It looks like Intel Corp. has gotten itself into a bit of a mess with its rollout of new microprocessors -- its core business.
Clearly the MPU kingpin -- and the PC community that is joined at the hip with Intel -- needs a jumpstart for the Pentium 4, which has had a sputtering start since its introduction last November.
Intel takes the position that every one of its new processor architectures has taken a little time to ramp up to take control of the market -- and Pentium 4 is no different. The company contends that the recent steep P4 price cuts, combined with a Taiwan vendor ramp-up of Pentium 4 motherboards, will make that chip the dominant PC processor.
But Intel's own shifting roadmap may be hampering the P4 takeoff.
The corporate market, a huge Intel customer base, isn't going to rush into Pentium 4 now, when many chip changes are swirling on the horizon. Business buyers want platform stability. With many known Pentium 4 revisions in the works, there is a lot of corporate IT reluctance to commit until Intel finally gets its act together, according to many analysts.
Memory.Because of its Direct Rambus legacy, Intel's current Willamette Pentium 4 can only use more costly RDRAM memory. Intel promises an 845 series Brookdale SDRAM chipset for P4 this quarter. But the 845 initially will only support single data rate PC133 SDRAMs, while DDR capability on the same chip is disabled.
People who have gotten their hands on early P4 motherboards from Taiwan with the 845 Brookdale generally are finding out the obvious: the restrained 1-gigbyte/second memory data transfer is a sad match with the 3.2-Gbyte/second rate of the processor.
Intel reiterates its oft-stated position that many potential P4 customers don't need great memory bandwidth and single data rate SDRAM Pentium 4s will be just fine for them. Time will tell if this rationale works against rival DDR-supported MPUs in the market, or even against Intel's own soon-to-be-enhanced Pentium III processors.
Corporate and power users -- the very market Intel is trying to entice with Pentium 4 -- aren't likely to be impressed. They may well sit on the sidelines until Intel, or third party chipset vendors, come out with DDR versions for Pentium 4 late this year or early 2002.
Rambus Inc., the Intel memory incumbent, isn't sitting still. Higher speed 1-GHz RDRAM chips, cheaper four-memory bank designs, less costly four-layer motherboards (vs. current six layer boards) are promised this fall. But this proliferation of memory types only adds to market confusion and the inclination of big customers to wait it out until some clear patterns emerge.
Pinout.The present 423-pin Willamette P4 changes to a 478-pin configuration this fall. That's because the follow-on 0.13-micron process Northwood P4 expected in Q4 '01 will use the new pinout, and Intel wants to start motherboard vendors transitioning early to the socket using the current Willamette MPU.
This is a Taiwan motherboard issue, where vendors don't want to tool up on 423-pin Willamette boards, which they have only nominally supported so far, only to do it all over again for the 478-pin P4s coming up.
Intel says the pinout issue will become moot this fall when all Pentium 4 motherboards in the market shift to the 478-pin socket. Until then....
Tualatin. Intel's new 0.13-micron process Pentium III chip, is code-named Tualatin. The Taiwan motherboard makers are making product now, and scattered OEMs are already announcing new systems, particularly for servers, using the new MPUs.
Again early tests suggest Intel may again be embarrassed -- this time by very good performance marks that you would expect from the shrink of a very mature Pentium III core. In fact, that was the original purpose of Tualatin: qualify and test out the new 0.13-micron and copper interconnect process on a proven MPU design, and then transition to the new Pentium 4 generation.
What Intel hadn't counted on is the apparent highly competitive performance of Tualatin, which many analysts believe will be very favorable against Pentium 4 in the mainstream PC market. Intel will try to contain the risk by launching Tualatin initially in the mobile space, where there is no P4 offering as yet.
But Intel is also introducing server Tualatin versions, taking advantage of the chip's higher Pentium III marks to compete aggressively against archrival AMD's sudden intrusion into the server market. Tualatin server processors are expected to be Intel's remedy for this week's embarrassing withdrawal of the 900-MHz Pentium III Xeon server MPUs, due to yet another intermittent Intel chip failure.
Tualatin server chip configurations most likely can be used for desktops as well, which could touch off comparisons to the mainstream PC Pentium 4, possibly muddying the waters for Intel.
Also awkward for the chip giant is the fact it has no DDR chipset of its own supporting Pentium III, thanks to its flawed Direct Rambus-only memory strategy. The 845 Brookdale SDRAM and DDR chipset only supports the new IA-32 Pentium 4 processors. Intel will be looking to Taiwan suppliers once again to pull its chestnuts out of the fire, this time providing DDR chipsets for Tualatin to compete against AMD's DDR-supported Athlon mobile processors.
There are reports in the market that factions within Intel are bitterly divided about whether the firm must bite the bullet and ship the Tualatin motherboards it makes itself with third party DDR chipsets.
Of course, DDR-enhanced Tualatins should perform only better, making the Pentium 4-comparison dilemma worse.
Intel says there are no company plans to use third party DDR chipsets on its own Tualatin boards. It claims Intel is waiting until Q1 '02 before activating the 845 DDR chipset capability, to be absolutely certain there are no problems with the new technology. But that delays the definitive Intel commitment to DDR that the market is seeking.
If Intel and the industry can hang on until late this year, the processor titan may finally get it all together. The 0.13-micron process Northwood Pentium 4 shrink with promise of much lower production costs and higher performance should hit the mainstream market. DDR chipsets for Pentium 4 should be available about this time. By then Intel can push Tualatin into the value PC segment, possibly recast as Celeron.
That may be what the corporate, volume and power customers are waiting for. Depending, of course, on how much AMD can take advantage of the Intel gyrations to take away further market share. Until then, everything seems up for grabs.