Yeah that's what I mean, since our 10.11 so far doesn't use metal and it has such better performance than 10.10, then 10.10 really sucks.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Metal, OpenCL/GL & GPU vendors
- Thread starter stevedusa
- Start date
- Sort by reaction score
Yeah that's what I mean, since our 10.11 so far doesn't use metal and it has such better performance than 10.10, then 10.10 really sucks.
Ah, okay, now I understand.
Well, 10.10 doesn't suck for me as hard as others, though I am hopeful for medal and After Effects, seeing the speed increase in their demo. As for the rest, well. I will be surprised probably.
That won't be more completely clear until Vulkan is done. If Vulkan ends up being something that could be wrapped around Metal then it would be relatively low impact if the GPUs vendors ( and/or Apple) stepped up to do the work. How long it is going to take Vulkan to get to the consensus point and production deployment is an issue that Apple wanted more control over, but doesn't practically have. OpenCL is an example where I suspect Apple wanted things to move faster than it did. Apple needed Metal soon, so did it themselves ( because it is faster when don't have to get consensus of a large group.).
One of the major groups pounding the table for Vulkan/Mantle/etc are the game engine folks. They want something to lay their own graphic API stack on. Apple has a major iOS game business. Games drive game engines so Apple needed to make the game engine folks happy. Once have Metal on iOS and hooked into the Core library frameworks , it is natural to have it on OS X which has very similar ( if not the same in a subset ) frameworks.
More than likely Apple saw or heard knew about the "slimmer glnext" (since in Khronos ) and had heard of Mantle ( AMD probably pitched it to them ) and perhaps even got wind of Direct 3d12. One of the primary groups
So Apple needed a solution sooner rather than later. Additionally, Apple's "trailing by 2 years" adoption of new OpenGL standards has its own priorities when it comes to the graphics stack. Apple does the "top" ( user facing) layer of OpenGL and the GPU vendors do the Apple designed bottom layer. Metal could represent a rethinking of that bottom layer.
Metal only has to cover Apple's four GPU vendors ( in order of magnitude: Imagination , Intel , and "others" ( Nvidia / AMD ) ). If the work on Metal helps Imagination and Intel get the things they need weaved into Vulkan then that helps with Vulkan viability on the OS X long term. If it turned out the function calls had different prefixes ( MTL , vk ) but had the same semantics it would be straightforward to build some adaptor "glue" that just present as different facade. I don't think it will be an easy "query replace prefix" effort, but it doesn't have to be a 100% semantic mismatch either.
Metal's multiple GPU support is only coming after Vulkan's initial multiple GPU refinements have been done. If Apple is looking over the Vulkan contributors' shoulders as they work it doesn't have to be a super wide gap.
Metal won't be standards compliant, but also wouldn't be standard hostile.
I think the game engine folks hope to build a base of games locked into their API and are willing to take the pain of porting the bottom layer. At least for now.... long term when targeting 15 different GPU specific implementations becomes a pain they may not be so happy.
Unless Vulkan picks up Android or Windows, just Linux/OS X is not really enough. Android + iOS + Windows puts those other two into the very low single digit or less territory. It would be a gross overstatement to claim some multi-platform title when that is the scope of the overall general market coverage.
I think Vulkan might end up a common wrapper around Metal/Direct3D/console APIs. Or possibly around GPU vendor specific drivers if OS folks have gross mismatches. Not quite "as thin as possible", but "thin" relative to OpenGL. Maybe in the "cure world hunger" category but if a future stack look like:
OpenGL ( all specific GPU arch targeted parts and legacy stuff tagged as deprecated )
Vulkan ( where did all the stuff kicked out above go? to this layer where don't have to 'hide'/'smooth' differences. )
OS/GPU specific micro graphics kernel
that would be nice. Normal "good enough" GPU performance programs use a more clearly defined OpenGL subset. The more performance critical graphics programs and engines use Vulkan. The extremely performance critical engines/programs use the OS specific stuff.
