AMD Kaveri Shipments Slip to 2014, Speculation Rampant on Unannounced Products
The story around AMD's upcoming Kaveri continues to evolve, but it's increasingly clear that the chip won't be available for retail purchase this year. If you recall, AMD initially promised that Kaveri would be available during 2013 and even published roadmaps earlier in May that show the chip shipping in the beginning of the fourth quarter. This was always surprising, but AMD adamantly stuck to the published roadmap and the Q4 2013 availability. Or at least, they did.
What the company is saying now, in the wake of a rather confused DigiTimes story (more on that in a moment) is that while Kaveri will ship to customers (read: OEMs) in late 2013, it won't actually hit shelves until 2014. Whether this is a delay depends on how seriously you believed the company would meet an end-Q3 / early Q4 ship date. Personally, we never really did. According to certain sources we spoke to within the company, Kaveri hadn't taped out as of November 2012.
The reason that's significant is because "tape-out" is the point at which a CPU design heads to the foundry for actual fabrication. GlobalFoundries and AMD collaborate at every level, but GlobalFoundries can't start ramping an architecture until it knows what the final silicon needs to look like. Ramping a chip from this point to physical production takes a certain amount of time, as does the following ramp from early manufacturing to full volume. In short, a chip that tapes out in, say, December 2012 is really stretching to be in volume production one year later. CPU manufacturing is just too difficult and there's always the chance of a late bug or errata that requires a further respin from the design team.
Given what we'd heard about Kaveri's late tape-out, I'm not sure it's accurate to call this a delay so much as an overly aggressive prediction. Modern semiconductor manufacturing is far more difficult than it used to be, and while GF is ramping 28nm in good order now, each node has proven more difficult than the last. If Kaveri ships early in Q1 rather than shipping for volume in Q3/Q4, it doesn't mean something went wrong. It means someone got a little too enthusiastic in the marketing department.
About That Desktop Kabini Rumor
The Kaveri story, however, was just one of DigiTimes' claims. The Taiwanese publication is also claiming that AMD's desktop Kabini with a TDP of 25W would be announced in February 2014 and ship in March 2014. This is apparently a "rescheduling." Unlike Kaveri, it's extremely difficult to understand how a part which hasn't been discussed, previewed, or marketed can actually now be considered delayed.
It's not that we haven't seen Kaveri desktop boards -- we have, from time to time -- but there's never been any indication that AMD intended to launch these as a separate desktop form factor. It's not even clear why the company would bother to do so. Kaveri as a "nettop" product on a small motherboard, perhaps paired with a PCI-E slot, makes sense. Kabini on a full-size socket with a line of drop-in replacements is the kind of product AMD would almost certainly discuss, not simply launch.
The stock took some hits on this today, possibly because people confused the already-launched notebook Kabini (which is shipping in hardware you can buy) with rumors of a delay to a desktop part that, to the best of our knowledge, no one has particularly expected. While Google has plenty of hits for the Socket FS1B that this chip would supposedly use, I'm going to chalk this one up to "Confused rumor" not "Meaningful hardware change."
What Does This Mean For AMD's Competitive Standing?
The short answer? Nothing.
The longer answer: Hopefully good things.
The reason Kaveri was late taping out, according to my sources, was that AMD kept the chip back to put some additional polish on its performance. Unlike Piledriver, which we knew would be a minor tweak to the core Bulldozer architecture, Steamroller is the first serious overhaul to that hardware. That means it's AMD's first chance to really fix things. Piledriver delivered improved clock speeds and power consumption, but CPU efficiency barely budged compared to 'Dozer. Steamroller needs to deliver on that front.
If holding Kaveri back a few months gave AMD the time it needed to further improve the core, and the end result is a chip that's 10-15% faster clock-for-clock than Piledriver as opposed to 8-11%, then there's no downside here. If it helps AMD ensure that the chip yields well at high frequency or low power consumption, that's all to the good. In the grand scheme of things, moving the launch from late November to early January isn't going to hurt the company's sales much, but it may lead to significant improvements in the final product.
