Gaming from the cloud: If Nvidia succeeds, I'll be suitably wowed

One of the myriad perks of a gig like mine is that I not only get regularly exposed to cool new products (and their underlying technologies), but also to the business strategies that drove development of those products and technologies, and to the personalities behind those strategies. Nvidia has been a particularly enjoyable company to watch over the years. One of the few survivors of the dozens (literally) of graphics processor companies that existed in the early era of my prior full-time career at EDN, the company has waged epic battles with ATI Technologies and others over the years. Plus, company co-founder, President and Chief Executive Officer Jen-Hsun Huang is extraordinarily entertaining.

For a while now, it's been clearly evident to industry onlookers such as myself that the era of standalone graphics ICs was sooner-or-later drawing to a close, save for in niche applications (gaming rigs, high-end workstations, etc). I distinctly remember when Intel unveiled the Real3D-partnered i740 graphics processor (more years ago than I'd care to admit at the moment). A lot of folks mocked its comparative capabilities versus the offerings of 3Dlabs, ATI, Matrox, Nvidia and others. I wasn't one of them, and I'm not attempting to pull off revisionist history in saying that. It was obvious to me from the start that Intel's long-term plan was to roll the GPU core first into the core logic chipset and (eventually) directly into the CPU. That's exactly how it's played out. And today, Intel's the largest graphics supplier to the PC platform (and maybe the largest graphics supplier, period; I haven't compared Intel's shipments to those of Imagination Technologies lately). Similarly, AMD bought ATI Technologies a few years back, and the Trinity "APU" (i.e. CPU+GPU), a notable improvement on the Llano precursor, is the latest output of that wise (albeit expensive) acquisition.

Where does this leave Nvidia? The company has attempted to broaden its graphics products' usefulness beyond pure polygon and pixel processing via various GPGPU attempts, but that aspiration has had limited-at-best success beyond a few high-profit but low-volume supercomputer design wins. For a while, Nvidia had reasonable success selling nForce core logic chipsets with integrated graphics capabilities (a vision first showcased in the first-generation Xbox), but AMD and Intel's chipset alternatives capped the potential market size for Nvidia's offerings. And the inability to secure a CPU front-side bus license from Intel (a failure hampered in no small part by Huang's seeming inability to resist publicly jabbing Intel at every possible opportunity) put the kibosh on the product line's evolutionary potential. Wisely, Nvidia bought PortalPlayer in early 2007, enabling it to diversify beyond graphics and develop ARM-based SOCs. The company's Tegra 2 and more recent Tegra 3 product lines have achieved modest success, although their most visible design wins, in Android-based tablets, have been substantially volume-overshadowed by Apple's iPad series. And in handsets, the company has formidable competitors, such as Apple, Qualcomm, Samsung, and TI.

The legacy PC graphics business is evaporating, and the ARM-based SOC business isn't (yet, at least) sufficiently robust to fiscally compensate. So what does Huang do to bridge the gap? He talks up cloud-based gaming, as both Maximum PC and VentureBeat noticed last week. And I scratch my head. For one thing, wireless broadband isn't yet sufficiently pervasive (nor inexpensive enough, for that matter) to deliver the reality behind the vision; Google's Chrome OS has the same fundamental limitation. So-called 'casual' games like the Angry Birds franchise (or any of the numerous time-wasters currently available on Facebook) might be able to pull off the concept, but they're adequately served by today's server hardware. So Huang is up-scaling the vision, touting graphics-rich titles, many of them of the fast-twitch variety...for which network coverage pervasiveness is the least of the concerns. Bandwidth and latency rule the roost here, and neither is sufficiently available even with today's (expensive) LTE services. In order to (partially) compensate, you'd need to load a notable percentage of the content directly on the network client. And in that case, why not just go all the way (as is largely done today), create a locally installable program, and avoid any offline- or deficient-network worries?

Don't get me wrong...I get the whole "cloud computing" concept: shifting cost and complexity away from the client to the server, simplifying software maintenance, preserving user data in spite of client hardware failure...yadda yadda yadda. And I'm amazed on a daily basis at the connectivity-enabled examples of its success, nearly 20 years after Larry Ellison first talked up the "network computer"...albeit today predominantly implemented on LAN-tethered and wired broadband-fed computers. But games? Specifically, fast-twitch and graphics-rich games? On cellular data-fed hardware? For Nvidia's sake, I hope I'm missing something. But until then, as the saying goes, I want some of what he's smoking.