Cuda's future

[Seero]

Read your own links.

I guess you are missing the point or simply trying to argue.

I said the article writer was bias. You used that link in some strange way to try and make it look like AMD was stupid for not implementing CUDA.

So is it Nvidia's fault developer chose to implement CUDA over Stream?

I don't know how you think you are deriving these numbers. As for the sabotage, while it is possible that nVidia wouldn't do it, it isn't guaranteed. AMD would be foolish to invest large amounts of money into a direct competitors technology.

I agree with your sabotage argument, but doesn't it also work on DirectCompute and Intel's CPUs/GPUs?

I never said anything about it being open. I said it was limited to nVidia only platforms. How is that beyond you, it is a true statement.

Yes, it is a true statement.

1)CUDA+NVidia: 100% efficiency
2)Stream+AMD : 100% efficiency
3)DirectCompute: 50% efficiency on both vendors, 70% on Intel.
4)OpenCL: 45% efficiency on all vendors.

I like option 3) best ^-^ and think the goal should be to make it as effective on both hardwares as possible. Even if its still beat by cuda/stream in effecienty, I think developers will go for it simply because of how many users, use windows and want a solution that ll work reguardless of whos card your useing.

Im probably biased because I use a ATI card, and Im not sure what stream can offer me as a casual user. I see cuda does have a few things for the casual user, and would love to see developers go with mircosofts direct compute and make that work so ati users could have that too.

Im rooting for mircrosoft to set the standart and win that market in the software GPGPU department. Since its a long term investment.... from mircosofts side... It seems likely that they ll gain more ground over time (there are alot of windows users).

Stream can offer you as much as CUDA can offer to Nvdia user. There are not a lot more programs that only benefits nvidia cards in terms of GPGPU. The biggest program is Adobe CS5. If you don't use it, than it won't rock your boat.

Now look at what is happening now. People are going crazy on smartphones. Some employees of some fruit company commit suicide because they cannot fill up the orders, while some get poisoned while making their product. I really don't see a brighter future of DirectCompute compare to CUDA. I will not be surprised if iOS takes over the crown in the future. If so, since CUDA is not OS dependent, it may actually have a brighter future.

 

[Bootleg Betty]

For example:
CUDA+NVidia: 100% efficiency
Stream+AMD : 100% efficiency
DirectCompute: 50% efficiency on both vendors, 70% on Intel.
OpenCL: 45% efficiency on all vendors.

I realize it's an example, but is there any way to get real numbers like these?

Because in benchmarks of older (non-profesional) cards, the performance of CUDA and OpenCL was mostly the same, however nvidia cards were better than amd cards. That was phoronix, though.

Also I'm not sure if AMD even has something comparable to tesla at the moment.

 

[Scali]

Also I'm not sure if AMD even has something comparable to tesla at the moment.

Basically:
GeForce -- Radeon
Quadro -- FirePro
Tesla -- FireStream

http://www.amd.com/us/products/workstation/firestream/firestream-9270/pages/firestream-9270.aspx

 

[Arkadrel]

http://www.brightsideofnews.com/new...nchmark-released2c-increases-gpgpu-tests.aspx


Sandra soft 2011 benchmark.

we spoke with Adrian and he told us that right now, H.264 Transcoding nor GPGPU Cryptography [hashing] do not work in nVidia's current Drivers. This is due to lack of exposure of some OpenCL features

Sounds like... nvidia are banking everything on Cuda, so arnt bothered with doing OpenCL drivers so they can have GPGPU support in this benchmark.

This might be a trend... we could see nvidia try to *avoid* doing OpenCL or DirectCompute support, hopeing that more will adopting cuda.

This might overall result in nvidia users missing out on things (if its not coded for cuda).

 

[Keysplayr]

http://www.brightsideofnews.com/new...nchmark-released2c-increases-gpgpu-tests.aspx


Sandra soft 2011 benchmark.



Sounds like... nvidia are banking everything on Cuda, so arnt bothered with doing OpenCL drivers so they can have GPGPU support in this benchmark.

This might be a trend... we could see nvidia try to *avoid* doing OpenCL or DirectCompute support, hopeing that more will adopting cuda.

This might overall result in nvidia users missing out on things (if its not coded for cuda).

Wrong company. And what you're alluding to here probably amounts to a simple driver bug in this particular bench, which is far more likely than your scenario.

 

[ModestGamer]

Ah yes, more nvidia is doomed conjecture.

