Refine search for Graphics

This CUDA Be The Start Of Something Great

Author Hugo Jobling
Published 7th Jun 2008
This CUDA Be The Start Of Something Great
Discuss this article  4 comments    Email this to a friend  Email this to a friend TrustedReviews NewslettersTrustedReviews Newsletters

Other forthcoming programs look set to show similar benefits. Adobe Photoshop Next, for example will feature GPU acceleration as (we're almost certain) will Premiere Pro.

There's one other really compelling CUDA-specific development that will pique a lot of interest, PhysX on the GPU - the step we all knew was coming when nVidia bought Ageia.

PhysX being integrated into CUDA is a hugely important step forwards for pushing game developers to use the potential of the PhysX SDK to full effect on the PC. The biggest problem Ageia had was developing an install base of cards able to use its technology. Real numbers aren't available, but even if 70 thousand PhysX cards were sold (and I honestly can't believe the figure is that high), that still means nVidia's CUDA implementation has one thousand times more available users - because every one of the 70-million odd GeForce 8- or 9-series GPU is CUDA capable. And, of course, every forthcoming card will be as well.

This means that game developers know that when they add PhysX effects into their games they can rest assured that a large proportion of the market is able to use them. While we can't talk specifics as yet due to various embargos, rest assured that come Christmas there are going to be PhysX games on the market able to take advantage of GPU acceleration. It's worth bearing in mind that on the console side, the software PhysX SDK already had a huge take-up as it posed a very preferable, free alternative to the only other rival, Havok. Unless said game maker wanted to code a proprietary engine - and only a company like Crytek has the budget for that.


One interesting fact is that CUDA PhysX doesn't require a dedicated card but instead runs in parallel with the graphics rendering. Preliminary figures we've been told suggest that about 10-15 per cent of the cards power would be dedicated to physics. Of course that would also mean losing a proportional amount of graphics grunt, but to those players favouring gameplay over visuals that won't be a terrible tradeoff.

Moreover, while a dedicated physics/PhysX card isn't required that isn't to say one can't be used. Current GPU owners could opt to buy a new high-end card for graphics and leave the older one for physics processing. Alternatively the older car could be sold to contribute partly towards the newer model, but with some of those funds going towards purchasing a cheaper, last-generation graphics card for physics. And, importantly for the graphics enthusiast, there's no performance deficit this way.

Considering that CUDA will even run on a £30-odd GeForce 8400, which should in turn offer more than enough power to run PhysX, it seems like a pretty compelling buy. We'll have to wait for the games to emerge first before a definitive decision can be made, but the future's bright.

The examples I've mentioned above are just a small sample of the development base CUDA has and there is no doubt in my mind that the situation is only going to improve. The more consumers are aware of the advantages the GPU can offer the faster this progression is likely to happen. There's no telling where the future could lead us - so kudos to CUDA and nVidia for helping, if only in small ways, to make out productive lives that bit simpler and more enjoyable.

 

Newsletters

Register to receive the latest Reviews and News Headlines directly to your Inbox every day, and enter our regular competitions. More Info.

Your Name


Email Address


Latest 4 of 4 Comments

Have your say: Leave a comment below about this article.

Comment yungchin said on 9th June 2008

Indeed AMD is lagging behind quite a bit in this respect. But they do offer something similar (in complexity) to CUDA now through BrookGPU (which is like C, unlike CTM which was mo... more

Comment Hugo said on 9th June 2008

The difference, as I said, is that CUDA works on any G80 or newer nVidia chipset. It isn't limited to a specific type of graphics card, unlike all AMD's solutions to date... more

Comment yungchin said on 10th June 2008

Isn't "R600 or newer" just about the same generation of cards as "G80 or newer"?

Comment Hugo said on 10th June 2008

Exactly the same in fact! Oh also, don't speak to soon about double precision - there's always a chance nVidia could add it with the next gen cards...

See all 4 comments on this article.

Add Comment Add your comment

You must be logged in to comment. Login or register here.