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

Geekbench Results 2017 m3 i5 i7

Ynk said:
There are a few results available for the i7 now, tossed in the 2015 m3 as well as that's what I'm upgrading from:

i7 4256/8364
i5 4058/7276
m3 3875/7062
2015 m3 2730/4891

i7 Open CL compute 19316
2015 m3 Open CL compute 11809

Full table: https://browser.primatelabs.com/v4/cpu/search?utf8=✓&q=Macbook10,1
Click to expand...


Just wow! Been dreaming about a macbook because of its lightness for a while, but haven't been able to justify the purcase yet. With these results, as soon as it gets tb3, I'm bying an i7 Macbook :D

Kind regards,
Owner of base nTB
 
  • Like
Reactions: borntrouble
Looking at the single-thread numbers, there's roughly a 5% jump between each model. Even going from m3 to i7 is only 10%, which is barely (if at all) noticeable. Save your money for something else.
 
  • Like
Reactions: lowkey
As expected, the (at least multi-core) score is significantly higher with the i7 but smaller with the i5. Again, they should just go with the i5 for the baser model. Stop confusion while also giving users power. They can still charged for a higher clocked i5. And they're effectively the same price for Apple.
 
Appleaker said:
As expected, the (at least multi-core) score is significantly higher with the i7 but smaller with the i5. Again, they should just go with the i5 for the baser model. Stop confusion while also giving users power. They can still charged for a higher clocked i5. And they're effectively the same price for Apple.
Click to expand...
Will you keep posting this in every.. single.. topic? :)
 
  • Like
Reactions: TripleMoxy
maerz001 said:
MacBook Pro (15-inch Late 2016) Intel Core i7-6920HQ @ 2.9 GHz (4 cores)
4287

so single core i7 MB is on par with the i7 MBP. Hard to believe
Click to expand...

If it is hard to believe, that's because it's false.

Guys, take Geekbench with a grain of healthy salt. It engages the Accelerate framework, which kicks the dedicated hardware accelerators (including the GPU) into taking over for the CPU for certain tasks.

Kaby Lake does include a new video accelerator for both decoding and encoding, and as Geekbench seemingly makes use of this, scores will be inflated for those tests.
 
  • Like
Reactions: maerz001
C64 said:
Will you keep posting this in every.. single.. topic? :)
Click to expand...
Yes because I feel so strongly about it. Just as strong as the branding is on the new Y-series i5 and i7 processors. Yet another reason why i5 should be in the base model! :p
 
bill-p said:
If it is hard to believe, that's because it's false.

Guys, take Geekbench with a grain of healthy salt. It engages the Accelerate framework, which kicks the dedicated hardware accelerators (including the GPU) into taking over for the CPU for certain tasks.

Kaby Lake does include a new video accelerator for both decoding and encoding, and as Geekbench seemingly makes use of this, scores will be inflated for those tests.
Click to expand...

This is false.
- GPU Is not used
- There are no video encoding or decoding benchmarks in geekbench.

I personally believe there should be a video encoding benchmark that uses the ffmeg library - it would make a great CPU benchmark. Applications like handbrake and many others make us of this library.
 
Last edited:
thunng8 said:
This is false.
- GPU Is not used
- There are no video encoding or decoding benchmarks in geekbench.

I personally believe there should be a video encoding benchmark that uses the ffmeg library - it would make a great CPU benchmark. Applications like handbrake and many others make us of this library.
Click to expand...

So then Geekbench is flawed. Plain and simple. It seems to favor certain types of tasks over other more objective measurements.

This has been shown before, too:
http://www.pcworld.com/article/3006...he-ipad-pro-really-isnt-as-fast-a-laptop.html

Again, what Geekbench shows does NOT correlate to real world performance.
 
I'm not surprised the rMB CPUs are hitting high scores as the chips can turbo pretty far. What would be interesting to know is how long they can sustain these speeds. Most people will only need these speeds for short periods (opening/installing apps, editing photos etc) and these chips will be absolutely fine. For people interested in video encoding, there are still questions about performance over longer time periods.
 
  • Like
Reactions: puh
kiranmk2 said:
I'm not surprised the rMB CPUs are hitting high scores as the chips can turbo pretty far. What would be interesting to know is how long they can sustain these speeds. Most people will only need these speeds for short periods (opening/installing apps, editing photos etc) and these chips will be absolutely fine. For people interested in video encoding, there are still questions about performance over longer time periods.
Click to expand...

For CPU intensive like multicore continuous Chess analysis it will run at just below the turboboost peak indefinitely as far as I can see. This thread has Intel Power Gadget screenshots after an hour at 350% CPU...it just seemed in a steady state mode. As pointed out in the thread tasks with more graphics load would change this.
 
bill-p said:
So then Geekbench is flawed. Plain and simple. It seems to favor certain types of tasks over other more objective measurements.

This has been shown before, too:
http://www.pcworld.com/article/3006...he-ipad-pro-really-isnt-as-fast-a-laptop.html

Again, what Geekbench shows does NOT correlate to real world performance.
Click to expand...
And somehow you think tabletmark correlates?

It is an obscure benchmark that no one mentions and the only CPU manufacturer that is a member of the group is Intel. AMD left the group because it believes Bapco is biased towards Intel. Apple, Qualcomm and Arm have not joined.
 
So anything that shows that the iPad is actually not as fast suddenly no longer "matters"?

