NVIDIA to License Kepler and Future GPU IP to 3rd Parties

bru · 19 Jun 2013 at 07:44

Wow certainly didn't see that coming, after reading the article it seems like a good idea for Nvidia, very forward thinking of them.

http://www.anandtech.com/show/7083/nvidia-to-license-kepler-and-future-gpu-ip-to-3rd-parties

Earlier today NVIDIA announced that it would begin licensing its Kepler GPU architecture to 3rd parties. This is a sensible next step for NVIDIA, but an unprecedented one among the two remaining discrete PC GPU suppliers.

In addition to its GPU architecture, NVIDIA is now also open to licensing its visual computing patents to 3rd parties. The visual computing patent portfolio includes all of NVIDIA’s 5500 patents in the area, as well as CUDA.

icehot · 19 Jun 2013 at 08:12

Great idea, will allow their partners to differentiate themselves even more so.

Optimis0r · 19 Jun 2013 at 08:49

Nice move Nvidia

pgi947 · 19 Jun 2013 at 09:17

So what does this mean for the consumer? Can someone give me an example of how this would work? Just woke up and am a tad confused

Doomedspeed · 19 Jun 2013 at 09:22

pgi947 said:
So what does this mean for the consumer? Can someone give me an example of how this would work? Just woke up and am a tad confused

Same here..

LtMatt · 19 Jun 2013 at 09:23

Nvidia is going to stop making chips and let others do it ?

kedge · 19 Jun 2013 at 09:54

LtMatt said:
Nvidia is going to stop making chips and let others do it ?

Could be right?, who knows?, but this would alleviate nvidia's responsibility for driver updates and not have to take huge flak for their poor drivers or problematic driver sets?. I'd like to see nvidia make a decent gaming console and give microsoft and sony a run for their money.

VIPER258 · 19 Jun 2013 at 09:56

Good move for Nvidia.

layte · 19 Jun 2013 at 10:16

License the tech to Intel? I bet a few AMD execs could have some sleepless nights over that idea. Nvidia's IP combined with Intel's fab capability......

~Dockie · 19 Jun 2013 at 11:33

layte said:
License the tech to Intel?

First thing that came to my mind too.

What will happen to OpenCL if every intel processor is capable of CUDA via its IGP?

Optimis0r · 19 Jun 2013 at 11:34

pgi947 said:
So what does this mean for the consumer? Can someone give me an example of how this would work? Just woke up and am a tad confused

basically means the kepler and future gpu architecture from Nvidia can be tweaked and manufactured by others such as for example Samsung/qualcomm etc.

Think of it like what ARM are doing with their cpu cores. Whos to say their wont be a custom chip from Samsung with ARM cores and Kepler cores for example.

Nvidia wont stop manufacturing high performance desktop graphics cards since its their bread and butter. Its just a way of making more money and to recoup research and dev cost of the kepler and future architecture.

edit: thinking about it.. if Nvidia is licensing their tech could it mean the return of the PhysX / PPU cards???

CAT-THE-FIFTH · 19 Jun 2013 at 12:03

layte said:
License the tech to Intel? I bet a few AMD execs could have some sleepless nights over that idea. Nvidia's IP combined with Intel's fab capability......

Then Intel can produce plug in cards which also means Nvidia execs might have sleepness nights too?

Wait,they already are!

IIRC,Intel does already license some IP from Nvidia already and this has not stopped Intel from trying to take on Nvidia cards in the SC field,which is a major source of its profits.

I have a feeling Nvidia would rather have companies like Samsung,Qualcomm,etc using its IP to cause problems for Intel.

If so its a cunning way of doing things which should help the company out in the longterm. Basically create a cloud of potential competitors to take on Intel and to a lesser degree AMD.

Even ARM was happy to send senior people to AMD events too:

http://techreport.com/news/21126/arm-fellow-takes-the-mic-at-amd-event

So even ARM is cunning too,by getting the last major X86 company outside Intel to start jumping ship,and AMD will be the first introduce an ARM 64 bit server CPU:

http://hexus.net/tech/news/cpu/56941-amd-will-launch-seattle-first-arm-chip-h2-2014/

~Dockie said:
First thing that came to my mind too.

What will happen to OpenCL if every intel processor is capable of CUDA via its IGP?

OpenCL was not initiated by AMD though - Apple did IIRC.

Why do you think Adobe CS for OS X uses OpenCL instead of CUDA and the Mac Pro uses AMD graphics cards??

Even,Intel has put more effort into the OpenCL performance of the latest Intel IGPs.

~Dockie · 19 Jun 2013 at 12:38

CAT-THE-FIFTH said:
OpenCL was not initiated by AMD though - Apple did IIRC.

