Thoughts and Tests on the 2.2GHz MacBook Pro

Discussion in 'MacBook Pro' started by DannEboE, Aug 9, 2018.

  1. DannEboE macrumors member

    DannEboE

    Joined:
    Oct 18, 2016
    #1
    Hello all, I know we’re all sort of over the throttling issue with the new MacBook, but I thought I’d contribute my own test results considering I don’t see too many people with my configuration. More data is always better haha. I specifically got the 2.2GHz model, thinking it would be less constrained thermally than the other 2 processor models. I did upgrade the Radeon Pro 555X to a 560X for the prospect of gaming, so that may have contributed some additional heat to these tests as well. I didn’t tinker around with any of the fan controls since the few tests I ran, it performed better in the beginning, but then stabilized to be equal near the end of the rounds.

    I also waited a few days after getting my MacBook and finished installing everything to make sure Spotlight wouldn’t be indexing anything while these tests are run. During idle, CPU utilization is under 0.5% which I felt was safe enough to conduct these tests. All tests were run on MacOS.

    My MacBook Model:
    15-inch MacBook Pro - Space Gray
    Hardware:

    2.2GHz 6‑core 8th‑generation Intel Core i7 processor, Turbo Boost up to 4.1GHz
    Radeon Pro 560X with 4GB of GDDR5 memory
    16GB 2400MHz DDR4 memory
    512GB SSD storage

    My MacBook spends 80% of its time plugged into my single cable setup, so I figure I would run all the tests using that single cable setup. 1 single cable will charge the laptop, power a 4k monitor, power keyboards and mouse, and a cooling fan. It also has a hard shell case on the laptop itself.

    There were 6 main rounds of testing, round 1 is unlimited where I didn’t touch any of the power limits so unlimited, round 2 is where I limited it to 50W, round 3 is where I limited to 45W. Rounds 4,5, and 6 are the same except tests are done after the patch is installed. I had considered running the processor at 40W, but at that level, I began to saw performance lost and halted any further testing on limiting it to 40W. The room temperature is 72 F (~22 C) and there was very little moisture in the air.

    I waited until the CPU temperature would drop to 47 C before starting the next round to maintain consistency between the rounds.

    Rounds 1-3 Testing
    It was quickly evident that without any thermal limits, the processor swinged back and forth in frequency and power dramatically. Its probably due to these swings that it ended performing the worse out of the 3. Limiting the processor to 45W proved to be the best before the patch was applied. It reduced temperatures dramatically compared to the other limits, and thus, it was able to maintain a 2.9 GHz speed during the entire testing. It is probably why it could maintain those high 900 values through 15 rounds of back to back testing.

    [​IMG]

    With those tests out of the way, I then downloaded and installed the patch. The whole process took about 20 minutes.

    I then waited for the laptop to cool once again and finish indexing before conducting the next rounds.

    Rounds 4-6 Testing
    Temperatures spiked up very quickly, but it was no longer ping ponging around from high to low in frequency. Patch was wonderful in what it did for the laptop performance.

    There were a few interesting results. On average of 15 runs, limiting power to 45W still returned best result since it maintain very equal scores. Unfortunately, it seems to be doing worse than it was before the patch. For initial spikes in performance, 50W seems to be the sweet spot for performance.

    [​IMG]

    On The Go Testing
    Now that I’ve determined, at least for my laptop when plugged in, that the most optimal performance occurs when it is thermally limited to 45W, I then sought out to determine what performance of the laptop would look like if I thermally limited even further. This scenario would most likely play out when I am on the go using battery power and want to conserve energy. I wanted to compare these values from my previous laptop as well. Optimally, I would want it to be at least 1.5X faster to make the purchase seem worth it and keep it fully usable.

    For reference, my previous laptop was:
    2017 13-inch MacBook Pro - Space Gray
    Hardware:

    3.5GHz dual-core Intel Core i7, Turbo Boost up to 4.0GHz
    Intel Iris Plus Graphics 650
    16GB 2133MHz LPDDR3 on board memory
    512GB SSD storage

    It averaged about 330 on Cinebench and for geekbench, would get 4741 single core and 9680 on multicore.

    All of these tests were run 3 times consecutively. I don’t imagine I would be doing very heavy work on the go anyway, so 3 tests, I felt, was sufficient to prove my point.