It has upsides of possibly assigning the OpenGL "top" layer to the OS vendors who will likely keep GPU specific stuff out of future iterations of OpenGL. The GPU vendors get a shorter stack to implement ( perhaps only some subset of Vulkan and probably most of the micro kernel level. ). The fights about getting "my arch" into high level stuff could be pulled out of the standards process and limited in the middle layer. The consensus spectrum have to get from highly competing parties goes down as you down the stack.[/QUOTE]
I think Apple's slides that show OpenGL disappearing completely from OS stack are a bit of "clean looking slide" convenience rather than a depiction of the whole graphics interfaces and stacks still present. OpenGL is still going to be around. They would break way too many programs if wiped it out completely. Metal and OpenGL apps are going to have to share the same screen. It isn't classic iOS where everything runs one full screen app at a time. At some level they have to get along because the GPU is a shared resource.
Aehm, yes, AFAIK each product which has 'Mac Edition' or similar in the name, takes an approval process with Apple. Once a card is approved it will be supported in all future OS X versions (till death by obsolescence).
Where did you get this info?
Based on what you said, the following has gone through an approval process with Apple and will be supported in all future OS X versions?
http://www.macvidcards.com/store/p42/Nvidia_GTX_980_4_GB_Mac_Edition.html
But the following does not mention anything about "Mac Edition"...
https://www.pny.com/NVIDIA_Quadro_K5000_for_Mac
Last edited:
Eh? Introduction title for the product on this page
"...
Professional Graphics Card for Mac ...
...
NVIDIA® Quadro® K5000 for Mac ... "
In the Specs tab the system requirements are for
- MacPro3,1; MacPro4,1; MacPro5,1 or later "
Apple is in the loop, but there is a financially viable aspect to that too. That means getting Apple testing done ( which costs money and resources ) and getting the driver bugs worked out ( working with AMD/Nvidia ) development and support queues costs money and resources.
When PNY, Sapphire , EVGA do all this work and folks come along and pirate the contributions they made to getting all this done, they also loose out on getting back that investment also. There is a crabs in the barrell impediment to the Mac GPU card market too. If there was substantially larger base it would be easier to brush off, but it isn't.
Last edited:
MVC cards are not official, and of course not approved cards, although I highly appreciate them (I've got two of them). That's why we have the odds with OS X updates breaking Nvidia web drivers with GTX 9xx cards. They are supported by NVidia web drivers, but not by OS X drivers. Until an official new 'Mac' Nvidia card is released, with integration of newest drivers in OS X.
Seen from that angle, they are approved by NVidia (and blessed with web drivers), but not by Apple.
I wrote 'Mac Edition' or similar. This one has 'for Mac' in the name, so yes, it's an approved one.
Last edited:
So, Apple is fine with any unofficial card to use the "Mac Edition" in their naming whether approved by them or not?
The fact that there is built in support for some Nvidia GPUs has nothing to do with approval. Where were there any official or approved Fermi video cards ever released? Yet, the drivers for them are there.
"or similar" pretty much covers everything. They certainly aren't going to market a video card made for Macs as "Linux Edition" or "for Linux".
My question remains. Where did you get this info. Or is it just a product of your imagination?
Also, you still haven't addressed why AMD doesn't make any new PCI-e video cards for Macs since the drivers for them are already built-in to OS X.
You continue to believe that these manufacturers are not selling Mac video cards because Apple told them not to???????
Last edited:
I think the GTX 680 Mac Edition was sold very well and brought very good revenues to EVGA. I bought one too. So they could have made a GTX 980 release cross-subsidized by the GTX 680 earnings.
Also the graphic cards manufacturers should be faster than in the past. GTX 680 Mac Edition appeared one year after the release of the PC version. They have a time window, in which they have to undertake the release.
If they don't MVC will do it.
And i'm glad.
Last edited:
"....
IMG: Who is primarily responsible for the drivers, Apple or AMD?
