- NVIDIA Toolsets Target GPU Acceleration of Deep Learning, Other Algorithms for High-Performance Computing
- Jeff Bier’s Impulse Response—Neural Network Processors: Has Their Time Come?
- Analog Devices' SHARC Doubles Up, Adds ARM Option
- CogniVue's "Opus" APEX Generation 3: Vision Processing With Implementation Flexibility
- Jeff Bier’s Impulse Response—What Does Semiconductor Industry Consolidation Mean for Embedded Systems Designers?
Jeff Bier’s Impulse Response: NVIDIA GPUs Turn Up the Heat
In October of 2007, I wrote a column called “When Worlds Collide,” which was about NVIDIA’s emerging strategy of offering “general-purpose GPUs.” At the time, I thought it was interesting that NVIDIA had begun to move beyond graphics applications to target “high-performance computing” (
Well, two years later, two things are now clear. First, NVIDIA is not only going after
The second is that—if I do say so myself—I was right about their edge. NVIDIA has done an excellent job in leveraging the ubiquity of their GPUs (which are included in a large percentage of PCs) to gain widespread adoption of their chips as parallel processing engines for non-graphics applications. NVIDIA’s strategy of making the CUDA development tools readily available for free was also a smart move, with the net result that CUDA is becoming the parallel processing platform of first resort in many universities and research labs.
The bottom line is that, over the last two years, NVIDIA has tapped into strong demand for higher-performance computing solutions, and has benefitted from the widespread confusion and chaos about how to tackle parallel programming.
I saw plenty of evidence of NVIDIA’s growing traction at the NVIDIA GPU Developers’ Conference a couple of months ago. I was really impressed by the number and diversity of applications that people have implemented on NVIDIA GPUs, and by the performance they’re able to achieve.
In high-performance applications, worlds are definitely colliding. GPUs are competing with DSPs and other embedded processors, which are themselves increasingly competing with FPGAs. I expect that companies like TI, Freescale, Intel, and Xilinx are already feeling the heat from NVIDIA in
Jeff Bier is the president of Berkeley Design Technology, Inc. (www.BDTI.com), a benchmarking and consulting firm focusing on embedded processing technology. Jennifer Eyre White of BDTI contributed to this column.