, correct?

Correct.

 

[Coolio1234]

Dunno for sure but probably not. But you can't get the 512 GB SSD you seem to want, if you get the m3. Minimum i5.


Correct.

Sorry, I meant i5. In reference to the deal I saw online, there's only a $70 price difference between the i5 and i7 with my particular specs. Think it's worth the upgrade or does the potential performance decrease negate the upgrade?

 

[Mike Boreham]

......once thermal throttling kicks in the MacBook would get really slow.

I hope you have picked up from the charts after you posted the above, that this is not what happens on these MacBooks.

It is a common misconception that they are fine for a few minutes before they throttle back to a crawl.

All three processors on the charts show that throttling back is more like 20% after an extended period.

And it depends what you are doing. I sometimes use mine for analysis of chess games, CPU 350% for an hour, and it hardly slows up at all, because it is a low GPU usage. See https://forums.macrumors.com/threads/common-misconception-about-cpu-throttling.2017759/

 

[EugW]

I hope you have picked up from the charts after you posted the above, that this is not what happens on these MacBooks.

It is a common misconception that they are fine for a few minutes before they throttle back to a crawl.

All three processors on the charts show that throttling back is more like 20% after an extended period.

And it depends what you are doing. I sometimes use mine for analysis of chess games, CPU 350% for an hour, and it hardly slows up at all, because it is a low GPU usage. See https://forums.macrumors.com/threads/common-misconception-about-cpu-throttling.2017759/

Actually, that graph shows something much worse. It shows that indeed, throttling on that particular m7 is really bad. With any sustained usage, it slows it down to well below i5 throttled speed. The 2016 m7 scores significantly lower than the 2016 m5 throughout the entire performance vs time curve in that graph, and it actually gets worse over time. The curve slightly diverges from the m5 curve. What you may have missed in the graph is that the middle curve is actually the m7. The top curve is the m5.

What it also shows is that particular m3 can maintain its maximum performance for far longer than the other two models.

This may not apply to other 2016 units as it's quite possible that it was an anomalous 2016 m7, but nonetheless that's what the graph shows.

BTW, here is the 2017 m3:

https://www.notebookcheck.net/Apple-MacBook-12-2017-Laptop-Review.230656.0.html

The good news is the 2017 m3 performs as fast as the 2016 m5 (which was last year's fastest). The bad news --- for this particular m3 --- is that the curve looks very different. After about 15 minutes of sustained load, throttling does indeed cause the speed fall sharply, to less than 60% of full load. But then it recovers to higher performance levels... until it falls even more sharply to extremely low performance levels briefly, running only at about 30% of peak speed.

Again, this may just be a peculiarity to this one specific unit, but things are not as simple as your testing with your own m5 suggests it may be.

 

[Mike Boreham]

Actually, that graph shows something much worse. It shows that indeed, throttling on that particular m7 is really bad. With any sustained usage, it slows it down to well below i5 throttled speed. The 2016 m7 scores significantly lower than the 2016 m5 throughout the entire performance vs time curve in that graph. What you may have missed in the graph is that the middle curve is actually the m7. The top curve is the m5.

View attachment 720656

What it also shows is that particular m3 can maintain its maximum performance for far longer than the other two models.

This may not apply to other 2016 units as it's quite possible that it was an anomalous 2016 m7, but nonetheless that's what the graph shows.

My point in my reply to Coolio1234 was a more general one about the fact that none of the processors fall off a cliff edge as thermal throttling kicks in.

About that chart, I am familiar with Notebookcheck test, and have discussed it in many threads here, probably with you!

I have felt from the start there was something anomalous about that test. I have not seen the same repeated anywhere else. I find it very difficult to believe that Apple was really shipping out M7s that were actually inferior in both initial performance and thermal degradation. Even if the M7 does have a tendency to fall off more with temperature it is not credible that it's first run should be inferior. Geekbench shows first M7 run results higher than M5.

I have no axe to grind (unlike some disappointed M7 owners) as I have an M5 and so was actually quite relieved by that test, as I had some regrets about not getting an M7. But I am still doubtful about it. For one thing my 2016 M5 will not score more than 234 on first run, well less than the M5 and M7 in that test. So I suspect the anomaly in the Notebookcheck test was actually that the M5 was too good. If the M5 line was shifted down to match my M5, it would put the M5 between the M7 and M3, where it should be. Incidentally my M5 Geekbench scores are always higher than the Geekbench average for the models dont think so I have a lemon.

Probably all shows treat benchmarks with caution!

