[Very Possibly Solved] Why the Macbook Pro 2016 is slower than 2015?

[leman]

Leman man, i'm not suggesting anything! This post is to explain why it's slow. Give an answer to the question only.

The problem is how you arrive at your evidence. Don't forget that the sources you take compare desktop RAM with different performance characteristics. I think that we should be very careful in extrapolating from there. For instance, here are some other results:

https://browser.primatelabs.com/v4/cpu/1077988
https://browser.primatelabs.com/v4/cpu/794505

One is the 2016 MBP (16Gb LPDDR3 2133), the other one is Dell XPS 15 (16GB DDR3 2133), using the same CPU. I am using Geekbench (even though I am very sceptical about it), as we don't have anything else to compare. I also picked two results from the top of the respective list (basically top results for each laptop/OS). Note that the MBP scores much better in RAM bandwidth. So what is the truth here? Is LPDDR3 faster after all? I honestly don't know. To find it out one needs to take a MBP and a comparable DDR4 laptop (e.g. the XPS 15), sit down, and do some careful memory testing. Everything else is just idle speculation.

 

[Charlesje]

Leman man, i'm not suggesting anything! This post is to explain why it's slow. Give an answer to the question only.

The title of this thread is quite misleading. The question you really ask is why the geekbench multicore results are lower than the previous generation. This should better be indicated as so. There is no fixed correlation between geekbench and actual performance, certainly not in mobile devices. Apart from geekbench, other tests point to the opposite. It's interesting however to dig deeper into the moperations/operations result.

Leman is right. The graph you provided shows the opposite of your statement.

We should be carefull in jumping to conclusions. Otherwise, what's the use of these conversations?

 

[Kris.K]

Ahhh got it.

Yeah I see where you're coming from now. Yeah we do need more proper benchmarks to conclude and what I asked for in the original post.

Yes I agree with what you're saying man and yes it is idle speculation. Until we get better benchmarks this isn't going anywhere.

And Samuelsan2001 you seem like a happy bloke lol. I'll leave you to your little bubble of happiness.

 

[Charlesje]

For your interest, I just did the Cinebench CPU multicore benchmark on the new 15-inch, 2.9 Ghz macbook pro and it returned a 744 score which is significantly higher when compared to last years macbook pro score (I found a result of 577 for the 2015 mid-tier, and 641 for the 2.8 Ghz model). I think Cinebench is one of most accurate tests for realworld cpu performance (and it's widely used for comparison). As expected, the new MBP doesn't seem to suffer from the throttling issues the latest generations had in real world use. This is also confirmed in a Logic Pro x benchmark (between 15 and 20% increase towards last gen's fastest model).

More real world performance tests should be brought to the table, but it seems that Skylake in notebooks, due to its much lower TDP, brings a significant improvement for CPU oriented users (DAW's, statistical modeling etc).

 

[was2be]

Today compared my Late 2013 MBP (2.8 GHz i7, 16 MB RAM, NVIDIA GeForce GT 750M 2048 MB, Intel Iris Pro 1536 MB, 1 TB Flash), with my new, out-of-the-box-yesterday 2016 MBP (2.9 HGz i7, 16 GB RAM, Intel HD Graphics 530 / Radeon Pro 460 graphics, 1 TB Flash drive). In other words, both MBPs were top-of-the-line MBPs when originally purchased (albeit with the new one with only a 1 TB drive). Since the new computer was set up from the old one, they are mirror images of each other in terms of software installed, etc., both running latest mac os.

Following restarting each computer, and with no other applications running, I ran some computationally heavy simulations (in NEURON), the new MBP was 7 to 15% SLOWER than my 3 year old MBP. This is really very disappointing. I feel like I was suckered for > $3K. What a bummer.

 

[was2be]

Today compared my Late 2013 MBP (2.8 GHz i7, 16 MB RAM, NVIDIA GeForce GT 750M 2048 MB, Intel Iris Pro 1536 MB, 1 TB Flash), with my new, out-of-the-box-yesterday 2016 MBP (2.9 HGz i7, 16 GB RAM, Intel HD Graphics 530 / Radeon Pro 460 graphics, 1 TB Flash drive). In other words, both MBPs were top-of-the-line MBPs when originally purchased (albeit with the new one with only a 1 TB drive). Since the new computer was set up from the old one, they are mirror images of each other in terms of software installed, etc., both running latest mac os.

Following restarting each computer, and with no other applications running, I ran some computationally heavy simulations (in NEURON), the new MBP was 7 to 15% SLOWER than my 3 year old MBP. This is really very disappointing. I feel like I was suckered for > $3K. What a bummer.

