As much as I want that to happen, it won't. If Nvidia supports AdaptiveSync, then G-sync would fail immediately and all the R&D and manufacturing would go to waste. CUDA is basically the reason why the professional market is exclusive to Nvidia. OpenCL is getting there but still nowhere near the status that Cuda is.
Open source is like a curse word for Nvidia.
Add the fact that DX12 and Vulkan will lower the operations of the CPU, PhysX would be nothing for a CPU.
I disagree. GFLOPS of a CPU are orders of magnitude smaller than those of GPUs, due to the benefits of parallelization. Physics calculations are best done in parallel, and often don't require that much single-threaded power.
Physics will still cripple modern CPUs. GPUs are where most of the future is in.
(Beginning tangent about possible future for GPUs in 3, 2, 1...)
And with low-powered applications (like Android and iOS apps) running on ARM processors that run at 1.2Ghz, we could see a super-parallelized OS running on a GPU, with similarly-powered apps as modern smartphone/tablet apps.
Desktop GPU speeds are very comparable to those of ARM CPUs in smartphones and tablets. And AMD CPUs can run OpenCL, so the question is whether they could be used to run virtualized ARM-based OSes (like Android and iOS) as a means of offering better computational efficiency than using CPUs to run virtual OSes?
(Weapons lock on incoming tanget. Fire!... target destroyed.)
True. I know Nvidia won't easily give up on G-Sync until Adaptive Sync is everywhere, and until the market is so angry at Nvidia for not offering Adaptive Sync on their GPUs that they'd lose more money from not supporting Adaptive Sync than they'd stand to gain than by keeping a closed ecosystem.
maybe not on the topic anymore but this also might help amd and intel with drivers...
It won't help AMD nor Intel. Here's why:
Nvidia owns all rights to all of the code. And any modifications to it, as well, including modifications to allow this code to run on AMD GPUs or Intel co-processor cards (like Xeon Phi discrete co-processors, or Intel integrated graphics).
This means Nvidia just wants some nice publicity. But the emperor is still naked, and they haven't made any changes to remove their walled-garden, closed-ecosystem mentality. They just say "Open-Source" because it sounds nice. In reality they just shared the source code with those who agree to the terms of their EULA, but without letting anyone modify or distribute their code without their approval and/or licensing agreements. Meaning that AMD would have to get permission and pay royalties to run PhysX, and only they'd only be able to run a modified version of PhysX for CPUs, not the full-featured CUDA version of PhysX.
ok, but hold one - even if they cannot edit their repo physx code they can modify their drivers.
Also nope. The license also says this:
That means including PhysX for CPUs in AMD drivers is also prohibited. PhysX for CPUs can run separately from drivers, and is normally included in a game's installer if said game uses PhysX for CPUs.
So nobody can modify, use or redistribute PhysX code without Nvidia permission and/or license agreement. AMD or Intel.
Intel and AMD could modify their drivers, yes. But they cannot use or modify PhysX code in any way. (That probably also means that not only could they not distribute PhysX for CPUs with their drivers, they also couldn't modify the code in RAM to make it run on their GPUs in real-time. The modification to code doesn't refer strictly to source code, it can also refer to running binary code in DRAM.)
So, no. This changes nothing, other than Nvidia hogging up publicity over the false notion that they're changing their Prime Directive. Jon Stewart from the Daily Show said it best:
"That's their prime directive. And unlike Captain Kirk, they f$%#ing stick to the prime directive. They don't just drop the protocol every time they feel like humping a green girl in a unitard."
I think you misunderstand what I wrote - knowing whats in the code they can optimize their drivers. They can do it without knowledge of nvidia.
Well, considering that the Nvidia PhysX for CPUs code will be running on x86 without being included on their drivers, without being modified or modifiable, the only thing AMD or Intel could do is try to make their drivers less CPU-intensive with DX12, Mantle or glNext (next-gen OpenGL). Aside from that, which Intel and AMD should be doing anyways, there's no optimization on the driver side of things they could do, because the code wouldn't run on their drivers or GPUs, but rather on their CPUs (which don't require drivers anyways).