CB: The responsibility is shared. Apple's driver stack has a plugin architecture, so Apple writes the top of the stack and vendors like AMD provide hardware-specific plugin modules. ... "
http://www.insidemacgames.com/features/view.php?ID=574
Pragmatically, there are two parts to the driver stack. Apple doesn't release every incremental drop of the vendor "plugin" updates to the public release queue. All of these plugins should have a sanity check to make sure they are in synch with what they are plugging into (e.g., if the kernel changes they should stop/abort. ). The kernel and plugin interface will change over time ( requirement change, security holes need to get closed , etc. no significantly broad/important interface is fixed in stone forever. )
It isn't just Apple's approval have to get, it is also the 2nd 'half' of this shared responsibility also to push out a card.
Some of the low lever driver stuff is not maximized for modularity. Both AMD and Nvidia attempt to supply Apple with a variety of GPUs. There are likely going to be internal design bake-offs where crude prototypes ( caseless, bare board(s), mock ups connected to a monitor ) are hooked up and benchmarked on a alpha version of OS X that is never going to see light of day in public. Aspects of those plugins will make it into the mainstream drivers. It isn't a surprise there is a superset buried inside. Long term the mobile GPUs and desktop GPUs tend to the same features so even a desktop GPU that is slightly ahead of the mobile space might get worked on in protoype mode were the mobile variant is the ultimate target ( e.g., GPUs in the 5K iMacs ... mostly highly clocked down desktop variants. )
You are correct in those edge cases are never going to get official Apple product support, but some folks around here equate "supported" with "happens to work for me". It is probably not a matter of Apple banning them as much as Apple saying you must invest/pay/resource this much to get an approval stamp. Not a completely ban, but not making it minimally expensive either.
I was specifically looking for the words "Mac Edition" as per Synchro3's suggestion that merely having a video card named "Mac Edition" meant Apple approved it.
Agree that it was a double edged sword.
I think the GTX 680 Mac Edition was sold very well and brought very good revenues to EVGA. I bought one too. So they could have made a GTX 980 release cross-subsidized by the GTX 680 earnings.
Also the graphic cards manufacturers should be faster than in the past. GTX 680 Mac Edition appeared one year after the release of the PC version. They have a time window, in which they have to undertake the release.
If they don't MVC will do it.
We can only speculate on how well or how poorly the official EVGA GTX 680 Mac Edition sold. My guess is that they didn't do too well, hence nothing after that was every released for Macs by them.
None of this confirms Synchro3's assertion that the few retail Mac video cards that have made it to market were somehow Apple approved. Neither does a card having the words "Mac Edition" mean that Apple approved of them.
If there were enough of a profit motive, manufacturers will sell video cards for Macs whether or not Apple likes it. They're not just going to decide to not make money if Apple tells them not to. According to his theory, companies will agree to not make money simply because Apple tells them not to.
the workstation market has just started the migration from driver based Gpu to more machine code Gpu drivers. This will separate video drivers from video processing. The thunderbolt video is pass through patch, metal will probably be the interface for external PCIE video cards.
The more efficient machine coding you do, the more the Gpu hardware has to be fixed. In the router world, when you moved from driver based process to more of a micro code process, there was a 10x improvement external to the processor.
The more efficient machine coding you do, the more the Gpu hardware has to be fixed. In the router world, when you moved from driver based process to more of a micro code process, there was a 10x improvement external to the processor.
I don't have detailed record of this, but I'm pretty sure it is correct. Almost every one of these "approved" cards was at one time sold in online Apple store. Here is a list of GPU cards that FCP X and Motion support.
https://support.apple.com/en-us/HT202239
Pretty sure there is only one entry there which has the tag "Only available from third-party resellers" ( not sure if it was never sold in Apple Online store or had been dropped by time of the first version of this support note). The rest Apple sold at some point. They have dropped most now though. All the products in the Apple store are approved by Apple. They go through at least some testing to see if they work well and meet Apple standards.
It is also extremely hard to believe that it was a coincidence that very few (if any? ) "for Mac" cards overlapped in base GPU. There would be a 7970 card... not 2-3. A 680 ... not 2-3. Apple and the GPU vendors did not coordinate this and it just happened that way over several years ........ or they did coordinate their activity? [ Remember Apple's contract where they told publishers how to price their ebooks , the Govt sued and Apple lost ..... yeah Apple doesn't coordinate. *cough* ] That coordination is an defacto approval.
