Contribute
Register

Nvidia Drivers for Mojave 10.14 Not Available, no WorkAround works well

Status
Not open for further replies.
Kepler support was always out of the box, because they used it in laptops and imacs

It is supported, but it has bugs. For example, the bugs with 650 GT where the graphics glitches or pixelated, still in Mojave.

Tbh, some of the Kepler cards are dead in the water since Metal released. Neither Apple or Nvidia cares fixing the issues. Time to move on to Intel or AMD card.
 
This all goes back to Nvidia being unable to write drivers that comply to what's required to be compatible with macOS. They need to do a better job at writing drivers that won't break things like Little Snitch.

Whether their drivers are Family1 or Family 2 compliant isn't a matter of opinion. Once the drivers meet certain criteria, they will be Family2 compatible.

They can't do anything about breaking things like Little Snitch if Apple does not approve the drivers. Without Apple's explicit approval, the drivers do not get access to the OS' built in library validation functionality and as such, will break apps like Little Snitch that use lib validation. This is not an nVidia problem, it's an Apple problem. Period.
 
They can't do anything about breaking things like Little Snitch if Apple does not approve the drivers. Without Apple's explicit approval, the drivers do not get access to the OS' built in library validation functionality and as such, will break apps like Little Snitch that use lib validation. This is not an nVidia problem, it's an Apple problem. Period.

What? Isn't library validation done in Xcode? Are you telling me that Apple needs to individually approve of every driver for it to work? When did IntelMausiEthernet.kext get approved?
 
IntelMausiEthernet isn't a graphics driver that needs to DRAW windows, so it's irrelevant. It's likely IntelMausiEthernet doesn't approach anything that'd get in way or if it does it's nothing we actually SEE happening.

furthermore, hackintosh users are already bypassing validation requirements anyways with clover injection, custom SIP in many cases and tools like lilu and whatevergreen that force macOS to accept them. You cannot take something that works on a hack and use it as a comparison.

Again, nvidia isn't here to serve just the hackintosh communities, their drivers need to comply with all apple policies and work with SIP fully enabled, without boot args that disable kext measures and without forced injection. the drivers are for real macs using nvidia hardware as well.
 
IntelMausiEthernet isn't a graphics driver that needs to DRAW windows, so it's irrelevant. It's likely IntelMausiEthernet doesn't approach anything that'd get in way or if it does it's nothing we actually SEE happening.

furthermore, hackintosh users are already bypassing validation requirements anyways with clover injection, custom SIP in many cases and tools like lilu and whatevergreen that force macOS to accept them. You cannot take something that works on a hack and use it as a comparison.

Again, nvidia isn't here to serve just the hackintosh communities, their drivers need to comply with all apple policies and work with SIP fully enabled, without boot args that disable kext measures and without forced injection. the drivers are for real macs using nvidia hardware as well.

My point is, release what they have unsigned and let users decide if they want to disable SIP.
 
For sure Nvidia already has the driver ready and working: it's not something done in the last minute, and it's already done for Kepler. I'd be amazed if a significant part of the code base is different. Probably it's just a commercial dispute. Given the size of the market of Mac users with Nvidia cards, I think neither Apple nor Nvidia care much about it.
 
Hello, friends!
I will tell you my experience to enable the very old nVidia video card on the latest MacOs 10.14.1 Mojave
Perhaps it will be interesting to someone.

I tried all possible values in the Config.plist file, offered a lot of boot options to the system, but the result was the same: my GeForce 9800 GT 512 mB is always defined by the system as nVidia Chip Model 8 mB.
Accordingly, there were very strong graphical interface artifacts, the system was very slow.

But then I copied the Extra folder from another Mac OS 10.6 Snow Leopard system into the boot volume and downloaded Mojave not with Clover, but with Chameleon, which I installed a long time ago.

Mojave booted perfectly using a very old Chameleon bootloader, and the video card immediately displayed correctly. Graphic brakes immediately disappeared.
In this case, I did not install any nVidia drivers into the system, only a bare system after a clean installation.
The only kext in the Extra folder is FakeSMC.
The only entry in the Chameleon bootloader about the video card is the key GraphicsEnabler = Yes

My very old computer hardware is in the signature.

It is strange that with the help of Clover the video card is not detected by any tricks, and with the help of Chameleon it is determined on the fly.


62d20f0a65f5.png
 
"What needs to be "approved" by Apple."

on 10.12, the nvidia drivers were signed, but they did not have approval. The drivers did certain things apple didn't like so they were denied library validation. Things like little snitch/etc would break. This is because apple would reject window access in applications using library validation to the unapproved nvidia drivers, signed or not.

in 10.13 nvidia rewrote the drivers, changing how driver worked to appease apple overlords and in return apple blessed them with approval to be treated as trusted kexts that are allowed library validation.

