OTOY Forums

Hi everyone !

I just came across an article about the next generation of nvidia graphic cards. The article can be found here : http://www.theinquirer.net/inquirer/new ... system-ram. It is from March I know, but according to a quick googling it still seems to be on its way. It also seems that bandwidth is quite high and with the upcoming DDR4.

So, as I understand it, it would mean that in about 6 months, cuda softwares will be able to address as much memory as needed from the system which could then finally be it, the end of a long fight with vram amount limitations. I understand that accessing system memory might slow things down a little, but would at least allow users to render scenes that were before not even imaginable.

If anyone got thoughts on that !

Have a good day !
Arnaud

a.boeglin wrote: So, as I understand it, it would mean that in about 6 months, cuda softwares will be able to address as much memory as needed from the system which could then finally be it, the end of a long fight with vram amount limitations.

I think the new MacPro can do that? Unfortunately, it's going to use AMD gpu's.

justavisitor wrote:

a.boeglin wrote: So, as I understand it, it would mean that in about 6 months, cuda softwares will be able to address as much memory as needed from the system which could then finally be it, the end of a long fight with vram amount limitations.

I think the new MacPro can do that? Unfortunately, it's going to use AMD gpu's.

bullshit.
the new mac has only Ati yet.
And u need Cuda!

And the new Maxwell Card will come 2014....how is the new Mac that have already???

C'mon, no need to be so unfriendly.

The Mac Pro has ATI in it's standard configuration, no one knows what the options are for a Built-To-Order MP.
It might as well be a SLI Titan configuration, right?

Let's hope NVIDIA will keep developing the CUDA technology

sorry....

My original point had actually nothing to do with mac pro guys .. I know ATI got a similar technology which will, according to my readings, be part of the playstation 4 coming out soon. I only refer to the next architecture of Nvidia cards here, which states that graphic cards to be released in 2014 ( Maxwell architecture, so presumably, 800 series and / or new titan, quadro, tesla ) will be able to address system memory. This should come at about the same time DDR4 will be available.

The question was rather aimed at people either developing octane or having a good understanding of Cuda / GPU devices. I just thought this might finally be an ending to the VRAM limitations we suffer from and might unleash new possibilities.

just for information :

CUDA 6 unifie l'architecture mémoire
http://www.tomshardware.fr/articles/nvi ... 46219.html
http://www.anandtech.com/show/7515/nvid ... y-for-cuda

good things are on the way.

Dom74 wrote:just for information :

CUDA 6 unifie l'architecture mémoire
http://www.tomshardware.fr/articles/nvi ... 46219.html
http://www.anandtech.com/show/7515/nvid ... y-for-cuda

good things are on the way.

Thanks for the links, CUDA 6 and Maxwell look really interesting...

one more link :
NVIDIA Dramatically Simplifies Parallel Programming With CUDA 6

http://www.techpowerup.com/194505/nvidi ... uda-6.html