Report your NVIDIA Web Driver and CUDA issues and I'll pass them on to the developers

[star-affinity]

Indeed interesting - seems to happen now with my system (did not happen when i still had the gtx 970 a few weeks ago). Oh well, at least someone else also had the same symptoms

Aha – so you didn't have the wake from sleep problem on your GTX 970? Because that's the card I have and I used to have wake from sleep problems before the driver supporting Pascal GPUs was released. Maybe the problem is still there for Pascal GPUs.

 

[star-affinity]

2) Geekbench CUDA and Metal compute tests all crash as soon as they start, a dev from Geekbench claimed it's WebDriver's fault.

Which are these tests? Can you point to the software used?

 

[elmacci]

Aha – so you didn't have the wake from sleep problem on your GTX 970? Because that's the card I have and I used to have wake from sleep problems before the driver supporting Pascal GPUs was released. Maybe the problem is still there for Pascal GPUs.

Correct - well, partly.
I DID have the problem with 970 as well, but I managed to solve it by using active Display-Port to HDMI2.0 Adapters (Club 3D DP2HDMI 4K 60Hz).

Somehow, these Adapters helped the GTX 970 to correctly identify the connected Monitors even after sleep.
Unfortunately, this method does not work anymore. My guess is that the pascal drivers somehow interfere with a correct "handshake" between GPU and Monitor - Maybe an EDID problem? Obviously the Webdriver does not recognize the monitors after sleep and gets stuck.

 

[elmacci]

Which are these tests? Can you point to the software used?

I also had this issue. The Software used is Geekbench 4, a benchmarking program. When running the "Compute Benchmark" with Compute Api set to OpenCL, the Benchmark would begin but abort at the step "Histogram Equalization" after just a few seconds.
In the Support forum of Geekbench (from a company called Primatelabs) the Developers claim that it is a problem with the webdriver.

After figuring it out and getting it to run i do believe to understand why the devs from primatelabs say it is a webdriver issue.
It is more or less a webdriver/apple issue.
You see, due to the way macos handles drivers it is not allowed for any third party (=nvidia) to use certain apis of macos.

It has something to to with LibraryValidation used by certain apps (for example ibooks). This LibraryValidation is only possible when the drivers are signed by Apple. Which the webdrivers are not (signed by nvidia). It is questionable if apple ever will sign nvidias Drivers as "Apple"-approved.

But - after installing the ibooks fix (from here - Lilu.kext needed) - Geekbench is running fine.
This fix validates the webdrivers as being signed by apple and does not only fix the transparancy bug in ibooks, but geekbench as well.

Sorry for my english, not native - any errors in my explanation might have been lost in translation

 

[MacPack]

If it's possible. HDMI 2.0 and DisplayPort 1.4 support for our Pascal cards.

Apple is waiting on Intel to integrate it into their CPUs, that won't be until 2019 at the earliest.
Hopefully with a new appleTV having 4k HDR support will find it's way into the shared codebase.

 

[dmrkillah]

I also have a problem with monitors waking from sleep / not even starting randomly on boot:

Hello everyone!

I have just recently upgraded from GTX 970 to Zotac GTX 1080 TI and that's when the problem started.

I've got a somewhat uncommon setup as I have got 4 displays connected to my hack:

1 - ASUS 27" 4K monitor via DP
2 - Samsung 27" 1080p monitor/tv via HDMI
3 - Manta 22" 1050p monitor via DVI
4 - Panasonic 4K tv via HDMI2.0/DP active adapter

The problem is that everytime after wake from sleep the connection to nr. 4 Panasonic 4k TV is lost. It isn't detected anymore as a display.

It sometimes randomly happens after restart too, although most of the time it works fine.

If I try to sleep again, then my hack completely freezes after wake and monitor nr 1 isn't turning on.

Does anyone have any idea for a fix/workaround for this issue?


Spec: Asus Maximus VIII Hero, i7 6700K, 16GB DDR4, Zotac GTX 1080Ti AMP Extreme
SMBIOS: iMac 14,2

Everything worked fine on GTX 970.

The whole thread: https://www.tonymacx86.com/threads/pascal-and-multi-monitors.227320/

I would be very grateful if you'll report this to nVidia

 

[nagha]

4k/HDPI via DP1.2 at 60hz is my dream wish for my Nvidia 950 GTX. Please please pretty please.

 

[star-affinity]

I also had this issue. The Software used is Geekbench 4, a benchmarking program. When running the "Compute Benchmark" with Compute Api set to OpenCL, the Benchmark would begin but abort at the step "Histogram Equalization" after just a few seconds.
In the Support forum of Geekbench (from a company called Primatelabs) the Developers claim that it is a problem with the webdriver.

After figuring it out and getting it to run i do believe to understand why the devs from primatelabs say it is a webdriver issue.
It is more or less a webdriver/apple issue.
You see, due to the way macos handles drivers it is not allowed for any third party (=nvidia) to use certain apis of macos.

It has something to to with LibraryValidation used by certain apps (for example ibooks). This LibraryValidation is only possible when the drivers are signed by Apple. Which the webdrivers are not (signed by nvidia). It is questionable if apple ever will sign nvidias Drivers as "Apple"-approved.

But - after installing the ibooks fix (from here - Lilu.kext needed) - Geekbench is running fine.
This fix validates the webdrivers as being signed by apple and does not only fix the transparancy bug in ibooks, but geekbench as well.

Sorry for my english, not native - any errors in my explanation might have been lost in translation

Interesting – I just tried it now on my real Mac Pro (Mid 2010) with a GTX 970 and the benchmark seemed to have gone through just fine.

https://browser.geekbench.com/v4/compute/1026200

Wonder if it would be different if I had a Pascal GPU… Maybe I have to get one.

No worries about your English – I understand you fine. I actually think you write quite well, but I'm not native either so what do I know.

 

[redfoxhound]

This sort of explains my issue where the driver is not signed by apple. With his new build I have GTX 1060 and certain apps would freeze the screen. Audio is still working when playing music as well as my mouse can still be moved. However everything else on screen is frozen.

I also have the issue with the 4k monitor where the UI get's jumbled up after wake. All I do is turn on/off the monitor and it goes away. Connected via display port.

So far I've experienced this freeze while launching Skype and Coda a code editor.

With Coda, the text editor would be fine but slowly the text will delay as I type. Then it will freeze the screen.

Hardware:
1060 GTX
i7700
Asus Z270e
Mac 14,2

 

[silvercircle]

Wake from sleep:
Never had an issue with my dual monitor setup, Dell monitor connected via DP (2560*1440) and Samsung DVI(1600*900)
Problems began when I swapped my samsung monitor with a 4k monitor Asus MX27uc via DP.
Sometimes the 4K monitor remains black, the other monitor wakes normal.
It is really random, most of the time both monitors wake up.
When I use HDMI for the 4k monitor it is better but then sometimes the 4k monitor shows a 'snow' screen.(like television in the old times when no cable was connected).
Turning the power off on the Asus after wake and turn it back on will fix it most of the times.

Card is a GTX 1080ti Asus 11GB, i7 6700k.