Lets be clear on this. This REQUIRED Apple's blessing. Once the drivers adhere to certain rules, macOS goes "ok, i blesseth thee"

So can nvidia just say "screw apple, we'll release drivers anyways." ? Sure, but you can be assured these drivers would not be blessed by apple and would not work properly for the INCREASING number of apps that use library validation. Just because hackintosh users can have https://github.com/acidanthera/WhateverGreen or nvidialibvalfix magically fix the problem doesn't change the fact that nvidia has a target audience that also includes a lot of mac pro users and even old imacs and macbook pros that still prefer the web driver over stock (web driver has MUCH higher opengl performance, metal performance about the same, but web driver is up to 30% faster still at opengl games).

Whatever pissing war is going on between apple and nvidia, i suspect nvidia is trying not to burn the bridges completely down just yet, because they do want apples approval to remain on their drivers so they can be used properly with SIP enabled and without hackery requirements if they aren't blessed.

I feel nvidia's approach of calling apple out is so they can encourage us as users to write to apple and show them that a relationship with nvdia still matters. That is what people should be doing at this point.

All this nonsense and conspiracy theories that nvidia is struggling with driver because of metal 2 and other stuff is just plain nonsense. I've talked with developers who have been working on metal for years now. Metal 2 is not a reinvention of wheel. it's literally just new feature extensions added to existing api that would not take that much effort to add. At most, the hardest part would be trying to get apple to approve certain nvidia gpus as Family2V1 feature set so every nvidia gpu didn't get locked into family1V4 and be locked out of using new metal features (hey there is that word again, APPROVE).

TL/DR. Nvidia has been good to us for years, I don't believe for one minute that they suddenly decided to drag their feet on purpose and make up stories because their engineers suddenly got stupid and couldn't figure out how to add 3 measly new api functions to their already existing and complex metal code. I bet drivers been done for at least 2 weeks if you want my honest opinion. This has apple written all over it.

I agree, for sure it's Apple that's not certifying it. Apple has always been like that: they won't certify anything that doesn't pass their standards. Certification processes exist for many things, for good reasons in most of the cases. I just don't understand how Nvidia's Kepler drivers would pass the certification, but not Pascal. For sure there are lots of stories here.
I think that stating that "couldn't figure out how to add 3 measly new api" is underestimating the problem. OpenGL has been deprecated and probably that's where Nvidia and Apple are disagreeing. Also, even not being a "reinvention of the wheel", it changes completely the rendering logic.

If I could put my bet: Nvidia's implementation of Metal 2 is just a wrapper to old rendering methods (a shortcut), and Apple isn't happy with that because the benchmark figures for Metal 2 would look just "slightly worse" than native OpenGL (remember early Vulkan benchmarks?), and you can only imagine how bad it would look for the marketing team.
 
I agree, for sure it's Apple that's not certifying it. Apple has always been like that: they won't certify anything that doesn't pass their standards. Certification processes exist for many things, for good reasons in most of the cases. I just don't understand how Nvidia's Kepler drivers would pass the certification, but not Pascal. For sure there are lots of stories here.
I think that stating that "couldn't figure out how to add 3 measly new api" is underestimating the problem. OpenGL has been deprecated and probably that's where Nvidia and Apple are disagreeing. Also, even not being a "reinvention of the wheel", it changes completely the rendering logic.

If I could put my bet: Nvidia's implementation of Metal 2 is just a wrapper to old rendering methods (a shortcut), and Apple isn't happy with that because the benchmark figures for Metal 2 would look just "slightly worse" than native OpenGL (remember early Vulkan benchmarks?), and you can only imagine how bad it would look for the marketing team.

The Kepler drivers are native in the OS from previous generation Mac hardware. Apple did say at launch of Mojave that they are working on making them better. The updates have shown improvement for the Kepler drivers.
Maybe Apple wants or needs to get those straightened out first so the hal(web) drivers will have something solid to latch on to?
 
Hello, friends!
I will tell you my experience to enable the very old nVidia video card on the latest MacOs 10.14.1 Mojave
Perhaps it will be interesting to someone.

I tried all possible values in the Config.plist file, offered a lot of boot options to the system, but the result was the same: my GeForce 9800 GT 512 mB is always defined by the system as nVidia Chip Model 8 mB.
Accordingly, there were very strong graphical interface artifacts, the system was very slow.

But then I copied the Extra folder from another Mac OS 10.6 Snow Leopard system into the boot volume and downloaded Mojave not with Clover, but with Chameleon, which I installed a long time ago.

Mojave booted perfectly using a very old Chameleon bootloader, and the video card immediately displayed correctly. Graphic brakes immediately disappeared.
In this case, I did not install any nVidia drivers into the system, only a bare system after a clean installation.
The only kext in the Extra folder is FakeSMC.
The only entry in the Chameleon bootloader about the video card is the key GraphicsEnabler = Yes

My very old computer hardware is in the signature.

It is strange that with the help of Clover the video card is not detected by any tricks, and with the help of Chameleon it is determined on the fly.


62d20f0a65f5.png
QE/CI enabled?
 
Status
Not open for further replies.
Back
Top