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.
AMD Discussion
- Thread starter koyoot
- Start date
- Sort by reaction score
Ok, openGL is now in rest home. Still alive, but maybe not after couple of years.
And I don't believe Apple will kill openGL just like that, but it will move as a layer on top of Metal. Someone could write a DX layer on top of Metal too, if they were nasty... Metal with its with low overhead is like a GPU itself... so you could write a driver for almost any GL tech there is. Vulkan included.
And I don't believe Apple will kill openGL just like that, but it will move as a layer on top of Metal. Someone could write a DX layer on top of Metal too, if they were nasty... Metal with its with low overhead is like a GPU itself... so you could write a driver for almost any GL tech there is. Vulkan included.
Last edited:
But don't underestimate the open source community... when there's a real need, and means... it will happen.
Chuckle ... and Wayland is going to displace X Windows when? Eventually, but X11 was created in the 80's. ( I used X10 for a short while) and nothing came even though there was lots of "need" complaining. The "and means" is critical. Nobody, including Apple, is going to throw a ton of money legacy OpenGL edge cases that don't fit well on top of Metal ( and/or Vulkan).
OpenGL doesn't need to as much "evolve" as much as clean up. What OpenGL really needs is a schedule for deprecating some of the evolutionary 'dead ends" that have clogged up the works over time. Folks who have legacy code can then start planning to make changes to get off. With few deadlines have apps that don't put the money into getting out of those. Open source's "collective consensus" is one of the problems OpenGL has had problems with. 'Open' doesn't magically uncork consensus problems.
[doublepost=1472098749][/doublepost]
There is little to indicate that 'Coherent Link' is not coupled to what is labelled on those die masks at "Global Memory Interconnect'.
http://www.fudzilla.com/news/processors/38402-amd-s-coherent-data-fabric-enables-100-gb-s
The Xeon E5 series products have an internal "ring bus" to hook cache/cores/uncore internal submodules on the same die. That bus appears to be this "Coherent Link". How that link is extended across dies is by the GMI
Unless this these are TARDIS DIMMs that doesn't work. DIMMs on the current Mac Pro are approximately the same length as the vertically placed boards. There is no internal volume for more DIMMs.
Multiple CPU package. Not APU.
In a different form factor? Possible yes. In the same general form factor as now? No.
10 TB v3 ports is just deeply misguided. It is a waste. Extremely high probability that a sizable fraction of those ports are never used in TB specific mode. Which would made it an added expense with extremely little value. Some plain USB sockets are far better design choice. Four TB v3 ports and some miniDP/HDMI, and USB ports would have much higher utterly.
single socket 96 cores are coming in a couple of years. Can do 72 in a month with a simple "off the shelf" buy:
http://www.anandtech.com/show/10584...erworkstation-systems-featuring-xeon-phi-x200
Some folks are going to buy these. I doubt Apple thinks that is a large enough pool of folks in that class to target.
The rack mount servers will go into render/compute clusters and the Mac Pro will largely work just fine visualizing the
data that comes out of those clusters.
There was many years where it was going to be default in next release. I run it now and have been for the last six months so it's here and will trickle in to the rest of the Linux world as they release new versions. Wayland is finally here and shipping*. The bigger question is the the Ubuntu's give up an Mir and adopt Wayland or do they continue to tilt windmills and if they opt for Mir do the derivatives break away?
*Slackware always excepted*
Last edited:
So you're saying Apple wont put openGL on top of Metal and they keep paying Nvidia, Intel and AMD for three drivers (with openCL) for every OS? I don't think so. They'll pay for Metal drivers only after couple of years.
Last edited:
They do not have to pay anyone for Metal drivers. Intel and AMD architectures are well documented and Apple can write them by themselves. Nvidia hardware on the other hand is not well documented, so Apple would have to pay them money for the drivers. But they are not going to bother, because of number of reasons.
NVidia don't need to document their hardware, they did excellent software API as long you attach to nVidia standards, and seems Apple lately don't like this word "standard" Metal it's just a pretext to not support industry standard Vulkan. (a very big mistake)
GL was created by a comitee, hence why it perform so badly compared to DX. Microsoft made the right decision at the time to leave that comitee and focus on a better solution.