In reality, its more like nvidia keeps on actively developing useful stuff and AMD and its fan's keep on whining about how its not open source, won't be accepted yada yada......open source is great but when its proprietary vs nothing its not really a problem.

IF cuda does get eventually supplanted, nvidia won't really be that upset. They support all the open standards better then AMD NOW anyway. they developed their own standard to get the ball rolling unlike AMD.

Now that microsoft has chosen a direction and a aPI. I fully expect AMD to embrace this.

Nvidia is getting push out of the market, yet again.

your reading this move the wrong way. I'd bet both AMD and Intel lobbied microsoft to set a standard. Now it is being delivered.

AMd and Intel will both very likely support it going forward, whereas the brigde to know where. Well it was built on cuda.

I expect in about 2 years cuda will be a has been.

 

[Keysplayr]

Now that microsoft has chosen a direction and a aPI. I fully expect AMD to embrace this.

Nvidia is getting push out of the market, yet again.

your reading this move the wrong way. I'd bet both AMD and Intel lobbied microsoft to set a standard. Now it is being delivered.

AMd and Intel will both very likely support it going forward, whereas the brigde to know where. Well it was built on cuda.

I expect in about 2 years cuda will be a has been.

Have you been given any indication that Nvidia isn't embracing it? And as far as I can tell, offers the better support for OpenCL/Directcompute over AMD and Intel? And I'm not saying this to insult, but, what are you thinking, or are you thinking?

 

[Genx87]

DirectCompute has been out for over a year. Why hasnt Adobe moved to it yet? What about the other projects being built on Cuda? You would think if DirectCompute is as potent as you claim. Cuda would had been shelved over a year ago in favor of the Microsoft standard.

Either way I dont think nvidia really cares whether they smoke AMDs product using Cuda or DirectCompute. At the end of the day the hardware and drivers still has to perform.

 

[ModestGamer]

Have you been given any indication that Nvidia isn't embracing it? And as far as I can tell, offers the better support for OpenCL/Directcompute over AMD and Intel? What are you thinking?

intel and AMD have 75% of the marketshare in graphics and the honest truth is, now that MS has derivived a standard we will see both intel and AMD introduce the needed chip level components to the gpu to make them compliant with the new microsoft standard therby leaving the propriatary stanrd of cuda in the dirt. BTW opencl performance etc isn't any better on either companys hardware and as software developers move to the MS API you'll see the cuda has no customers. Why develop for one piece of hardware when a single API makes the application far more deployeable.

I certainly wouldn't support cuda if I was a software vendor. I'd support that which gives me the largest customer base. I would work hard to increase my efficiencys to.

 

[ModestGamer]

DirectCompute has been out for over a year. Why hasnt Adobe moved to it yet? What about the other projects being built on Cuda? You would think if DirectCompute is as potent as you claim. Cuda would had been shelved over a year ago in favor of the Microsoft standard.

Either way I dont think nvidia really cares whether they smoke AMDs product using Cuda or DirectCompute. At the end of the day the hardware and drivers still has to perform.

I am willing to bet that the new 69xx cards on the new arch from AMd " i bet intel gpu will be as well" will be 100% directcompute commpliant if not have 100% conformance.

You think this standard was developed without guidance from the hardware vendors ?

LOLZ

 

[Keysplayr]

intel and AMD have 75% of the marketshare in graphics and the honest truth is, now that MS has derivived a standard we will see both intel and AMD introduce the needed chip level components to the gpu to make them compliant with the new microsoft standard therby leaving the propriatary stanrd of cuda in the dirt. BTW opencl performance etc isn't any better on either companys hardware and as software developers move to the MS API you'll see the cuda has no customers. Why develop for one piece of hardware when a single API makes the application far more deployeable.

I certainly wouldn't support cuda if I was a software vendor. I'd support that which gives me the largest customer base. I would work hard to increase my efficiencys to.

Defo on the wrong train of thought there partner.

 

[ModestGamer]

Defo on the wrong train of thought there partner.

yes becuase my prediction about market direction have been so wrong in the previous months including nvidias new cashflow pinch.

being a bussiness owner give me a far different perspective then a enthuasits or engineer.

when AMD and intel control the graphics market collectively, they get to set the standards and let me explain something to you, windows is 95% of the installed OS user base, when they set the standard hardware vendors bend to their will.

your just not grasping the reality.

 

[Genx87]

I am willing to bet that the new 69xx cards on the new arch from AMd " i bet intel gpu will be as well" will be 100% directcompute commpliant if not have 100% conformance.

You think this standard was developed without guidance from the hardware vendors ?

