Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
All for what? Data privacy on device? Reducing the cost of inference in the cloud? Preventing Nvidia from getting into mobile?

[SNIP]

Once the cost of inference is dirt cheap and 5G becomes ubiquitous there will be little reason to do on device besides trivial models. I don't think Apple can win the ML game against Nvidia.
First of all, no I do not work in AI/ML. I've always been interested, though, and would like to dabble next year. Something simple, like recognize cat faces, so I can see for myself how this is done.

Don't you think "data privacy on device and preventing Nvidia from getting into mobile?" are really good reasons?
This thread is so interesting to me. To the guys battling it out in here, what exactly do you use the workstation for? MacPro users usually buy the workstation for the power under the hood, and are usually Video editors, composers, graphic artists, and the like with a general understanding of what configurations are the most bang for the buck. It sounds like you guys can speak to Apple's technical roadmap to a philosophical degree. Again, what do you guys use the workstation for?
I'd say there's also a large segment who have no idea of bang for the buck and simply use what their employer hands them. We're a bunch of nerds on here. Most people don't care as long as it isn't slow and doesn't crash.

Yes this thread is in an awesome phase right now. I started out doing system administration / managed IT services. I always have a fairly current Mac Pro & MacBook Pro. I also enjoy the occasional game at max settings but not willing to bootcamp. There's an HP Z820 for VR.

Increasingly I've been doing stuff with Mac forensic software which requires GPU power. Since Apple crippled the 6,1s eGPU capabilities for so long, I hired someone to Hackintosh several HP Z820s. All they do is sit in a server room, crunching numbers at high CPU / GPU utilization. 3 GPUs each, mostly GTX 1080s. They've been flawless and cheap. Any 5,1 owner who doesn't like the $6k price should seriously consider getting an HP Z series professionally Hackintoshed or set up with some kind of Hypervisor.

Also, I use a 6,1 as a dedicated Plex server. It handles 15TB and growing smoothly.
 
This thread is so interesting to me. To the guys battling it out in here, what exactly do you use the workstation for? MacPro users usually buy the workstation for the power under the hood, and are usually Video editors, composers, graphic artists, and the like with a general understanding of what configurations are the most bang for the buck. It sounds like you guys can speak to Apple's technical roadmap to a philosophical degree. Again, what do you guys use the workstation for?
They also buy them for internal expandability / upgradability. Many who like / need this aspect of the Mac Pro may not need the ultra high capabilities this new Mac Pro will offer. In the past it wasn't an issue to purchase the high end system to get this expansion / upgradability because, in their opinion, the price was acceptable. With the jump in price they find having to buy a high end capable system to get expandability / upgradability no longer palatable.

It's my observation they're not faulting this new Mac Pro per se but rather the lack of a lower cost alternative which provides what they can only get in the Mac Pro.
 
  • Like
Reactions: Korican100
Vulkan is fantastic. You can still use OpenGL & OpenCL. Machine learning is actually doable

Except Vulkan and OpenGL are still on the Mac. There are open source third party versions that work great and are being used in deployed apps. iOS is also using an OpenGL-on-Metal implementation that keeps OpenGL going without relying on OpenGL drivers.

It's also important to note: Apple isn't stopping Nvidia from shipping CUDA on the Mac. They're only withholding Nvidia's ability to ship Metal or OpenGL drivers. The Metal and OpenGL driver architecture is not public and requires a special agreement with Apple. But there is no restriction on the CUDA drivers. The CUDA drivers are written using public, Apple supplied tools and resources.

So Apple isn't withholding CUDA. Nvidia is withholding CUDA as leverage to get their Metal and OpenGL drivers back. Apple has no restrictions on accelerator card drivers, just graphics drivers.
 
Except Vulkan and OpenGL are still on the Mac. There are open source third party versions that work great and are being used in deployed apps. iOS is also using an OpenGL-on-Metal implementation that keeps OpenGL going without relying on OpenGL drivers.

It's also important to note: Apple isn't stopping Nvidia from shipping CUDA on the Mac. They're only withholding Nvidia's ability to ship Metal or OpenGL drivers. The Metal and OpenGL driver architecture is not public and requires a special agreement with Apple. But there is no restriction on the CUDA drivers. The CUDA drivers are written using public, Apple supplied tools and resources.

So Apple isn't withholding CUDA. Nvidia is withholding CUDA as leverage to get their Metal and OpenGL drivers back. Apple has no restrictions on accelerator card drivers, just graphics drivers.

