Bookmark and Share


Roy Taylor, a former high-ranking executive from Nvidia Corp. who now heads channel sales at Advanced Micro Devices, said in an interview that proprietary technologies like Nvidia's CUDA and PhysX were doomed. The industry has come a long way with open-standards GPU compute technologies and nowadays it makes no sense to continue pushing proprietary technologies in a bid to create unique experience.

"I think CUDA is doomed. Our industry does not like proprietary standards. PhysX is an utter failure because it is proprietary. Nobody wants it. You do not want it, I do not want it, gamers do not want it. Analysts do not want it. In the early days of our industry, you could get away with it and it worked. We have all had enough of it. They are unhealthy," said Mr. Taylor in an interview with VR-Zone web-site.

Mr. Taylor, who spent more than a decade at Nvidia and was instrumental in promoting numerous proprietary technologies, admits that early in the days of general-purpose computing on graphics processing units (GPGPU technologies) there was an urgent need for technologies that simply worked, no matter that they only worked on certain hardware. Nowadays, GPGPU is supported by industry-standard OpenCL and Microsoft’s DirectCompute, it makes no sense for developers to program anything on low-level for particular GPUs.

“Nvidia should be congratulated for its invention. As a trend, GPGPU is absolutely fantastic and fabulous. But that was then, this is now. Now, collectively our industry does not want a proprietary standard. That is why people are migrating to OpenCL,” said Mr. Taylor.

While Nvidia will likely continue to promote its proprietary CUDA, PhysX and 3DVision technologies among its sponsored PC titles, it is unlikely that those will actually be more or less popular on the market. Both next-generation video game consoles – Microsoft Corp.’s Xbox One and Sony Corp.’s PlayStation 4 – rely on AMD Fusion hardware and will most likely take advantage of open-standards. Titles developed for the consoles will migrate to PCs and therefore make OpenCL and DirectCompute de-facto standards for GPGPU technologies on this platform.

Last year AMD teamed up with a number of industry leaders to form heterogeneous system architecture foundation, which purpose is to enable hardware and software developers to take advantage of industry-standard computing on different types of cores (general-purpose cores, highly-parallel graphics processing units, etc.). While loads of companies, including ARM, Imagination, Qualcomm, Samsung, Sony, LG and others, have joined the alliance, the two companies that decided to distance themselves from the group are Intel Corp. and Nvidia.

Tags: Nvidia, CUDA, PhysX, Tesla, Tegra, Quadro, OpenCL, GPGPU


Comments currently: 13
Discussion started: 08/08/13 01:06:55 AM
Latest comment: 11/20/13 07:28:34 PM
Expand all threads | Collapse all threads


two companies that decided to distance themselves from the group are Intel Corp. and Nvidia

well, what do you know...
0 0 [Posted by: Yorgos  | Date: 08/08/13 01:06:55 AM]

I hope never to see that PhysX crap again.
1 1 [Posted by: TAViX  | Date: 08/08/13 02:07:28 AM]
- collapse thread

You know you can run PhysX on AMD systems too right? You just don't get hardware accelerated particles.

And as a physics engine, it's not all that different than Havok(and actually has a better licensing agreement!) or Bullet, but it's currently the only stable hardware accelerated physics engine freely available. (Havok and Bullet have hardware accelerated pipelines in the works, but not market ready)
0 1 [Posted by: basroil  | Date: 08/08/13 07:23:11 PM]
You can, but your game becomes unplayable if you use the full effects. Ex: Mirror's Edge, Batman, Metro 2,Borderlands 2, Mafia 2, etc, etc.
0 0 [Posted by: TAViX  | Date: 08/09/13 08:42:35 AM]
So you think that you will not slow down if Havok have the same full effect like physx?
0 0 [Posted by: Sagia86  | Date: 11/20/13 07:28:34 PM]

Nvidia is doomed.

