QUESTION: What are your thoughts on the 1080Ti vs 580? Which do you prefer and why? Why would you need both, other than for testing/developing purposes?
On cMP, RX580 is a better choice most of the time. Unless the software can utilise CUDA. If the user don’t need CUDA, and doesn’t really need high end GPU’s power. I usually recommend the RX580. Mostly because the RX580 isn’t that much slower on the cMP for most operation, but way cheaper, and has native support, much easier to maintain.
I don’t really need any of them. I am a lucky guy, has reasonable affordability, so that I can buy both to have some fun.
My 1080Ti is now in a Hackintosh, but mainly boot to Windows (becasue my cMP still working, that Hackintosh is just my backup MacOS computer). No professional usage at all, mainly use it for gaming and video encoding (Windows). I really love the 1080Ti’s video engine, it’s super fast and high quality. But that’s not available in MacOS. This is the main reason why this Hackintosh mainly boot to Windows.
The RX580 is in the cMP now for day to day use (my cMP run 24/7). TBH, both 1080Ti (with V387 driver) and RX580 provide the same smoothness on my monitor for almost everything. The monitor is Samsung CHG 90, I let the UI render at 7680x2160 @144Hz. So, obviously both cards are way over power for normal usage even up to 4k 60Hz.
I also do some video editing in FCPX, in this case, RX580 of course way faster due to much better optimisation. And this is the main reason why I buy the RX580 when I decided to move my 1080Ti to the Hackintosh. My Apple 4870 can still drive the UI at 3840x1080 (I couldn’t remember if it can run at 120Hz, but 60Hz is definitely not an issue at all). However, I want a better card for FCPX (I still occasionally do some 4k video editing. And I expect will pick that up more in future), also able to render better UI.
On the other hand, when I perform some Photo de-noise by the Neat noise removal filter. I can easily tell that the 1080Ti is significantly faster.
I rarely use PP, or DV now. I tried, but for my own usage, FCPX really the better choice. The 1080Ti performed quite well in PP or DV, but I don’t know how good the RX580 is in these software, no time to test it yet.
It’s easier to fit the Sapphire PULSE RX580 in the cMP due to the single 8pin design. My PNY XLR8 1080Ti is 100% reference, it can run anything (including Furmark, but not CUDA-Z heavy test), with just the mini 6pins. But obviously power spike can be a protential issue (CUDA-Z proved that).
I moved the 1080Ti from cMP to the Hackintosh because it fit better there (which is a 8700K machine). The Hackintosh won’t have the black screen issue (if do it right), so running a 1080Ti is a piece of cake. Also, the 8700K can better release the 1080Ti’s power. Even just for video encoding, as long as my encoding parameter need CPU assist, the 1080Ti can give me 50% more FPS in the Hackintosh than in the cMP due to CPU single thread performance differences.
Of course, gaming on the 1080Ti (Windows) is better than RX580.
Even though the 1080Ti is a better card in almost all areas. However, on the cMP (MacOS), it’s ability is suppressed by the CPU performance and the OS / driver support. So, unless the target is its CUDA performance. For most other usage, 1080Ti can hardly be a better choice on cMP if we also put price into the consideration (the RX580 has much better performance to cost ratio in MacOS most of the time).