    [​IMG]

    Looking at these values and comparing how fast my old MacBook was, setting the thermal limit to 25W is optimal for battery life while still retaining the performance I’m used too. I’ve tested the MacBook at this voltage and I easily see 14 hours battery life remaining with medium usage and 75% brightness! Fantastic!
    Overall very satisfied with the product.
     
  2. roland.g macrumors 603

    roland.g

    Joined:
    Apr 11, 2005
    Location:
    One mile up and soaring
    #2
    I don't know much about limiting voltage and had not thought of that in terms of what it could produce and that one might choose to do that consistently.

    1. Can you set it to run at two separate voltages automatically for battery vs plugged in?

    2. Would you run it at the lower voltage all the time to keep heat down? I plan to use mine on my lap a lot and don't want it to get hot but don't expect to push it all that hard. A cooler laptop with longer battery would be great.

    3. Any long term damage to the machine in doing this?

    I am looking at a 2.2 555X 16GB 1TB. I'd love to see 14 hour battery life.
     
  3. DannEboE thread starter macrumors member

    DannEboE

    Joined:
    Oct 18, 2016
    #3
    Hello!
    With the application I'm using, it won't automatically set the power limits depending on when it is plugged in or on battery. It is something I would have to do manually.

    When it is on battery, I plan on running it at 25W limit. At that power limit, it is more than fast enough for my usage while remaining really cool. I don't think I've seen it go past 45 C running Netflix and other tabs with that power limit.

    No long term damage at all! It's perfectly safe to set a power limit. It helps increase battery life and reduces temperatures. :)
     
  4. Thysanoptera macrumors regular

    Joined:
    Jun 12, 2018
    Location:
    Pittsburgh, PA
    #4
    Thank you for the tests. Just couple of remarks: your results are quite low, on my 2.2 I'm getting 1070 on first run and all the next ones stabilize at around 1020 on cinebench (1010 to 1030), clock stays at 3.2 GHz flat. No Volta, no cooling fan, just sitting on a table. I think it is due to the 4k external screen requiring more from the integrated UHD 630 and in turn heating the CPU cores and also hard shell trapping heat, you could remove it and just run the tests on the internal display, that would be more representative for the rest of us.

    And are you powering all your setup with the 87W charger? You may be starving your MacBook of power.

    Also, since you have 2.2 + 560x I bet a lot of people (me including) would be interested in some GPU tests, I'll be happy to repeat the same with 2.2 + 555x, but without shell, no Volta and internal display only.
     
  5. DannEboE thread starter macrumors member

    DannEboE

    Joined:
    Oct 18, 2016
    #5
    Yeah almost certainly. Running the tests without the monitor easily result in about 70 more points across the board from the CPU. The external monitor unfortunately is forcing the dedicated GPU to be active so there is a lot more heat than if it was just the integrated graphics.

    It's powering the laptop at 60W due to the adapter I used. The entire time, the battery never once dropped from 100% though, so that might be okay.

    Is there any GPU tests in particular? I can definitely complete some. :)
     
  6. Thysanoptera macrumors regular

    Joined:
    Jun 12, 2018
    Location:
    Pittsburgh, PA
    #6
    The three benchmarks from Unigine (Valley, Heaven and Superposition) plus 3dmark's firestrike, sky diver and time spy. In Windows obviously. Do you have any games that have built-in benchmarks? Something from this list:

    https://xsreviews.co.uk/news/games-with-built-in-benchmarks-2017/
     
  7. Vazza macrumors 6502

    Joined:
    Nov 25, 2006
    Location:
    London, UK
    #7
    Would definitely appreciate some GPU tests comparing your 560x to the 555x of @Thysanoptera
     
  8. joshi3joshi macrumors regular

    Joined:
    Oct 1, 2007
    #8
    Thanks for the tests, I have the same configuration as you. Would you mind putting up a little bit of a guide to run the MacBook at 45W (or 25w) as you have it? Is it with the Volta app? Was wondering if you give some instruction since I’m interested in testing it out to get that kind of battery life!
     
  9. Adeel_ macrumors member

    Adeel_

    Joined:
    Aug 1, 2018
    #9
    So basically your old MacBook Pro is faster than the 2018 2.2 model?
     
