AMD Radeon RX 580 works in macOS

Discussion in 'Mac Pro' started by theitsage, Apr 19, 2017.

  1. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #51
    Correct, this control panel won’t work in Windows 10 due to unable to read the APFS partition.
     
  2. SoyCapitanSoyCapitan macrumors 68040

    SoyCapitanSoyCapitan

    Joined:
    Jul 4, 2015
    #52
    Only on cMP? Because High Sierra docs say everything works as normal. The control panel talks to the NVRAM.
     
  3. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #53
    Not sure, but seems affect all Mac.

    For other Macs, doesn’t really matter, option boot always available.
     
  4. pierrox macrumors regular

    pierrox

    Joined:
    Jun 19, 2015
    Location:
    Paris, France
    #54
  5. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #55
  6. babyjoe, Oct 12, 2017
    Last edited: Oct 12, 2017

    babyjoe macrumors member

    Joined:
    Sep 1, 2017
    Location:
    Sweden
    #56
    I tried a MSI RX 580 Armor 8G today and me and friend made a benchmark test comparing it to a GTX970 which is supposed to be equal to the 580. Unfortunately the 580 had a much lower score than the 970.
    Screen Shot 2017-10-12 at 21.06.44.png V__ACC8(1).jpg
     
  7. jeanlain macrumors 65816

    Joined:
    Mar 14, 2009
    #57
    You're comparing different OSes. Your test is invalid.
    Also, Unigine is perhaps the game engine that disfavours AMD cards the most on macOS. Even the lowly GTX 775m on my iMac can score close to the Radeon Pro 580 on this benchmark. But it's utterly destroyed in games. And since no game uses Unigine...
     
  8. fendersrule macrumors 6502

    Joined:
    Oct 9, 2008
    #58
    You will never win for the following reasons:

    Heaven is geared towards nVidia cards
    Heaven in more optimized in Windows for OpenGL than MacOS
    Heaven OpenGL vs Direct 3D is no contest. Direct 3D will win by a landslide every time.
    MacOS GPU drivers for RX 580 is pretty laughable. Only a minor increase, if any, from a 7950.

    You have all of those going against you.

    It's best you run these in D3D in Windows for a more fair comparison. You'll gain 35+ FPS immediately.

    btw, the RX 580 is better than the GTX 970. The nVidia contender for the RX 580 is the GTX 980 (non-Ti) and the GTX 1060.
     
  9. babyjoe macrumors member

    Joined:
    Sep 1, 2017
    Location:
    Sweden
    #59
    Well, it's not totally invalid. It shows that the 970 has much better performance in a PC than a 580 in a cMP. But I'll try later to take his card into my computer and run the same test again. Maybe you have a better suggestion for another benchmark test?
    --- Post Merged, Oct 12, 2017 ---
    Ok, thanks for the info! Do you know any better way to compare the different cards?
     
  10. fendersrule macrumors 6502

    Joined:
    Oct 9, 2008
    #60
    It's very invalid. You guys are not even using the same APIs, nor the same OSes!

    Yes, taking his card and putting it in your Mac Pro will be more "valid" with the asterisk of knowing that Heaven favors nVidia cards, and the MacOS is probably not running your card at OC speeds and with a stock power limit, and zero optimization. It's going to be very painful to do it this way because you have to deal with web drivers.

    A much more valid (and easier) test would be booting into windows 10, downloading afterburner and making sure your power limit is cranked (and your card's factory OC speeds are applied) and re-running with D3D.

    I'll post what my RX 580 8GB does in D3D in Windows 10 for you if you'd like, with a 5,1 X5690 MP. My RX 580 is undervolted with a clock speed of 3860MHz, at +50 power limit.
     
  11. babyjoe macrumors member

    Joined:
    Sep 1, 2017
    Location:
    Sweden
    #61
    The thing is that I'm trying to choose what GPU is best for me as a cMP owner, and I would prefer to run the cMP with OSX and not Windows (maybe you got some suggestion?). But it would be really interesting with a test from your system aswell :)
     
  12. fendersrule, Oct 12, 2017
    Last edited: Oct 12, 2017

    fendersrule macrumors 6502

    Joined:
    Oct 9, 2008
    #62
    Well, I don't think any card out there is going to allow you to compete with him. MacOS for gaming is gimped.