[doublepost=1472129225][/doublepost]
There is no relation between Mantle and DX12. DX12 doesn't share one line of code with Mantle. Both do the same thing like GL and D3D did the same thing but nothing more. This as already been demonstrated to you but you keep repeating this once the dust settle.
Because people constantly resist the facts.
Read here: https://community.amd.com/community/gaming/blog/2015/05/12/on-apis-and-the-future-of-mantle
Mantle 1.0 is base of DirectX 12 and Vulkan. However you are free to be entitled to your opinions despite the fact they contradict the reality.
Another blog post...
Being inspired by doesn't mean using the code of.
There is no Mantle code in DX12. Mantle is the basis for Vulkan, not DX12. Microsoft implemented their own code to do the same thing as Mantle & vulkan, just like they did with D3D in regard to OGL. But, hey, keep posting some more anonymoous blog post from the AMD forum no less to try to make your point...
For god's sake...
The functionality of all three APIs is exactly the same as feature set of Mantle was. Low CPU overhead, HSA 2.0 capabilities, explicit multithreading, and explicit control over the hardware.
Argue with that however you want to. Fact is, Mantle is the functional base of all of the modern most known APIs in the industry. They do differ. Of course they do. But they have "few" common things that are coming from their source. Which was Mantle, and what AMD have specifically said in that blog post you are not admitting to be factual.
And for last time. It was NOT inspired by. Feature set is EXACTLY the same for all three APIs. Because they share the same "chassis", to use this motoring analogy. On the same chassis has been built Volkswagen Golf, Skoda Superb, and Audi RS3. You get this right now?
Are we really having this argument again? In the other thread we had slides from AMD showing that Vulkan is partially derived from Mantle and AMD claims that DX12 and Metal were inspired by Mantle (i.e. no code directly from mantle).
Vulkan is not partially derived from Mantle. It IS Mantle. Same as LiquidVR.
Feature sets are exactly the same for every one of the APIs. If you would use logic, rather than argue with reality, you would ask yourself why it is possible for MoltenVK, which is attempt to bring Vulkan to not only macOS, but also iOS(!) to be used with Metal on both Apple platforms, desktop and mobile.
Because all you have to do is figure out how to bring the lacking capabilities from Vulkan to OS X. Feature sets are exactly the same.
But after all, you are free to be entitled to your opinion.
For god's sake indeed...
Both a GM car and Ford car have the same feature yet neither use part from the other company! Is that too hard to comprehend?
NO CODE FROM MANTLE ARE PART OF DX12. Microsoft wrote an API that does the same thing than Mantle, they didn't use Mantle. The same goes for Direct 3D in regard to OGL. Mantle was rejected in the Mircrosoft ecosphere and flopped.
And for the last time, having the same feature set doesn't equate to using the same code to do it. I can write my own OS and copy the feature set of Windows or Mac OS and yet never use a single byte of code from either source. Kapish!
[doublepost=1472135605][/doublepost]
SG cede it to the Kronos group, which is a comitee of hardware manufacturer.
Come on, do you really think M$ wrote DX12 from scratch? Dream on. They worked with AMD and that's where it came from. this is my understanding of course, but it's not that hard to believe.
I don't mean to say I know all the facts but I'd be willing to bet most of it was just renaming strings...
You can of course grab the code and tweak it a bit just to make it different, even if just to say it's not the same thing, but if you started off of it, well...
But let's not get into this discussion again, please.
I don't mean to say I know all the facts but I'd be willing to bet most of it was just renaming strings...
You can of course grab the code and tweak it a bit just to make it different, even if just to say it's not the same thing, but if you started off of it, well...
But let's not get into this discussion again, please.
Why it was rejected? Because Microsoft was using Mantle as the base of DX12. What is Mantle is essence? What is DX12 in essence?
Mantle is this: Low CPU overhead, Asynchronous Compute, HSA 2.0 capabilities, explicit multithreading, and direct control over hardware, graphics and compute united into single queue.
What is DirectX 12? All of above plus Direct3D 12.
What is Vulkan? All of above plus OpenGL.
What is Metal? All of above, plus bits and pieces from other APIs like OpenCL, D3D10, D3D11.
If the functionality is the same, the code must be also the same in some parts.
Lets stop this, its Zen thread.
Last edited:
No one said that DX12 has to have a line of code from Mantle. There could be, but it is not mandatory. Koyoot said the rest.