Quote Semiaccurate: "The path Nvidia has locked itself into is unsustainable and they are unquestionably aware of this, their reactive roadmap changes are short-term holding actions but fatal long-term.... 100% of next generation game engine development is AMD architecture focused".(Dated 2013-08-08)
2 3 [Posted by: linuxlowdown  | Date: 08/08/13 02:53:49 AM]

4. is unlikely that those will actually be more or less popular on the market.

can we get a copywriter in here stat ?
0 0 [Posted by: p3ngwin  | Date: 08/08/13 04:14:09 AM]

Interesting... Last I checked the top 500 supercomputer list has just as many openCL computers as CUDA ones. And only one with AMD openCL! CUDA is available in open source compilers too. As for proprietary meaning hardware being proprietary... load of BS there, considering AMD always includes proprietary components, and the top 500 is almost entirely proprietary (infiniband is proprietary after all).

As for " it makes no sense for developers to program anything on low-level for particular GPUs.", AMD and Sony are pushing for exactly that with the PS4, since without that it's an underpowered piece of crap!
0 1 [Posted by: basroil  | Date: 08/08/13 07:13:38 PM]
- collapse thread

Nice Job taking information out of context.

Obviously this article isn't about running a physics engine on a supercomputer or a console. Both of those examples are MADE for specific purposes that basicly demand that they run close to the rails and usually running some sort of proprietary software.

I agree with the point of the article; the fact that PhysX will disappear and that Nvidia will have a serious software disadvantage for the upcoming console generation. It will be hard for Nvidia to overcome AMD's hardware when every dev. will most likely have better optimization for the competitor.
1 1 [Posted by: evernessince  | Date: 08/09/13 09:34:12 PM]
No, the article is about one ex-dev that has no clue about what his own programs were for. CUDA is almost entirely a high performance computing language, it's not supposed to replace CPUs and does not have a lot of the implementation restrictions OpenCL does (one of the reason AMD's own Bullet engine was first ported to CUDA then stalled for four years while OpenCL hardware and drivers got up to speed, same with Blender). CUDA was never developed with the consumer market in mind, and corporate markets actually prefer proprietary things (means easier support)

As for PhysX, it actually has a higher AAA game use base than Bullet, and only Havok is higher up. Havok, like PhysX, is proprietary and unlike PhysX entirely behind an NDA (you can't even know what methods it uses to calculate physic, and even benchmarks are entirely banned!). PhysX was already ported to PS4, so it's not like it's out of the running yet. They are probably avoiding the xboxone though because they would end up doing compute with AMP++/DX11, which would mean desktop games would also have hardware physics on AMD.
0 1 [Posted by: basroil  | Date: 08/10/13 03:05:21 AM]

Is there an alternative to PhysX? Not at the moment, I believe.

PhysX can completely change the gameplay. Just look at Borderlands 2, it is a completely different experience with PhysX.

I would like a universal API like PhysX that would work on all platforms, but until then PhysX will still be with us.
0 1 [Posted by: Harry Lloyd  | Date: 08/13/13 04:08:51 AM]


Add your Comment

Related news

Latest News

Wednesday, November 5, 2014

10:48 pm | LG’s Unique Ultra-Wide Curved 34” Display Finally Hits the Market. LG 34UC97 Available in the U.S. and the U.K.

Wednesday, October 8, 2014

12:52 pm | Lisa Su Appointed as New CEO of Advanced Micro Devices. Rory Read Steps Down, Lisa Su Becomes New CEO of AMD

Thursday, August 28, 2014

4:22 am | AMD Has No Plans to Reconsider Recommended Prices of Radeon R9 Graphics Cards. AMD Will Not Lower Recommended Prices of Radeon R9 Graphics Solutions

Wednesday, August 27, 2014

1:09 pm | Samsung Begins to Produce 2.13GHz 64GB DDR4 Memory Modules. Samsung Uses TSV DRAMs for 64GB DDR4 RDIMMs

Tuesday, August 26, 2014

10:41 am | AMD Quietly Reveals Third Iteration of GCN Architecture with Tonga GPU. AMD Unleashes Radeon R9 285 Graphics Cards, Tonga GPU, GCN 1.2 Architecture