OpenGL support on iOS/MacOS is garbage. The driver is no longer maintained and deprecated. Things don't work like they are supposed to. I can personally attest to this.

As for MoltenVK, it's just a runtime on top of Metal thats cumbersome, messy and doesn't support the full features of Vulkan. I'd hardly qualify it as real Vulkan on the Mac. Then you are also at the complete mercy of getting rejected from AppStore for using it.


The CUDA driver relies on low-level functionality that certain components of the graphics driver provide... CUDA needs the graphics driver.


What we found was support inside the Spaceship for the idea, but a lack of will to allow Nvidia GPUs. We've spoken with several dozen developers inside Apple, obviously not authorized to speak on behalf of the company, who feel that support for Nvidia's higher-end cards would be welcome, but disallowed quietly at higher levels of the company.

"It's not like we have any real work to do on it, Nvidia has great engineers," said one developer in a sentiment echoed by nearly all of the Apple staff we spoke with. "It's not like Metal 2 can't be moved to Nvidia with great performance. Somebody just doesn't want it there."
 
  • Like
Reactions: ssgbryan
There seems to be a bubble around some Mac users because they aren't aware of the amazing technology outside the ecosystem.
It's a mixed bag, 8 years ago macOS was the true alternative to Wincrash, and only *nix with GUI that doesn't sucks, but things evolved faster than the now fat overfeeded Apple's CEOs, now Linux it's easy to install and doesn't sucks, and while you haven't *nice* stuff like touch bar, Siri and continuity, you have the most solid OS for heavy computational work, on the other hand if you can deal with instability and security holes and you love whistles and bells windows is the champion on useless nice looking stuff, and the best OS for non crítical applications like gaming...
.
I bet that very very few will use the

Quite safe bet, no chance to lose.
, I don't see an alternative solution for having two Retina displays, enough ports, 10 Gbit Ethernet and a proper GPU.

What about the Mac mini 6 core i7 ? It has 10G options, has 4 thunderbolt 3 ports and even HDMI output, easy you can connect 3x 4k display and a eGPU and while have not 8-28 core options, for programming 6 core are fairly good.

And then think to yourself why can't this happen on MacOS... oh because Apple is petty.

You love or hate apple, I loved job's apple, but I hate Cook's apple, I love the iPad but I hate the iPhone, I love mostly all Macs but I hate not having nVidia inside
 
Last edited:
  • Like
Reactions: OkiRun
and you were right to do it. I can testimony that windows crashes even today, not just 8 years ago.
Search this website for "kernel panic", and tell me what you find. (Also try "kernal panic" and "colonel panic" - sometimes spell-checkers aren't enough.)

If Windows or Apple OSX regularly crash, there's a strong possibility that there is a hardware problem.
 
  • Like
Reactions: H2SO4 and Mago
Search this website for "kernel panic", and tell me what you find. (Also try "kernal panic" and "colonel panic" - sometimes spell-checkers aren't enough.)

lots of threads about hackintoshes and cMPs with unsupported hardware/software?

edit to add: Colonel Panic sounds like a great name for a WWE character :)
 
As for MoltenVK, it's just a runtime on top of Metal thats cumbersome, messy and doesn't support the full features of Vulkan. I'd hardly qualify it as real Vulkan on the Mac. Then you are also at the complete mercy of getting rejected from AppStore for using it.

Apple allows (and encourages) both MoltenVK and MoltenGL. It's not banned software.

The CUDA driver relies on low-level functionality that certain components of the graphics driver provide... CUDA needs the graphics driver.

And Nvidia can do whatever they want with the GPU as long as they don't touch Metal or OpenGL. Which CUDA does not need. Metal and OpenGL is specifically what Apple is preventing them from touching. If they want to bring it that low level functionality without bringing in Metal and OpenGL support they totally can.

But Nvidia also doesn't want their card mixing with other hardware. They're withholding CUDA because they don't want a CUDA card being used with an AMD card for graphics, in addition to holding something over Apple.

All Apple is doing is preventing the card from being used as a device to drive displays. That's it. Nvidia is not allowed to provide a Metal or OpenGL accelerated frame buffer. CUDA does not rely on a Metal or OpenGL frame buffer.
 
  • Like
Reactions: OkiRun
Apple allows (and encourages) both MoltenVK and MoltenGL. It's not banned software.

And Nvidia can do whatever they want with the GPU as long as they don't touch Metal or OpenGL. Which CUDA does not need. Metal and OpenGL is specifically what Apple is preventing them from touching. If they want to bring it that low level functionality without bringing in Metal and OpenGL support they totally can.

