3dfx bought their own fab. Look how that turned out for them.

Nvidia gave the following presentation at the International Trade Partner Conference (ITPC) forum last November. Many of the company’s complaints regarding its current partnership with TSMC are exactly what you’d expect given the manufacturing problems the entire industry is facing. What’s surprising are Nvidia’s remarks concerning TSMC’s current cost curves and manufacturing ramps. This is normally the sort of information discussed quietly between a foundry and its customers or by the press with help from various anonymous sources. Discussing the problems publicly is a sign of just how frustrated the company has become.

The GK104 is great, but it doesn’t change the nature or severity of the underlying problems. As for whether Nvidia’s unhappiness with TSMC heralds a potential alliance with GlobalFoundries, we’re dubious. Not only has GF only recently ironed out its own 28nm issues, the nature of the foundry business doesn’t allow for quick shifts. Indeed, part of the reason that manufacturers like TSMC have historically exercised such control over their partners’ PR releases is because once you’ve committed to a foundry, you’re locked in for a substantial period of time. The fact that there’s now two foundries available with cutting-edge technology doesn’t change that, and the Common Platform Alliance favored by IBM, Samsung, and GloFo only mitigates some of the problems with moving a design from foundry to foundry, it doesn’t remove them.

The real question, at least for TSMC’s other customers, is whether the graphs and charts Nvidia has shown are specific to the company’s own products or reflect universal trends. There’s good reason to suspect the latter; Nvidia may have had more trouble than some of TSMC’s other customers, but our analysis of semiconductor industry roadmaps revealed a great deal of uncertainty about the road forward. Nvidia opted to aggressively optimize GK104 precisely because the old strategy of bolting on more cores and ratcheting up transistor counts isn’t sustainable.

Further evidence for the accuracy of NV’s presentation comes, ironically, from the company’s primary GPU competitor. At AMD’s Financial Analyst Day, CEO Rory Read made a point of saying that the company no longer intends to aggressively transition to new process nodes given the diminishing marginal returns from doing so.

Change the color scheme, and Nvidia’s graphs could’ve dropped right into AMD’s presentations in early February.

Nvidia’s willingness to stand up and talk about these problems is an “Emperor’s new clothes” sort of moment. The long-term repercussions, if any, are still unclear.

As for whether Nvidia’s unhappiness with TSMC heralds a potential alliance with GlobalFoundries, we’re dubious.

Both Tegra 4 and Kepler are going to be built using the 28nm process at TSMC and the biggest worry for the company is the constrained supply. Nobody can get enough supply, and the battle is on between AMD, NVIDIA and Qualcomm are scraping for each available 28nm wafer.

The problem Qualcomm faced with TSMC was done swiftly and the company is now utilizing alternative manufacturers as well, such as GlobalFoundries. However, the problem with the Common Platform Manufacturing Alliance (IBM, GlobalFoundries, and Samsung) is that they are using "Gate First" approach for the High-K Metal Gate. The Common Platform alliance announced that they are switching to "Gate Last" approach with the 20nm process. Intel and TSMC on the other hand, utilize "HKMG Gate Last" for their manufacturing processes.

While even Intel insiders agree that Gate First is a more efficient approach, it also has more risk-prone and Intel and TSMC just didn’t want to take the plunge into a temporary advantage. At 22/20nm, Gate Last is the approach to have.

Second problem was the fact that according to our sources close to heart of the matter, NVIDIA did not want to deal with GlobalFoundries while AMD was the shareholder "for competitive purposes". NVIDIA’s executives obviously didn’t have the memo in which AMD-GF relationship was strained (Llano is a good example).
Thus, what could NVIDIA do? The company talked with the Common Platform for quite some time, and the choice was obvious. When 28nm and 20nm processes started to reach early phase, NVIDIA talked to Samsung about a trial run of Tegra chips. The engineering work required for this task took a lot of effort from NVIDIA and Samsung, but we received word that very recently, NVIDIA received chips from Samsung.

Once 20nm process matures AMD, NVIDIA, and Qualcomm and just about everybody else will be able to switch foundries from TSMC to Common Platform and back, which are what all companies need to battle Intel’s "transistor excellence".

Qualcomm already uses Common Platform through GlobalFoundries and it looks NVIDIA is joining the Common Platform through the small doors with Samsung. With AMD expanding from GlobalFoundries to TSMC, it looks like the foundry battle will seriously heat up at 20nm.