 

[Coolio1234]

Actually, that graph shows something much worse. It shows that indeed, throttling on that particular m7 is really bad. With any sustained usage, it slows it down to well below i5 throttled speed. The 2016 m7 scores significantly lower than the 2016 m5 throughout the entire performance vs time curve in that graph, and it actually gets worse over time. The curve slightly diverges from the m5 curve. What you may have missed in the graph is that the middle curve is actually the m7. The top curve is the m5.

View attachment 720656

What it also shows is that particular m3 can maintain its maximum performance for far longer than the other two models.

This may not apply to other 2016 units as it's quite possible that it was an anomalous 2016 m7, but nonetheless that's what the graph shows.

BTW, here is the 2017 m3:

View attachment 720657

https://www.notebookcheck.net/Apple-MacBook-12-2017-Laptop-Review.230656.0.html

The good news is the 2017 m3 performs as fast as the 2016 m5 (which was last year's fastest). The bad news --- for this particular m3 --- is that the curve looks very different. After about 15 minutes of sustained load, throttling does indeed cause the speed fall sharply, to less than 60% of full load. But then it recovers to higher performance levels... until it falls even more sharply to extremely low performance levels briefly, running only at about 30% of peak speed.

Again, this may just be a peculiarity to this one specific unit, but things are not as simple as your testing with your own m5 suggests it may be.

If my understanding of processors is correct, is it possible this year's i7 is more capable of handling those small drops in performance of the m3? Only due to it having more power. Of course, this is purely hypothetical, but is it possible that the i7 has the best of both worlds this year? What I mean by saying that is on the low intensity tasks, it is snappier, but on the heavy ones it is able to minimize those drops in performance and maintain a steady (but higher overall output)?

I'm also curious to see whether that same test for the m3 can be replicated on an m7 2017 model.
[doublepost=1506255705][/doublepost]

I hope you have picked up from the charts after you posted the above, that this is not what happens on these MacBooks.

It is a common misconception that they are fine for a few minutes before they throttle back to a crawl.

All three processors on the charts show that throttling back is more like 20% after an extended period.

And it depends what you are doing. I sometimes use mine for analysis of chess games, CPU 350% for an hour, and it hardly slows up at all, because it is a low GPU usage. See https://forums.macrumors.com/threads/common-misconception-about-cpu-throttling.2017759/

This is some really encouraging news. If the throttling isn't as much of an issue as I thought, then the higher processors should benefit both light and heavy use; making the former snappier via more powerful short bursts and the latter more capable of sustaining/handling powerful tasks.

(Not to the extent of MBP's or anything, but the best of the best when it comes to MB's)

To be honest, I've searched through the threads here and couldn't find anything referencing the m7 2017 MB having those same problems. Headed to the Apple Store today, perhaps they might know a thing or two (especially when I mention that I won't be buying directly from them )

 

[EugW]

My point in my reply to Coolio1234 was a more general one about the fact that none of the processors fall off a cliff edge as thermal throttling kicks in.

About that chart, I am familiar with Notebookcheck test, and have discussed it in many threads here, probably with you!

I have felt from the start there was something anomalous about that test. I have not seen the same repeated anywhere else. I find it very difficult to believe that Apple was really shipping out M7s that were actually inferior in both initial performance and thermal degradation. Even if the M7 does have a tendency to fall off more with temperature it is not credible that it's first run should be inferior. Geekbench shows first M7 run results higher than M5.

I have no axe to grind (unlike some disappointed M7 owners) as I have an M5 and so was actually quite relieved by that test, as I had some regrets about not getting an M7. But I am still doubtful about it. For one thing my 2016 M5 will not score more than 234 on first run, well less than the M5 and M7 in that test. So I suspect the anomaly in the Notebookcheck test was actually that the M5 was too good. If the M5 line was shifted down to match my M5, it would put the M5 between the M7 and M3, where it should be. Incidentally my M5 Geekbench scores are always higher than the Geekbench average for the models dont think so I have a lemon.

Probably all shows treat benchmarks with caution!

I too have wondered about the repeatability of that test. However, 260ish for the 2016 m5 has been reported elsewhere for Cinebench R15.

https://arstechnica.com/gadgets/2017/06/mini-review-the-2017-macbook-comes-into-its-own/

The 2017 in this graph is an m3 and the 2016 is an m5. (Apple was shipping m5 review units last year, but this year they're shipping the m3. I guess Apple feels happy with the m3 this year.)

Furthermore, IIRC, 260 is about what I got for my 2017 m3.

