Sun Delays "Rock" Sparc Machines Until 2H 2009

Published: February 14, 2008

by Timothy Prickett Morgan

If there are not product delays, then it can't be the processor business. Every chip designer spends some time in the dog house with its sales reps, reseller partners, and customers for making promises that for one reason or another do not get delivered upon or that do not get into the field in the expected time frame. The list of delayed and ill-fated processors is too long to go over, but you can add Sun Microsystems' "Rock" UltraSparc RK processors to the list.

Processor designs have become exceedingly complex and the shrinking of transistors that is required to keep adding features and boosting performance of processors has become increasingly difficult. It may be sacrilege to say so, but in 2008, Moore's Law should really be renamed to Moore's Estimation, since processor power is not doubling every 12 to 18 months, but rather core counts are doubling every 24 months or so. Processor performance is staying more or less where it was because clock speeds for chips that do out-of-order execution have hit a wall at around 3 GHz and those that use a simpler pipeline, like IBM's Power6, are stuck at around 5 GHz or maybe 6 GHz if IBM transitions to new processes. I am not even sure we should call them processors any more, since all the extra stuff like caches, memory controllers, co-processors for encryption and math, and other features that used to be implemented separately in a chip package or, further back in time, on a motherboard have all been crammed onto the processor. These things really are becoming systems on a chip, which is great for simplifying server designs.

It is also a disaster when the processor design hits a snag, as we all know. Remember the several generations of Intel's Itanium chips, starting with the original "Merced" designs? And how many times did Intel change or re-use the code names on subsequent Itaniums? Sun Microsystems' UltraSparc-III processors were 18 to 24 months late coming into full ramp, and that delay and a bad economy pushed out the UltraSparc-IV chips and killed off the "Millennium" UltraSparc-V processors. IBM's PowerPC 630 never made it out of the chip fab alive, and IBM had delays with the Power3, Power5+, and Power6 processors, too. Fujitsu-Siemens was about a year late last year getting the dual-core Sparc64-VI chips to market. Hewlett-Packard and the formerly independent Digital Equipment had their share of delays with their PA-RISC and Alpha processors, too.

Sun taped out the UltraSparc Rock processor, a 16-core chip with plenty of innovations packed into it, in January 2007. By May last year, Sun had booted Solaris on the chip, only six weeks after getting back first silicon on the Rock chips from its fab partner, Texas Instruments. At the time, Sun remained committed to its scheduled delivery for machines based on the Rock processors, originally being vague and saying some time in 2008 (which the company said in 2006 and early 2007) and then firming it up a little to the second half of 2008 as 2007 began to wind down. Something has clearly gone wrong--or perhaps many things, given the fact that the Rock chips deploy a number of new technologies, including scout threads and transactional memory--because now Sun is telling partners and customers not to expect Rock machines until sometime in the second half of 2009.

That is a pretty disappointing delay, particularly with IBM ramping up Power6 machines and Intel possibly getting four-core "Tukwila" Itanium processors in the field in late 2008--maybe.

Here's Sun's official statement on the matter:

"Sun's current Sun Fire systems portfolio is robust and its UltraSparc T2-based and SPARC Enterprise M-series-based systems are delivering the performance, scalability, extensive virtualization capabilities and Solaris compatibility required in the enterprise. In Q4CY07, Sun experienced continued growth in Sun's UltraSparc server business with particular strength in Sun's OPL systems. Sun continues to evolve its systems portfolio on multiple fronts and early Rock chips are testing well in Sun Labs. Rock is an entirely new design and given its uniqueness and complexity, Sun is investing heavily to fully validate the chip and to do advanced testing on the entire hardware and software stack of future systems to be powered by Rock processors. Systems based on the Rock processor are expected to be available in the second half of calendar year 2009."

I am working to get some time with Sun's top brass in the Systems division and Microelectronics division to understand the nature of the delay and what Sun and its server partner, Fujitsu-Siemens, plan to do. I would guess that the future quad-core Sparc64-VII is getting a look over by Sun right now. Of course, these quad-core Sparc64-VII chips were originally slated for delivery in mid-2006 . . . . and heaven only knows when the fabs at Fujitsu will really deliver them.

RELATED STORIES

Rock and Tukwila Are the Stars of ISSCC This Week

Sun Boots Solaris 10 on "Rock" Sparc Processors

Schwartz Blogs a Bit About the Dud Rock Chip on His Desk

Yen Explains Sun's Chip Strategy

Sun Boosts Performance of UltraSparc-IV+ Chips

Chip Makers Strut Their Stuff at ISSCC

Sun Details Server Chip Roadmaps at Analyst Summit

Sun Tapes Out Rock Sparc Chip, Gooses Clocks on Niagara Sparc T1

The X Factor: High-End Chips Draw Even, Vendors Prepare to Differentiate

Sun Merges Server Units, Taps Key Exec for Storage

Sun Firms Up Its Sparc Chip Plans

Sun Allies with Fujitsu for Future Joint Sparc Platform

Sun Pulls the Plug on Future Millennium, Gemini Chips

                     Post this story to del.icio.us
               Post this story to Digg
    Post this story to Slashdot