Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Hmm, not sure why your estimate on the m395x is so low, my scores were considerably higher. 107.75 when in osx and 139 when in windows. Here's the bootcamp screenshot if you need it.
 

Attachments

  • Capture.PNG
    Capture.PNG
    70.7 KB · Views: 386
Hmm, not sure why your estimate on the m395x is so low, my scores were considerably higher. 107.75 when in osx and 139 when in windows. Here's the bootcamp screenshot if you need it.

I always saw a big difference on the opengl score in Cinebench but is your cpu score similar?
 
System: iMac 27" Late 2013
CPU: core i7 4771 3.5 Ghz
RAM: 8GB
HDD: Samsung SSD SM0512F PCIe
GPU: Nvidia geforce gtx 780m 4GB
OS: OS X El Capitan 10.11.1
Benchmarks with native driver
cinebench macosx native driver.png heaven macosx native driver.png
Blender with cuda driver for mac 7.5.21
blender bmw27 native driver.png i got 3:00:86
 
  • Like
Reactions: phi7
System: iMac 27" Late 2013
CPU: core i7 4771 3.5 Ghz
RAM: 8GB
HDD: Samsung SSD SM0512F PCIe
GPU: Nvidia geforce gtx 780m 4GB
OS: OS X El Capitan 10.11.1
Benchmarks with nvidia web driver 346.03.03f02
cinebench macosx nvidia webdriver.png heaven macosx webdriver.png blender bmw27 webdriver.png
OS: Windows 10 PRO x64
Nvidia driver 358.50; GPU Stock Freq ( 784 Mhz core clock, 2500Mhz Memory Clock)
Cinebench GPU Stock.png heaven gpu stock.jpg blender bmw27 GPU Stock.png
OS: Windows 10 PRO x64
Nvidia driver 358.50;GPU overclocked with MSI afterburner 4.1.1 ( 985Mhz Core clock, 2810Mhz Memory Clock)
Cinebench GPU with MSI aftebrurner.png heaven Gpu msi afterburner.jpg blender bmw27 GPU with msi after burner.png
 
System: iMac 27" Late 2013
CPU: core i7 4771 3.5 Ghz
RAM: 8GB
HDD: Samsung SSD SM0512F PCIe
GPU: Nvidia geforce gtx 780m 4GB
OS: OS X El Capitan 10.11.1
Benchmarks with nvidia web driver 346.03.03f02
View attachment 597446 View attachment 597447 View attachment 597448
OS: Windows 10 PRO x64
Nvidia driver 358.50; GPU Stock Freq ( 784 Mhz core clock, 2500Mhz Memory Clock)
View attachment 597450 View attachment 597451 View attachment 597452
OS: Windows 10 PRO x64
Nvidia driver 358.50;GPU overclocked with MSI afterburner 4.1.1 ( 985Mhz Core clock, 2810Mhz Memory Clock)
View attachment 597453 View attachment 597454 View attachment 597455
It's always amused me that my 680MX also overclocks almost just as well and hardly gets any hotter. When the 680/780MX is overclocked the m295/m395x is either equaled or beaten except in vram limited situations (which only really applies to the 2gb 680MX in mine)
 
System: iMac 27" Late 2013
GPU: Nvidia geforce gtx 780m 4GB
OS: OS X El Capitan 10.11.1
Blender with cuda driver for mac 7.5.21
View attachment 597443 i got 3:00:86

Thank you for doing the Blender Benchmark.
Now would someone be so kind and do the same with an AMD GPU..

Reference:
CPU: Intel i7 3.8Ghz
GPU: Nvidia GTX 770
OS: Windows 7 64bit
Time: 1 min 50 seconds (GPU - CUDA)
 
last time I tried the blender benchmark with an amd card, it crashed. Maybe the successes with nvidia cards have relied on a more mature cuda codebase?
 
