Contribute
Register

Post OS X Luxmark OpenCL Benchmarks

Status
Not open for further replies.
Yes, at $250, it's a lot of bang for the buck especially for FCPX users. I got in on a pretty good deal on my GTX 1080 towards the beginning of the year for $430. I liked it's low TDP and how cool it ran. A single 8-pin connector powers it and I don't think I've ever seen it go much beyond 50C.
 
Yes, at $250, it's a lot of bang for the buck especially for FCPX users. I got in on a pretty good deal on my GTX 1080 towards the beginning of the year for $430. I liked it's low TDP and how cool it ran. A single 8-pin connector powers it and I don't think I've ever seen it go much beyond 50C.

Right around the time of the 1080ti release, perfect time to buy, also before the mining craze took off.
 
Those AMD cards are really awesome for OpenCL but still trail a little in OpenGL, at least in macOS. For comparison's sake, here's Valley on my GTX 1080.
View attachment 284278
Not trying to off topic but this shows how conservative WebDriver on macOS is. My 980ti with Voltage&Power unlocked plus some tweaks almost catches up a stock 1080. Still, not at it's full potential like on Windows(1.318GHz vs 1.57GHz)
螢幕快照 2017-10-10 下午4.25.04.png


Also here's my Luxmark result on Windows, even with extremely OC, still can't beat RX480 on openCL. Apple doesn't choose AMD for no reason.
圖片-2017-10-10~16-54-34.png
 
From what I remember, LuxMark scores between Windows and macOS are pretty close.
 
From what I remember, LuxMark scores between Windows and macOS are pretty close.
That's correct if the card is running at the same clock speed. But from what I've experienced, WebDriver on macOS doesn't support nVidia's "GPU Boost".

So in my case, I set 1.57GHz on my card for absolute max speed in vBios. On Windows it runs exact as I set, but on macOS it won't go any higher than 1.318GHz.

Maybe this is not the same case on AMD platform or Pascal architecture?
 
I'm not sure how GPU boost works in macOS. I believe the only way to get working GPU overclock is by modding the video card's firmware.
 
AMD knocks nVidia out of the park.. in OpenCL. But what Nvidia lose in OpenCL they gain in CUDA.. by a lot. Check this out:

http://barefeats.com/vega_resolve.html

Vega 64 beats 1080 in OpenCL, yes. In DaVinci Resolve Candle light test by 5 FPS. In Geekbench GPU Compute tests by roughly 15000 points.

But in another test, they try CUDA performance of 1080ti and 1080:
http://barefeats.com/early_vega.html

1080 gets 36 FPS, same as Vega 64, and 1080ti achieves 54 FPS. That is a huge gain. And those tests were run under Sierra, as opposed to the newer test that was on High Sierra, where the Nvidia cards saw a large OpenCL increase. It is possible there is a CUDA increase in score as well in High Sierra, not tested.

And CUDA GPU Compute tests from 1080 gan easily hit towards 200.000.

I am NOT a fanboy (people staying "loyal" to companies are **** tards) but there is evidence that the large lead AMD gets in OpenCL only apps is not fair.. because most apps can be run with CUDA, and CUDA yields far better results for Nvidia.
 
Status
Not open for further replies.
Back
Top