iMac 2019 i9 - perfect machine, when App nap is turned off !

Discussion in 'iMac' started by Melleken, Apr 26, 2019.

  1. Melleken, Apr 26, 2019
    Last edited: Apr 28, 2019

    Melleken macrumors newbie

    Joined:
    Apr 3, 2019
    Location:
    Ghent, Belgium
    #1
    edit of startpost after finding real problem and solutions:

    i though my machine was thermal throttling, but the only problem was app nap.
    Really problematic 'feature' when you are some who renders (video or images) on the backround, will working in other programs in the meanwhile !)

    See the end for more info, but the beginning of this thread is little bit pointless.

    ---------------------------------------------------------


    Hello everybody,

    (sorry for my english writing and mistakes, it is not my native language)

    ordered maxed out 2019 iMac i9 with the vega graphics + 64GB memory installed.
    Use for the machine = 3D modeling and a lot of rendering with Vray.
    So long sustained use of all the cores with HT.

    I was very aware and scared of thermal throttling before buying this.
    But the first reviews and experiences were positive;

    The iMac has to replace an 12-core 3,34Ghz Mac Pro 2010 with 48GB.
    Especially the single core performance began to be a large problem in modeling larger and larger, complexer and complexer 3D models.
    We upgraded this Mac Pro 1,5 year ago to this maximum CPU 12-core configuration, and the plan was to wait and use this machine until the new modular Mac Pro 2019-2020 came out.

    But the irritation grow per week and when this 'monster' iMac 2019 i9 came out, i was very interested.
    The iMac pro would be an option, but we need a least 64GB of ram and then the iMac pro becomes very expensive, to much for us at this point + the need for extra ram grows per project and we NEED an option to self upgrade the ram, a minimal expectation for a pro desktop computer and the iMac deliveres that, where the iMac pro fails.

    Long story short, after couple of weeks reading reviews, i was confident enough to buy the iMac 2019.
    And the first impression was awesome - easy install of the 64GB ram, fantastic screen, super quiet, very fast single core performance and we the first test - very good multi core performance. On a the exact same render the iMac did 13minutes, and our old Mac Pro 5,1 24 minutes. So a much more then expected performance gain.

    Then the holiday came, and we were waiting for an thunderbay that only arrived with us 12 days after ordering. Yesterday arrived, so i was busy al evening to change all the drives, test the thunderbay and finished the setup of the iMac 2019 - and finally ready to make the change. The mac pro was already on the second hand market and i have a couple of very interested buyers...

    But this night i planned a final stress test - an render that took 2hours and 39 minutes on the Mac Pro 5,1, was setup and started on the iMac 2019 i9 this night, hoping and expecting to see may be the iMac finishing in less the 1,5hours.

    I think you already know what we saw this morning:
    render took 3hours and 50 minutes on the iMac 2019. Much longer then on the Mac Pro.
    I have istat and intel power gadget running and this is what is saw:

    first 15 minutes > 4,6 gHz (turbo boost mode)
    second 15 minutes > 3,8 gHz (base clock speed)
    then 1,5 hour > around 2,3 gHz (=start thermal throttling - very bad)
    then the last 2 hr > around 1,5 gHz (=unbelievable performance - thermal throttling to te level of a MacBook air)

    (see also screenshot)

    CONCLUSION - immediately called for a return, (a luckily no problem, is was 2days overdue, but their was no issue) so on Monday the machine goes back. This is really no option and unacceptable performance.

    I will do so more testing today, but that one experience is enough for me.
    (may be i am to fast with my conclusion, but for now the disappointment speaks)

    This machine throttles BIG TIME, just the first half hour is oké.
    So the i9 can not be used under sustained load !!!!


    So very big disappointment - luckily we get complete refund, but i wanted to share this experience !

    Michael from Belgium.
     

    Attached Files:

  2. MacRS4 macrumors regular

    Joined:
    Aug 18, 2010
    Location:
    London, UK
    #2
    That's not been our experience on any of them. I've not once seen any of them drop below clock when under load.
     
  3. kaintxu macrumors regular

    Joined:
    Jul 9, 2018
    #3
    Have you done intensive and heavy testing like him? Not doubting you, just curious
     
  4. MacRS4 macrumors regular

    Joined:
    Aug 18, 2010
    Location:
    London, UK
    #4
    Yup, they're pretty heavily loaded for most of the working day.
     
  5. Shivetya macrumors 68000

    Joined:
    Jan 16, 2008
    #5
    Please provide some proof, like the tests and charts accompanying them. Also, whether or not the tests are a long continuous process and typical work day loads which come and go
     
  6. MacRS4 macrumors regular

    Joined:
    Aug 18, 2010
    Location:
    London, UK
    #6
    Sigh. Well, you can see on my YouTube channel here my original tests. Don't particularly care if you believe me or not! I've no axe to grind.

