Intel's Atom is on a collision course with Core—and everybody loses
Intel has made it clear that its fourth-generation Core or “Haswell” processors will power future PCs. But at the low end of the notebook market, the company's branding strategy will be far muddier—and that's bad for Intel, considering it needs to tell consumers exactly what to buy.
On one hand, Intel needs to present a clear alternative to ARM, whose tablet partners dominate the mobile space and threaten traditional PCs. From a technical standpoint, Intel seems well on its way to solving that problem. Intel’s Haswell chip will eventually migrate to as little as 6 watts, Intel said Tuesday, meaning that its Core chips could eventually power tablets, too. But in the meantime, Intel's upcoming “Bay Trail” Atom chips are waiting in the wings—and could offer the performance buyers want at a fraction of the price.
This fall, buyers looking to pick up an inexpensive Windows convertible or notebook may have to sort through several processor brands, including Atom, Core, Pentium, and Celeron. And that’s without even factoring in the code names: Bay Trail, Silvermont, and Haswell, among others. Intel may have the right chip at the right price, but because its processor line-up is so crowded, it's in real danger of screwing the whole thing up.
If there’s one thing Intel did “wrong,” it’s that it made the Bay Trail Atom chip so good—too good for some Wall Street analysts, who worry that a $60 Atom chip will drag down profit margins at a company used to selling Core chips for hundreds of dollars.
While the first-generation of “Clover Trail” chips powering Windows 8 convertible tablets were underwhelming, the “Bay Trail” generation, featuring quad-core processors, will power touch-enabled thin notebooks with “really good performance” that will hit $300 price points, former chief executive Paul Otellini said in April. So who’s going to buy a $600 Core notebook if a machine half the price will do the job?
"Basically, in Atom they have a part that’s going to be a problem," summed up Dean McCarron, principal analyst with Mercury Research.
Branding just makes the problem worse
Intel tried to solve the problem on Friday by bringing the Bay Trail chip under the Pentium and Celeron brands. Technically, this makes sense, because Bay Trail is PC-ready, with support for PC components like PCI Express.
Intel sources also said Friday that Intel has begun phasing out the Atom brand in discussions with OEMs, replacing it with the “Silvermont” name. That’s the name of the architectural refresh for the upcoming Bay Trail chip. Whether “Silvermont” appears on the outside of a box is unclear, but for clarity’s sake, let’s hope not. (Clarification: While Intel sources may have said that the Atom brand is being downplayed in discussions with OEMs, in favor of the "Silvermont" microarchitecture, Atom is still an official brand within Intel and is not being replaced with Siilvermont, an Intel spokeswoman said.)
But in trying to associate the Bay Trail chips with Intel’s PC brands, Intel’s marketing gurus have created a hot mess. On one hand, we have the Pentium, the name for Intel’s flagship part—it was launched 20 years ago, in 1993. The Celeron isn’t much better. That’s the name Intel gave to its cut-rate, low-performance parts. If you wanted a cheap PC that could barely get the job done, you bought a Celeron. Granted, Intel’s Atom doesn’t have that much brand equity, either—it powered the first generation of Google TVs, for example, and the “Clover Trail” version of the chip produced middle-of-the-range performance in convertibles like the Samsung ATIV Smart PC.
But what consumers will be asking is a simple question: OK, so if I buy an ARM tablet, I get this: whatever Qualcomm or Nvidia or Samsung offer. So what can you do for me, Intel?
Instead of offering a simple answer—“This is our ARM killer, Atom”—it appears that Intel’s response will be something like, “Oh, you’ll want a Pentium, then. No, not the old one. The new one. See, there’s the model number right there. Oh, and you can go with a Celeron, too. No, they’re better. See, they’re part of the Silvermont family. What’s a Silvermont? Here, take some brochures...”
Put another way: Intel lists fifty-eight different Pentium models—alone. How is a consumer going to know if he or she is buying the good "Bay Trail" Pentium, or one of the older, crappier ones?
Tablets will use “Bay Trail"—but will the Surface?
Sources within Intel say that Intel also plans to provide a “Bay Trail” derivative for tablets, dubbed “Bay Trail-T,” that will be out by the holidays. The good news for Intel? It delivers “kick-ass quad-core performance,” the source said. The bad news? It delivers “kick-ass quad-core performance”.
The leader in the tablet market is ARM, which powers, well, almost everything in the tablet and mobile space. Intel has tried to make inroads via a partnership with Lenovo, and its Atom-powered K800 phone that’s so far been sold exclusively in Asia. But that’s a drop in the ocean compared to ARM.
ARM has already claimed that it will offer better performance per watt than the the Silvermont architecture driving Bay Trail, and Microsoft said last year that it will work with ARM to port its Windows 8 OS over to the first generation of ARM’s 64-bit chips, due to sample later this year. But Windows RT on ARM has largely failed, and Samsung has chosen an Atom chip to power its Galaxy Tab 3 tablet (albeit the older Clover Trail+ variant). Intel appears to have a windows of opportunity to exploit, if it can.
Pessimists would say that any success in tablets means that Intel will sell fewer, higher-margin Core chips. Optimists will respond by noting that there’s no real proof that buyers favor tablets over notebooks, and that some consumers will buy both—or even favor an Intel-powered Windows tablet if the performance is high enough. And analysts like Nathan Brookwood of Insight 64 note that the “Haswell” Core parts should eventually be offered in power envelopes as low as 6 watts, muddying the water further.
And there’s one more possibility, too: If Windows RT has failed and Microsoft tosses in the towel, it will still want a low-power chip to power the cheaper version of the Surface. And what could be a better option than Bay Trail? (While sources close to Intel say that’s a likely possibility, my source has remained mum on the matter.)
Could Atom be crippled?
Last week, the real question on the lips of Intel watchers was simple: Would Intel “cripple” the Bay Trail chip? Doing do would be relatively easy. Intel could artificially limit the clock speed or graphics performance of the chip, creating an artificial separation between the Core and Atom families.
It’s not clear if Intel plans to go this route, although it appears that it might be leaning that way. Company executives seem to be hoping that rebranding Bay Trail solves the problem.
A new family of cheap, low-power, high-performance convertible tablets could be just the thing the “Wintel” partnership of Intel and Microsoft needs for the holidays. Let’s just hope that Intel cares more about giving customers the shiny new toys we’ve been asking for, even if it might mean coal in the stockings of its investors.