GTX560/570 user reports w/new Nvidia Drivers for OS X 10.9.2 (Aaron's post updated 3/15)
Feedback including some w/problems using the new driver update in Monday's news. I didn't see problems mentioned from GTX-285 owners (yet at least), so not sure if this is related to these (PC) models or just another case of Displayport problems in OS X 10.9.2. (The GTX560/570 was not in the supported cards list from nvidia, but previous drivers worked with them. However this may just be another case of DisplayPort problems in 10.9.2 - if possible, see if using the DVI port works.)
"GTX570 and Mac Pro 3,1
I am using an unflashed (Out of the Box) GTX570. (so no boot screen until drivers load) I had to reset my NVRAM after loading the new CUDA and web Nvidia driver update and have seen no problems. I did have it drop to one monitor. (You mean 1 monitor lost video? Did the new driver load?)
I use dual monitors, both 1080P and use DVI. All is good for me.
(previous posts from Tuesday follow)
"Nvidia OSX Driver Release 331.01.01 experience
I have a MacPro 5,1 with a GTX 560 ti card. (PC/not flashed so no video until OS loads) Works fine with standard 10.9.2 drivers, but won't boot with the new driver.
(Does it really fail to boot or just no display output? What display/connection type used?)
So warn everyone that if they don't have one of the specific cards called out, they may have trouble. Luckily, I keep a Old Mac Radeon card in the MacPro to drive a second monitor so I was able to recover by pulling the Nvidia card, going into the Nvidia pref pane and reverting to the Apple supplied driver. Put the card back in and it works fine. Was able to get cuda 5.5.47 installed though.
(he later wrote with more info.)
Follow up to my post about the Nvidia 560 ti. It is an unflashed standard PC (PCI-E) card, with it in the MacPro all I see on either monitor is a white screen (I have one screen hooked to a Mac EFI Radeon card (DVI) and the Nvidia card is hooked via DVI (DVI-I port) through a VGA adapter to a Samsung LED display (1080p). As soon as I pulled the nvidia card out it booted right up. I could tell it hadn't booted as I was unable to access it via the Apple Remote Desktop tool (didn't even see it on the network).
Thanks for your excellent site. You've saved me a number of times over the years.
Thanks Jeff, and appreciate the follow-up. And another reader that said he's also seen problems (flashed GTX570 card using DisplayPort):
"Back in 2012 I sent you a review of a MacVidCards-flashed nVidia GTX 570 running in a 2008 Mac Pro. I'm still using this card and thought I should update the review with some less-positive findings. But first, I'm having a problem with the nVidia web drivers and wanted to see if anyone else was having the same issue.
nVidia just released new Mac Pro web drivers for Mavericks (10.9.2 only) a couple days ago. The version is 331.01.01f03f01. (corrected, see his later comments below.) This is the first set of web drivers they've released since Mavericks came out. I'd like to run them, but unfortunately I'm seeing a show-stopping issue: with these drivers enabled, if I shut off my monitor, the graphics card shuts off image output and never turns it back on when the monitor is started again. In order to regain the display, I need to hard shut down my computer and then restart it. Console doesn't display any kernel panic or other relevant error messages, so the computer isn't crashing. Because of this, I've had to revert to the Apple drivers. Is anyone else seeing this problem? (see above, there's also been many cases of DisplayPort problems w/10.9.2 - some feel displayport 1.2 was changed/broken in 10.9.2.) I use a Dell 3008WFP connected via DisplayPort. I haven't tried DVI to see if that shows the same issue. And I never had trouble like this with the Mountain Lion web drivers.
Now, regarding the MacVidCards (flashed/mod for macs) GTX 570, I would now say that I would not recommend getting one of these flashed cards... if you want an upgrade, I'd recommend going for the EVGA card. Even though it's expensive, it's decently fast and at least you'll get a (manufacturer) warranty with it.
As far as the problems I experienced, it's mostly been general instability. One symptom that drove me batty was the computer occasionally getting into a mode where it (or sometimes just the video) would suddenly shut down without warning. (any info in system logs of related errors/kernel panic, shutdown cause, etc?) Once it started, it would keep happening at high frequency. I tried various things and eventually concluded that the problem seemed to be a flaky connection with one of the auxiliary power cables. Wiggling it or re-seating it would stop the problem, but it seems that heat cycling in the case (or something like that) will eventually trigger the problem again. This doesn't seem to be a problem with the card itself, though I was unsure about this for months.
Another problem but is pretty rare to encounter... Software that produces very high GPU power draw, like FurMark, appears to cause (mac pro) power supply to suddenly shut down the computer to protect itself and the motherboard. (Or is it a kernel panic/restart? Is it a total shutdown/power off?) I think I've only seen this once outside of FurMark. MacVidCards did convert the card from using 1 6-pin and 1 8-pin power connector to two 6-pin connectors. (Since 2006-12 mac pro has only two 6-pin aux power connectors for PCIe cards.)
