GlobalFoundries Denies Report of 14nm Delay, Claims Technology Ramp Is On Schedule
Earlier today, Barron's published an article speculating that GlobalFoundries would delay its 14nm product ramp by several quarters as it struggled with unspecified production problems. GlobalFoundries has denied any such issues in a statement to Hot Hardware, and has specifically told us that its 14nm ramp plans remain on-track.
The state of GlobalFoundries 14nm ramp is important for multiple reasons. Earlier this year, the company announced that it would drop its own plans for a hybrid 14nm/20nm hybrid node (14nm XM) and would instead adopt Samsung's 14nm process technology. At least some of this capacity would be dedicated to Samsung's production, making GF a second-source facility for the Korean chip manufacturer. Given that Samsung has been working hard to win Apple's business for its 14nm nodes, a delay at GF could jeopardize that process.
These conclusions, however, may be premature. In a statement to Hot Hardware, Senior Manager of Technology Communications at GlobalFoundries, Jason Gorss, told us:
Right now, Samsung is expected to ramp 14nm production in 2015, with GlobalFoundries following suit. It's not clear which companies will make the shift to 14nm first; AMD has a longstanding agreement with GloFo that stipulated the use of GF for certain products and processors below the 28nm node, but whether those agreements are still in place is uncertain. Apple will likely be one of the first customers to deploy 14nm hardware, just as it's led the industry on 20nm.
There's always a certain amount of arbitrariness to the various node designations, but 14nm is expected to deliver die size and frequency advantages that are greater than the 28nm - 20nm shift. Companies like AMD appear to be pursuing a bifurcated strategy, with some of the company's products debuting on 20nm (GPUs, the upcoming Skybridge APU), but others skipping that node in favor of a 14nm FinFET deployment at a later date.
The state of GlobalFoundries 14nm ramp is important for multiple reasons. Earlier this year, the company announced that it would drop its own plans for a hybrid 14nm/20nm hybrid node (14nm XM) and would instead adopt Samsung's 14nm process technology. At least some of this capacity would be dedicated to Samsung's production, making GF a second-source facility for the Korean chip manufacturer. Given that Samsung has been working hard to win Apple's business for its 14nm nodes, a delay at GF could jeopardize that process.
These conclusions, however, may be premature. In a statement to Hot Hardware, Senior Manager of Technology Communications at GlobalFoundries, Jason Gorss, told us:
Our 14nm plan has not changed. A key part of our strategy is to order tools ahead of facility readiness to enable the fastest possible ramp. Due to the large number of tools coming in, we have our vendors stage these tools at a nearby warehouse to facilitate a fast install. This logistical move is in no way related to yield challenges or a delay in our technology ramp and is, in fact, quite the opposite. Our Fab 8 ramp is on track and we have yielding customer product on our 14nm technology.As statements go, this is a fairly strong one. It's extremely unlikely that Apple would change its plans to ramp the A9 at Samsung based on issues at GlobalFoundries; Apple tends to demand strong agreements from its suppliers and doesn't often leap before it understands the shape of the land (the recent GTAT debacle notwithstanding).
Right now, Samsung is expected to ramp 14nm production in 2015, with GlobalFoundries following suit. It's not clear which companies will make the shift to 14nm first; AMD has a longstanding agreement with GloFo that stipulated the use of GF for certain products and processors below the 28nm node, but whether those agreements are still in place is uncertain. Apple will likely be one of the first customers to deploy 14nm hardware, just as it's led the industry on 20nm.
There's always a certain amount of arbitrariness to the various node designations, but 14nm is expected to deliver die size and frequency advantages that are greater than the 28nm - 20nm shift. Companies like AMD appear to be pursuing a bifurcated strategy, with some of the company's products debuting on 20nm (GPUs, the upcoming Skybridge APU), but others skipping that node in favor of a 14nm FinFET deployment at a later date.
