CUDA vs OpenCL needs a reality check...

[Lonbjerg]

Indeed.
OpenCL and CUDA are not equal.
OpenCL: 100% of modern PC GPUs support it by sales
CUDA: 16% of modern PC GPUs by sales

http://vr-zone.com/articles/jon-ped...-results-nvidia-share-loss-reduced/16037.html

Nice way to cherry pick.
I guess that must mean tha CUDA have less the 16% of the market, right?

If not...your post is useless

FACT:
You can do stuff TODAY in CUDA...that OpenCL either struggles to do...or cannot do.

Comparing the 2 eco-systems cleary shows that.

Thanks for playing...neeeext!!!

 

[Lonyo]

Bring on OpenCL, someone, anyone, please, we can't pay you, do it for free, it's open source, do it for me!!!

Bring on multi-threading, anyone, please, we can't pay you, do it for free, it's the future, do it for me!!!

 

[BallaTheFeared]

That doesn't make any sense.

 

[Lonbjerg]

It sounds like you're upset that an open standard will soon usurp nvidia's product, and all those years and millions in research, development, and marketing will be snuffed. It was inevitable.

What World do you live in?
CUDA is leading OpenCL...can you please do a joke post more?

 

[Lonbjerg]

Bring on multi-threading, anyone, please, we can't pay you, do it for free, it's the future, do it for me!!!

So why is CUDA leading and OpenCL playing "wannabee"? ^^

 

[Lonyo]

Nice way to cherry pick.
I guess that must mean tha CUDA have less the 16% of the market, right?

If not...your post is useless

FACT:
You can do stuff TODAY in CUDA...that OpenCL either struggles to do...or cannot do.

Comparing the 2 eco-systems cleary shows that.

Thanks for playing...neeeext!!!

Why is my post useless because CUDA doesn't have 15.1% of the market?
I am pointing out that going forward, there will be a far far far larger hardware ecosystem for OpenCL than there will for a closed CUDA.
If you think that a massive hardware ecosystem won't help in getting OpenCL out there, that's your choice.
Currently CUDA may be more refined, and have more support in some areas, but why would I care about something that locks me to specific hardware? Personally, I don't care, because I don't want to be locked to specific hardware based on the software I use. Choice is a nice thing.

I am also not a "high performance" requiring person (e.g. someone using OpenCL or CUDA in a professional context). I would rather have a 50% speedup on 100% of hardware, than a 100% speedup on 16% of hardware. That's a personal preference thing, but to me, 50% on anything I choose is better than 100% with no choice.

Clearly you don't value choice and are happy to be locked in to specific hardware in order to obtain currently higher performance. That doesn't mean everyone shares that view.

 

[Lonyo]

So why is CUDA leading and OpenCL playing "wannabee"? ^^

Because no one has thrown money at it.
Build it and they will come vs build it because we'll give you money.

 

[Cloudfire777]

Does anyone here even use Premiere Pro? :ninja:

 

[Lonbjerg]

Because no one has thrown money at it.
Build it and they will come vs build it because we'll give you money.

Now you are the one sounding mad :awe:
You could suggest AMD that they start making a profit...so they can do all the NVIDIA does...but then again...better to lean on "what if's"...i get it :sneaky:

 

[Lonyo]

Now you are the one sounding mad :awe:
You could suggest AMD that they start making a profit...so they can do all the NVIDIA does...but then again...better to lean on "what if's"...i get it :sneaky:

Pretty sure AMD doesn't control OpenCL.

Here, read this:
http://en.wikipedia.org/wiki/OpenCL

 

[Dankk]

Does anyone here even use Premiere Pro? :ninja:

Yes, I use Premiere Pro. I don't understand how it benefits in performance from "Mercury Playback Engine-supported" graphics cards. I can open, play, and edit multiple HD 1080p timelines flawlessly with no choppiness or anything. Even with multiple effects enabled. My main PC currently does not have a Nvidia card.

Call me ignorant. If I were having performance problems, I would maybe investigate getting a CUDA card. But I'm not.

 

[Olikan]

So why is CUDA leading and OpenCL playing "wannabee"? ^^

just for pure interest.... but what programs is CUDA supported, that OpenCL don't?
because lately, i hear\read nothing about cuda accelareted :|

 

[cmdrdredd]

Does anyone here even use Premiere Pro? :ninja:

I have it as part of CS6, but I don't have any clue how to do anything much with it lol.

Really though...what I wanna understand is why people claim open source stuff is the best thing ever when it doesn't catch on that much? Linux, OpenGL, OpenCL all open source and yet the market for them is comparatively small. I understand the money involved but if it(openCL in this case) was able to be used by all hardware including APUs, iGPUs, and all discreet graphics cards, why has it taken so long for someone to make their product stand out by using it and making it work? There has to be something I am missing. Does it require different programming for each vendor? Meaning Nvidia cards need specific code to function vs AMD and Intel iGPUs different than an APU on an AMD system?

 

[lambchops511]

just for pure interest.... but what programs is CUDA supported, that OpenCL don't?
because lately, i hear\read nothing about cuda accelareted :|

Pretty much any non-consumer custom software. OpenCL is relatively hard to work w/ compared to CUDA. Most of the PhDs graduating are trained to work w/ CUDA, its only natural when they enter the industry to use CUDA as well. For example many custom software such as oil & gas exploration, monte carlo derivative pricing, energy nuclear simulation, artificial intelligence and drug simulation. While it is true some of these do support OpenCL, the majority do use CUDA. This can be seen in the top500 supercomputers, e.g., when US DOE / DOD buys a $1B supercomputer, chances are it supports CUDA w/ NVDA chips rather than AMD because of enterprise support and simulation software support.