    I do about my MBP throttling, but that's a whole different story.

    EDITed to add - one thing one of techs noticed (and I haven't verified this) is that they seem to top out at around 95 watts, so they may be restricted from that perspective. We haven't seen any drop below clock however....which is why we're buying them over the iMPro's now.

    EDITed again to add, they're used to run some fairly chunky Virtual Machines in Parallels. Average util is >80% for most of the working day.
     
  7. willmtaylor macrumors G3

    willmtaylor

    Joined:
    Oct 31, 2009
    Location:
    A Natural State
    #7
    Linus from LTT also noticed throttled from a power perspective as well in his review.
     
  8. MacRS4 macrumors regular

    Joined:
    Aug 18, 2010
    Location:
    London, UK
    #8
    Ah ok - that makes a bit more sense tbh. I thought they were higher TDP chips, but ours seems to settle out at 95 watts.
     
  9. Zdigital2015 macrumors 65816

    Joined:
    Jul 14, 2015
    Location:
    East Coast, United States
    #9
  10. joema2 macrumors 68000

    joema2

    Joined:
    Sep 3, 2013
    #10
    This is an interesting situation because most machines (Mac or PC) reach thermal equilibrium fairly quickly under a sustained load. They don't usually get progressively slower or hotter after 1 hr, 2 hr or 5 hr.

    I'm not sure this has been reported before, maybe because I don't think many people do 4 hour high-intensity testing while watching CPU and thermal performance.

    It would be interesting if other people ran 4 hr of the "yes" stress test while logging CPU clock performance with Intel Power Gadget. Is Melleken's machine an anomaly or are all i9 iMacs like this? Maybe previous iMacs do this but few people have done 4 hr instrumented CPU tests.

    Yes stress test: http://osxdaily.com/2012/10/02/stress-test-mac-cpu/
    Installing Intel Power Gadget on Mojave: http://osxdaily.com/2019/03/03/install-intel-power-gadget-mac/
     
  11. Zdigital2015, Apr 26, 2019
    Last edited: Apr 26, 2019

    Zdigital2015 macrumors 65816

    Joined:
    Jul 14, 2015
    Location:
    East Coast, United States
    #11
    I am curious why you wouldn’t take the money that you spent on the 2019 iMac and build a render node for V-Ray or implement a better GPU (or 2) in the Mac Pro? A mid-2010 is perfectly capable of running Mojave with a Metal 2-supported GPU like the Vega 56 or 64.

    Did you buy your 2019 iMac with an SSD or a Fusion Drive?

    I would also assume that V-Ray uses OpenGL, which is deprecated in macOS Mojave along with OpenCL. Apps will still run, but I would think there might be some issues with older versions of applications.

    I noticed from the Wikipedia page that the last stable release for V-Ray is being listed at version 3.6 (July 1st, 2017). That cannot be correct? I tried looking on the Chaos Group website to see when the last version was released and I am not finding that information. Looked on Twitter and didn’t see anything either. Is it possible that the version of V-Ray that you are using needs to be updated and is not optimized for newer Intel CPUs and macOS Mojave? Or are you running this under Windows Bootcamp?

    It is surely possible that you got a defective unit, but those lower clock speeds seem to indicate either a hardware issue from the factory or a software update is needed. Hope it works out for you.
     
  12. MacRS4 macrumors regular

    Joined:
    Aug 18, 2010
    Location:
    London, UK
    #12
    The 'yes' test is what we used to test the throttling - left it running for 2-3 hours. You can see a shorter version in my youtube video. There was a fair bit of resistance to not buying the pros...You know what techs are like with their shiny stuff :)

    But after the performance and throttle testing the i9 units were accepted. Performance on our workloads is within a few % of the base-spec iMac Pro, but at a cheaper price point.
     
  13. Shivetya macrumors 68000

    Joined:
    Jan 16, 2008
    #13
    the linked video is short duration test which doesn't amount to a hill of beans. the OP test ran for three hours straight, that is the kind of testing that someone crunching vids/pics or just gaming will care about.
     
  14. MacRS4, Apr 26, 2019
    Last edited: Apr 26, 2019

    MacRS4 macrumors regular

    Joined:
    Aug 18, 2010
    Location:
    London, UK
    #14
    The main tests we did before buying a few of them was a lot, lot longer. Again, no real interest if you believe me or not. Showing a 3+ hour 'yes' test in a video is not particularly interesting.
     
