netkas.org forum
July 22, 2018, 12:17:19 AM *
Welcome, Guest. Please login or register.
Did you miss your activation email?

Login with username, password and session length
News: Information for registering users http://forum.netkas.org/index.php/topic,2246.0.html
 
   Home   Help Search Login Register  
Pages: 1 ... 14 15 [16]
  Print  
Author Topic: High Sierra and The Nvidia Web Driver  (Read 39147 times)
mysticalos
Hero Member
*****
Offline Offline

Posts: 630


« Reply #225 on: April 19, 2018, 12:00:14 AM »

Based on comments with https://github.com/lvs1974/NvidiaGraphicsFixup/commit/9ddc07558737d5085f665002dda94b636958e47b

It seems the new web driver is forcing legacy opengl path internally? I hope that doesn't mean those who weren't having issues are suddenly using opengl instead of metal for windowserver? Hard to really say, I don't recall how to test for that. I also hope it's not a perm work around but just one nvidia is applying for now until they get the metal code more solid.
Logged
Asgorath
Sr. Member
****
Offline Offline

Posts: 295


« Reply #226 on: April 19, 2018, 03:43:29 AM »

No, that's not what it means, based on my reading of the diffs.  The driver hack was apparently taking an older implementation of some internal C++ function, and substituting it for the newer one.  The latest driver appears to have gone back to this older implementation (or something like it) along with a bunch of other binary changes, so the author is claiming that he no longer needs to make the swap in his kext.  The "force OpenGL" thing was separate and unrelated to this particular part of the hack.

I would be very surprised if the fix for the stuttering was contained in just that one function, my bet is that there was some driver path that had issues and thus NVIDIA had to turn it off.  The kext was hacking the driver by swapping back in the older version that had the path enabled.  While this fixed the stuttering/lag issues for most people, I would bet that there were still other issues that people may or may not have been hitting.  If NVIDIA fixed the actual underlying problem, then they could turn the original path back on and everyone is happy.
Logged
Rominator
Hero Member
*****
Offline Offline

Posts: 2315



« Reply #227 on: April 22, 2018, 10:06:33 AM »

Hopefully the color bug is fixed.

I can report that the "lower card has driver issues" bug remains.

Can be triggered by a GT640 in slot 3 and a Pascal card in slot 1.

Display output on lower card has opaque dock and poor driver response, but CUDA works.

With any luck, they can fix in next update.
Logged

Before asking a question, check your "Personal Settings" and be sure that you have "Brain Services" set to "On".
ClassicMartini
Not Newbie
*
Offline Offline

Posts: 23


« Reply #228 on: April 23, 2018, 07:28:30 AM »

Reading about update on reddit it fixes the lag issues too

Important to note about person recommending lilu and fixup. FYI the workaround they were using for lag was to force opengl mode instead of metal on windowserver. If the drivers fix bug, you WANT to be using metal not opengl. I generally avoid lilu in general too cause every macos update people update to new version and start having kernel panics until lilu is updated. It's a pretty low level kernel hack so it always needs updates too before updating macOS (and always someone doesn't listen).

It also taints nvidia driver so libval is no longer valid (which is also why fixup also does libval work around)

TLDR though, if you don't NEED lilu or any of fixups, it's best to run as vanilla as possible so libval is clean and nvidia drivers run the optimized metal path. Only use work arounds if it doesn't work otherwise, Don't use them cause everyone on and their grandma on tonymac said so.

I run a hack and it's very clean.

Can't lie: I'm the 'culprit'. :-)

Got the tip from one of the clover devs - Apianti - in a recent "clover-general-discussion" thread at IM.

IM is not as beginner friendly as TM: they tend towards the "how" (once its been sorted/perfected at IM) rather than the "why".

Thanks for your insights. Perhaps you can go one better and explicitly/comprehensively state what "architecture" worked for you in newbie friendly terms?

Maybe just a link to your build? (So we don't derail this thread?).

Cheers.
« Last Edit: April 23, 2018, 07:30:37 AM by ClassicMartini » Logged
lowrider
Sr. Member
****
Offline Offline

Posts: 362


« Reply #229 on: April 26, 2018, 12:23:27 AM »

New Web Driver for 10.13.4 (build 17E202) is here"

https://images.nvidia.com/mac/pkg/387/WebDriver-387.10.10.10.30.107.pkg

Lou
Logged
lowrider
Sr. Member
****
Offline Offline

Posts: 362


« Reply #230 on: May 17, 2018, 05:06:48 PM »

New Cuda Driver out this morning:

http://www.nvidia.com/object/macosx-cuda-396.64-driver.html

Lou
Logged
lowrider
Sr. Member
****
Offline Offline

Posts: 362


« Reply #231 on: June 02, 2018, 12:42:55 AM »

OS 10.13.5 is now released Here:

https://support.apple.com/kb/DL1965?viewlocale=en_US&locale=en_US

Nvidia Web Driver is Here:

https://images.nvidia.com/mac/pkg/387/WebDriver-387.10.10.10.35.106.pkg

Lou

Logged
lowrider
Sr. Member
****
Offline Offline

Posts: 362


« Reply #232 on: July 09, 2018, 05:16:34 PM »

New Cuda Driver out this morning:

http://www.nvidia.com/object/macosx-cuda-396.148-driver.html

Lou
Logged
lowrider
Sr. Member
****
Offline Offline

Posts: 362


« Reply #233 on: July 10, 2018, 04:52:57 AM »

10.13.6 out.  New Nvidia Web Driver here:

https://images.nvidia.com/mac/pkg/387/WebDriver-387.10.10.10.40.105.pkg

Lou
« Last Edit: July 10, 2018, 04:54:53 AM by lowrider » Logged
apple pie
Newbie

Offline Offline

Posts: 1


« Reply #234 on: July 19, 2018, 04:32:44 AM »

Has anyone made progress getting any driver after 378.10.10.10.25.106 to fully work (OpenGL/CL) with dual GPU's? Thanks
« Last Edit: July 19, 2018, 04:34:18 AM by apple pie » Logged
Pages: 1 ... 14 15 [16]
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.21 | SMF © 2015, Simple Machines Valid XHTML 1.0! Valid CSS!