Acceleration technology uses graphics processors

Technology is linearly scalable to several hundred desktops, making it possible to complete OPC or verification tasks within a few hours.

Gauda has developed a breakthrough technology that can accelerate OPC (optical proximity correction) and OPV (optimal proximity verification) up to 200 times faster than similar products, running on today's desktop computers.

Gauda achieves this acceleration without any specialised hardware or FPGAs, but by developing a new breed of algorithms using CPUs and GPUs (graphical processing units) that are traditionally used for gaming.

As design is moving to the 45 and 32nm technology nodes, requisite computations increase by two to three times per year.

Unfortunately, CPUs are not addressing this need, managing only a 30% increase in computation growth per year.

To solve RET (resolution enhancement technologies) problems, the only current solution is to keep increasing the size of CPU clusters, which often number in the thousands.

Tool vendors are at an inflection point where the time it takes to rewrite tools is lagging behind the changes in process nodes.

Fortunately, Gauda has anticipated this inflection point, realising an attractive speed advantage for OPC using GPUs.

Gauda engineers started by studying the available computational resources commonly found in typical desktop computers and optimised those resources to speed up the RET-related computations without sacrificing accuracy.

From the beginning, the company decided not to use any specialised hardware.

With Gauda technology and approximately 10 all-commodity desktop computers, a large 45nm full-chip layout can be decorated overnight.

Similarly, one single desktop computer is sufficient to complete the verification overnight.

"Gauda is in an enviable position".

"Its strength is a technology that spans all parallel software development", says Gary Smith, founder and Chief Analyst at Gary Smith EDA.

"Gauda is aiming its technology at an application that is on the leading edge of parallel computing".

"Once the company gains traction and the reputation that goes along with it, Gauda will be able to shift its attention to larger markets".

In stealth mode since its inception in late 2005, the company now has demonstrated technology that is linearly scalable to several hundred desktops, making it possible to complete OPC or verification tasks within a few hours.

Gauda technology runs on standard desktop computers equipped with graphics cards usually sold by companies such as Nvidia and ATI.

"With the growth of GPU computational capability doubling almost every year, our technology not only fits today's needs, but also has the potential to address future computational requirements", said Dr Ahmet Karakas, Gauda's founder, President and CEO.

"Whenever the GPU companies introduce an improved solution, all we need to do is to plug in the new hardware and immediately benefit from the performance boost".

"This and many other features of our technology make it a perfect candidate to enable the DFM efforts, since our solution is available with minimal or no additional hardware cost".

Gauda raised seed funding from a world-renowned, high profile group of investors in mid-2006, and is actively developing its first product.

The company was founded by Dr Ahmet Karakas and Ilhami Torunoglu, a veteran of the EDA industry who also has extensive chip design experience, and is staffed by developers with advanced degrees from Caltech and Stanford University.

"Creating this technology was not a simple mapping problem for us", says Ilhami Torunoglu, Gauda's CTO and cofounder.

"In order to reach the speeds we've achieved, we ended up developing our own parallel algorithms from scratch".

"Our solution is not limited to GPUs; we currently use CPU+GPU optimally as a resource in a typical desktop computer but the software has the potential to expand to other parallel compute engines".

Back to top