Terrible performance 2017 (3.1, 2TB) vs Jan 2016 (2.8, 1TB)

[sface]

The 2017 I just received was a grand more and I expected to get at least as good performance (but obviously it should be better) as my Jan 2016 model but I'm having some serious issues and wonder if you think it's possible it's just the one I was shipped or if this may be a symptom of something else (for instance, the CPU% being used for tasks seems to be considerably lower on the 2017)?

On some very basic tasks the 2017 is much slower:

Task 1: Optimize a 2min 13sec video at 720p in iMovie
Old MacBook pro: 1 min 26 sec to complete optimization/share
New MacBook pro: 1 min 36 sec to complete optimization/share
(for some reason CPU% for the task was only 220% on the 2017 vs 300% on the 2016)

Task 2: Import a 55 min mp4 (3.78 gig) into Audacity for editing
Old MacBook pro: 19 seconds
New MacBook pro: 27 seconds
(for some reason CPU% for the task was only 70% on the 2017 vs 100% on the 2016)

Both computers were restarted before tasks were attempted and they are both running the exact same programs and have about the exact same amount of storage space % being used. I did a number of tests over and over and the results are consistently coming out like the above examples. The 2017 is clearly performing poorly in comparison to a supposedly inferior early 2016 model.

Here are the specs of the two I'm using:

Also, as mentioned in another thread, the SSD is not performing as stated by Apple (they claim
(2200 and 3100 respectively, but results are coming out considerably lower):

https://forums.macrumors.com/threads/2017-ssd-performance.2055680/

Anyone know if the culprit is simply less CPU being used by the 2017 or any other suggestions before I return it?

 

[ixxx69]

Anyone know if the culprit is simply less CPU being used by the 2017 or any other suggestions before I return it?

That is indeed kind of weird. Have you had the 2017 long enough that it's finished all the spotlight indexing, has all the updates installed, etc. Have you checked your processes list to see if something unexpected is running?

I'd suggest doing a geekbench run on each to see how it matches up with what's expected for each model.

 

[Charlesje]

Could you run cinebench r15 cpu test iteratively while measuring temps/clock? I'm having suspicions the kaby lake mbp throttles quite a bit when sustained performance is needed. There was someone witnessing this behaviour on this forum. I just saw additional YouTube videos where the 7920hq scored about 700 cb on the cinebench cpu render test (eg the 6920hq of the 2016 mbp scores up to 750 cb on the same test). It seems counterintuituve, but the skylake cpu's (mid and high tier) seem to perform better than the kaby lake cpu's due to the mbp's thermal profile.

 

[sface]

(yes, everything is fully indexed and nothing else was using CPU while doing my tests above, FYI)

I'm not very technically inclined but am doing the things mentioned so far, thank you. This is the first one (will now find where I can compare mine to what is expected):

https://browser.geekbench.com/v4/cpu/3365161

 

[Queen6]

Check the thermals, if the 2017 is throttling due to hitting the thermal envelope it will explain the difference. Sadly sometimes more is truly less...

Q-6

 

[ixxx69]

Could you run cinebench r15 cpu test iteratively while measuring temps/clock? I'm having suspicions the kaby lake mbp throttles quite a bit when sustained performance is needed. There was someone witnessing this behaviour on this forum. I just saw additional YouTube videos where the 7920hq scored about 700 cb on the cinebench cpu render test (eg the 6920hq of the 2016 mbp scores up to 750 cb on the same test). It seems counterintuituve, but the skylake cpu's (mid and high tier) seem to perform better than the kaby lake cpu's due to the mbp's thermal profile.

Not that it still couldn't be similar issue, but FTR, it's 2015 Haswell vs 2017 KBL.

I'm not very technically inclined but am doing the things mentioned so far, thank you. This is the first one (will now find where I can compare mine to what is expected):

https://browser.geekbench.com/v4/cpu/3365161

http://browser.geekbench.com/mac-benchmarks

If that's the 2017, it would appear it's doing very well. Have you tested the 2015 yet?

 

[sface]

The 2015:

https://browser.geekbench.com/v4/cpu/3365264

 

[Charlesje]

Geekbench is a synthetic benchmark, not really trustable in predicting actual performance. We saw this in the 2016 vs 2015 mbp.

Sface: you van easilu find cinebench on the web,
If you want to run the app you need to ctrl click -> open package contents, macos(sierra ? ) folder

 

[ixxx69]

Geekbench is a synthetic benchmark, not really trustable in predicting actual performance. We saw this in the 2016 vs 2015 mbp.