    Anyhow, this is about what you'd get in Heaven. Our MSI cards are similarly clocked too. Ignore the bug in Heaven that reads my card as having only 4GB of vRam:

    Boom.PNG

    If you bench him on some other tool that isn't more optimized for nVidia, you would pull away.

    Btw, this is aweful resolution/settings for our first gen i7s. The CPU exhibits some micro-stutters once in awhile because it is not able to keep up with the GPU.

    I'd recommend getting a higher res monitor or maxing your settings at this resolution for this bench. You need to put more load on the GPU.
     
  13. babyjoe macrumors member

    Joined:
    Sep 1, 2017
    Location:
    Sweden
    #63
    Very interesting!! I'm gonna play some, but mostly I'm actually into video editing with Premiere, and that is the primarily purpose for the GPU I'm buying. Do you think I should stay with the 580 or maybe it's better to go NVIDIA, maybe a 980ti?
     
  14. jeanlain macrumors 65816

    Joined:
    Mar 14, 2009
    #64
    Not in my experience. The difference between cards is as expected, even though performance is behind Windows in OpenGL. I don't think it's worse than other cards.
     
  15. fendersrule, Oct 12, 2017
    Last edited: Oct 12, 2017

    fendersrule macrumors 6502

    Joined:
    Oct 9, 2008
    #65
    Perhaps I overplay it some, but the RX 580 should be at the absolute minimum, 50% faster then the 7950. Usually, about 70% faster is expected in real world. This can be validated using several benchmarks on relative performance on the web for the 7950/280 vs RX 580.

    My MacOS Heaven score shows the RX580 being 39.38% faster than the 7950 @ 1440p in Mac OS.

    Windows on the other hand shows the RX580 being 80% faster than the 7950 @ 1440p in Windows.

    The RX 580 underperforms in MacOS. There's no doubt about it. It probably underperforms because OpenGL is just outdated as hell (hitting a ceiling effect) and Apple has done nothing in the past several years to further optimize or improve. It could also be that as a user, you get no control for how much the RX 580 is throttling out-of-the-box, and are locked out from raising up the power limit. It "MAY" be unlikely that you are seeing any factory OC speeds either, since many GPU manufacturers are doing this via software. This is highly likely, being that my 5,1 Mac Pro in MacOS pulled 2FPS faster numbers than a user here which had an MSI Gaming X RX580, but as soon as we tested each other in windows, he clearly pulled away.

    So the difference between the cards, is NOT as expected in MacOS if you go by the % gains that others are getting in Windows from the same upgrade. Going from 25.9 FPS to 36.1 FPS in Heaven @ 1440p in MacOS is significantly less than 50-70% faster.

    So while saying me saying "if any" is a gross misrepresentation, it isn't a gross misrepresentation to say "not enough faster". Which leads you to the same conclusion: Game in Windows.

    If MacOS is where you live and breathe for gaming, the RX580 is a worthy $150-175 upgrade, max. I can't help but shake my head at people still wanting to hunt out the $400+ pulse editions. Heh.

    My .02, take it or leave it.
     
  16. Flint Ironstag macrumors 6502a

    Flint Ironstag

    Joined:
    Dec 1, 2013
    Location:
    Houston, TX USA
    #66
    All the more shame they never enabled SLI. Maybe with all the tweaks they're doing for eGPU, we'll finally see some multi-GPU aware games on Mac.
     
  17. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #67
    Then ask your friend to install MacOS on his PC, make it a Hackintosh. Then it can be a valid comparison. If you want to know which GPU is better in MacOS, you have to done all the test / benchmark in MacOS.

    If comparing different hardware in different OS with different API is still a valid test, then why not make your benchmark run at 480p and the other one at 4k, so that you have a “valid” result which shows your RX580 is way faster than the 970?
     
  18. babyjoe macrumors member

    Joined:
    Sep 1, 2017
    Location:
    Sweden
    #68
    Don't see how a test with different resolution would give me anything. The tests that I made gave some reasons to believe that the 580 underperforms in OSX.
     