Top500: http://top500.org/list/2012/11/

In the top100, there were 16 using CUDA chips (#1, #8, #12, #17), while only 1 computer using AMD (#52 King Abdulaziz City for Science and Technology).

 

[Lonbjerg]

Pretty much any non-consumer custom software. OpenCL is relatively hard to work w/ compared to CUDA. Most of the PhDs graduating are trained to work w/ CUDA, its only natural when they enter the industry to use CUDA as well. For example many custom software such as oil & gas exploration, monte carlo derivative pricing, energy nuclear simulation, artificial intelligence and drug simulation. While it is true some of these do support OpenCL, the majority do use CUDA. This can be seen in the top500 supercomputers, e.g., when US DOE / DOD buys a $1B supercomputer, chances are it supports CUDA w/ NVDA chips rather than AMD because of enterprise support and simulation software support.

Top500: http://top500.org/list/2012/11/

In the top100, there were 16 using CUDA chips (#1, #8, #12, #17), while only 1 computer using AMD (#52 King Abdulaziz City for Science and Technology).

But, but...bit...didn't you get the memo?

OpenCL is BETTER than CUDA!!!
It's Opensource too!!!!

PLEEEeeeeeaaaassee.....

UUUSSEEEEeee....

IIIItttt....

pretty please?

*sarcasm could be applied*

 

[Greenlepricon]

But, but...bit...didn't you get the memo?

OpenCL is BETTER than CUDA!!!
It's Opensource too!!!!

PLEEEeeeeeaaaassee.....

UUUSSEEEEeee....

IIIItttt....

pretty please?

*sarcasm could be applied*

Not even once in there did it say openCL or CUDA is inferior to the other. Just one is more used. If that starts to change (which is what I see happening), then we'll really get to see which is superior. This thread is pretty pointless until Premier Pro starts using it. I have a friend who uses the program and is running a 7950, so once this update comes I'll ask him how it turned out.

 

[Dankk]

But, but...bit...didn't you get the memo?

OpenCL is BETTER than CUDA!!!
It's Opensource too!!!!

PLEEEeeeeeaaaassee.....

UUUSSEEEEeee....

IIIItttt....

pretty please?

*sarcasm could be applied*

Lonbjerg, I think you've been on the internet for long enough today. Go outside and relax for a bit.

 

[Arkadrel]

Not even once in there did it say openCL or CUDA is inferior to the other. Just one is more used. If that starts to change (which is what I see happening), then we'll really get to see which is superior. This thread is pretty pointless until Premier Pro starts using it. I have a friend who uses the program and is running a 7950, so once this update comes I'll ask him how it turned out.

I read there was a trend of CUDA getting less and less adaption rates, for programs
(dispite this its still used in more stuff than OpenCL is currently).

OpenCL is on the rise though. If OpenCL was so bad, there wouldnt be more new programs comeing out with OpenCL support than CUDA ones.

CUDA is king of the hill when it comes to research though.

 

[Lonbjerg]

Not even once in there did it say openCL or CUDA is inferior to the other. Just one is more used. If that starts to change (which is what I see happening), then we'll really get to see which is superior. This thread is pretty pointless until Premier Pro starts using it. I have a friend who uses the program and is running a 7950, so once this update comes I'll ask him how it turned out.

Try reading the links in the OP, before you waste my time again, thanks.

 

[Lonbjerg]

I read there was a trend of CUDA getting less and less adaption rates, for programs
(dispite this its still used in more stuff than OpenCL is currently).

OpenCL is on the rise though. If OpenCL was so bad, there wouldnt be more new programs comeing out with OpenCL support than CUDA ones.

CUDA is king of the hill when it comes to research though.

Because AMD is trying to push OpenCL...guess WHO is behind the push for OpenCL in Adobe?

Hint: It's not Adobe.

 

[Greenlepricon]

Try reading the links in the OP, before you waste my time again, thanks.

Which one? The ones saying that openCL is getting support and should do well against CUDA, or should I travel back in time three years to refresh my history knowledge? I never said that openCL will beat CUDA or even be good, but your links are pointless, as are all your arguments. We have no proof of how this will perform. Get that through your head already.

 

[Arkadrel]

Somehow I dont think AMD is pushing development that hard for OpenCL.

The reason more programs made 2012+, for OpenCL instead of CUDA, is because the program designers want to make something that runs on more PCs.

 

[Erenhardt]

Because AMD is trying to push OpenCL...guess WHO is behind the push for OpenCL in Adobe?

Hint: It's not Adobe.

Uhh...smart...
And CUDA is pushing itself... go back to cave troll!

 

[3DVagabond]

Rvenger works for "an Apple authorized b2b reseller", and pointed out that the comment Cloudfire777 made was incorrect. I tend to take people at their words, and believe someone of his position would know best
(the products he sells).

Why is he getting s**t for that?

Because he didn't bow down to all that is nVidia.

Adobe doesn't need nVidia or CUDA. Adobe, and Adobe's customers, would much prefer that their software didn't even need a discrete GPU. If it runs efficiently on nothing more than an AMD APU or Intel processor, that's a win/win for both Adobe and their customer.

 

[3DVagabond]

I have it as part of CS6, but I don't have any clue how to do anything much with it lol.

Really though...what I wanna understand is why people claim open source stuff is the best thing ever when it doesn't catch on that much? Linux, OpenGL, OpenCL all open source and yet the market for them is comparatively small. I understand the money involved but if it(openCL in this case) was able to be used by all hardware including APUs, iGPUs, and all discreet graphics cards, why has it taken so long for someone to make their product stand out by using it and making it work? There has to be something I am missing. Does it require different programming for each vendor? Meaning Nvidia cards need specific code to function vs AMD and Intel iGPUs different than an APU on an AMD system?

No offense, but this needs a new thread. It has nothing to do with CUDA vs. OpenCL. It is important IMO though.