But Nvidia also doesn't want their card mixing with other hardware. They're withholding CUDA because they don't want a CUDA card being used with an AMD card for graphics, in addition to holding something over Apple.

All Apple is doing is preventing the card from being used as a device to drive displays. That's it. Nvidia is not allowed to provide a Metal or OpenGL accelerated frame buffer. CUDA does not rely on a Metal or OpenGL frame buffer.

Just because it's okay now, doesn't mean they won't change their mind later. And as I've said, it's not the full version of Vulkan. Go implement a ray tracer with the Vulkan API that runs on AMD & Nvidia. Then port it to MoltenVK. Oh boy it doesn't work....

Anyone with sufficient skill could take CUDA and implement OpenGL. Literally write their own rasterization engine in CUDA that does what OpenGL does. OpenGL and Metal are nothing special. Even if CUDA was on a Mac without a display frame buffer. You could still use the Nvidia GPU to render and then transfer that buffer to the GPU connected to a display. Might suffer some latency, but totally doable.. Your reason is ridiculous.
 
This thread is so interesting to me. To the guys battling it out in here, what exactly do you use the workstation for? MacPro users usually buy the workstation for the power under the hood, and are usually Video editors, composers, graphic artists, and the like with a general understanding of what configurations are the most bang for the buck. It sounds like you guys can speak to Apple's technical roadmap to a philosophical degree. Again, what do you guys use the workstation for?

Apple doesn't have a road map - road maps imply long term planning, and Apple doesn't appear to be doing that with the Mac Pro (at least since Tim took over).

I do 3d art. I am not a professional, I am a hobbyist. The Mac Pro was the only product that Apple made that can do 3d art without massive throttling. (I have a LOT of dead Apple products because Sir Idiot Boy never truly grasped the 2nd law of thermodynamics).

Workflow consists of Hexagon (for modeling - it is a garbage modeler, but I have memorized the function keys), Poser Pro 11 (Where most of my work is done), Daz Studio (to get DS assets out of DS and into Poser), Adobe Acrobat (to assemble PDFs), Photoshop (Postwork), ZBrush (to rework digital assets), Vue (outdoor scenes), Bryce (also outdoor scenes), Luxrender (although I an retiring this, since Poser has a PBR render engine.) And Blender, because because at some point, I'll retire Hexagon.

And..... Since none of my Adobe products (CS2 and Adobe Acrobat 9) will run on the 7,1 - that is another reason to move to windows - all of it still runs on windows 10.
 
Apple doesn't have a road map - road maps imply long term planning, and Apple doesn't appear to be doing that with the Mac Pro (at least since Tim took over).

I do 3d art. I am not a professional, I am a hobbyist. The Mac Pro was the only product that Apple made that can do 3d art without massive throttling. (I have a LOT of dead Apple products because Sir Idiot Boy never truly grasped the 2nd law of thermodynamics).

Workflow consists of Hexagon (for modeling - it is a garbage modeler, but I have memorized the function keys, Poser Pro 11 (Where most of my work is done), Daz Studio (to get DS assets out of DS and into Poser), Adobe Acrobat (to assemble PDFs), Photoshop (Postwork), ZBrush (to rework digital assets), Vue (outdoor scenes), Bryce (also outdoor scenes), Luxrender (although I an retiring this, since Poser has a PBR render engine.) And Blender, because because at some point, I'll retire Hexagon.

And..... Since none of my Adobe products (CS2 and Adobe Acrobat 9) will run on the 7,1 - that is another reason to move to windows - all of it still runs on windows 10.

The experiences of my team of professional 3D modelers doesn't align with yours. They are churning out 3D work on a daily basis on their iMac Pros.
 
  • Like
Reactions: OkiRun and sirio76
Why?

And since Tesla cards don't have display hardware, why doesn't Apple allow Tesla drivers (which by definition can't be connected to displays)?

Your argument is absurd.

They do. That's what I'm saying.

The only thing being withheld from Nvidia is the Metal/OpenGL stack.

Apple isn't preventing Tesla drivers. Nvidia is withholding them as leverage. A Tesla is just a generic card to macOS. You can write drivers for it just like any other card. Again, only the Metal/OpenGL driver stack and the window server integration is the private bit.
 
  • Like
Reactions: Zdigital2015
Why?

And since Tesla cards don't have display hardware, why doesn't Apple allow Tesla drivers (which by definition can't be connected to displays)?

