Just switched to the Green Team. After discuss with another member here. I agree that I should choose a reference 1080Ti. That should have least problem. Because Nvidia test their web driver with their own reference card.
However, the actual Nvidia brand 1080Ti is considered very expensive, really hard to find in my town, and with the lowest performance cooler. Therefore, I finally go for the PNY 1080Ti XLR8.
View attachment 733616
This should be the 2nd safest option. Which has a PCB that 100% identical to the Nvidia reference one. When I go for this route, I found that there are only 2 cards has reference PCB with an after market cooler. The Galax EXOC and this PNY XLR8. In fact, the EXOC's PCB is made by Nvidia.
View attachment 733619
But I end up still pick the PNY card. It's also a 100% reference PCB, so the chance of having incompatibility issue should be also very low. I quite like its 3 fan design. Even though it's not the quietest one, but it's not that loud anyway. Also, I quite like its "always on" fan. May be some users prefer the new "silent" feature, but I am still pretty old fashion, and prefer to keep the card cooler as long as the fan noise is not significant. (The cMP is not a silent machine at the beginning anyway).
The card is very long, but perfectly fit inside the cMP's PCIe compartment slot 1 (with the back plate ON). And this card is "just" a dual slot card. Most of the after market 1080Ti is 2.5 slot card, which I would like to avoid.
View attachment 733731
View attachment 733729
The web driver enable PCIe 2.0 as expected.
View attachment 753861
My previous R9 380 locked the refresh rate up to 120Hz, and the 1080Ti go up to 144Hz straight away.
View attachment 753859
Perform quite well in Luxmark and GB Metal test (not sure why it says the card ran at 1240MHz, but I don't really care as long as it perform).
View attachment 753860
View attachment 753862
This card works fine with just dual mini 6pin including Furmark. For some reason, even though I use a bridge between the card and the mini 6pins. The power draw is not that perfectly even (but it clearly help a lot). From my observation, the difference vary between 2-10W (update: these power draw difference is slot related, the cMP count some of the slot 1 power is drawer from Booster 2). So, since one of the mini 6pin shows the power draw is varying at ~95W under Furmark. I suspect the other one is fluctuating up to around 105W (just a bit above the display limit 97W). With another 60W from the slot. Totally ~260W, pretty much as expected. And since Luxmark and Unigine benchmarks shows the power draw rarely go above 90W from either mini 6pin. That's pretty much representing normal daily very high demand situation. Therefore, even though CUDA-Z heavy test can still shutdown my Mac (I tested it myself). I am still very happy with the power draw (and the distribution). It's well within my own comfort zone.
View attachment 753755
On Windows' side, the driver / software doing a better job. Super high performance.
View attachment 733739
And the power management setting make sure the card always stay within the decided 250W TDP.
I keep SIP and gatekeeper completely off, played few games, tried Safari, Chrome, some basic Photoshop etc, all work flawlessly at this moment. No glitch yet. Not sure if that's because I pick the right card or just lucky. Anyway, I am happy with the switching. The next challenge will be OS update. But I have the 7950 Mac Editing card on hand, nothing can go too wrong.
P.S. This card is definitely not for FCPX, BruceX 35s, and the power draw is about 50-60W in total during the test. Obviously FCPX cannot utilise the card (at least not in this particular situation).