It's a good starting place for troubleshooting because it's super simple, takes only a few minutes, and if the numbers aren't close to what's posted, you know immediately that there is something really going wrong.

The 2015:

https://browser.geekbench.com/v4/cpu/3365264

So, you're basically getting expected results. Time to try other suggested benchmarks, and good idea to download a temp utility to check the thermals while you're benchmarking (even if it's your own tests, you can see if the CPU is topping out temp wise).

Best of luck figuring out what's going on.

 

[sface]

Here are the basic tests for the cinebench where my new computer is clearly outperforming the old (unlike my own tests). The Open GL test was 2 seconds faster on the 2017 vs the 2015. The CPU test was 14 seconds faster. So cinebench shows my new laptop as outperforming the old even though my own testing continues to show less CPU being used for my own tasks and those tasks taking longer to complete.

 

[Charlesje]

Did you try to rerun the test (about 5 If you have the time 10 times one after another) to check If the scores drop?

And you could use Intel power gadget to check temps and clock in the meantime...

It's a good starting place for troubleshooting because it's super simple, takes only a few minutes, and if the numbers aren't close to what's posted, you know immediately that there is something really going wrong.

Indeed, it s good for troubleshooting.

 

[sface]

Did you try to rerun the test (about 10 times one after another) to check If the scores drop?
[doublepost=1499630154][/doublepost]And you could use Intel power gadget to check temps and clock in the meantime...

I appreciate the patience shown to me here as I'm new and I realize it's frustrating to deal with someone who doesn't know what they're doing/the proper terminology. I'll get there. I'm not sure what I'm really looking for, but after about 10-15 tests of each CPU and Open GL, the following:

2017 Open GL: started around 81fps and ended around 72fps after around 15 tests. It incrementally gets worse but I'm not sure if that isn't normal (my 2015 also dropped small increments during the testing). It's not dropping incredible amounts (or maybe it is, I have no idea what the numbers I'm posting really mean). While running intel power gadget, the temp of computer stayed around 75 degrees during the Open GL tests.

2017 CPU: started around 730 and varied through 10 tests but didn't keep going down. Computer temp remained around 98 degrees during the tests.

The 2015 temperatures were also around 98 for CPU testing but were about 10-15 degrees higher than the 2017 for the Open GL testing. But I'm not really concerned with the 2015, other than the fact that it seems to be processing certain tasks like the ones I listed in my OP faster than my new, 2017 for some reason. Should I be looking at other variables in the Intel gadget? Thanks again.

 

[Charlesje]

The scores you have for the opengl and cpu tests seem to point to (limited) thermal throttling of the cpu and gpu. You could check the cpu frequency as well in power gadget to confirm this. It should be close to 3,7 GHz when no throttling occurs while running cinebench. Someone on macrumors mentioned a drop in clock frequency to 2,7 GHz.

 

[sface]

The scores you have for the opengl and cpu tests seem to point to (limited) thermal throttling of the cpu and gpu. You could check the cpu frequency as well in power gadget to confirm this. It should be close to 3,7 GHz when no throttling occurs while running cinebench. Someone on macrumors mentioned a drop in clock frequency to 2,7 GHz.

This is exactly what happens. Starts around 3.7, hovers around 2.7 suddenly for most of the Open GL test. My 2015 starts around 3.7 and lowers, as well, but seemingly not below around 3. So, is this the most likely reason certain things are taking longer on the 2017? CPU throttling? Is this to assure less heat in the unit?

What a disappointment (in addition to all the other major disappointments that have been mentioned dozens of times on the site, most likely). Is this something that Apple might adjust with an update or is it likely going to stay like this?

 

[jerryk]

This is exactly what happens. Starts around 3.7, hovers around 2.7 suddenly for most of the Open GL test. My 2015 starts around 3.7 and lowers, as well, but seemingly not below around 3. So, is this the most likely reason certain things are taking longer on the 2017? CPU throttling? Is this to assure less heat in the unit?

What a disappointment (in addition to all the other major disappointments that have been mentioned dozens of times on the site, most likely). Is this something that Apple might adjust with an update or is it likely going to stay like this?

My guess it is inherent in the way Intel builds the Kaby Lake. One of the changes they made was to make the Kaby Lake switch to turbo mode sooner. This is means it handles short burst of performance well, but must throttle sooner to manage temps. This usually works out in most situations because most work is controlled by user interaction, ex. launch app, and then there is a period of non activity while the user decides what to do next.