  10. DannEboE, Aug 10, 2018
    Last edited: Aug 10, 2018

    DannEboE thread starter macrumors member

    DannEboE

    Joined:
    Oct 18, 2016
    #10
    Unigine Valley and Heaven. (Top is with 4k monitor, bottom is without monitor)

    [​IMG]

    Superposition
    [​IMG]

    Firestrike
    [​IMG]

    Sky Diver
    [​IMG]

    Time Spy
    [​IMG]

    I didn't notice much difference with the 3Dreal benchmarks with the 4k monitor or built in screen.
    I am not entirely sure on what levels I should be testing. My only point of comparison was my old laptop and these values beat them haha.


    FFXV Benchmark as well
    [​IMG]

    http://benchmark.finalfantasyxv.com...b6d4a8fddf83&Resolution=1280x720&Quality=High

    Hello! Yeah! I did use the Volta app. I used the installation guide followed by this URL.
    http://volta.garymathews.com/installation.html

    Once it is properly installed, I just moved the power limit to 25W (or whatever number you like). You can then check Intel Power Gadget. There will be a gray line as to where the limit is now set (you can also run benchmarks and see if the power goes past the number you set. :)

    No! Haha, the 2018 model is MUCH faster than my old MacBook Pro. It is only when I want to conserve battery and purposely reducing the power does it become equal in power. In reality, it is much faster. 6 cores vs 2 cores.
     
  11. maflynn Moderator

    maflynn

    Staff Member

    Joined:
    May 3, 2009
    Location:
    Boston
    #11
    Given the poor performance of the 555x/560x, GPU benchmarks aren't all that useful, only because the MBP is not a gaming laptop, no where near it tbh.

    The Dell which has the same chipset is not considered a gaming laptop and it produces better results with the 1050 GPU. The issue with the Dell, is that in gaming it will throttle pretty quickly, and given that the MBP is even thinner, its going to perform even worse.

    One thing I miss about the Razer 15" that I returned is the awesome GPU performance that I was seeing. I still had to manage it thermally for most tasks, and it couldn't play games on your lap. I do think for my needs the MBP is a better overall choice over the razer.
     
  12. Thysanoptera macrumors regular

    Joined:
    Jun 12, 2018
    Location:
    Pittsburgh, PA
    #12
    I never really understood that mantra about MBP not being a gaming laptop, or that you're not supposed to use a laptop in any other way that is not marketed by a laptop manufacturer. And 555x/560x has plenty of power, more than 960m which was a staple of gaming laptops just two years ago. My other gaming desktop at home has 1Mhz CPU and can display 320x200 screen with 16 colors tops and my kids have a blast with it, as I do.

    @DannEboE - Thanks for the benches, I'll post mine shortly, I forgot to mention that you should use 'presets', not custom modes, because in Valley/Heaven I don't really know what settings were used.
     
  13. DannEboE thread starter macrumors member

    DannEboE

    Joined:
    Oct 18, 2016
    #13
    My bad. I just ran those tests without changing any settings, just whatever it came with.

    Honestly with these laptops, having different expectations makes all the difference. I don't expect to do VR or game in 4K. I gamed before on the Intel 650, so this machine itself is a massive upgrade and thus, benchmarks are still worth it imo.
     
  14. Thysanoptera macrumors regular

    Joined:
    Jun 12, 2018
    Location:
    Pittsburgh, PA
    #14
    Don’t worry about it then, I can run it in the same settings. Just my kids get in the way, in unigine benchmarks you can enable ‘walk’ mode and they nagging me to let them play.
     
  15. maflynn Moderator

    maflynn

    Staff Member

    Joined:
    May 3, 2009
    Location:
    Boston
    #15
    No, that's not my point. You can use the laptop any way as you see fit. My point is its not a gaming laptop so its going to perform poorly on most top tier games. If you or anyone else is ok with the reducing detail and slow frame rates, who am I to say you can't. That doesn't change the fact that its not the best tool for the job.
    --- Post Merged, Aug 10, 2018 ---
    Here's a for instance. Just one benchmark. The MBP vs the Razerblade and you can the numbers are quite telling. In both cases, the benchmark was run on a newly purchased laptop with nothing much done in terms of tweaks or updates.
    2018-07-27_07-18-20.png
     
  16. Thysanoptera macrumors regular

    Joined:
    Jun 12, 2018
    Location:
    Pittsburgh, PA
    #16
    Well of course nobody buys MBP and expects dedicated Windows machine performance in games. Or strictly for gaming as primary use. But MBP with dGPU is still quite capable in this regard for occasional session when you’re bored in a hotel room.

