As if OEMs don't have enough of a headache picking from a jumble of competing DRAM products, controversy swirling around a new DDR400 chip being sampled by the top two memory component suppliers threatens to further compound their problem.
It's not helping that DRAM makers are sharply divided over the direction of the higher-speed 400MHz memory chip, which was demonstrated last week in Santa Clara, Calif., at the first Jedex conference of JEDEC, the semiconductor industry's standards-setting organization.
Conference attendees were split down the middle over two versions of the 400MHz memory chip. Jumping ahead of the pack, Samsung Electronics Co. Ltd. and Micron Technology Inc., the first- and second-largest memory suppliers, respectively, are touting new DDR400 versions that both have started sampling, using the existing DDR-I industry standard.
The introduction of DDR-I chips by Micron and Samsung "makes no sense," said executives at other DRAM companies. Elpida, Hynix Semiconductor, and Infineon Technologies said they would prefer to wait for the next-generation DDR-II standard for their 400MHz double-data-rate chips. Trying to achieve 400MHz with all its high-speed challenges with the existing DDR-I standard could prove difficult, said Joo Sun Choi, director of application marketing at Hynix Semiconductor America, San Jose.
And even if chip companies can make 400MHz DDR-I devices, it isn't clear that module vendors or motherboard suppliers would support such a memory, said Katsuyuki Sato, deputy general manager of the technical marketing division at Elpida Memory (USA) Inc., Santa Clara, Calif.
The harsh words mask a fundamental challenge to OEM procurement executives as the memory market continues to exhibit product frag- mentation. DRAM customers in the coming year will face a potpourri of memory devices, compounded by the DDR400 split, according to industry executives and analysts.
"This is a natural corollary of customer diversification in the market," said Sherry Garber, an analyst at Semico Research Corp., Phoenix. "Server and communications and networking customers are growing in size and importance and want memory more specifically tailored to their different needs."
The DDR400 controversy exemplifies this split, Garber said. DDR-I 400MHz chips will find a market with a few high-performance server customers that can pay premium prices and qualify the non-JEDEC standard devices, she said.
Finding its niche
However, the option provided by Micron and Samsung may benefit OEMs in specialized fields. Unlike PC OEMs, these companies make their own chipsets and are not dependent on third-party memory controller vendors. In the past, PC OEMs dictated what memory types would be sold, but a new class of OEMs with their own peculiar memory requirements have come of age, creating markets for non-standardized products, Garber said. It is this group of OEMs that Samsung and Micron are targeting, she said.
"Samsung and Micron will push DDR400 in a DDR-I standard because it will have much higher profit margins," Garber said. By contrast, "PC OEMs need standard-validated DDR400 modules and chipsets, so they'll likely wait on DDR400 DDR-II devices."
Samsung and Micron shouldn't expect any compromise from JEDEC, according to Bill Gervasi, director of technology analysis at Transmeta Corp., Santa Clara. Gervasi, who is also chairman of the JEDEC memory parametrics committee, told the Jedex conference that the standards body likely will not have the time or resources to amend its DDR-I spec for chips and modules to handle new Micron and Samsung DDR400 devices.
In any case, "the [DDR-I] DDR400 will be limited only to a few boutique applications for customers willing to pay for a specialized memory chip," Gervasi told EBN at the conference.
Not so, said Samsung and Micron, which unveiled DDR400 (DDR-I) chips within days of each other earlier this month. "Naysayers said PC133 memory would only go into boutique applications when those devices were first introduced," said a spokesman for Samsung's semiconductor subsidiary in San Jose.
Waiting for DDR-II
DDR-II chips probably won't draw the same heated exchanges. The chips should be attractive because of their potential to lower costs by reducing parts in motherboards and systems. Also, all sides to the controversy agree that DDR-II will offer a lower 1.8V capability in a new chip-scale package with on-die termination and programmable impedance control to handle the much faster memory speeds.
Executives at the Jedex session said samples of DDR-II will start hitting the market late in 2002 or early 2003, with volume production starting in 2004. The new devices will have on-chip termination (OCT), which boosts signal integrity by reducing impedance and noise that increasingly play havoc with higher-speed DRAMs.
Jeff Janzen, senior DRAM applications engineer at Micron, said OCT can also reduce parts on motherboards by eliminating series resistors and voltage regulators and filters for a savings per board of $3 to $8.
But even as Jedex concentrates on DDR-II, the first hint was given of DDR-III architecture slated for development in 2004. DDR-III will continue to lower memory voltage levels, initially to 1.5V and then to 1.2V, by 2006, according to Elpida's Sato.