 

[Immiddx]

Leman man, i'm not suggesting anything!

Stop copping out.

 

[monkeydax]

Today compared my Late 2013 MBP (2.8 GHz i7, 16 MB RAM, NVIDIA GeForce GT 750M 2048 MB, Intel Iris Pro 1536 MB, 1 TB Flash), with my new, out-of-the-box-yesterday 2016 MBP (2.9 HGz i7, 16 GB RAM, Intel HD Graphics 530 / Radeon Pro 460 graphics, 1 TB Flash drive). In other words, both MBPs were top-of-the-line MBPs when originally purchased (albeit with the new one with only a 1 TB drive). Since the new computer was set up from the old one, they are mirror images of each other in terms of software installed, etc., both running latest mac os.

Following restarting each computer, and with no other applications running, I ran some computationally heavy simulations (in NEURON), the new MBP was 7 to 15% SLOWER than my 3 year old MBP. This is really very disappointing. I feel like I was suckered for > $3K. What a bummer.

How did you arrive at those relative numbers of 7-15%? Are the results consistent across any other applications?

 

[linguist]

seems like the 2016" more aggressive power draw limiting in action, not all 100% cpu workload draw same power.

result would be inconsistent across different application, because power draw limit is different from previous generation.

if it runs much cooler and less noisy as a result, I say is good tradeoff, noise and heat can only go so far for a given design package.

 

[Kris.K]

Stop copping out.

The hero has arrived.

 

[Immiddx]

The hero has arrived.

Sarcastic response instead of defending your points. What a huge surprise.

 

[evec]

For your interest, I just did the Cinebench CPU multicore benchmark on the new 15-inch, 2.9 Ghz macbook pro and it returned a 744 score which is significantly higher when compared to last years macbook pro score (I found a result of 577 for the 2015 mid-tier, and 641 for the 2.8 Ghz model). I think Cinebench is one of most accurate tests for realworld cpu performance (and it's widely used for comparison). As expected, the new MBP doesn't seem to suffer from the throttling issues the latest generations had in real world use. This is also confirmed in a Logic Pro x benchmark (between 15 and 20% increase towards last gen's fastest model).

More real world performance tests should be brought to the table, but it seems that Skylake in notebooks, due to its much lower TDP, brings a significant improvement for CPU oriented users (DAW's, statistical modeling etc).

Intel try to optimum famous benchmark rather that real world performance and their turbo is design for make benchmark high.

Such as some non famous benchmark for example Xcode benchmark may tell you another story:
https://github.com/ashfurrow/xcode-hardware-performance

 

[MH01]

I just wanted to say thank you for the effort you went to to put this together, we need more of this on MR. I'll install windows on my maxed out 2015 model and get your the memory info

 

[Charlesje]

Intel try to optimum famous benchmark rather that real world performance and their turbo is design for make benchmark high.

Such as some non famous benchmark for example Xcode benchmark may tell you another story:
https://github.com/ashfurrow/xcode-hardware-performance

Thanks for the information but I tested the performance in various real world apps for music production use as logic x, cubase and ableton live. The performance gain is consistent with the cinebench results. Cinebench is simple polygon mathematics so i don't really see how intel would 'specifically' try to influence cinebench results but i'm still curious to learn.

I run intel power gadget at the same time and found that under heavy load, this processor keeps working at 3,4 ghz so doesn't throttle. I ran consecutive tests for 30 minutes to confirm this. This is probably the difference to last years 2,8 model which speedsteps to roughly 3,1 when under load (after a few seconds).

It would be interesting to have info about temps and processor speedstepping on your project?

 

[jackoatmon]

people just running around saying the sky is under their feet and potatoes are an animal

there's absolutely no measure by which these machines are slower and anyone who says different is either trolling or wildly ignorant about computer hardware

 

[was2be]

How did you arrive at those relative numbers of 7-15%? Are the results consistent across any other applications?

Those numbers were based on actual time to complete a simulation (~40 to 46 min each). However, results are NOT consistent across applications or even within applications, and vary an awful lot trial-to-trial. I can say that, consistently, start-up is several seconds faster on my 2013 MBP. Also, in some simulations where the 2013 MBP seems to be working faster (as measured by estimated time to complete), when really loaded down and the fans kick in on the 2013 machine, it becomes much slower than the new MBP (this may reflect 3-years of lint stuffing up the cooling system). But otherwise the 2013 matches or beats the 2016 speed. Disappointing.