Your argument is absurd.
while Tesla drivers aren't for Display cards, it's still needs kernel mode drivers to improve sys ram read write, so nVidia still registers it as kernel peripheral, it even can render graphics for a main GPU with display output, CUDA 10 give cues a pure user mode drivers could come later for Linux servers, drivers with kernel access are hard to release given it can compromise the entire system.
[automerge]1571836105[/automerge]
The only thing being withheld from Nvidia is the Metal
There are any version about why apple don't allow nVidia, One I've read is apple requires all compute rendering thru Metal 2, banning CUDA ir require CUDA to run on the top of metal (like moltenvk with vulkan), this breaks CUDA, apple old argument is to not to allowing redundancy or duplicated functions.

Of course nVidia still could release a graphic-disabled user mode drivers and all the compute stuff loaded on it, but still faces apple's duplicity ban, but these case a simple lawyer could solve this as there is no reason to ban a device not interfering the system operation (a STD peripheral) it's like to ban a 10$ Bluetooth keyboard, you have right to alternative as long you don't interfering the core system or 3rd party apps.

A user-mode Cuda driver won't work with legacy Cuda apps tho.
[automerge]1571836222[/automerge]
The only thing being withheld from Nvidia is the Metal
There are any version about why apple don't allow nVidia, One I've read is apple requires all compute rendering thru Metal 2, banning CUDA ir require CUDA to run on the top of metal (like moltenvk with vulkan), this breaks CUDA, apple old argument is to not to allowing redundancy or duplicated functions.

Of course nVidia still could release a graphic-disabled user mode drivers and all the compute stuff loaded on it, but still faces apple's duplicity ban, but these case a simple lawyer could solve this as there is no reason to ban a device not interfering the system operation (a STD peripheral) it's like to ban a 10$ Bluetooth keyboard, you have right to alternative as long you don't interfering the core system or 3rd party apps.

User mode Cuda driver won't work with old legacy Cuda apps maybe only with CUDA 10
 
Last edited:
The experiences of my team of professional 3D modelers doesn't align with yours. They are churning out 3D work on a daily basis on their iMac Pros.

Good for you. I am sure that they are doing that - but I am also sure that they not doing their production rendering on an iMac Pro.
 
I'm sorry if this has been discussed already, but I think not only did Apple out-price many for the new Mac Pros, it appears that they are using propriety PCI-E slots. No one is ever going to make GPU cards that will work in this system except for Apple. That means of all the graphics cards out there, you can only install what Apple makes.

Has this been discussed already?
 
while Tesla drivers aren't for Display cards, it's still needs kernel mode drivers to improve sys ram read write

That's allowed and publicly accessible.

There are any version about why apple don't allow nVidia, One I've read is apple requires all compute rendering thru Metal 2, banning CUDA ir require CUDA to run on the top of metal (like moltenvk with vulkan), this breaks CUDA, apple old argument is to not to allowing redundancy or duplicated functions.

Apple does not release public specs on the accelerated frame buffer part of the kernel, which is what interacts with Metal and OpenGL. That's what's being withheld from Nvidia since Mojave. Mojave had changes to the accelerated frame buffer API that weren't provided to Nvidia.

Writing Metal/OpenGL drivers requires being part of a special program with Apple that they've opted to no longer include Nvidia in. There's no public documentation, and Apple doesn't release the source in their open source repos.

Of course nVidia still could release a graphic-disabled user mode drivers and all the compute stuff loaded on it

They don't even have to do user mode. The kernel mode stuff is open.

But yes, they could even do it user mode.

but still faces apple's duplicity ban

Which is exactly what I said, they're withholding CUDA as leverage.

A user-mode Cuda driver won't work with legacy Cuda apps tho.

They can do it kernel mode. That part of the kernel is open source, public, and accessible and shippable by any developer. Unlike the Metal/OpenGL stacks.
 
  • Like
Reactions: OkiRun
No, FYI kernel mode drivers has to be digitally signed by apple to be loaded

Apple will provide a signing certificate and the developer signs. An accelerator card is a allowed case for a signing certificate. The Avid HDX cards are a good example.

Gaining a Kext signing certificate does not allow access to the Metal/OpenGL stack though, that's a different program. Signing certificates will not be provided for GPU drivers.

once the new driverKIY framework replace old kext, only non-kernel mode (user space) devices will work with unsigned drivers.

Yes, once kexts go away (they have not yet), drivers will have to move to user space with DriverKit. As you've noted, this is also not a serious problem for Nvidia either. CUDA is completely capable of running in user space.

Not only that, but running in user space would remove Nvidia completely from the kext signing process.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.