 

[Mike Boreham]

I am familiar with that 2017 test as well and find it very difficult to imagine what could be happening at those huge dips, and why it recovers next run as if nothing had happened. That doesn't sound like a thermal effect. I suspect something else is going on. You can see the thermal effect in all these tests, as a steady decline. I think it much more likely that those dips were caused by something else happening briefly on the computer during the test, something being done by the OS, invisible to the user.

 

[EugW]

I am familiar with that 2017 test as well and find it very difficult to imagine what could be happening at those huge dips, and why it recovers next run as if nothing had happened. That doesn't sound like a thermal effect. I suspect something else is going on. You can see the thermal effect in all these tests, as a steady decline. I think it much more likely that those dips were caused by something else happening briefly on the computer during the test, something being done by the OS, invisible to the user.

Now you're just wildly speculating.

And we already know that 250-260ish is normal for 2017 m5 for Cinebench R15. If you're not getting that for your machine then perhaps you should re-evaluate your setup.

Now for some speculation on my part: Perhaps something is holding your max speed back due your install. Do you have anti-virus running? Do you have other small services running?

 

[Mike Boreham]

Of course, but there has to be an engineering explanation for test results, and I am simply saying that a sudden dip and equally sudden full recovery is not typical of a thermal effect.
[doublepost=1506258332][/doublepost]

However, 260ish for the 2016 m5 has been reported elsewhere for Cinebench R15.

Perhaps I have got a lemon after all

Just ran it again, first time for a while, 237.

 

[EugW]

I just tried this again for confirmation and I get 262 in Cinebench R15 on my 2017 m3:

 

[Mike Boreham]

And we already know that 250-260ish is normal for 2017 m5 for Cinebench R15. If you're not getting that for your machine then perhaps you should re-evaluate your setup.

Now for some speculation on my part: Perhaps something is holding your max speed back due your install. Do you have anti-virus running? Do you have other small services running?

Yes, I have lots of utilities running in the background (not AV), also Filevault, though don't know if this would affect Cinebench (it certainly hits disk r/w speeds). I just turned off the ones that were easy to turn off and got a Cinebench score 240.

BTW Techradar's 2016 M5 scored 237.

I am not too bothered by my score. I expect if I fresh installed without my useful utilities I would score more but that isn't the point is it! They are too useful to dump.

 

[EugW]

I would not suggest shutting down such services for real use of course, but keeping them on for benchmark testing and throttling testing may affect accuracy of those tests, at least when comparing against notebookcheck.

My guess is having them active will, if anything, reduce the likelihood of severe throttling as compared to say pure video encoding or repeated Cinebench.

 

[Coolio1234]

Are there any known tests for the m7 2017 MacBook similar to the ones you all are posting?

And better yet, if I were to get one would it be easy to replicate these tests so we can finally solve the mystery surrounding this thermal throttle/CPU usage on higher power processors?

 

[EugW]

Are there any known tests for the m7 2017 MacBook similar to the ones you all are posting?

And better yet, if I were to get one would it be easy to replicate these tests so we can finally solve the mystery surrounding this thermal throttle/CPU usage on higher power processors?

I haven't seen anyone test the 2017 i7 this way yet. Just some YES testing, but I'm not sure the Intel Power Gadget software for the Mac is reading the CPUs correctly, since it has not been updated since early 2016, and for every user the CPU throttles to the same 2.5/2.6 GHz regardless of which model 2017 CPU it is.

 

[Coolio1234]

Maxed it out, how do I run those tests to verify? May as well contribute to the community!

 

[Mike Boreham]

Maxed it out, how do I run those tests to verify? May as well contribute to the community!

Would be great!

Download free Cinebench from https://www.maxon.net/en/products/cinebench/

Put the whole CINEBENCH_R15 folder in your apps folder.

Launch CINEBENCH OSX app (which is in the folder)

Run the CPU test repeatedly ie click the CPU Run button as soon as previous finishes (each takes about 2 mins)

Plot or list the results.

 

[Eldar Gezalov]

Would be great!

Download free Cinebench from https://www.maxon.net/en/products/cinebench/

Put the whole CINEBENCH_R15 folder in your apps folder.

Launch CINEBENCH OSX app (which is in the folder)

Run the CPU test repeatedly ie click the CPU Run button as soon as previous finishes (each takes about 2 mins)

Plot or list the results.

Hey, I have done this test when comparing the maxed out 12-inch MacBook to 13-inch MBP a while ago. I didn’t get any throttling issues at all. Take a look at screenshot or check out this page https://eldargezalov.com/news/12-inch-macbook-vs-13-inch-macbook-pro