Sign in with
Sign up | Sign in
Your question
Closed

Nvidia Announces Carma: Cuda ARM Development Kit

Last response: in News comments
Share
December 20, 2011 9:29:55 PM

Very cool!
It is really interesting to see how ARM does compared to x86 as software becomes more optimized for it. Sure the x86 can do more per core... but for the price you can buy a ton of ARM cores which (if programmed properly) can handle a much larger overall workload. I wonder too if this new threat will make Intel's 'many core' architectures to become a reality instead of just a really interesting lab experiment. However, I do not see ARM doing that well in the laptop/desktop market like they are shooting for. x86 (at this point) just makes more sense for such applications. but who knows, with Win8 for ARM it could turn the whole world upside-down.
Score
4
December 20, 2011 9:57:05 PM

Cuda + ARM = CARMa

I C what U did there! :p 

All jokes aside, it does look interesting.
Score
10
Related resources
Can't find your answer ? Ask !
December 20, 2011 10:13:29 PM

I know what i'll be doing for this summer project :) 
Score
2
December 20, 2011 11:28:10 PM

CUDA can run on everything platform and device,, except AMD hardware... isnt that funny?
Score
-4
December 20, 2011 11:34:07 PM

As long as it can run ALL my programs, I'm all for it.
Score
2
December 20, 2011 11:36:44 PM

atikkurCUDA can run on everything platform and device,, except AMD hardware... isnt that funny?


@ atikkur, it runs on AMD platform too, CUDA is a C++ extension for NV GPU,

sweet, most of my computation is already ported and working on CUDA/NV GPU, now with this i can even use less power on CPU and off load everything on GPU, leave ARMs free
Score
2
December 20, 2011 11:52:06 PM

allenpan@ atikkur, it runs on AMD platform too, CUDA is a C++ extension for NV GPU, sweet, most of my computation is already ported and working on CUDA/NV GPU, now with this i can even use less power on CPU and off load everything on GPU, leave ARMs free


i mean AMD GPU,, not processor. can you run CUDA powered program on AMD GPU?
Score
2
December 21, 2011 12:08:04 AM

atikkuri mean AMD GPU,, not processor. can you run CUDA powered program on AMD GPU?

Nope, not that I'm aware of.
Score
2
a b Î Nvidia
December 21, 2011 1:27:06 AM

atikkuri mean AMD GPU,, not processor. can you run CUDA powered program on AMD GPU?


AFAIK AMD have their own version of CUDA. they called it ATi Stream (now APP).
Score
2
Anonymous
December 21, 2011 1:31:18 AM

Just need a company called CORE now.

So ARM and CORE can face off once again.
Score
1
December 21, 2011 2:15:36 AM

I can't see who would buy a laptop using the current ARM technology. Even the 4 core Tegra 3 is what a 1.3 ghz ARM core..... I mean how would that compete with a quad core 3 ghz X86 ? (another fairly low cost option from AMD).
Score
-3
December 21, 2011 2:34:14 AM

Interesting, though my 580 pushes way over 1581 Giga flops (About 1.6 tera flops)
Score
-2
December 21, 2011 3:15:00 AM

southernsharkI can't see who would buy a laptop using the current ARM technology. Even the 4 core Tegra 3 is what a 1.3 ghz ARM core..... I mean how would that compete with a quad core 3 ghz X86 ? (another fairly low cost option from AMD).

Yes, Im sure the non arm system will be a hell of a lot faster, but how many times is it gonna need to be recharged before the arm based system does?
Score
1
December 21, 2011 8:17:06 AM

This is the advantage of NVIDIA over AMD, they learned from their mistake on Fermi in which AMD overtook them into Directx 11.

Now they are moving aggressively for early releases, specially in the mobile area.
Score
3
December 21, 2011 9:25:50 PM

renz496AFAIK AMD have their own version of CUDA. they called it ATi Stream (now APP).

You can program in OpenCL if you want to cover all your bases. Also note that with OpenCL you can work on a Fusion chip or a Tegra3, or on Desktop/Servers with GPU's.
I personally like CUDA best since Nvidia is much more far ahead in programmability than AMD at this point (Also more scientific Flops, did I mention I'm into science not games ? ). And lets face it, programmability is very important if your GPU is means to an end.
Score
2
!