Actually, that might explain a lot. But I digress.

The point is that Nvidia, unlike Intel, is acutely tuned in to what people are saying about it — and can be quick to respond.

I well recall how, at last year’s Nvision event, one fledgling journalist received a stern dressing down from PR director Derek Perez mere hours after she’d posted an online article that cheekily – but accurately – reported his impression of CEO Jen-Hsun Huang’s keynote address. (For the record, it was “dull and boring.”)

So I wasn’t wholly surprised when this morning, at the conclusion of my meeting with CUDA general manager Sanford Russell, I was ushered into a luxurious suite on the twentieth floor of San Jose’s Fairmont Hotel for an impromptu chat with Tony Tamasi, Nvidia’s Senior VP for content and technology, on the subject of my last blog post.

A billion dollars on CUDA

It turned out that Tamasi wanted to respond to two points. The first was the doubt I had expressed over whether CUDA could ever be a real money-maker for Nvidia.

“Supercomputing,” he assured me, “is a billion-dollar market.”

This I could not deny; but given CUDA’s apparent focus on academia, it seemed a surprisingly ambitious figure for Nvidia to be bandying about.

So I asked: “Have you made a billion dollars from it this year?”

“Of course not,” Tamasi laughed. “But we believe in the potential. We’ve been investing heavily in that for years. And when the market arrives, we’re going to be at the head of it.”

He showed me a slide demonstrating how Nvidia expects its investment to pay off:

— and I could only agree that – if PowerPoint was to be believed – there did appear to be a lot of money out there for the taking.

“So you’re going to sell ten thousand GPUs to the Department of Defence?”

“At least!” he declared confidently.

There’s no telling how much this market will really turn out to be worth to Nvidia. But the company’s investments in research have indeed positioned it well to “trickle up” into industry and government; and with CUDA now programmable in C++, the company’s ambitions are sounding increasingly credible. I wouldn’t bet against the technology growing quickly in these areas – in the short term, at least.

Sizing up Larrabee

But what of the longer game? I was half joking when I suggested that Larrabee might displace CUDA, but Tamasi agreed that it was a possibility.

“Intel is… not very excited when they see a researcher talking about porting code from Intel CPUs to Nvidia GPUs and getting a hundred-fold speed-up,” he predicted.

“And those are the super-high-margin juicy CPUs for Intel.”

“So Intel is defending their computing front. And I agree with you that Larabee is at least partly an effort to try to keep applications from going to the GPU style of parallelism.”

Is Nvidia worried about the long-term challenge?

“Nobody knows how good Larrabee is,” Tamasi mused. “Probably Intel doesn’t know how good Larrabee is. So we take them incredibly seriously.”

“There are strengths and weaknesses to their style of architecture. And I just don’t know how that’s going to play out.”

In my last post I suggested that DirectX 11’s extensive GPGPU support could mark the end of the road for CUDA. And I do expect that mass market GPU applications will quickly move to DirectX rather than restricting themselves to a single architecture.

But the other day I was discussing DX11 with Bit-Tech editor Tim Smalley, and I found him very reluctant to write CUDA off just yet. He pointed out that CUDA retains one big advantage over DX11, in that developers can knock up CUDA routines directly in C – or Fortran or even Matlab – without having to deal with the DirectX API.

Two different markets

Thinking about this, I’ve realised that there are in fact two wholly separate markets for GPU computing. As a mainstream technology, it’s a great way for application developers to wring extra performance from whatever hardware the user happens to already own. In this market, it makes sense to write code that will benefit as many users as possible, which today means DirectX.

But the truth is that there aren’t many desktop tasks that really benefit all that much from GPU acceleration – everyone talks about video transcoding, physics simulations and AI for games, but once you move past those very specific applications it’s slim pickings. For now, the real potential of massively parallel computing appears to lie in its ability to accelerate scientific research.

An academic question

And when scientists and engineers are choosing a research platform, they don’t really care about issues like market share. Their code only needs to run on a handful of machines, and it’s no problem to design those machines to suit the task, rather than vice versa. Here, CUDA is a no-brainer, because it lets researchers program in familiar languages, producing code that can be maintained and expanded without having to learn a new API.

Nvidia realises this, of course, and rather than continuing to talk about games, it’s been carefully positioning CUDA as a friend of academia. As the photo above shows, the hallways at GTC this morning were filled with boards – not whiteboards, like at IDF, but display boards showing summaries of research projects. Some focused on graphical techniques; others targeted problems in biology, physics or engineering. But all of them had something in common…

It’s a bold display. It cleverly makes CUDA look like serious business while ATI is still worrying about copying DVDs onto iPods.

And, what’s more, it’s persuaded me that Mr Smalley does have a point. Clearly, CUDA isn’t going to vanish from these environments overnight. Why would it, when DirectX offers no advantage?

The road ahead

But while academia may be a respectable market, you have to question how much actual revenue Nvidia sees from each of these projects. And though the use of (more or less) industry standard C is working for CUDA right now, it will work against it when a real rival comes along – a rival, perhaps, that can offer even greater programmability, backed up by the kind of muscle that doesn’t need to worry about revenue.

So here’s my crazy prediction. Some form of high-level GPGPU interface will survive alongside DX11 for the foreseeable future. And for the time being that will be CUDA as a matter of default. But I don’t think CUDA will ever be a real money-maker for Nvidia. And within five years I predict that Nvidia’s share of the GPU computing market will be swallowed up… by Larrabee.

(The picture, in case you’re wondering, is a strange engine-type affair that’s been set up at the entrance to the delegates’ hotel, apparently to welcome us as we arrive. I guess that’s how they communicate with one another down here in the Valley.)

Under pressure

Yet despite the company’s outward confidence, you have to wonder whether the green team really enjoys its 72 hours under the spotlight. Last year the recurrent themes of the event were faulty notebook GPUs and Larrabee – hardly the issues Nvidia will have wanted to focus on.

This year the notebook GPU issue seems to have died away, and it’s become clear that Larrabee is no threat for the immediate future. Yet Nvidia is still on the back foot in several fights. In the desktop market, it’s been playing catch-up with ATI for over a year: it’s rumoured that a new generation of GeForce cards may be unveiled in the coming days, but they’ll have to be very impressive to compete with the ATI Radeon HD 5870.

Challenged by new technology

In the mobile and lightweight markets, meanwhile, the company’s Tegra and Ion platforms have attracted praise, but will soon be challenged by Intel’s two new Atom architectures – Moorestown for handhelds and the CE4100 for media devices. Nvidia did well to get to these markets first, but can it really cling onto the territory now Intel’s tanks are rolling into town?

And lastly there’s CUDA. This time last year, Nvidia was touting the flexibility of CUDA as a unique benefit of Nvidia hardware, and since then the technology has shown its potential in some real-world applications – notably the built-in CUDA acceleration for Premiere Pro CS4. But now that DirectX 11 is here, its extensive platform-agnostic GPGPU functions make CUDA’s exclusivity look like a liability.

With all this in mind, I think it will be very interesting to see what the company has to say for itself over the coming days. Stay tuned and I’ll report back with all the news and nuggets that emerge…