Here we go.
Heaven under OS X 10.11: 21.2 fps
Heaven under Win 10: 32.4 fps
Cinebench 15 under OS X 10.11: 107.04 fps
 

Attachments

  • Screen Shot 2015-11-02 at 8.39.44 PM.png
    Screen Shot 2015-11-02 at 8.39.44 PM.png
    181.8 KB · Views: 240
  • Screen Shot 2015-11-02 at 8.40.57 PM.png
    Screen Shot 2015-11-02 at 8.40.57 PM.png
    193.1 KB · Views: 219
  • Screen Shot 2015-11-02 at 8.37.58 PM.png
    Screen Shot 2015-11-02 at 8.37.58 PM.png
    227.8 KB · Views: 225
  • Like
Reactions: iemcj
Seems that windows runs things consistently around 30% -35% better. Which seems surprising cuz you'd expect OSX to be optimized for it's own hardware but who knows. Those are actually quite good scores in Heaven, really only beat out by high end gaming specific gpus.
 
you'd expect OSX to be optimized for it's own hardware

You'd expect that, yeah.
Apple is typically several versions of OpenGL behind Windows. And to the extent the spend any time at all optimizing their GPU drivers, it's to make pro apps, UI, and GPGPU stuff faster. They have zero ****s to give about gaming on OS X, which is why those trying to get the most gaming out of a Mac typically dual-boot.
 
You'd expect that, yeah.
Apple is typically several versions of OpenGL behind Windows. And to the extent the spend any time at all optimizing their GPU drivers, it's to make pro apps, UI, and GPGPU stuff faster. They have zero ****s to give about gaming on OS X, which is why those trying to get the most gaming out of a Mac typically dual-boot.

phoronix did some testing, it's not just opengl, running linux bench on a mac is also faster than OSX.

link

OSX was way slower ..
 
phoronix did some testing, it's not just opengl, running linux bench on a mac is also faster than OSX.

Yeah that's the point I was trying to make - everyone has better optimized drivers than Apple, and they're also using an older version of OpenGL...

I'm hopeful that further development of Metal will improve things, but I'm not optimistic that the effort will yield much improvement in OS X gaming.
 
Yeah that's the point I was trying to make - everyone has better optimized drivers than Apple, and they're also using an older version of OpenGL...

I'm hopeful that further development of Metal will improve things, but I'm not optimistic that the effort will yield much improvement in OS X gaming.

indeed! but with Metal, it will optimize the graphics layer but they will still lag in pure computing power. I hope they work on that one day.
 
My results from the heaven benchmark in bootcamp.
32.2fps on the settings requested in the original post. So far higher than what the original poster got in osx.
 
Last edited:
How do you get benchmark version 4.0?

edit: please disregard. Posting heaven benchmark soon for m395.

7r5fHCn.png
 
Last edited:
Heaven 4.0 as opposed to Valley 1.0, I should think.

Anyway, assuming that no one has a m290 to test, the benchmarking is pretty much complete..

Except that OP hasn't updated the first post with some significantly better - and consistent - results...
 
Nvidia card of course can drive 5k display, it works in the Mac Pro. So, I am sure they can work in the iMac if Apple want to do it.

http://barefeats.com/imac5k17.html

Desktop GPU's can with dual DP 1.2 but Mobile series cant, single DP 1.2. I'm sure that Apple did make an offer to Nvidia also for iMac's and they would get 970M/980M to work with 5k. AMD just was cheaper and better with OpenGL/CL.

Bsa1i7O.jpg


Desktop:

http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-980/specifications
 
Last edited:
Also, the m380 has 12 compute units; the m390, 16; the m395, 28; the m395x, 32. If you need openCL, it's kind of a clear progression.

Not if you look at the Benchmarks on Barefeats.
The openCL difference is only 10% between the 390 and 395, while it is about 40% between the 380-390 and 18% between the 395-395x. So the difference isn't correlating to the number of compute units.
 
The 395x is going to make a bigger difference in something like video editing than it will in gaming.
 
The 395x is going to make a bigger difference in something like video editing than it will in gaming.

Depends on the game and settings. Granted this test was on a windows machine, but Heaven benchmark w/ 8xAA on 4k was consuming just shy of 3GB of VRAM. Higher res gaming will demand more. And then you have some poorly optimized games like (modded) Skyrim that just chew through VRAM. It could be the difference in being able to have certain settings/mods, or not.

Edit: Some AAA titles even in 1440p can consume quite a bit: http://www.tweaktown.com/tweakipedia/90/much-vram-need-1080p-1440p-4k-aa-enabled/index.html
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.