It’s now been over three months since Nvidia’s CEO Jen-Hsun Huang first paraded what turned out to be a fake Fermi card on stage. At that point, we thought we might see the first Fermi products in 2009, but this then evolved into the beginning of 2010, and now looks set to stretch even further into the year.
During a conference call to discuss Nvidia’s financial results for the fourth quarter of fiscal 2010, the company’s CEO Jen-Hsun Huang revealed that “Q2 is going to be the quarter when Fermi is hitting the full stride.” In this case, Q2 refers to the second quarter of fiscal year 2011, which will span the May to July period of 2010.
Of course, you can read what you like into the phrase “full stride”, but to us it looks as though Nvidia may be able to get a few high-end samples of Fermi out to the press in the next couple of months, but that it won’t be getting big volumes of Fermi cards into the shops until much later on in the year.
According to Nvidia, the reason for the hold-up is a severe lack of 40nm chips. During the conference call, the company explained that the lack of 40nm silicon had resulted in a “hand-to-mouth” system of supplying laptop manufacturers over the last year, who were apparently demanding far more 40nm chips than Nvidia could hand out.
In fact, Jen-Hsun Huang even estimated that the company had already missed out on between $100 million and $200 million US as a direct result of the supply problems. However, he later back-tracked on these figures, adding that he “probably shouldn’t have given any number at all, because we really don’t know.”
Huang was keen to point out that there’s no bad blood between Nvidia and TSMC [Nvidia’s manufacturer of 40nm chips] over the supply issue, and noted that “TSMC is doing a fabulous job improving their yields.” He also noted that “40nm supply is constrained for the world,” and that it doesn’t just affect Nvidia. Let’s face it; it hasn’t been easy for gamers to pick up 40nm Radeon HD 5800-series cards either. ATI uses the same 40nm line at TSMC as Nvidia.
During the call, Huang also said that the 40nm supply problems could hold up the production of Fermi-based cards at the lower end of the scale too. Although the GTX 470 and 480 have are likely to make it out of the doors fairly soon, we may have to wait much longer for mainstream and low-end Fermi products.
“All of that just depends on 40nm supply,” said Huang, adding that, “for the entry-level products, the truth is that the new architectures - at the very entry level – GPUs - are probably not extremely well appreciated anyhow. The reasons why people buy the new architectures are they tend to be early adopters, they tend to be game enthusiasts or workstation designers or creative artists.”
“Our current-generation GPUs are fabulous at all the things that mainstream consumers would use,” said Huang, citing “high definition videos” and, bizarrely, “3D Blu-Ray” as features that everyday customers can already achieve with current-generation mainstream graphics cards.
“My sense is that they’re going to do quite nicely in the marketplace,” said Huang, adding “and then we’ll just transition as fast as we can,” presumably meaning a transition to mainstream Fermi-based products.
Thanks to X-bit Labs for prompting us to look into this.