BTW, most of the price difference between the two systems in the SSD size increase. I bet in 2015 if you could have gotten a 2TB system it would have been several hundred more than what you paid for your 2017

 

[sface]

BTW, most of the price difference between the two systems in the SSD size increase.

Yeah, absolutely true, but regardless of that I was hoping to get an improvement in the overall power of the computer vs the model I bought 18 months ago. Right now, considering the issues I'm having with the programs I use most taking longer to render things/import/etc compared to my 2015, combined with lack of magsafe, a horrible keyboard that is so loud I'll be waking up guests in hotel rooms next to mine, multiple dongle needs, and the other issues mentioned on the site, it just isn't worth upgrading.

Right now the only benefit I can pinpoint in comparing the two models is the ability to use touch ID rather than typing in a long password to log in from sleep. I mean, I can't think of a single other benefit off the top of my head (for my needs), considering the revelation that the 2017 will work more slowly because of this throttling thing (if that's what is causing the problem). (I do like the space grey look and the general size of it compared to the older model)

 

[New_Mac_Smell]

Yeah, absolutely true, but regardless of that I was hoping to get an improvement in the overall power of the computer vs the model I bought 18 months ago. Right now, considering the issues I'm having with the programs I use most taking longer to render things/import/etc compared to my 2015, combined with lack of magsafe, a horrible keyboard that is so loud I'll be waking up guests in hotel rooms next to mine, multiple dongle needs, and the other issues mentioned on the site, it just isn't worth upgrading.

Right now the only benefit I can pinpoint in comparing the two models is the ability to use touch ID rather than typing in a long password to log in from sleep. I mean, I can't think of a single other benefit off the top of my head (for my needs), considering the revelation that the 2017 will work more slowly because of this throttling thing (if that's what is causing the problem). (I do like the space grey look and the general size of it compared to the older model)

It's so loud it's waking up people in another room, seriously? How hard are people typing on these things! No wonder so many are breaking if they're bashing them ham-fisted like.

I don't know how much of a leap you were expecting in performance but it's only a couple of years, 2015 Turbos up to 4Ghz, 2017 up to 4.1Ghz, so in any side by side benchmark they're going to be very similar. The 2017 is a whole 300Mhz faster under sustained load than the 2015. Again, not really sure anyone would notice 300Mhz difference.

The 2017 is faster across the board, it's just not by very much if you pick each and every component. Not sure why renders are taking longer on the 2017, I would make sure that you haven't opened it out of the box and migrated the OS then immediately tested. You'll want to make sure the 2017 has been fully updated, has completed encryption and background processing, and anything else before giving them a side by side comparison.

Apple's claims of 2000/3000 R/W are ambitious at best, and you shouldn't demand you achieve them as it's not going to happen outside of specific lab tests. They do consistently get exceptionally good speeds which is good enough for people, but transferring a 3GB file or a 3MB file is different and will result in a different speed each time you run the test.

So yeah, the processors are very similar clock speeds, the RAM is very similar core speeds, the SSD is a lot faster. Overall you should be expecting at most 20% faster, real world you won't notice anything. But overall the system should feel snappier. Just make sure everything's updated.

 

[sface]

It's so loud it's waking up people in another room, seriously? How hard are people typing on these things! No wonder so many are breaking if they're bashing them ham-fisted like.

I don't know how much of a leap you were expecting in performance but it's only a couple of years, 2015 Turbos up to 4Ghz, 2017 up to 4.1Ghz, so in any side by side benchmark they're going to be very similar. The 2017 is a whole 300Mhz faster under sustained load than the 2015. Again, not really sure anyone would notice 300Mhz difference.

The 2017 is faster across the board, it's just not by very much if you pick each and every component. Not sure why renders are taking longer on the 2017, I would make sure that you haven't opened it out of the box and migrated the OS then immediately tested. You'll want to make sure the 2017 has been fully updated, has completed encryption and background processing, and anything else before giving them a side by side comparison.

Apple's claims of 2000/3000 R/W are ambitious at best, and you shouldn't demand you achieve them as it's not going to happen outside of specific lab tests. They do consistently get exceptionally good speeds which is good enough for people, but transferring a 3GB file or a 3MB file is different and will result in a different speed each time you run the test.

So yeah, the processors are very similar clock speeds, the RAM is very similar core speeds, the SSD is a lot faster. Overall you should be expecting at most 20% faster, real world you won't notice anything. But overall the system should feel snappier. Just make sure everything's updated.