    @DannEboE - I’ve run all the test but I think we’re doing something wrong. 560x comes out 15-35% faster, which is a lot more than what I expected, but still possible. But all the physics/cpu scores I get are a whooping 50% faster on your machine, and that doesn’t make any sense, we have the same CPUs. Do you use anything to limit CPU, like Intel XTU or something? What are your drivers version? And did you buy advanced version of 3dmark? The base/free version doesn’t have all the graphs etc. and just to be sure - you have your windows desktop set to native retina resultion? I guess so based on heaven/valley.
     
  17. DannEboE, Aug 10, 2018
    Last edited: Aug 10, 2018

    DannEboE thread starter macrumors member

    DannEboE

    Joined:
    Oct 18, 2016
    #17
    On Windows side, I am not using anything to limit CPU. I have Intel Power Gadget to monitor temperatures and they seem to be holding pretty well (relatively). I can check driver versions when I get home from work, but they are default from bootcamp installation + the latest windows update. I have Windows 10 Pro if that does anything. I should only have the basic version of 3dmark. Yeah, all resolutions are set as native.

    It is possible I also received a slightly slower variant of the processor. Maybe I got the lower end of the stick for what was acceptable for that processor model.
    --- Post Merged, Aug 10, 2018 ---
    At least looking at specs on paper, those benchmarks seem pretty accurate

    http://gpu.userbenchmark.com/Compare/Nvidia-GTX-1060-Mobile-vs-AMD-Radeon-Pro-560X/m164336vsm541446

    With a power limit of 80 watts compared to Apple's 35 watts, I am not at all surprised by the disparity of the results. Though I do recognize that Nvidia is dominating and winning the gaming space right now. Even a comparable power limit GPU, 1050 Max-Q, that card still dominates the 560x in the MacBook. It's just a shame Windows laptops don't seem to hold up as well as a MacBook.
     
  18. MBX macrumors 68000

    Joined:
    Sep 14, 2006
    #18
    I have the 2.2ghz model. Super happy. Don't think the more expensive ones would be much faster in most cases. Because it speeds up to 4.1ghz anyway in turboboost if necessary.
     
  19. maflynn Moderator

    maflynn

    Staff Member

    Joined:
    May 3, 2009
    Location:
    Boston
    #19
    Me too, I'm so happy I went with the base model, and its been fast enough for my humble needs.
     
  20. Thysanoptera macrumors regular

    Joined:
    Jun 12, 2018
    Location:
    Pittsburgh, PA
    #20
    It’s the other way around, your CPU is 50% faster. Do you still have that cooling fan?
     
  21. DannEboE thread starter macrumors member

    DannEboE

    Joined:
    Oct 18, 2016
    #21
    Oh jk! I misunderstand. Yeah, the cooling fan sort of sucks/doesn't really help. I honestly keep it because it slightly elevates the laptop to allow more airflow and it has pretty blue LED lights.

    I wouldn't imagine a cooling pad being able to produce that much better results. It's also weird too because your cinebench results were higher than mine.
     
  22. Thysanoptera macrumors regular

    Joined:
    Jun 12, 2018
    Location:
    Pittsburgh, PA
    #22
    @DannEboE - you have the advanced edition of 3dmark, I just noticed this on the fire strike screen, did you get it from steam? Would you be able to download the basic one and run those? I mean I have no idea why those cpu scores are so much different, other than the different version of the benchmark.
     
  23. DannEboE thread starter macrumors member

    DannEboE

    Joined:
    Oct 18, 2016
    #23
    That's odd. I don't recall ever purchasing 3dMark haha. Yeah its in my steam.
    I can try running basic ones when I get home in a few hours.
     
  24. Thysanoptera macrumors regular

    Joined:
    Jun 12, 2018
    Location:
    Pittsburgh, PA
    #24
    Ok, wait, I’ll just get the steam version and run it again.
    --- Post Merged, Aug 10, 2018 ---
    I'm sorry everyone, I forgot I've set the cpu limit in energy settings to 99%, which disables turbo boost... I have to run all that crap again.
     
  25. DannEboE thread starter macrumors member

    DannEboE

    Joined:
    Oct 18, 2016
    #25
    Haha! Those benchmarks take awhile with all the demos included. At least for me, the time it took for the benchmark to identify my system was as long as the actual benchmark itself.
     

Share This Page