Because ultimately, that's what this is about. AMD needs Kaveri to hit, and hit big. While there's no chance of closing the gap between Piledriver and Haswell in a single leap, a 10-15% IPC gain would put Piledriver's efficiency back at Thuban/K10 levels (but at a significantly higher clock speed). At that point, AMD's quad-cores stop being outperformed on a regular basis by Intel dual-core chips and the CPU side of the APU equation gets a decided kick in the pants.
So here's hoping.
What the company is saying now, in the wake of a rather confused DigiTimes story (more on that in a moment) is that while Kaveri will ship to customers (read: OEMs) in late 2013, it won't actually hit shelves until 2014. Whether this is a delay depends on how seriously you believed the company would meet an end-Q3 / early Q4 ship date. Personally, we never really did. According to certain sources we spoke to within the company, Kaveri hadn't taped out as of November 2012.
The reason that's significant is because "tape-out" is the point at which a CPU design heads to the foundry for actual fabrication. GlobalFoundries and AMD collaborate at every level, but GlobalFoundries can't start ramping an architecture until it knows what the final silicon needs to look like. Ramping a chip from this point to physical production takes a certain amount of time, as does the following ramp from early manufacturing to full volume. In short, a chip that tapes out in, say, December 2012 is really stretching to be in volume production one year later. CPU manufacturing is just too difficult and there's always the chance of a late bug or errata that requires a further respin from the design team.
Given what we'd heard about Kaveri's late tape-out, I'm not sure it's accurate to call this a delay so much as an overly aggressive prediction. Modern semiconductor manufacturing is far more difficult than it used to be, and while GF is ramping 28nm in good order now, each node has proven more difficult than the last. If Kaveri ships early in Q1 rather than shipping for volume in Q3/Q4, it doesn't mean something went wrong. It means someone got a little too enthusiastic in the marketing department.
About That Desktop Kabini Rumor
The Kaveri story, however, was just one of DigiTimes' claims. The Taiwanese publication is also claiming that AMD's desktop Kabini with a TDP of 25W would be announced in February 2014 and ship in March 2014. This is apparently a "rescheduling." Unlike Kaveri, it's extremely difficult to understand how a part which hasn't been discussed, previewed, or marketed can actually now be considered delayed.
It's not that we haven't seen Kaveri desktop boards -- we have, from time to time -- but there's never been any indication that AMD intended to launch these as a separate desktop form factor. It's not even clear why the company would bother to do so. Kaveri as a "nettop" product on a small motherboard, perhaps paired with a PCI-E slot, makes sense. Kabini on a full-size socket with a line of drop-in replacements is the kind of product AMD would almost certainly discuss, not simply launch.
The stock took some hits on this today, possibly because people confused the already-launched notebook Kabini (which is shipping in hardware you can buy) with rumors of a delay to a desktop part that, to the best of our knowledge, no one has particularly expected. While Google has plenty of hits for the Socket FS1B that this chip would supposedly use, I'm going to chalk this one up to "Confused rumor" not "Meaningful hardware change."
What Does This Mean For AMD's Competitive Standing?
The short answer? Nothing.
The longer answer: Hopefully good things.
The reason Kaveri was late taping out, according to my sources, was that AMD kept the chip back to put some additional polish on its performance. Unlike Piledriver, which we knew would be a minor tweak to the core Bulldozer architecture, Steamroller is the first serious overhaul to that hardware. That means it's AMD's first chance to really fix things. Piledriver delivered improved clock speeds and power consumption, but CPU efficiency barely budged compared to 'Dozer. Steamroller needs to deliver on that front.
If holding Kaveri back a few months gave AMD the time it needed to further improve the core, and the end result is a chip that's 10-15% faster clock-for-clock than Piledriver as opposed to 8-11%, then there's no downside here. If it helps AMD ensure that the chip yields well at high frequency or low power consumption, that's all to the good. In the grand scheme of things, moving the launch from late November to early January isn't going to hurt the company's sales much, but it may lead to significant improvements in the final product.
Because ultimately, that's what this is about. AMD needs Kaveri to hit, and hit big. While there's no chance of closing the gap between Piledriver and Haswell in a single leap, a 10-15% IPC gain would put Piledriver's efficiency back at Thuban/K10 levels (but at a significantly higher clock speed). At that point, AMD's quad-cores stop being outperformed on a regular basis by Intel dual-core chips and the CPU side of the APU equation gets a decided kick in the pants.
So here's hoping.