 

[Morten_Hjort]

people just running around saying the sky is under their feet and potatoes are an animal

there's absolutely no measure by which these machines are slower and anyone who says different is either trolling or wildly ignorant about computer hardware

My 2016 tMBP 13" is way slower than my old MBP 15" from 2011 at almost every task. Haven't measured anything, but its quite clear for me in my everyday use.

 

[ntnskrds]

My 2016 tMBP 13" is way slower than my old MBP 15" from 2011 at almost every task. Haven't measured anything, but its quite clear for me in my everyday use.

This is so insightful !!! You are comparing a machine with a dual-core CPU to one with a quad-core CPU and finding the dual core one to be slower !!!

 

[Morten_Hjort]

This is so insightful !!! You are comparing a machine with a dual-core CPU to one with a quad-core CPU and finding the dual core one to be slower !!!

For opening simple documents? Of course I am. Why should I not? Heard about RAM and SSD? AFAIK quad-core only kicks in for more heavy usage. If I'm wrong, sorry I ruined your day...

 

[ntnskrds]

For opening simple documents? Of course I am. Why should I not? Heard about RAM and SSD? AFAIK quad-core only kicks in for more heavy usage. If I'm wrong, sorry I ruined your day...

You haven't ruined my day, of course. But you are wrong (the part that I highlighted). Today's software (be it applications, core OS or the OS'es services) can and do benefit from more cores regardless of whether you are running something you consider "heavy" or not...

 

[Morten_Hjort]

You haven't ruined my day, of course. But you are wrong (the part that I highlighted). Today's software (be it applications, core OS or the OS'es services) can and do benefit from more cores regardless of whether you are running something you consider "heavy" or not...

Thanks for a more "normal" language and also a good answer Would saving documents in Photoshop really use that much GPU or CPU? It takes like 5x longer than the old one and I kinda expected that a new integrated GPU and all the other updates to be at least at the level of a 5 year old computer. :l

 

[ntnskrds]

Thanks for a more "normal" language and also a good answer Would saving documents in Photoshop really use that much GPU or CPU? It takes like 5x longer than the old one and I kinda expected that a new integrated GPU and all the other updates to be at least at the level of a 5 year old computer. :l

I don't know what "saving in Photoshop" means. From the language you used ("saving") it seems more like a storage-subsystem bound operation so given the speeds of the 2016 MBP SSD I would think it would be faster. But if this involves processing (like flattening the image, compositing hundreds of layers etc) the CPU and GPU time could be significant. Again, I am not familiar with Photoshop so I can't be sure but a 5x delay doesn't seem normal.

 

[Zwopple]

Thanks for a more "normal" language and also a good answer Would saving documents in Photoshop really use that much GPU or CPU? It takes like 5x longer than the old one and I kinda expected that a new integrated GPU and all the other updates to be at least at the level of a 5 year old computer. :l

Photoshop will be drastically faster on a quad core and dedicated GPU over a dual core integrated GPU. File opening/saving probably involves a lot of image operations which will be explicitly able to utilise all 4 cores and the dGPU efficiently.

I'm curious if more Late 2016 15" owners could run the rough Xcode test here https://github.com/ashfurrow/xcode-hardware-performance/ because as far as the results show the 2016 is significantly slower at Xcode compilations (about 40%) which is really bad considering as a developer I compile projects a good 1000x a day.

 

[Sanpete]

Photoshop will be drastically faster on a quad core and dedicated GPU over a dual core integrated GPU. File opening/saving probably involves a lot of image operations which will be explicitly able to utilise all 4 cores and the dGPU efficiently.

I'm curious if more Late 2016 15" owners could run the rough Xcode test here https://github.com/ashfurrow/xcode-hardware-performance/ because as far as the results show the 2016 is significantly slower at Xcode compilations (about 40%) which is really bad considering as a developer I compile projects a good 1000x a day.

Hard to tell how reliable these numbers are. There are two listings for a 2016 15" 2.9 MBP with very different results. The comparative results seem at odds with those from another source, which finds the new 2.9 2016 faster than the 2.3 2012, while the github source finds the opposite.

https://hackernoon.com/on-the-macbook-pro-2016-and-why-i-returned-mine-a9d8ccf5c7b9#.lm646gvkf

So it would be interesting to see more measurements. On a related note, the 2015 15" is known to be a bit faster at some CPU tasks than the new one, though slower at others. But if you read this thread you know that.

 

[HumpYourWayUp]

What a WORTHLESSSSS THREAD. UNBEARABLE TO READ - LOL
WTF is going on?