In an interview with NVIDIA CEO Jen-Hsun Huang, Digitimes asked some probing questions about the company’s stance in the GPU market, NVIDIA’s plans for surviving in the world of Intel’s Larrabee and more.  One interesting comment came when Huang was asked about how both Intel and AMD planning to integrate GPU functionality into their CPUs would affect NVIDIA.  His response was that “as for whether CPUs with integrated GPUs will affect Nvidia’s market share, there is no direct relationship between the two” and even “our competitors’ integrated GPU solutions will just be a waste of resources”.

As for NVIDIA’s plans for an x86 processor, Huang still claims that they have no plans to move into that market:

Q: There are rumors Nvidia is heading into x86 CPU design, your comments?

A: Currently, we have no plans for that. We are very busy already and have no time to cross over to our competitors’ field. The most important thing to remember is that Intel is the leader in the CPU market and so it’s better that Nvidia focuses on what we do best. To cut in to the x86 CPU market would just be a waste of time and resources.

Huang also admits to something we have been saying for some time: single GPU solutions are almost always better than dual-GPU solutions in terms of efficiency and cost.  When asked about AMD’s recently strategy to build a GPU for the mainstream market and then double it up for their high end part, Huang said “I don’t agree with the strategy since single-GPU designs are definitely more efficient than dual-GPU ones. If a single GPU can perform 100%, a dual-GPU card can only achieve 130-180% performance, while costs are much higher.”

There are lot of other great comments and quotes from the story, so give it a read.  I have included some more highlights below.
Digitimes interviews NVIDIA CEO: GPU on CPU 'a waste' - Graphics Cards 2

Q: Does Nvidia have any plans to own a fab?

A: If you ask Intel, they will tell you that owning your own fab is the best choice since it helps you lower manufacturing costs and control process advancement. However, if you ask AMD, they will likely tell you that they don’t want to own any fabs. Nvidia has never owned a fab, and yet we are going strong.

Consumers don’t care about whether a company owns its own fabs or outsources production to other makers. Consumers only want to buy products bearing the Nvidia name. We have good relationships with Taiwan Semiconductor Manufacturing Company (TSMC) and United Microelectronics Corporation (UMC) and this cooperation won’t see any changes in the next few years.


Q: Will Intel’s Larrabee impact Nvidia?

A: Intel has been talking about GPU development since they are worried that the importance of GPUs will surpass that of the CPU. However, since Intel has not brought any product to market, I can’t really comment about it.

However, I would like to comment on Intel’s push to using x86 architecture for GPUs. High-end graphics technology is not something you can achieve by just stepping through the door. Intel’s hope of changing the whole GPU industry is not going to be something easily achieved.


Q: So, with CPU makers starting to integrate different functionalities such as memory controllers and GPUs on-die or in-package, will we start to see GPUs also begin to integrate different kinds of technologies or cores?

A: GPUs were born for graphics calculations, they are not made to compete with CPUs and the two should live side by side. Our Cuda technology isn’t meant to enable a GPU to completely replace a CPU. I believe GPUs don’t need to integrate different kinds of cores.


Q: AMD’s current strategy for GPUs is to focus resources on mainstream products in order to save costs and time, and then target the high-end market with dual-GPU designs, what is your opinion on this?

A: Basically, I don’t agree with the strategy since single-GPU designs are definitely more efficient than dual-GPU ones. If a single GPU can perform 100%, a dual-GPU card can only achieve 130-180% performance, while costs are much higher.

As for whether dual-GPU designs help save development time, I think this is a question of the research team’s ability. Since Nvidia’s team is much stronger than our competitor’s, I believe we are able to achieve time-to-market and yet still maintain design excellence.


Q: In addition to GPUs, chipsets and mobile device processors, does Nvidia have plans for new product lines?

A: I am a person that likes to complete one task before heading into another; Nvidia’s GPU journey is not yet complete. I believe the GPU market can continue to grow for another 15 years. Currently, I only want to focus on our current product lines.

I expect the mobile device market to have large business potential in the future. Currently, consumers have at least one handset and with internet applications continuing to expand, multifunction mobile devices which integrate telecommunication capabilities will become standard equipment in the future.