Why do you think Adobe CS for OS X uses OpenCL instead of CUDA and the Mac Pro uses AMD graphics cards??

Even,Intel has put more effort into the OpenCL performance of the latest Intel IGPs.

My point was, isn't CUDA the preferred platform to most people over OpenCL? if everyone was getting intel CPU's which could potentially be bundled with a CUDA enabled IGP, why would people bother getting an AIB for OpenCL?

I may be wrong though, but i was under the impression people preferred CUDA.

CAT-THE-FIFTH · 19 Jun 2013 at 12:48

~Dockie said:
My point was, isn't CUDA the preferred platform to most people over OpenCL? if everyone was getting intel CPU's which could potentially be bundled with a CUDA enabled IGP, why would people bother getting an AIB for OpenCL?

I may be wrong though, but i was under the impression people preferred CUDA.

http://hsafoundation.com/f-a-q/

HSA is an optimized platform that will be leveraged to enhance the OpenCL runtime. HSA is not an alternative to OpenCL. HSA benefits OpenCL by removing memory copies, bringing low latency dispatch, and improving memory model and pointers shared between the CPU and GPU.

The HSA Foundation will also work with all software foundations and open initiatives for the promotion of respective language runtimes and libraries around the advancement of heterogeneous computing.

Adobe has already made OpenCL the primary path over CUDA for OS X CS,and is already going to introduce it for the Windows version. Software like HandBrake either uses OpenCL or Quick Sync and has no CUDA optimised version.

Anyway,OpenCL works on Nvidia cards.

If anything it is the scientific computing field not the consumer field where CUDA is most prevalent. Intel is attacking Nvidia in this very field and Nvidia needs to respond to this.

Apple is a major backer of OpenCL:

http://en.wikipedia.org/wiki/OpenCL

It means it will be increasingly leveraged accross all of their own products alone.

Look at the number of members:

http://hsafoundation.com/

humbug · 19 Jun 2013 at 15:53

Does this mean others can make their own discrete GPU's by simply paying a small fee to Nvidia?

Surely that's not really going to work for Nvidia in practice with a market flooded with competing GPU's.

humbug · 19 Jun 2013 at 16:09

~Dockie said:
First thing that came to my mind too.

What will happen to OpenCL if every intel processor is capable of CUDA via its IGP?

Doesn't really mater, AMD can just as easily licence CUDA and integrate it into existing GPU's, make them both OpenCL and CUDA ready, may the best format win........

It means 3'parties / developers are free to use both without the problem of users being locked to Nvidia GPU's, Nvidia probably did that realising locking CUDA to themselves wasn't going to work, OpenCL is increasingly being taken up instead of CUDA. and sometimes along with it AMD's GPU's with them being more powerfull OpenCL engines. (Apple)
"format lockout fail"

This is a good thing for us.

andybird123 · 19 Jun 2013 at 16:21

humbug said:
Doesn't really mater, AMD can just as easily licence CUDA and integrate it into existing GPU's, make them both OpenCL and CUDA ready, may the best format win........

that's assuming that nvidia would licence anything to AMD, which of course they wouldn't lol

PorkPieCat · 19 Jun 2013 at 16:21

I guess it's like Creative and Auzen with their sound cards. Other manufacturers will just create slightly modified cards, or in some cases boutique/niche designs like the GTX Titan of today. And they won't be called "Nvidia GTX 980" (for example), but say "Gigabyte Windstream 1.1 (powered by Nvidia)", and they'll differentiate themselves by using completely different board and component designs, and perhaps firmware and hardware tweaks here and there for the chip. And it sounds like there'll be a few more Nvidia products on the market to every one AMD product, for which AMD had to do all the research, all the modifications to reach all price points, all the marketing, etc. Sounds like... trés bon pour Nvidia.

Martini1991 · 19 Jun 2013 at 16:22

humbug said:
Does this mean others can make their own discrete GPU's by simply paying a small fee to Nvidia?

Surely that's not really going to work for Nvidia in practice with a market flooded with competing GPU's.

It could be very similar to Power VR now.

humbug · 19 Jun 2013 at 16:36

andybird123 said:
that's assuming that nvidia would licence anything to AMD, which of course they wouldn't lol

Then nothing has changed, Developers are still locked to Nvidia in using CUDA.

Martini1991 said:
It could be very similar to Power VR now.

Its the bit that I don't understand, allowing anyone to use your GPU technology will mean anyone can compete with you on a level footing.
That would hurt Nvidia more than it would Intel and AMD given that they are nowhere near as dependant on GPU sales as Nvidia are.