  19. jeanlain, Oct 12, 2017
    Last edited: Oct 12, 2017

    jeanlain macrumors 65816

    Joined:
    Mar 14, 2009
    #69
    That could be easily explained by CPU being the bottleneck on macOS more than on Windows because of openGL driver overhead. Remember than GPU drivers "live" on the CPU.
    This is typically why high-end GPUs bring little to no performance gain at low resolution. It's not just the 580. The fact that Unigine uses a single CPU core makes it worse. It's also why difference between OS X and Windows is typically larger at low resolution. At highest resolution, performance is limited by the GPU hardware, which is the same between OSes.
    To compare GPUs, you need to use the fastest CPU core available and/or play at highest resolution (increasing resolution only affects GPU performance).
    For instance, the radeon 580 is only 12-13% faster than the R9 M395X in Hitman at 1080p on macOS. At 1440p, it's already 37% faster. And the difference should be even larger at 4K.
    https://forums.macrumors.com/threads/f1-2017-performance-comparison.2063652/page-2
    If you can't go above 1440p, check CPU usage and GPU usage. If CPU usage is the same for both cards, the comparison is invalid (it should be higher for the faster card). Is GPU usage is lower for the faster card, the comparison is also invalid (should be 95-100% for both).
    --- Post Merged, Oct 12, 2017 ---
    Compared to what? You have shown no evidence that GTX cards do better in the same conditions. In fact, several of the latest macOS games simply don't work on nVidia cards (they are listed as incompatible). In these games, one could say that performance of nVidia cards is catastrophic on OS X. That would be a more valid comparison than what you posted.
    I would not recommend getting an nVidia card to use on macOS. It's been 4 years since we saw a new Mac with an nVidia GPU. Apple don't want to work with nVidia, hence current drivers aren't good. Several Metal games are listed as incompatible.
     
  20. babyjoe macrumors member

    Joined:
    Sep 1, 2017
    Location:
    Sweden
    #70
    Not any evidence, but some indications to believe the 580 might have some issues. This indication made me post in this thread to give me more information. The user fendersrule made another test with a similar system as mine, but with Windows instead of MacOSX, and it gave furthermore reasons to believe that the 580 underperforms in MacOSX. Of course I don't know anything about how a GTX-card would perform compared to the 580. I'd be very glad if you could provide some more info on that issue. Maybe it's better to move over to Windows?
     
  21. fendersrule, Oct 13, 2017
    Last edited: Oct 13, 2017

    fendersrule macrumors 6502

    Joined:
    Oct 9, 2008
    #71
    All semi-modern to modern cards underperform in MacOS. OpenGL is slow. Jean stated why above.

    But I stand by my statement that even in MacOS, the rx580 is not putting everything out on the table. The simple fact that they are heavy throttlers and require maxing the power limit can account for 7 FPS ALONE @ 1440p makes a significant difference. See my VR thread for data. More interesting data would be how MacOS handles all the rx 580s that are out there. Power limit set to what? Core clock set to what?

    1440p IMO is the sweet spot res for gaming today and in the near future, even though I would bet that the vast majority are still at 1080p for gaming. I consider 1080p low res and quite frankly, Im tired of seeing 1080p benchmarks. Btw, I think we're also mostly saying the same thing. But I find it hard to believe that you would only expect 39% faster per your "as expected" comment if you shoveled out big money for one. Anyways, I made my point, which is data-driven. the RX580 is not seeing the gains as expected in MacOS when it comes to gaming. I saw huge gains from a 5770 to a 7950. I see much less gains from a 7950 to a RX580. In fact, some games seem unaffected, and others show a lower minimum FPS. The same would likely be true with an nVidia GPU. That's not a criticism to the 580, but to Apple.

    Babyjoe, it's no secret to anyone that gaming in MacOS is half the speed of Windows at this point. Many of us will have dual boot systems to remedy this.

    Though, I do find myself in Windows 10 more than MacOS these days. Windows 10 is faster for everything, even common things like boot times, time for apps to open, etc than MacOS is for a 5,1. :)

    But I do like me some Apple ecosystem though. And as a designer, I find myself more at home with MacOS. But Apple, please add more UI horsepower, k thx.
     
  22. babyjoe macrumors member

    Joined:
    Sep 1, 2017
    Location:
    Sweden
    #72
    So, if this is true, are there any reason to keep using MacOS?
     
  23. jeanlain macrumors 65816

    Joined:
    Mar 14, 2009
    #73
    Metal fixes poor drivers. MacOS can perform close to windows in Metal games.
     
  24. babyjoe macrumors member

    Joined:
    Sep 1, 2017
    Location:
    Sweden
    #74
    Ok, cool! I wonder how it is with metal nowadays when it comes to rendering in Premiere or After Effects...
     
  25. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #75
    That’s exactly we try to point out!

    Compare the 580 in MacOS with OpenGL and 970 in Windows with DirectX is meaningless.
     

Share This Page