Today UK startup brytlyt announced an enterprise data platform using Nvidia’s Tegra K1 superchip that offers faster, more efficient and more flexible data processing and analytics.
brytlyt’s approach uses a network adaptor physically located on the same board as the Nvidia Tegra K1 system on a chip. With the GPU and CPU now on the same piece of silicon in the K1 SoC, both processors can access a shared memory at a very high speed, eliminating the current bottleneck of GPU accelerated databases. This novel configuration increases network bandwidth up to 25x, resulting in a faster, less expensive and more scalable solution.
According to brytlyt, the real magic is in the software that presents a virtualized database to the end user. brytlyt can now control the entire stack from hardware and network configuration through to software and user experience.
We’re confident this is a real paradigm-shift in data processing” said Richard Heyns, Founder and CEO of Brytlyt.
At an estimated cost of $400 per device, a thousand of these network-attached GPUs could mine Facebook’s 750TB dataset for valuable insight in under 15 minutes at a cost of $400,000. Current alternative solutions providing this kind of performance cost millions of dollars.
Brytlyt will demonstrate their technology at the GPU Technology Conference this week in San Jose.
Leave a Comment