Way to debate, good sir.
 
Geekbench is valid for what it’s used to show. However, it’s still a rediculous benchmark to use to make a purchasing decision. A lot of people will base their choice on its results, without realizing that it’s a nonsense indicator of real world performance, as the MB will throttle due to temperature long before the MacBook Pro will (passive vs active cooling.) The geekbench tests do nothing to test long running workloads in that regard.

Takeaway from this is if you’re only going to be a casual user, with small bursts of CPU activity, then the MB will be fine. This would include tasks like listening to music, watching videos, photo editing, web browsing, word processing, presentations, and emails. If you know you’re going to be using the CPU heavily for extended periods (ie: handbrake encoding as just one example,) then it’s most likely not the machine for you ...or at least the MBP would be much better option, and a desktop even better.
 
Last edited:
duervo said:
Geekbench is valid for what it’s used to show. However, it’s still a rediculous benchmark to use to make a purchasing decision. A lot of people will base their choice on its results, without realizing that it’s a nonsense indicator of real world performance, as the MB will throttle due to temperature long before the MacBook Pro will (passive vs active cooling.) The geekbench tests do nothing to test long running workloads in that regard.

Takeaway from this is if you’re only going to be a casual user, with small bursts of CPU activity, then the MB will be fine. This would include tasks like listening to music, watching videos, photo editing, web browsing, word processing, presentations, and emails. If you know you’re going to be using the CPU heavily for extended periods (ie: handbrake encoding as just one example,) then it’s most likely not the machine for you ...or at least the MBP would be much better option, and a desktop even better.
Click to expand...

Did you have a look at post #18 above and the link in that?
My M5 will run indefinitely at 350% CPU throttled back from the peak overboots 2.4 to 2.2.
GPU seems to be what is more likely to cause throttle back, but I have only seen reports not hard data.
 
Mike Boreham said:
Did you have a look at post #18 above and the link in that?
My M5 will run indefinitely at 350% CPU throttled back from the peak overboots 2.4 to 2.2.
GPU seems to be what is more likely to cause throttle back, but I have only seen reports not hard data.
Click to expand...

Before we go any further, we should come to an understanding about terminology. When I state “CPU”, I am including the integrated GPU in that CPU term, as it’s all in the same package. This is also why I mentioned handbrake video encoding as one example where the MB would likely not hold its own against a MBP.
 
duervo said:
Before we go any further, we should come to an understanding about terminology. When I state “CPU”, I am including the integrated GPU in that CPU term, as it’s all in the same package. This is also why I mentioned handbrake video encoding as one example where the MB would likely not hold its own against a MBP.
Click to expand...

Indeed terminology is important.

I don't normally use Handbrake on my MB but just did a 47 minute encode. Looks like the CPU/GPU throttles back to about 1.8. This screenshot was near the end. Note the temperature is only 80 degC so not sure why it has throttled back. Throttle back from 2.4 to 1.8 after 45 minutes is not as drastic as often implied. It dropped fairly uniformly during this period.

I am not claiming the MB is a fast machine (my quad core 15" did the same encode in 16 mins)....only that it doesn't throttle back as much as often implied. Certainly won't hold its own against an MBP, although it would be interesting to know how long a dual core 13"MBP would have taken for this same encode.

Nor do I claim to understand as much about the subject as the experts here....just making simple user observations and measurements.

Screen Shot 2017-06-14 at 07.20.18.png
 
Last edited:
The following values are the median (I think it is a better reference than the average) of all the current data from Primate Labs' Geekbench Browser:

Benchmark 4
Single-Core Score Multi-Core Score​
MacBook10,1 Intel Core m3-7Y32 1200 MHz (2 cores) 3655 6856.5
MacBook10,1 Intel Core i5-7Y54 1300 MHz (2 cores) 3726 7239
MacBook10,1 Intel Core i7-7Y75 1400 MHz (2 cores) 3776 7551

Percentage difference between each CPU
MacBook10,1 Intel Core m3-7Y32 1200 MHz (2 cores) 3655 6856.5
MacBook10,1 Intel Core i5-7Y54 1300 MHz (2 cores) 3726 7239 +1.94% +5.58%
MacBook10,1 Intel Core i7-7Y75 1400 MHz (2 cores) 3776 7551 +1.34% +4.31%

Percentage difference between the less powerful (m3) and the most powerful (i7)
MacBook10,1 Intel Core m3-7Y32 1200 MHz (2 cores) 3655 6856.5
MacBook10,1 Intel Core i7-7Y75 1400 MHz (2 cores) 3776 7551 +3.31% +10.13%

Benchmark 3 (I would not trust Benchmark 3 because there is not enough data at this time)
MacBook10,1 Intel Core m3-7Y32 1200 MHz (2 cores) 2908.5 5807
MacBook10,1 Intel Core m3-7Y32 1200 MHz (2 cores) 2918 6040
MacBook10,1 Intel Core i7-7Y75 1400 MHz (2 cores) 3147 6781

I am very surprised that there is only 3,31% difference in Single-Core and just 10,13% in Multi-Core between the m3 and i7. I have been buying i7 since 2011 but I wonder if is worth to buy an i7 over an m3 in this case, not just because of the price (I am willing to pay the difference) but because of the fan noise (is there actually a fan on the i5 and i7?) and battery life..
 
Last edited:
  • Like
Reactions: EugW
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back