  15. mj_ macrumors 6502

    mj_

    Joined:
    May 18, 2017
    Location:
    Austin, TX
    #15
    It's not about belief or disbelief, it's about different test scenarios yielding different results. You can run the yes test for 6 hours straight without noticing any throttling while others run a render stressing both CPU and GPU to the max that causes thermal throttling after a few minutes already. In OP's case it seems that the throttling really only started after 1.5 hours and turned real ugly another 30 minutes later. Which is interesting considering that most media outlets reviewing hardware run their stress test for no longer than an hour.
     
  16. MacRS4 macrumors regular

    Joined:
    Aug 18, 2010
    Location:
    London, UK
    #16
    Agreed. Like I say, with our workloads we've not noticed any < clock speed throttling. Very little graphics load in those, it's mostly transactional data processing. Same workload has my MBP throttling inside 2-3 minutes...Which is infuriatingly annoying.
     
  17. Zdigital2015 macrumors 65816

    Joined:
    Jul 14, 2015
    Location:
    East Coast, United States
    #17
    The tester also left out some important information that might provide a basis for why the Core i9 iMac throttled as much as it did, or he may have simply gotten a bad unit from the factory. Others on these forums have had very good experiences with the Core i9 iMac and the 2019 model in general versus the fan noise and thermal issues the 2017 iMac seems to have seared into everyone's consciousness.

    What I don't know is what version of macOS his 2010 Mac Pro was running, what GPU was in it, what version of V-Ray was installed, et al. I did a cursory glance at the Chaos Group website and I didn't see a lot of front facing information about what version is the current stable version and I am loathe to sign up for any more free trials of anything. The Wikipedia page says the last stable version was in July 1 of 2017, which I just cannot believe is the last time they updated. If it is, then perhaps the Mac Pro is running an older version of macOS and V-Ray is not optimized for Mojave. We are six months on since Mojave's release and Avid still hasn't released a Mojave compatible version of Pro Tools...so the same may apply here.

    I would think that V-Ray might simply work better under an NVIDIA GPU, as Chaos Group emphasizes in their marketing information. I suggested a render node might be money better spent than on the iMac since the Mac Pro he has is Mojave compatible. Perhaps there is an issue with OpenGL rendering under macOS Mojave with V-Ray, perhaps not.

    My point is that while he had a bad experience, the thread title stating - WARNING> does thermal throttle!!! doesn't mean every 2019 iMac is a ticking time bomb. You are taking the post at face value without asking the OP any critical questions while asking for charts and graphs from someone who has not had the same experience and simply expresses his opinion.
     
  18. dcpmark macrumors 6502a

    Joined:
    Oct 20, 2009
    #18
    Well, I for one am grateful for the OP’s info, and have learned a valuable lesson.....never buy a consumer-level machine expecting it to do professional, workstation-level 3D model rendering.

    Thanks.
     
  19. xgman macrumors 601

    xgman

    Joined:
    Aug 6, 2007
    #19
    I would like a curious follow up by the Op as to "now what"?
     
  20. joema2 macrumors 68000

    joema2

    Joined:
    Sep 3, 2013
    #20
    I have never seen anybody report thermal throttling that suddenly degraded performance to 39% at the two hour point of a four hour test. Normally systems reach thermal equilibrium long before that.

    Anybody can run combined CPU and GPU tests. Just use "yes" combined with the Furmark GPU test: https://www.geeks3d.com/gputest/

    It would be interesting if any other user on any other Mac from any production year could reproduce a dramatic thermal slowdown which worsens dramatically after two hours of constant uniform stress.
     
  21. Melleken, Apr 26, 2019
    Last edited: Apr 26, 2019

    Melleken thread starter macrumors newbie

    Joined:
    Apr 3, 2019
    Location:
    Ghent, Belgium
    #21
    Oké, just got home from work. This is what we learned from further testing today.
    (i have quickly scanned all the replies, my post has made some dust wave by... but that probably is not an english expression... :) - i will answer some questions later)

    I seems i was wrong. (luckily and may be not so surprisingly...) but in my defense, i surely looked like thermal throttling.
    My girlfriend (who uses the computer for her work - architectural 3D visualisations) had rendered and worked all that long, and the throttling seemed gone. Then she went to get the children, 'throttling' came back.

    Long story short - is seems that there is an issue between mojave - sketchup & Vray, that when the screen goes into 'sluimer' modus (do not know the english word: sleep mode?) so just the screen, not the mac itself, there is an effect that looks like thermal throttling, but clearly is not. The mac keeps on rendering the file, but slows gradually down - do not know why yet, or have not a solution (for one person on the internet even having the setting screen 'never go into sleep modus" it still happened) but maybe i has something to do that the fans go into some kind of sleep mode, when the screen does, so by default you will have throttling, but not caused by 'thermal throttling' in the way we speek about it.

    I hope the above is somewhat clear, because i feel that it is hard to explain it clearly, also because i do not fully understand the problem yet.

