Ok, I have results for 1 x 1080, 1 x 1070 and 2 x 1070
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Webdriver for GTX 1080&1070
- Thread starter chenfromchina
- Start date
- Sort by reaction score
hmm, no performance gain for current webdriver for Pascal... cpu or driver bottleneck under osx? we will see
980Ti in cMP scored between 1600 - 1800 in Heaven and Titan X (Maxwell) scored around 18000 in Heaven. So there is a performance increase from Maxwell to Pascal in the cMP. In your Hackintosh machines with i7's it may be different but in my 3.46GHz 12C setup Pascal is better. Once they release the CUDA driver performance in Pro Apps I would also expect to be an improvement.
The other bit of good news is you can easily install 2 1070 or 1080's off internal power (I now await many messages saying this isn't safe as each GPU requires 1 x 8pin). As you can see in the LuxMark test 2 x 1070's score over 30000. The nMP with dual D700's scores just over 20000 in LuxMark.
I'm receiving an eGPU rig in the next week I'm looking forward to running these tests with the 2.7GHz 12C nMP to see what improvements I get in GPU performance.
The other bit of good news is you can easily install 2 1070 or 1080's off internal power (I now await many messages saying this isn't safe as each GPU requires 1 x 8pin). As you can see in the LuxMark test 2 x 1070's score over 30000. The nMP with dual D700's scores just over 20000 in LuxMark.
I'm receiving an eGPU rig in the next week I'm looking forward to running these tests with the 2.7GHz 12C nMP to see what improvements I get in GPU performance.
Try running Furmark with two GTX 1080s.
GpuTest - Cross-Platform GPU Stress Test and OpenGL Benchmark for Windows, Linux and OS X | Geeks3D.com
There is a performance gain for Pascal over Maxwell. At the moment it looks like the 1080 is the best card to choose but if the driver can be improved or you need more VRAM maybe Ti and Titan Xp would be an option.
[doublepost=1491922683][/doublepost]
I'm unfamiliar with that test but I have run a permutation
Furmark is a good test of stability in regards to power. Extremely power hungry GPUs have been known to shut down Mac Pros when running this benchmark.
C'mon guys, are you really running Unigine Heaven at 1600x900??? This is going to be absolutely 100% CPU limited on macOS. Why don't you try running it at 4K, or even 2560x1600 or something. This might give the GPU a chance to stretch its legs despite being limited by Apple's terrible OpenGL framework design. Or, focus on Metal benchmarks since those should run much, much better due to the low overhead nature of the API.
This CUDA driver should work. Has anyone tried it? I have to wait until tonight to get my GTX 1080 up and running in my hack.
With Apple-Shift-4 you can make screenshots of a selected area on OS X. Don't have to photograph.
Last edited:
I read more about the bottleneck issue. Now I understand. Your are right sir. All cards are bottlenecked somehow. If I'm correct, all that matters is full potential of the gpu (saturation of the pcie for pro apps) and stable 60fps in games. More demanding games or applications will use more potential of the card that is available. High, 200+ fps only matters on high frequency monitors. So I'm gonna buy 1080ti, it will be the most future proof gpu.
Right, this is why my FAQ says "do not use Cinebench as a GPU benchmark". At some point, the CPU simply cannot keep up with the GPU. This happens very quickly when you have a slow CPU (e.g. in the cMP) or a very, very fast GPU (e.g. a high-end Pascal card). If you were to run Unigine Heaven/Valley at 4K resolution, the GPU would have a ton more work to do (since it's rendering more than 5x the number of pixels compared with 16x9). You'd be able to very clearly see that the GPU is mostly idle if you ran the OpenGL Driver Monitor and enabled the "GPU Core Utilization" stat in Linear display mode while running a CPU-limited benchmark like low-res Unigine or Cinebench.
Yep, that's it. 4k vs 1080p should clearly show that even on cMP I would get stable 60fps and CUDA will saturate all my PCIe bandwidth. I don't need 200+ fps, so I will be happy to have stable 60fps on 1080p for many years and 4k in the future. All that matters is performance in Octane for me. I just wanted to check really if the 1080ti would hit bottleneck in LuxMark or Octane, so performance would be the same as 980ti, so it will be more optimal to buy 980ti for less.
Why are so many people doing 1600x900 windowed? Is that the default setting? A common standard for comparison?
Do you guys think a GTX 1050 Ti or GTX 1060 will be a worthwhile upgrade, for me, $$$-wise, coming from an RX 460, if, all I care about is performance in FCP X?
You need 2 monitors, each attach to one of the GPU, and run two Furmark (at the same time), each on one screen. Otherwise, only one card will be stressed, but not both at the same time. Also, Furmark will warm up the card slowly. And the card will draw more power once warmed up. So, for stability test, you need to run Furmark x2, and keep them run at full screen (on both screen), for at least 15 min to reach the stabilised point in order to determine if the setup is stable enough.
[doublepost=1491946908][/doublepost]
For FCPX, usually AMD card works better (for the same cost).
So, you're saying keep using the RX 460? I'm only asking because I do notice that my system is completely not smooth editing 60P 28M (PS) 1080P video from a Sony NEX-6 camera. Not sure, if, it's because I upgraded to a 1440P monitor. But, I do notice FCP X not keeping up with me as I work on the video. But, it's workable. It's just not smooth all the time. Do you know what I mean?
I was just wondering, if I need a little more "juice" in the GPU department to make it smooth again.
PS-My system is a mid-2010 Mac Pro, 3.33 6-core, 16MB RAM, RX460, latest MacOS Sierra and FCPX... I also have an SSD for my scratch drive....
My understand is that your monitor's resolution is quite irreverent to the video editing performance.
And since you emphasis the NEX-6. So, I assume normally your Mac Pro edit 1080P video just fine. If that's the case, may be it's the codec's problem. Did you ever try to use proxy?
+1
"Edit" can mean different things. If you're applying effects and doing color correction then you should use optimization (ProRes) or even proxy files for some of it.
"Edit" can mean different things. If you're applying effects and doing color correction then you should use optimization (ProRes) or even proxy files for some of it.
