The company expects another record financial evolution in the professional segment

Feb 18, 2012 09:36 GMT  ·  By

NVIDIA already has some design wins in the supercomputing market segment, but the upcoming, newer GPUs are expected to seriously boost graphics accelerator adoption in this field.

According to X-bit Labs, the number of HPC design wins scored by Fermi-based GPU compute accelerators will be dwarfed by Kepler's popularity.

The new GPUs will be launched in the second quarter of the year, probably in April.

The report states that, according to NVIDIA, builders of high-performance computing clusters are anticipating the first shipments.

"Our professional solutions business is expected to have another record year. Maximus enables us to sell more than one GPU on to a workstation, and new supercomputer centers around the world are anticipating the shipment of Kepler," said Rob Csongor, vice president of investor relations at Nvidia, during the most recent conference call with financial analysts.

The Santa Clara, California-based company has been making Kepler chips since last year (2011), hoping to amass a quantity large enough to divide between the consumer, enterprise and HPC segments.

The reason NVIDIA sees supercomputers as such a ripe market is the double precision GFLOPS performance per watt of Kepler (twice that of Fermi).

There is also the virtual memory space technology to consider, which will let the CPUs and GPUs use the “unified” virtual memory. Pre-emption and improved GPU autonomy are part of the list of new assets as well.

Unfortunately, the GPU maker may have a hard time attaining its goals, mostly due to the low chip yields of Taiwan Semiconductor Manufacturing Company (TSMC).

TSMC makes all its 28nm GPUs, just like it was, and still is, the exclusive provider of 40nm units for the past few years.

It turns out that the foundry is not having the easiest time with the manufacturing process, just like with the 40nm in 2009 and 2010.