And, of course, my "unsupported configuration" is always cited by any software developer who wants an excuse not to think about the problem.
One thing (probably unrelated to the card) that I can verify is a noticeable drop in performance after the upgrade from Mountain Lion to Mavericks. This was across the board and not limited solely to OpenGL -- the Luxmark OpenCL benchmarks also all saw a drop in performance. The mode of performance degradation is a bit strange, though. You're unlikely to notice it in static or simple scenes, like in most benchmarks (though there is some degradation -- the Trine 2 title screen at my usual settings went from 46 fps in ML to 40 fps in Mavericks). But periods in games that have a lot of action on screen have become noticeably more choppy and, in some cases, unplayable. In addition, reducing graphical complexity often does NOT have a strong effect to remedy the problem. The performance is just all-around bad and it doesn't really matter what your settings are (so you might as well turn them all the way up!).
This means that games like Bioshock Infinite and Borderlands 2 can get literally *double* the fps in Boot Camp compared to OS X at the same settings, plus they will support additional features (like higher graphics settings in Bioshock Infinite and PhysX support in Borderlands 2). I think this is a real shame. We're seeing so many great games coming to the Mac these days, but they perform so poorly that I only play them in Windows. Thanks for continuing to keep up with your site! I'd appreciate if you'd post at least the question about the nVidia web drivers on the main page.
Thanks Aaron. You're not alone it seems as far as OS X 10.9.2 and video/display (esp. displayport) problems. If you're able to try the DVI port let me know. (And also if you have any 3rd party display related addons.) Some Sapphire (Mac Edition) Radeon 7950 owners reported Displayport not working in OS X 10.9.2 (standard drivers) if the card bios switch was not set to EFI (Mac position, vs UEFI for PC). Displayport had worked fine with previous 10.9.1 drivers. Also some DisplayPort monitor users reported problems after updating to 10.9.2 - some Dell monitor owners said disabling DisplayPort v1.2 support in their monitor helped so perhaps there was some change in Displayport/1.2 in OS X 10.9.2. (Not sure if your Dell monitor has that option but just a FYI.)
(reply from Aaron, added 3/15/2014)
Here's a quick update in response to your request for me to test using DVI.
Yesterday nVidia released an update to the web drivers: 331.01.01f02. (FYI: Nvidia has a new URL to OS X driver 331.01.01F04 download, dated 4/24/2014 for OS X 10.9.2 Build 13C1021, the build number after installing Security update 2014-002 from 4/22/2014.-M) Note that my original statement about the first driver release was incorrect, I was using the f01 drivers, not f03. (now corrected in your original post above.-M) So they did a very minor update and I don't know what the differences are. I can verify though that when hooked up with DisplayPort, turning off my monitor permanently shuts off the video output from my GPU (including from DVI) until the computer is restarted. However I have also verified that the problem DOES NOT occur with DVI. And here's another interesting tidbit. If I follow this procedure:
1. Restart the computer with the monitor connected via DVI.
2. Connect the DisplayPort cable to the computer and monitor (now two simultaneous connections, OS X will detect this as two different monitors and go into extended desktop mode).
3. Switch the monitor input to DisplayPort.
4. Disconnect the DVI cable from the computer (OS X goes back into single screen mode).
Then that prevents the black-screen problem from occurring until the next restart! Very strange.
Thanks for the follow-up. (I had asked about trying the DVI port after many earlier posts (not related to these drivers/not nvidia specific) on DisplayPort problems w/10.9.2. There was also a thread on DisplayPort monitor problems where some (Dell) users disabled DisplayPort 1.2 mode support in the Monitor's OSD/menu using this tip.)
(FYI: I checked on 3/5/2014 and finally saw the download page for Nvidia CUDA for OS X v5.5.47. Yesterday's news had info on that update (via nvidia control panel) but the quadro/geforce driver update d/l page was still reverting to the previous 5.5.28 version from Oct. 2013.)
With every OS X update release there's reminders about possible problems with 3rd-party addons but 10.9.2 seems worse than usual for video/display related items. Reportedly there were notes in the last beta for devs to check for display/graphics related impacts but many feel that 10.9.2 was rushed out due to the urgency of included security patches. (This is why I think all OS X security updates should be separate releases instead of rolled into other updates. That should also reduce delays in releasing important patches and minimize impacts of other changes on users that only want/need the security fixes.) I realize sometimes security patches can also have other effects/dependencies (and there was also an Nvidia driver security fix in the 10.9.2 update) but at least as separate releases the user isn't also forced to accept other updates/changes that may have their own problems with some configurations or simply bugs in the update. (Just my opinion)