Are you a real person? The keyboard is extremely loud, no matter how soft you type (it's a glaring difference). And you're really trying to justify the slow SSD speed as something customers "shouldn't demand"... or my own render tests as having "opened it out of the box and migrated the OS then immediately tested.". Come on, man. Stop the nonsense.

 

[New_Mac_Smell]

Are you a real person? The keyboard is extremely loud, no matter how soft you type. And you're really trying to justify the slow SSD speed or my own tests as having "opened it out of the box and migrated the OS then immediately tested.". Come on, man. Stop the nonsense.

Am I real? Seriously? The keyboard isn't loud at all when I type! I genuinely have no idea how people are complaining so much, if I type hard on my old Magic Keyboard it's loud too, so I can only guess as to why so many people here complain about the newer ones. Are you genuinely claiming it is so loud it wakes people up in the next room?? Because that sounds like one huge over-statement.

And no, I don't believe I said anything about the SSD speeds being 'slow' because of migration, I was referring to render tasks. I made a note of mentioning the SSD speeds being obtained compared to 'Apples stated speeds' being about right, did you read it? I guarantee if you run BM enough you'll see those speeds that you want, just make sure you screen shot at the right moment. Or take a look at the overall average and know that files are different and the speed of processing is a very complex thing to merely look at a piece of software and claim it's 'slow'.

 

[Ries]

Am I real? Seriously? The keyboard isn't loud at all when I type! I genuinely have no idea how people are complaining so much, if I type hard on my old Magic Keyboard it's loud too, so I can only guess as to why so many people here complain about the newer ones. Are you genuinely claiming it is so loud it wakes people up in the next room?? Because that sounds like one huge over-statement.

"You" don't have any idea, why tons of "Others" have a problem. Maybe you should consider which side is the problem.

 

[New_Mac_Smell]

"You" don't have any idea, why tons of "Others" have a problem. Maybe you should consider which side is the problem.

Come on, are you seriously suggesting the keyboard is so loud it is capable of waking people up in a separate room? Seems one heck of an over-exaggeration no matter which 'side' you're on. The keyboard is as loud as you type on it, if you're heavy handed and bashing it all day long then yes it's going to make noise, as any keyboard would. You wouldn't slam a door shut then complain it's loud would you? You'd gently close it like a normal person.

 

[workerbee]

Not trying to fan up the flames any more here, but: while I would not want to sleep in a Hotel where walls are so flimsy as to let typing on the new keyboard in the next room make me wake up, I think it is pretty fair to say that those butterfly keyboards are, in fact, quite a bit noisier than the old MacBook Pro keyboards, regardless of how one is typing.

Actually, the noisy keyboard and the not-very-much-faster-overall-speeds (except for the SSD) are what make me hesitate no end about buying a TouchBar MBP and sell my trusty mid-2012 rMBP.

 

[Charlesje]

My guess it is inherent in the way Intel builds the Kaby Lake. One of the changes they made was to make the Kaby Lake switch to turbo mode sooner. This is means it handles short burst of performance well, but must throttle sooner to manage temps. This usually works out in most situations because most work is controlled by user interaction, ex. launch app, and then there is a period of non activity while the user decides what to do next.

I think Apple made the design decision to give the mbps a thermal profile that allows for a more 'average' use of it's cpus or cpus, just s it did before the 2016 mbp. This means e.g. the high tier cpu versions aren't meant to be pushed to the hardest in sustained use. The low tier model doesn't seem to refrain to throttling as seen in the notebookcheck review. Short burst performance will however be a bit faster on the high tier, but sustained performance could actually be lower and less continious (regarding clock speed) then on the low tier mbp.
The maxed out 2016 seems to be the best performer in sustained cpu use. Makes me speculate that the 2016 was actually designed with the kaby lake in mind. This could help explain its short introduction (and speculating even further, its low, not really well prepared Geekbench score?).

This is news for performance users. Developers, audio/video enthusiasts, statistical modelers, etc should really be aware.

 

[Queen6]

You should all take a look at this MR thread here:

automated-tool-to-reveal-throttling-and-overheating-github.1731178

You will see that the 15" MBP has frequently struggled with it's thermal envelope, with some models performing better than others...

Q-6

 

[whg]

Do you know about this: "Kaby Lake Macbook Pro VS 4 Windows Laptops, Way Slower", see here

The shocking result of a rendering test can be seen at about 4:40 into the video

I found this very strange, but there are many similar tests to be found on you tube.