eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
Suns top microprocessor executive said Tuesday that the UltraSPARC V may come to market after all, after the company “focused” its roadmap at the end of last week.
David Yen, executive vice president of processor and network products for Sun Microsystems Inc., said the company also may explore closer ties to Fujitsu Ltd., a licensee of its SPARC architecture.
At the end of last week, Sun of Mountain View, Calif., said it would lay off 3,300 employees and cancel the UltraSPARC V and “Gemini” processor, a chip that contained two cores and was designed for servers.
Rivals seized on Suns reworked roadmap as an indication that the systems vendor had run out of steam and that the UltraSPARC line had nowhere left to go. Companies such as Intel Corp. publish general roadmaps of their processor direction, which allow customers to plot their own expansion paths.
But Yen said this week that the companys roadmap remains “focused” and that the company remains absolutely committed to the UltraSPARC line.
“With the UltraSPARC V announcement, there was a misleading perception we are killing future generations of SPARC,” Yen said. “Instead, we are not canceling future generations of SPARC, we are adjusting direction to be better served with a new breed of processors and corresponding platforms.”
Sun may not even kill off the UltraSPARC V brand, simply applying the label to another core. “There is an internal debate [whether] to skip the number,” Yen acknowledged.
Yen confirmed that the design teams responsible for the UltraSPARC V (aka “Millennium”) and Gemini programs had been part of the layoffs but was prevented from commenting further by a Sun spokeswoman, who said the company is in its quiet period. More details will be revealed Thursday when Sun reports its quarterly earnings to Wall Street analysts, she said.
Sun recently introduced its UltraSPARC IV chip, and the company plans a move to 90-nm geometries, dubbed UltraSPARC IV+, next year.
From there, Sun will use expertise applied from its July 2002 acquisition of Afara Websystems Inc. to its “Niagara” line in 2006, which will increase application performance with a technology called “chip-level multithreading.” Subsequently, its on to “Rock”, Yen said, which combines the multithreading technology with virtual cores that may or may not be dedicated to application-specific functions.
“If you look at this from our perspective, Suns processor roadmap will be a very clean one and in our opinion a very effective one,” Yen said.
The “incremental growth” that the UltraSPARC V offered will instead be replaced by a generational leap with the Niagara, according to Yen. “Incremental growth is not necessary,” he said. “At worst, youre losing time, leaving things on the table.”
Analysts observed that the UltraSPARC V was expected to be somewhat of a stopgap between the enhanced UltraSPARC IV and Niagara. According to Yen, some of the “resources” involved in both the UltraSPARC V and Gemini have been shifted to Gemini and Rock, a move that will allow the company to potentially release both chips earlier than expected.
“Thats the right approach,” said Nathan Brookwood, an analyst with Insight64 in Saratoga, Calif. “Why continue to invest in UltraSPARC V, which will have a relatively short lifetime, when you can rapidly invest in the Rock stuff? The only danger I see there is if they hit a development snag with Rock, a la the problem they had with the UltraSPARC III.”
Rivals, naturally, said there were too many unanswered questions in Suns roadmap. “Whats been communicated so far has been the factual aspect, that theyre going to stop moving forward on UltraSPARC V and beyond,” said Mark Hudson, vice president of marketing for business-critical systems within the global business unit of Hewlett-Packard Co. of Palo Alto, Calif.
He suggested that the performance of the UltraSPARC IV wont keep up with rival architectures. “What I havent seen yet is whats the implications of that, first and foremost from a customer perspective.”
But thats not the way one customer saw it. Michael Hodges, manager of systems services at the University of Hawaii, said he applauds the moves that Sun is making, both in refocusing its SPARC development efforts and in entering the lower-cost x86 space with Advanced Micro Devices Inc.s Opteron processor.
“Sun is most definitely retrenching, and that is a good thing,” Hodges said in Honolulu. “I see many similarities between Sun and that 80s stellar company, [Digital Equipment Corp.], except that Sun has more options than DEC given its middleware stack, [Java Enterprise System]. Sun is also changing the focus from a fixation on raw performance to manageability and utilization of resources.”
The university runs a large server farm that is managed by a small staff, requiring a lot of repetitive work to manage the servers. “We are not getting every ounce of potential CPU cycles out of the servers,” he said. “Suns prioritization of these issues will better help us.”
The school is rethinking its IT platform in terms of tiers, with its mission-critical Student Information System running on an UltraSparc III/Solaris platform. Later, the university will roll out its lighter, Web-enabled applications on x86 systems running Linux “so that we can aggressively contain costs,” Hodges said. “Here, too, Sun is playing into our needs with their diversification.”
Next Page: Sun, Fujitsu rejigger SPARC relationship?
Sun, Fujitsu Rejigger SPARC
Relationship?”>
Meanwhile, Sun is continuing to examine production and technology options for its UltraSPARC IV processor line.
One possibility is that Sun may work more closely with Fujitsu Ltd.s Microelectronics business, which also designs its own 64-bit SPARC processors. Suns own UltraSPARC IV, for example, runs at speeds of 1.05GHz and 1.2GHz; faster speeds are expected next year with the UltraSPARC IV+, due when Sun shifts to its foundry partner Texas Instruments Inc.s 90-nm processor.
Quinn Jacobsen, the chief architect of the UltraSPARC IV, said in October that future enhancements to the USIV line would include dual cores, larger Level-2 and new Level-3 caches, and a faster front-side bus interface.
For his part, Yen this week said the company “is not ready to talk about” a so-called UltraSPARC IV++ chip.
Yen said Sun currently enjoys a “very good relationship” with Texas Instruments. But he left himself room to explore new options for foundry partnerships.
“As Ive said, we have a very good relationship with TI, more than just a buyer and seller with a foundry partner. On the other hand, both companies have their own goals, and as a fabless design house were constantly monitoring the industry.”
Fujitsu, meanwhile, plans to ship a 1.5GHz version of its SPARC64 V chip inside of its PrimePower server line later this year, and has plans to shift to a 90nm, 2.0GHz or greater performing chip by the end of 2004. The SPARC64 VI is due late in 2005, boasting a pair of processor cores and 6MB of on-chip cache.
“Anything is possible,” Yen said. “Fujitsu and (Sun) have been SPARC partners for more than 10 years. There are regular discussions on regular levels. On a technical level, we have a very close relationship. If any opportunities arise where we can offer joint customers a better product line or better customer service, we will pursue them.”
But Sun and Fujitsu currently use incompatible packaging and bus schemes, so a replacement or an addition to the high-end UltraSPARC line with SPARC64 chips also would require a corresponding change in chipsets and motherboards, what HPs Hudson called a “box swap.”
In addition, Sun has adopted the AMD Opteron processor for many of its volume systems, Brookwood pointed out.
Yen said there was little chance that Sun would open up its SPARC chips in much the same way IBM did with its POWER architecture. “Its more or less a bottoms-up approach,” Yen said of IBMs strategy. “This type of approach tends to be incremental. I dont know how far it can go.”
“Sun is a smaller company than IBM,” Yen said. “We really dont have to seek additional growth; theres so much we have to strengthen before we seek additional business opportunities. Its back to the basics; thats the important thing.”
Additional reporting by Jeff Burt of eWEEK.