CUDA reverse engineered to work on Intel and AMD.

[zacRupnow]

 

SOURCE: http://venturebeat.com/2016/03/09/otoy-breakthrough-lets-game-developers-run-the-best-graphics-software-across-platforms/

 

Quote

It may be very arcane to most of us, but graphics startup Otoy has come up with a breakthrough that should help game developers create much more beautiful games that can run across different hardware platforms.

In a nutshell, Otoy reverse-engineered Nvidia’s general purpose graphics processing unit (GPGPU) software, known as CUDA, to run on non-Nvidia hardware. That means that programs written in the CUDA language are no longer exclusive to Nvidia graphics chips. Now they can run on GPUs from Advanced Micro Devices, ARM, and Intel. That means a CUDA program written for the PC could run on a PlayStation 4 or an Apple iPad.

Otoy has ported the CUDA language to run on AMD, Intel, and ARM GPUs for the first time ever. That means any application written in CUDA can run, without modification, on all major chipsets and platforms. While there is an independent GPGPU standard dubbed OpenCL, it isn’t necessarily as good as CUDA, Otoy believes.

I won't pretend to know alot about CUDA but I know it's a big deal for modeling and architecture. I don't know about its application in games.

Broad compatibility is always good, and there is not many reasons to need to lock software to hardware outside of the live lighting production world.

 

 

Quote

“We have been able to do this without changing a line of CUDA code, and it runs on AMD chips,” Urbach said. “You can now program once and take CUDA everywhere. AMD has never really been able to provide an alternative.”

Now I see a few responses NVIDIA may have.

1. They say nothing. -EDIT- And also do either 3 or 4.
2. They say that Otoy's re-work is an unofficial modification and they will not be responsible for anything going wrong. This is of course obvious but the statement would be meant to denounce or discourage the use of Otoy's CUDA.
3. They update their code to be exclusive again along with their next feature update.
4. They immediately release an update.

 

[Tech_Dreamer]

 Nvidia right now...

 

[TrigrH]

this is only for octane though

[iamdarkyoshi]

This is going to lead to nvidia attacking AMD with dedicated hair processing software and AMD shooting fire from a 290x back at them in defense, while intel drives off in a prius

[Dabombinable]

Doesn't Intel already have the license to use CUDA, they just haven't yet?

11 minutes ago, Tech_Dreamer said:

 Nvidia right now...

 

Nah, that's for when you use an Nvidia card for PhysX with an AMD one, or SLI on non SLI certified boards 

[QueenDemetria]

1 minute ago, iamdarkyoshi said:

while intel drives off in a prius

Intel is the type of company that would drive BMW i3's and i5's, they would probably pay BMW to rename the i8 to i7 too. 

[Sors]

20 minutes ago, zacRupnow said:

 

Now I see a few responses NVIDIA may have.

1. They say nothing.
2. They say that Otoy's re-work is an unofficial modification and they will not be responsible for anything going wrong.
3. This is of course obvious but the statement would be meant to denounce or discourage the use of Otoy's CUDA.
4. They update their code to be exclusive again along with their next feature update.
5. They immediately release an update.

 

2,3,4 and 5 will in all probability be what happens.

[zacRupnow]

1 minute ago, Sors said:

2,3,4 and 5 will in all probability be what happens.

That's what I think but they might do the updates without saying anything.

[Dabombinable]

2 minutes ago, Sors said:

2,3,4 and 5 will in all probability be what happens.

Doinf 4 would create a game of cat and mouse similar to what happened when people figured out how to enable SLI on motherboards with the right PCIe slot configuration but no Nvidia chipset.

http://www.overclock.net/t/670810/howto-sli-on-a-non-sli-motherboard-gtx5xx-compatible-1-0-final-released

 

Should get 2x GTX 580 and run them on my P5Q Deluxe+Xeon X5450 (3.6GHz, 4-4.4GHz when I get a better cooler).

[Citadelen]

I'm interested to see how this will affect AMDs project HIP.

[Bensemus]

Didn't AMD recently do something that gave them access to CUDA?

[Prysin]

5 minutes ago, Bensemus said:

Didn't AMD recently do something that gave them access to CUDA?

supposedly so, yes

 

[deviant88]

who said nvidia is going to allow it? they are going to court unless they signed anything with n-fucking-vidia

[AlwaysFSX]

Yeah CUDA requires a license for a reason, anyone using this in their product is immediately going to eat shit in court. AMD and Intel could get licenses for CUDA, but refuse not to. Their problem, not Nvidia's.

[UnknownEngineer]

Not only will anyone using this face serious legal consequences, but the performance is not guaranteed, all they say is the "it will run". They aren't saying that it will run without performance degradation/overhead, or whatever you want to call it. 

[AlexGoesHigh]

2 hours ago, AlwaysFSX said:

Yeah CUDA requires a license for a reason, anyone using this in their product is immediately going to eat shit in court. AMD and Intel could get licenses for CUDA, but refuse not to. Their problem, not Nvidia's.

Intel has a CUDA license but i think they only use it for Xeon Phi, not sure if you can get official CUDA support into a iGPU, and AMD has actively refused to get a CUDA license. From what i read on the forum whenever CUDA is part of the discussion is that there was a point where Nvidia offered them the license for no cost or ultra cheap or something like that, but AMD for whatever reason always refused, though recently they announced a way that CUDA code can work on their GPU's but is not exactly official CUDA is more like a converter of sorts, that turns it into C++ or another non proprietary language but either way CUDA in some form should work on AMD at some point.

[zMeul]

NGOHQ did this years ago and they got slapped by nVidia with cease and desists 

[Fetzie]

25 minutes ago, zMeul said:

NGOHQ did this years ago and they got slapped by nVidia with cease and desists 

Yeah, that tends to happen when somebody reverse-engineers proprietary software and distributes it without the permission of the owner. It'll happen here too.

[ApolloEleven]

Death to proprietary software! Hello open source. 

[Megahurt]

If I'm not mistaken, AMD got a CUDA license not too long ago, and I think the reason they didn't have support for it is that it would be patent infringement rather than them not being able to figure out how it works, and that applies to the rest of the proprietary nvidia stuff.

[DXMember]

kewl I guess...

AMD announced CUDA support to come a while ago

but can't we just use ACE?

[byalexandr]

Haha, the one thing Nvidia had over AMD at rendering is now fucked over! Screw you Nvidia!

[Jahramika]

4 hours ago, AlwaysFSX said:

Yeah CUDA requires a license for a reason, anyone using this in their product is immediately going to eat shit in court. AMD and Intel could get licenses for CUDA, but refuse not to. Their problem, not Nvidia's.

No it is Nvidia everything they do is closed.

Edited March 12, 2016 by Godlygamer23
Clean up.

[AlwaysFSX]

4 minutes ago, Jahramika said:

No it is Nvidia everything they do is closed.

Good, prevents idiots from fucking up software because they think they know what they're doing.

Edited March 12, 2016 by Godlygamer23
Clean up.

[Majestic]

7 hours ago, iamdarkyoshi said:

 while intel drives off in a prius

lol, i'd not rule out intel that easily. They've been gaining quite the momentum with their iGPU's. I'm suspecting those Iris Pro 580's will be quite competitive, if placed in affordable CPU packages.

 

Believe they're specced for 1,2Tflops, which is GTX 750/750TI territory.