Recently, Qualcomm has been the subject of many rumors on fabbing and capacity. Luckily, SemiAccurate can clear all of it up, they are moving from all TSMC to majority production at Global Foundries this year.
As we reported a month ago, Qualcomm is rather peeved at TSMC’s inability to supply wafers. A lot of people are seeing this as a yield problem, but SemiAccurate’s sources are saying that while yields are not perfect, they aren’t that bad either. Unless you are Nvidia, but that is self-inflicted on 28nm, and does not carry over to anyone else using the process.
Educational note: Low yields do not lead to wafer shortages unless they are so low that they cause a company to buy lots more wafers to make up for a shortfall. Given yields which are abysmally low, a company isn’t likely to invest in more expensive but low yielding wafer starts if they do not have to in order to meet customer demand. On the flip side, shortages of wafer starts do not indicate anything about yields. Please do not confuse yourself by calling yields = capacity constraints.
The lack of capacity has lead Qualcomm to move the majority of their 28nm production later in 2012 to Global Foundries, something we said more than a month ago. GloFo has tons of capacity, and although the 28nm process is very very late, it looks to be in much better shape than TSMC’s version of 28nm, especially on the HKMG variants that most want.
Qualcomm is also coming out with a quad-core S4 Snapdragon code named Krait, and this version will have an Imagination GPU to placate Microsoft. Both of these two additions mean a much bigger die than the dual core S4, and likely a massive bump in required wafer starts. TSMC has basically said they can’t do more than they have already promised, leaving Qualcomm very little option but to go elsewhere.
As we previously said, GloFo is there, has capacity to spare, and a process that at long last looks really good. It is good enough for AMD, good enough for Qualcomm, and given the smiles at the last Common Platform conference, probably good enough for a host of others. How things change in a mere six months.S|A
Latest posts by Charlie Demerjian (see all)
- Thing go bump(gate) in the night for Nvidia’s GP100 Pascal GPU - May 3, 2016
- Cavium’s Octeon TX blends compute and packet moving - May 2, 2016
- Qualcomm releases Zeroth API to developers - May 2, 2016
- Another detail about Qualcomm server SoCs revealed - Apr 27, 2016
- AMD finally really honestly launches the dual Fury - Apr 26, 2016