The market here is small. Apple knows that 2-3 players aiming at the same sub-sub-sub niche is going to turn out bad for almost everyone. Some customers might get cards at firesale prices when it doesn't work, but it is not a healthy ecosystem.
If Apple says I'm not going to sell your card ( being one of the biggest Mac accessory retailers ) and they aren't going to put your card in the support compatibility support matrix testing.... then have some "go to market" problems. Yeah if the market is worth $500M (and $50M pure profit ) and have to plow past those speed bumps... sure. If the margins are thinner then probably not.
Similar too if Apple say no and vendor says "hey I'll give you a $10M cut of the action"... Apple may change their mind. So yeah it comes down to market viability.
Revenue doesn't play a role. It is profits. If it is high revenue and lower profit than other products they could possible do for around the same money it won't get done. if the licensing , material , sales/marketing, and support (staff , phone, returns, defect ) costs eat up the profits then there is low return on investment.
Products that need subsidies have a market viability problem. A pool of money to jump start the product yes. The product never returns more money than was was put in is a problem.
I don't think it was an was an accident that the 7950 and 680 came around the same time that Apple said it was pulling the Mac Pro from the EU market ( January-February 2013 ) and there was no new Mac pro in sight.
Couple that to complete dead in the water for 2 years standard config video cards and had a market for both 2012 and 2009-2010 ( some previous ) owners that were all on borrowed time.
https://www.macrumors.com/2013/03/1...d-7950-mac-edition-graphics-card-for-mac-pro/
https://www.macrumors.com/2013/04/0...tx-680-mac-edition-graphics-card-for-mac-pro/
That's a bump but it isn't sustainable. Even if got enough to do a next round basically playing a game of music chairs. By June 2013 it is know that standard PCIe GPU card market is a dead-ender. (probably a relatively short term bump in Mac 2012 sales then too. ) That number of Mac Pros to sell into is only going to get smaller at that point as they are retired/decommissioned. On top of that more of the market is going to be dominated by the "I just buy old trailing edge technology from original owners at bargain prices " Old Mac Pros and cards being resold on online markets. Little of that generates any new "blood" revenue for original equipment vendors.
If they don't MVC will do it.
Folks clap and cheer at a game of musical chairs too. It is not a good long term model for a healthy ecosystem. At some point the game is over.[/quote][/QUOTE]
It would make sense that before selling a 3rd party product in the Apple Store, Apple makes sure it works on Macs.
There has been many products that have had "for Mac" or similar as part of their naming. Until I see evidence otherwise, I refuse that all of them have gone through testing at Apple for approval.
As far as I know, Apple has never had a program like the MFi program for Mac accessories.
https://developer.apple.com/programs/mfi/
I don't buy in to the conspiracy theories. I believe that the limited selection of 3rd party video cards was due to how small of a market it was and it didn't make financial sense to for most to do it and those who tried, I suspect, was met with very limited success.
If the conspiracy theories were true, then it would only make even more sense for Apple to have offered these "approved" cards as BTO options.
This has been my point all along. If these video card manufacturers felt that there was enough money to be made, nothing, not even Apple telling them not to, would have stopped them from selling their video cards for Macs.
Whether is is a snazzy label to slap on your product or not doesn't mean Apple doesn't certify devices. This KB document outlines that they are. Similar the 4K monitor support.
https://support.apple.com/en-us/HT202856
Some monitors are on the list. Alot aren't. Doesn't mean the don't work, but defacto Apple has put a blessing ("approved") on a subset. This happens. Apple is never going to test every device under the Sun. They will do a wide set with the MFI program because they get paid more money to do it. MFI is as much a structured payment system to fund a huge testing program as quality control gate. It is also a stronger mechanism by which they can sue and repressing the excessive "race to the bottom" pirates throw quality under the bus to get paid.
It is far more so collusion not conspiracy here. Apple knows how many Mac Pros they sell. All they have to do is put that on the table with an NDA with the card vendors. Apple could sit down individually with a couple and say "it would be nice of you could do a XXX based card for the new Mac Pro. If you do will help with with marketing and visibility by putting it in the store and we'll run though the support matrix for some of our high visibility apps that can help drive sales". If the vendor says ... we'd like to YYY which is an overlap all Apple has to do is hint that this isn't as interesting.