LOLZ

They may be 100% compliant but that doesnt mean the drivers will be up to snuff nor they will perform as well as Nvidia's solution. Nvidia will support directcompute just as well.

 

[ModestGamer]

They may be 100% compliant but that doesnt mean the drivers will be up to snuff nor they will perform as well as Nvidia's solution. Nvidia will support directcompute just as well and they have.

honestly it won't matter. all that matters is deployment. They can deploy easily on 100% of hardware whereas currently cuda deploys only on 90% of the professional market " read small market" and 25% of home pc's "thats being generous depends on how new the hardware is".

lets see who gets adopted first.

I think the answer is obvious to even a catatonic person.

 

[Arkadrel]

I too think we ll see more and more of DirectCompute in the future. That might mean less and less cuda, unless Nvidia throws alot of money into helping software developers also include cuda. If not... Nvidia simply start useing DirectCompute too (and cuda only for places where DirectCompute cant work).

 

[ModestGamer]

I too think we ll see more and more of DirectCompute in the future. That might mean less and less cuda, unless Nvidia throws alot of money into helping software developers also include cuda. If not... Nvidia simply start useing DirectCompute too (and cuda only for places where DirectCompute cant work).

I don't think nvidia as a company can make it to the end of next year to be flatly honest.

 

[Genx87]

honestly it won't matter. all that matters is deployment. They can deploy easily on 100% of hardware whereas currently cuda deploys only on 90% of the professional market " read small market" and 25% of home pc's "thats being generous depends on how new the hardware is".

lets see who gets adopted first.

I think the answer is obvious to even a catatonic person.

Uh lets not throw around insults when you make a statement about what is being adopted first. Clearly Cuda over DirectCompute at this point.

That said I am not really sure what your point is right now. Nvidia will support whatever they need to support to have their hardware sell. They support Cuda, DirectCompute, and OpenCL far better than AMD.

I know you want to take some credit for predicting a cash flow problem in the middle of the biggest recession the world has seen in 35 years. But even somebody in a catatonic state could see that one coming.

 

[ModestGamer]

I am not even going to bother correcting how absolutely fucking flawed your thought process is, thank god you aren't in charge of anything important.

BTW AMD is doing fine on cashflow.

Nvidia got pinched on cashflow becuase intel and AMD are shutting them out of integrated graphics. The CASHFLOW generator for nvidia that kept the workforce employed.

Now back to reality.

nvidia has lost alot of product sales, even though it was low profit, they are not getting squeezed on a GPU API that microsoft decided to implement.

AMD and Intel will both have superior performance in direct compute becuase intel and amd likely wrote the damn standard for hardware they were developing.

BTW direct compute is getting 100% adoption.

get over it.


BTW if they ban me for telling you that your just not thinking and your head is jammed up your ass. I really don't fucking care.

Uh lets not throw around insults when you make a statement about what is being adopted first. Clearly Cuda over DirectCompute at this point.

That said I am not really sure what your point is right now. Nvidia will support whatever they need to support to have their hardware sell. They support Cuda, DirectCompute, and OpenCL far better than AMD.

I know you want to take some credit for predicting a cash flow problem in the middle of the biggest recession the world has seen in 35 years. But even somebody in a catatonic state could see that one coming.

profanities are not allowed in the tech forums. -Admin DrPizza

 

[Genx87]

Clearly logic based thinking with lots of evidence to back it up.

 

[Keysplayr]

Clearly logic based thinking with lots of evidence to back it up.

And very, very nasty. No reason for that sort of crap.

 

[SickBeast]

I think that whatever Sandy Bridge uses for their GPGPU processing is going to in large part define what the industry does going forward, provided that they use an open standard. If Intel has their own proprietary "CUDA" platform, then things could get messy and who knows which standard will take over.

While I applaud Nvidia for their pioneering work in this realm, IMO they need to open-source the CUDA platform for their own good.

 

[Dark Shroud]

I think that whatever Sandy Bridge uses for their GPGPU processing is going to in large part define what the industry does going forward, provided that they use an open standard. If Intel has their own proprietary "CUDA" platform, then things could get messy and who knows which standard will take over.

Intel Sandy Bridge supports Direct Compute. I'm not sure if OpenCL has been added yet. If not it will probably get added in with DX11, hopefully.

Also for anyone interesed in Adobe's stance:
http://blogs.adobe.com/genesisproject/2009/11/technology_sneek_peek_adobe_me.html

It's about a year old but it talks about why they used CUDA and the fact that they are looking at OpenCL.