Intel’s 10nm Volume Ramp and Cannonlake Microarchitecture Delayed Due to Yield Problems
In what appears to be a case of deja vu, Intel has delayed its 10nm platform in its entirety. Needless to say, the Cannonlake Processor Architecture has been delayed as well, with the Kaby Lake intermediary platform taking its place. Unless I am very much mistaken, this could have interesting implications for Moore’s Law as we know it, unless TSMC or Samsung can do something about it.
Intel delays ramp to 10nm and consequently Cannonlake – again
The official reason for the delay is poor yields. Simply put the process isn’t mature enough for ramping to high volume production anytime soon. Intel has been following its tick tock cadence for quite some years now and has managed to stick to Moore’s Law by the skin of its teeth. On the 14nm node for example, the Core M chip qualified as the entrant to ushering in 14nnm in 2015, but that was only after the onset of unexpected delay after delay. The problem is, even according to their own projections 14nm was supposed to be a walk in the park, and that took much longer than expected.
So when 10nm, an admittedly complicated node to handle, gets delayed, that is something that everyone was expecting to happen. The only question is – how does it affect the Industry and why is the delay happening.If we take Intel’s explanation at face value then that means that yields are too low for Intel to ramp. But the economic feasibility of the yield percentage and process maturity should be compared against the loss of value as a business Intel risks to incur if TSMC or Samsung overtakes it.
Both Samsung and TSMC plan to start ramping 10nm in 2015 (via Semi Wiki). And if they hit the node before Intel – the market won’t care about the quality of the chips. They will go oh hey, Intel got beaten to 10nm. They wont look at the Gate pitch & Fin Pitch, and if Intel seems to have lost its technological lead (however inaccurate that might be) the company could see its value decrease over night. Intel and Samsung can chose to hit the ground running by compensating for bad yield by massively increasing production – but while that would be a viable business strategy for Samsung, it doesn’t really seem like Intel’s modus operandi.
That leads me to another alternative: Intel is deliberately slowing spending as a business decision. Its no secret that Intel is the only company in the world capable of manufacturing sub 20nm circuits that can withstand high clock rates. It can theoretically afford to wait and take its time on the 14nm node and this could very well be a business decision. Evidence of this could be suggested by the fact that Intel was supposed to buy over $6 Billion worth of equipment for its fabrication plant (Fab 28) in Kiryat Gat, Israel. However, the purchase was put off and has now been delayed to 2016 (via KitGuru) – needless to say, without this technological upgrade, the 10nm process ramp cannot be undertaken.