That company X wants to work more closely with company Y happens all the time in the tech industry. It is not illegal. If company X tries to dictate to company Y who else they can and cannot work with and mandates they need to have specific language in their contracts with not just company X but anyone else ... that where get into the illegal zone.
The deal thing is normal business. Apple slides into the zone of thinking that is so normal that they should dictate to everyone just who should be working with whom and how. That's the part that is illegal.
That problem with your view is that the "market" magically came to the actions all by themselves independently with no sharing of data/knowledge.
No it doesn't. They can control tighter inventories if they are not. If card doesn't sell well they can drop it from the store. Dropping items from BTO ( until next system upgrade ) is a non option. Those standard config cards are all made by contractors too. The "approved" cards are either out of upgrade cycle available or possible cards that didn't quite make it in the design bake off but Apple thinks have high enough differentiation (e.g., very expensive "Pro Graphics class" card) and volume rates too low for Online store inventory target thresholds.
It also blows up their policy of presenting an easy to comprehend set of options. (i.e.. don't blow up user's short term memory with a bazillion choices. ). If don't like any of them then pick the cheapest one and replace later if that is an option. Simple.
I think the disconnect is your looping into the "theory" that there is no interaction in the "go to market" decision whether Apple approves at all. What is being proposed in part is that what Apple says/approves matters. It does.
The issue more is whether they talk about that fact together or not. If the vendor and Apple have a conversation about doing a product and Apple effective says "yes, we think that is a good idea" then pragmatically they have approved the product.
"Mac" is an Apple trademark. If they really don't like what you are doing then product labeling something "for the Mac" is probably going to get some pit bull lawyers unleashed on you.
That "if ... was enough money" is probably pretty close to being just a theoretical exercise in this context. If piss off Apple then they aren't going to work with you anymore long term. ( Apple started to unwind their relationship with Samsung which is waaaaaaaaaaay deeper than any video card vendor could ever possibly hope for. ) They can even set it up so their major contractors won't work with you anymore on Apple related projects. It is a rare corner case where can piss of the primary controller of the ecosystem working inside of and still make money long term.
If Apple says "we don't think that is a good idea. we aren't going to help you" is one thing. If Apple says very clearly that "x is a bad idea. We strongly disapprove" is another. Short term might make some bucks, but long term damaging a relationship with Apple creates an adversary with giant buckets of money and resources. Long term, excessive greed tends to have blowback.
There are a tons of vendors out making limited money on crappy, pirate Lighting cables and the like, but they are probably never going to work on a lucrative Apple funded contract in the future. Surviving in the outer fridge of the iOS ecosystem may work because it is so big. The Mac ecosystem is much, much smaller. The desktop subset even smaller.
All of this stems back to Synchro3's comment that having "Mac Edition" on something means that it had Apple's approval. I don't believe this to be true and neither of you have proven to me otherwise.
Again, as far as I know, there's no "Made For Mac" program to officially approve any Mac accessories.
I'm not going to stray further from the original comment made by Synchro3 and go in to why the MFi program exists.
Here's more conjecture. Unless there's any evidence of this ever happening, I'll assume it never did.
Obviously, there had to be some sort of research in to what potential sales for their product might be like before actually entering the market. There's little doubt that they inquired with Apple about how many Mac Pros had actually been sold.
No sensible company would magically decided that they will sell their video cards for Mac Pros.
The bottom line is, if there's money to be made, they will enter the market whether they have Apple's blessings or not.
C'mon. Are you serious??? How else will they let customers know their products are Mac compatible???
That's an excellent example. I'm sure Apple expressed their displeasure with Samsung when they started making their Galaxy phones. It did nothing to stop them from going ahead and doing it because there was money to be made by Samsung.
Jeez. That was a long post.
Checked Intel iGPU, guis seems to utilize Metal - http://forum.netkas.org/index.php/topic,11144.msg30993.html#msg30993