    So anyway i have to change the title of this thread!

    I will do that first
    --- Post Merged, Apr 26, 2019 ---
    oke, do not find how to change title, i will be simple, but can anyone quickly help?
    very new here, and i have to say this forum has some complexity to it :)

    and of course , stopped looking and found it !
    --- Post Merged, Apr 26, 2019 ---
    When i read reactions, first start thinking, may be i do have a 'monday morning' model.
    So glad that it had to do with sleep modus - there will be solutions for that.

    So first the results of the tests today > see pictures.

    Some drops in clock speed during the day, but pretty stable performance at 3,8 Ghz.
    The CPU got hot, 90° Celsius. But then fans kick in, and it goes down to around 60°.
    The second screenshot shows when she went out to get the kids, around 17:47 the performance drops badly, to go up immediately when she came home and looked at the screen. That's when she knew it had something to do with the screen going to sleep.

    -------------------------------------------------------------

    Some info about how the mac is used.

    if Els (my girlfriend) is working, the mac is continuously under max load, she is always rendering a scene + modeling in sketchup or working in photoshop. And rendering means = al the cores maxed out, all the time.

    CPU is king in her workflow. GPU is not important at all in sketchup for mac, but still went for the overpriced vega upgrade because it supposed to be cooler)

    So anyway, i was very very suspicious of an normal old iMac & an 8-core i9 CPU. But after reading everything that was available (youtube, google, this forum) it looked like the thermal design of the iMac was just good enough for the i9 to keep it on base clock speed under sustained load. That is why is was so shocked this morning that the render time of the 4hour stress test was longer on the iMac 2019 compared to our 5,1 12-core...

    A iMac pro would be 'perfect' for us, but she really needs 64GB and even more over time, and that makes iMac pro very expensive + not upgradable;

    --------------------------

    iMac is with 512 SSD, never had and never will use fusion drives.

    --------------------------

    so conclusion at this point

    - no thermal throttling (sorry for my post but also hoera!)
    - no defective unit
    - just an software issue - something between mojave, sketchup and Vray.

    --------------------------

    we will go on doing test and work on iMac.
    This night same 4 hour render goes on + hopefully a solution for the sleep thing / not really want to have a screen always on at night.. not good for nothing !
    --- Post Merged, Apr 26, 2019 ---
    Little update > same thing happens when screen goes into screensaver modus.
     

    Attached Files:

  22. michaelrdk macrumors newbie

    Joined:
    Feb 3, 2014
    #22
    Wild guess, but could this be App Nap kicking in?

    You can read about it here: What Is App Nap? Is it Slowing Down My Mac Apps?

    However, that feature was added in 10.9 Mavericks so I guess it should have been a problem on your previous system too, unless you turned it off in the past.
     
  23. joema2 macrumors 68000

    joema2

    Joined:
    Sep 3, 2013
    #23
    That is more plausible. You are not alone. I and other people have had various problems with sleep mode and even screensaver mode. I recently had problems with an upload to DropBox slowing down when the screensaver activated, despite all other iMac power saving features being disabled. In system prefs I had "prevent computer from sleeping automatically when the display is off", yet it still happened. The only way to stop it was totally disable screen saver and I turned down the brightness for long unattended uploads.

    Your best test is disable all that as a trial. That means everything -- screen saver and all the power saving checkboxes in system prefs>energy saver. The "Turn off display after..." slider should be set to "never", and the "prevent computer from sleeping when display is off" should be checked.

    There are anti-sleep utilities like caffeinate. For initial troubleshooting I recommend not using those but just manually turn off all power saving features.

    For long runs you can just manually turn down the screen brightness. This is only for a test. Once you verify whether the throttling problem goes away, then you can selectively re-enable the power saving features one at a time for each subsequent run until you find which is responsible.

    I understand the need to save electric power, but misbehaving power saving features have cost tremendous amounts of troubleshooting time. At this stage of technology it should work with 100% reliability and every behavioral aspect should be clearly documented by Apple.

    Please report your results after trying this. It is understandable the symptoms you initially observed could appear like thermal throttling, esp if you only monitor the CPU frequency and not the temperature.
     
  24. ZipZap macrumors 603

    Joined:
    Dec 14, 2007
    #24
    The OP/Tester cannot leave out what he does not know. Why would he think his iMac is defective?
    --- Post Merged, Apr 26, 2019 ---
    So your workload is nothing like the OP's ?
     
  25. mmomega macrumors 68030

    mmomega

    Joined:
    Dec 30, 2009
    Location:
    DFW, TX
    #25
    If it didn't work for your particular case then it sounds like you made the correct decision to return it and look for something else.
     

Share This Page

35 April 26, 2019