Feeds

back to article Nvidia stretches CUDA coding to ARM chips

Nvidia's new CUDA 5.5 release aims to overcome the inherent mathematical suckiness of the ARM architecture by unleashing the powers of GPUs working in conjunction with those popular low-power chips – and not just ones from Nvidia. Nvidia is working on its own implementation of a 64-bit ARM processor, code-named Project Denver, …

COMMENTS

This topic is closed for new posts.
Silver badge
Paris Hilton

That market share graph is hard to understand

It is not Larry Ellison's market value expressed in South Sea Islands?

NO! It is the relative share of ARM Open vs. x86....

But in what market? The problem is that market exhibited 100% x86 way back when.

Embedded? Sure not. Supercomputers? No. PCs? I don't see many ARMs here. Mobile Market? No..

1
0

Re: That market share graph is hard to understand

Back in the 90's Apple were running PowerPC chips as well, sure the share wasn't big, but it wasn't zero.

0
0
Silver badge
Devil

And I have to say I foresee really impressive screensavers in the future.

0
0
h3
Bronze badge

Nvidia use pretty nasty tricks to try and create lock in they should just make the best gpu's and use opencl.

5
0
Silver badge
Holmes

Sure, OpenCL would be nice, but OpenACC seems to be rather open-ish. It is also very much higher-level (compiler does boring jobs for you) than OpenCL/CUDA (manual work required).

Looks like there is no support for gcc or clang.

Compiler_support

"Support of OpenACC is available in commercial compilers from PGI (from version 12.6), Cray, and CAPS. An open source compiler, accULL, is also developed by the University of La Laguna (C language only)."

There is work to do then.

I remember that special C for the Connection Machine...

1
0
Silver badge

For the HCC crowd it doesn't really matter as they buy, build or have built whatever hardware best suits their requirements at the time. Both CUDA and OpenCL have increased their options in this.

Academics with the necessary skills will no doubt continue to push for architecture neutral systems. I think nVidia understands this which is why they will be pushing chip + compiler, using the added value to promote their products in a commodified market which will at some point no doubt include clusters of AMD and nVidia based chips tuned for different tasks.

0
0
Bronze badge

Why don't they...

...put ALL of the stuff on a single chip and eliminate this co-processor stuff. Just have the CPU-GPU combo chip and hang some memory, and it generates nice video (HDMI/Display port) output. Add an USB port (or two) and you have about a 3-5 chip solution. It would be even better if the BIOS were contained in flash inside the combo chip as well.

Get with it, NVidia!!

p.s. Don't make the BIOS do that stupid UFEI signed stuff, I want to run a nice operating system!

0
2
Silver badge
Thumb Down

Re: Why don't they...

Why? because it's inefficient. Many use cases don't need FLOPs which is one of the reasons why ARM chips are so small and popular.

nVidia isn't building systems so it doesn't do any BIOS. Not sure what you mean by a "nice operating system". Choice is all that matters.

4
0
Silver badge

Many ARM chips

There are many more ARM chips sold even if you don't count smartphones etc. and just count PCs.

The typical PC has one Intel-based CPU chip, but the typical hard drive has two or three ARM cores in it, then there are Wifi modules, bluetooth modules, CD drives, etc etc...

Intel Inside... or should that really be ARM Inside?

3
0
Anonymous Coward

Re: Many ARM chips

Many ARMs make light work.

[Ducks and runs for the door]

0
0
Joke

nothing open about openacc

openacc doesn't support amd gpu or intel phi. how is that open? openmp is a true open standard that supports nvidia gpu, amd gpu, intel phi. timothy is clueless.

0
1
This topic is closed for new posts.