Sun Cancels UltraSPARC V, Gemini, But Not Future Processor Development

The headlines about Sun Microsystems killing its UltraSPARC V and Gemini processor development may have given somes VARs and customers, already on the edge with all the other bad news coming from the vendor, the impression that Sun had no processor roadmap going forward.

Sun, however, remains strongly committed to future processor development, says David Yen, executive vice president of Sun's Processor Group. He said any confusion stems from a single oversight on the part of the vendor. "We didn't do a good job with a marketing label," he said.

Yen confirmed the company did cancel development on a future processor, code-named "Millennium," based on what he called "conventional architectural wisdom," including the ability to issue six instructions per cycle and perform out-of-order execution, in order to shift resources towards the development of two new processor platforms, code-named Niagara and Rock.

However, unlike canceled projects at other vendors, including chip designers, Sun made the mistake of calling Millennium "UltraSPARC V" early on, thereby causing confusion among people who thought canceling the project signaled an end to the UltraSPARC platform.

The beginnings of the Millennium project go back several years, and while Sun publicly started talking about UltraSPARC V a year-and-a-half ago, the company started realizing over two years ago that there were latency and power issues that needed to be overcome, Yen said. So the company decided to discontinue Millennium and shift resources to new throughput computing processor projects.

Sun's Throughput Computing strategy is aimed at using its UltraSPARC processors to increase real-world application performance while cutting the cost and complexity of network computing.

Sun also decided to cancel its Gemini project several months ago, said Yen. Like UltraSPARC IV, which is based on two UltraSPARC III cores, Gemini was envisioned to be built on two UltraSPARC II cores. It was already developed to the point at which units were working in sample systems, said Yen.

"It was as fast as UltraSPARC IIIi in single-thread operations, and had a bit faster overall performance," he said. "But Sun already has two UltraSPARC product lines, and is working with AMD on entry-level processors. So we decided that for the slight delta in performance, Gemini was too confusing."

Furthermore, Sun will quickly move to 90-nanometer technology for its processors from the current 120-nanometer for both UltraSPARC lines, at which time the UltraSPARC IIIi performance will get another boost, Yen said.

Going forward, Sun is working on two other projects for its future processors.

Niagara, expected to be available in servers aimed at network handling and Web tier services by 2006, is based in part on technology Sun got with the acquisition of Afara Web Systems over a year ago, said Yen.

Afara is a start-up founded by the original architect of Sun's UltraSPARC 1 processor. Niagara will leverage much of the design work already done by Afara, Yen said.

Following the introduction of Niagara will be Rock, which will be aimed at midrange and high-end servers, said Yen. Rock will also leverage Afara technology, along with Sun's own designs, including the MAJC graphic processor, he said.

Both Niagara and Rock will be compliant with SPARC V9 standards, an open standard published by Sun against which Sun, Afara, and Fujitsu build compatible processors, Yen said. Compliance with SPARC V9 standards ensures that the processors will be binary compatible with previous UltraSPARC-based software.

Rob Wolfe, president and CEO of AvcomEast, a Silver Spring, Md.-based Sun solution provider, said he is so focused on what Sun business he can close in the next six months to 18 months that cancellation of forward-looking projects is no big deal.

Wolfe understands where customers might be confused. He has already been hearing about how other vendors are using the UltraSPARC cancellation to spread concern among Sun partners and customers. "It's more FUD (fear, uncertainty and doubt) for the other guys," he said.

However, Wolfe said, the news is actually positive for Sun.

"Sun is paring down some of their R&D projects, picking their battles," he said. "Are they canceling two chip projects? Or focusing on two chip projects?"