CUDA reverse engineered to work on Intel and AMD.

[Sauron]

1 hour ago, dalekphalm said:

Source?

No idea, I just thought I read that somewhere

[zacRupnow]

1 hour ago, Kloaked said:

Are you using the "Insert Other Media" option at the bottom right of the edit box?

Thanks, it was in the top with all the formatting tools on the old forum. Ever time I posted a picture link it just appeared so I never even thought to look at that... Derp.

[Lotus]

"AMD has been unable to provide an alternative"

 

...except for OpenCL which anyone can use, and in editing suites that have implemented both show that the lower priced R9 390 running OpenCL performs better than the more expensive GTX 980 running Cuda.

[spartaman64]

15 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.

putting money into the bank of your competitor is generally a bad idea

[AlwaysFSX]

1 hour ago, spartaman64 said:

putting money into the bank of your competitor is generally a bad idea

Yeah, that's why companies around the world do it every day. 

[Kobathor]

17 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.

honestly tho they probably want way too much for CUDA licenses

hundreds of millions, maybe

I mean obviously the code works on GPU's other than nVidia ones, so it's just the code they own

[spartaman64]

16 minutes ago, AlwaysFSX said:

Yeah, that's why companies around the world do it every day. 

ikr companies support each other's standards all the time thats why nvidia support adaptive sync and opengpu when amd makes them an open standard

[AlwaysFSX]

11 minutes ago, Kobathor said:

honestly tho they probably want way too much for CUDA licenses

hundreds of millions, maybe

I mean obviously the code works on GPU's other than nVidia ones, so it's just the code they own

Oh I highly doubt that. I'm super rusty but there are anti-competitive laws that prevent Nvidia charging someone more for a license to Company A compared to what they charge Company B just 'cause of some people having a "fuck that guy" mentality.

2 minutes ago, spartaman64 said:

ikr companies support each other's standards all the time thats why nvidia support adaptive sync and opengpu when amd makes them an open standard

Hmm, maybe if VESA added the Adaptive Sync standard to monitors long before Nvidia decided to work on it themselves this wouldn't be a discussion. Don't give Nvidia shit for doing things when others are slacking it.

[Heesleemer]

9 hours ago, Lotus said:

"AMD has been unable to provide an alternative"

 

...except for OpenCL which anyone can use, and in editing suites that have implemented both show that the lower priced R9 390 running OpenCL performs better than the more expensive GTX 980 running Cuda.

Benchmarks in Adobe CC please? That is one thing that has prevented me from jumping onto AMD is open cl support on adobe. 

[Lotus]

11 minutes ago, Heesleemer said:

Benchmarks in Adobe CC please? That is one thing that has prevented me from jumping onto AMD is open cl support on adobe. 

 

Disregard. I couldn't find benchmarks, and it's pissing me off. My google-fu is failing me.

[Kobathor]

4 minutes ago, AlwaysFSX said:

Oh I highly doubt that. I'm super rusty but there are anti-competitive laws that prevent Nvidia charging someone more for a license to Company A compared to what they charge Company B just 'cause of some people having a "fuck that guy" mentality.

I mean they'll charge anybody tens or hundreds of millions, not just AMD or Intel. It's a "good" proprietary system, why wouldn't they charge out the ass for it?

[Heesleemer]

3 minutes ago, Lotus said:

 

Now remember that AMD GPUs have more computation horsepower than NVidia GPUs by a very wide margin.

yeah but Open CL performance on specific software such as adobe has been pretty shit. 

[AlwaysFSX]

4 minutes ago, Kobathor said:

I mean they'll charge anybody tens or hundreds of millions, not just AMD or Intel. It's a "good" proprietary system, why wouldn't they charge out the ass for it?

Because then nobody pays the money for it. Whether you guys like it or not, Nvidia is very good with business.

[Kobathor]

1 minute ago, AlwaysFSX said:

Because then nobody pays the money for it. Whether you guys like it or not, Nvidia is very good with business.

Yeah, that's why it's still nVidia only, because nobody is going to pay the money for it. That's why people should develop with OpenCL instead of CUDA..

[AlwaysFSX]

8 minutes ago, Kobathor said:

Yeah, that's why it's still nVidia only, because nobody is going to pay the money for it. That's why people should develop with OpenCL instead of CUDA..

I dunno if you know this, but Nvidia is rolling in money from CUDA licenses.

[Kobathor]

3 minutes ago, AlwaysFSX said:

I dunno if you know this, but Nvidia is rolling in money from CUDA licenses.

I don't mean from the people developing apps for it (there are so many CUDA-compatible apps,) I mean large manufacturers (AMD/Intel/ARM) paying to make it work with their processors. Or has that happened already?

I'm asking in a serious manner, I'm not up-to-date on CUDA stuff.

[AlwaysFSX]

6 minutes ago, Kobathor said:

I don't mean from the people developing apps for it (there are so many CUDA-compatible apps,) I mean large manufacturers (AMD/Intel/ARM) paying to make it work with their processors. Or has that happened already?

I'm asking in a serious manner, I'm not up-to-date on CUDA stuff.

AMD refuses to or they would have by now to make any semblance of a standard with PhsyX for gaming, Intel has (had? don't remember) a CUDA license they play around with, ARM is their own little secluded project of craziness that doesn't poke things outside of their bubble too much.