I'm going a bit back and forth between hope and despair reading these latest post... can someone summarise if we've seen a real change since 10.12.1? Is the RX 480 a valid choice now? Did the name show up correctly by editing the kext?
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.
Dual RX 480s in Mac Pro 5,1 running MacOS (w/ How-to)
- Thread starter theitsage
- Start date
- Sort by reaction score
Also a little confused, as far as i can tell it looks solid on the latest Sierra build, just missing boot screen?
Theres a thread over at the blackmagic forum claiming the RX 480 outperforms the Titan X in Resolve: https://forum.blackmagicdesign.com/viewtopic.php?f=21&t=49713
Can that be right!? Test is on a windows machine so would we see the same on a Mac Pro 5.1?
Theres a thread over at the blackmagic forum claiming the RX 480 outperforms the Titan X in Resolve: https://forum.blackmagicdesign.com/viewtopic.php?f=21&t=49713
Can that be right!? Test is on a windows machine so would we see the same on a Mac Pro 5.1?
It is possible. RX 480 has 5.8 TFLOPs of compute power and Titan X 6.1 TFLOPs, so the difference is meaningless. Properly coded software will reflect this. In this situation: major role can play drivers, or CPU overhead, or simply CPU bottlenecking the Titan X.
Compute power has always been less than Windows due to Mac drivers not being updated or optimised regularly. APIs are also older on macOS.
I'm very confident support for Polaris GPUs is here to stay. 10.12.1 official release will most likely come with new hardware. I poked around in AMD9500Controller.kext last night to change AMD RX 480 name in "About this Mac" and saw tons of references to Polaris 10 and Polaris 11.
CS6 introduced OpenCL for some operations. It doesn't have the improved graphics acceleration and further optimisations of CC such as the Mercury Engine. At most you will see improvements for batch image resizing.
It would be a good test to see if the Baffin driver can perform complicated OpenCL actions without error. The Nvidia web driver can't.
[doublepost=1475067298][/doublepost]
The question is will you have to keep hacking kexts with every OS update. It was a chore enough when I had to deal with web drivers and Maxwell every time there was an update. If Apple adopts a variant of 460/470 you might still have to hack in the 480.
Phillip here from the forum has done tests on Windows and he has got a YouTube video up. The 'real world' chart in that specific test is a bit off though. Here's a corrected one:
We will see less than that on macOS. If application Metal support or newer CL comes then we will see closer to Windows performance.Phillip here from the forum has done tests on Windows and he has got a YouTube video up. The 'real world' chart in that specific test is a bit off though. Here's a corrected one:
I got an R9 Nano a few days ago. It's a better card and build quality is impressive for such a small package. I can't say the same about performance relative to cost though. Same kext edit in order to make use of Fiji GPUs in Sierra. More info on post #98.
weird that the 1070 and 1080 seem to outperform the Titan X in relation to the link i pointed to earlier (https://forum.blackmagicdesign.com/viewtopic.php?f=21&t=49713)Phillip here from the forum has done tests on Windows and he has got a YouTube video up. The 'real world' chart in that specific test is a bit off though. Here's a corrected one:
Or is that not weird?
Ignore me i didnt know much about the 1070 and 1080 but have done a bit of research and can see its not weird at all that they would outperform a Maxwell Titan X. Im a filmmaker just trying to build a resolve machine, new to this!
Im gonna push the button on the RX 480 for my Mac Pro 5.1 rather than a used Titan X for double the price. Can always sell it if it isnt working out.
Why it would not outperform it? GTX 1070 - 6.5 TFLOPs vs 6.1 TFLOPs Maxwell Titan X. GTX 1080 - 9.3 TFLOPs of compute power vs 6.1 TFLOPs of Titan X. Why it would not outperform the 28 nm Maxwell top-end part?
thanks soycapitan, just asked as CS6 is not supported now so things can brake.
at the mo iv got a 660 2gb which works for what id do, Ill wait till support/drivers & osx12 is more concrete before i make the jump (if the RX 470/480 is flakey in CS6 premier & AE then ill have to get a GTX 770 4GB
).
I gess as you mentioned there will be a benefit from vram in timeline but not a massive speedup & as shown in benchmarks in 1080p you only need 2GB it's not till 4K that you start seeing real gains from GPU power.
at the mo iv got a 660 2gb which works for what id do, Ill wait till support/drivers & osx12 is more concrete before i make the jump (if the RX 470/480 is flakey in CS6 premier & AE then ill have to get a GTX 770 4GB
).I gess as you mentioned there will be a benefit from vram in timeline but not a massive speedup & as shown in benchmarks in 1080p you only need 2GB it's not till 4K that you start seeing real gains from GPU power.
I received PMs for instructions so I wrote this Mac Pro Radeon RX 480 How-to.
Thanks for the guide. Just a little suggestion, may be it's better to be more precise that we have to Hex edit the Controller file inside the AMDxxxx.kext/contents/MacOS to edit the "About my Mac" into.
...aaaaand, with my card inbound.... is there anything stopping me from doing the prep work already now? These should be sticky settings, right—even after rebooting and such?
I also don't see this conflicting with my current 280X.
It would be nice to just drop the new card in and GO.
I also don't see this conflicting with my current 280X.
It would be nice to just drop the new card in and GO.
You can edit the required kext file now, it won't affect your current setup. And make the 480 become plug and play.
Are there are any OpenCL error calculations in the Luxmark report? Also, how is the Metal test looking in GFXBench?