nVidia (NASDAQ: NVDA) is dismissing reports that it is looking to acquire Taiwanese chipmaker VIA Technologies — despite persistent rumors about changes in strategy ahead for the company.
The most recent reports in Taiwan’s DigiTimes claimed the companies discussed three possible scenarios: a strategic alliance, a complete acquisition or the purchase of just VIA’s processor division.
According to the publication, which cited sources at motherboard vendors, the talks fell apart because VIA wanted too much money.
It’s unclear what sort of benefits such a deal might offer to nVidia, best known as a manufacturer of graphics processing units (GPUs). VIA develops a variety of semiconductors, from CPUs to motherboard chipsets to memory. It acquired two x86 clone developers, Cyrix and Centaur Technology, the latter of which still operates independently.
Rather than try to compete directly with Intel (NASDAQ: INTC) and AMD (NYSE: AMD), VIA has focused on very low-power x86 chips for the handheld and embedded markets. Its most recent is the ultra-low power “Isaiah” processor.
VIA was at one time a major player in the PC motherboard chipset market but made some blunders along the way, eventually losing out to Intel, AMD and nVidia. Now the chipset market is slowly disappearing: Chipsets consisted of external controllers and graphics, but slowly that’s all being moved onto the CPU.
No one at VIA’s Fremont, Calif., office was available to comment on the rumors. Mike Hara, vice president of investor relations at nVidia, dismissed the rumor out of hand.
“Why would we need them?” he asked. “We’re a GPU company, and we don’t see any need soon to burden it with a CPU because [a CPU] doesn’t do anything we need.”
Yet with AMD’s acquisition of graphics vendor ATI — and its subsequent promise of combined CPU/GPU efforts, coupled with Intel’s persistent dominance share of the motherboard-integrated graphics market — industry-watchers have long wondered whether nVidia will change its strategy. (Perennial rumors of an Intel-nVidia tie-up only add to the rumor mill.)
If nVidia wants to move into CPUs, allying itself with VIA may not be the ideal way for nVidia to proceed, observers said.
“Is it worth [nVidia] trying to go after general-purpose processors? Only on the low power side, but nVidia already got the APX2500 there,” said Jim McGregor, research director for semiconductors at In-Stat. nVidia recently announced the APX2500 processor for ultraportable and handheld devices, which will likely go up against Intel’s Silverthorne processor.
“It’s not like the Centaur group is going after the PC market — they are not,” McGregor told InternetNews.com. “You’re not going to buy the PC group from VIA and go after Intel, especially in the general-purpose market with processors like Nehalem coming out.”
Another reason VIA seems unlikely to interest nVidia is that VIA plays in the low-end of the performance spectrum — while nVidia has staked a claim on the bleeding edge.
“We want to build things that add value for the experience of the customer,” Hara said. “One metric we do not use is cheaper. Price is not our game.”
But Nathan Brookwood, research fellow with Insight 64, thinks VIA might bring some value for nVidia.
“The chipset isn’t worth it,” Brookwood said. “nVidia is already in the chipset business, and that market is going away as chipsets are being integrated into processors.”
“But if Intel and AMD are right — that the future is a combination of an x86 core with high-performance graphics on the same piece of silicon — then VIA is one of the last independent players with an x86 core,” he added. “They have a development team that’s small, but it’s very mature [and] has been able to put together power-efficient cores for over a decade now.”
For instance, Centaur’s president and chief engineer, Glenn Henry, is one of the most respected x86 engineers outside of Intel or AMD. Consequently, he and his team might make for a good pickup for any semiconductor maker.
However, McGregor thinks nVidia can get along just fine with its CUDA compiler, which lets developers write applications using a C++ syntax and the nVidia GPU like a standard CPU.
Hara, meanwhile, dismisses it all: “Our strategy ought to be stick to what we know best.”
