This telephone has too many shortcomings to be seriously considered as a means of communication - Western Union memo, 1876
CHIP DESIGNER Nvidia has said that it will have Kepler based K20 Tesla boards in supercomputers in time for November's Top 500 list.
Nvidia recently launched its Kepler K10 Tesla boards that focus on single-precision floating point performance, with the firm claiming its Tesla K20 boards will appear in the fourth quarter. Now the firm has said it expects supercomputer clusters featuring its Tesla K20 boards will appear in the November 2012 Top 500 list.
When The INQUIRER asked Nvidia if its Kepler-based accelerators will make a big splash in the June Top 500 list Sumit Gupta, GM of Nvidia's HPC business unit said that was "highly unlikely". Gupta continued, "A lot of the supercomputing centres running Linpack need double-precision [floating point] and our double-precision product is going to be available in the Q4 time frame, so its not going to impact Top 500 in the June announcement. But we expect to make a huge impact in the November announcement."
Although Gupta said K20 boards are penciled in for the fourth quarter, given that the Top 500 list will be released in November, K20 boards should be available a few months before then, which could push Intel to move forward its Xeon Phi launch plans.
Nvidia claims its K20 Tesla boards will provide three times the double-precision floating point performance of its previous generation Fermi M2090 boards, which in theory would put the board in the 1.5TFLOP region, significantly higher than AMD and Intel are currently claiming with their respective Firepro and Xeon Phi products expected around the same time.
While Nvidia might not improve on its already impressive Top 500 positions, the fact that it expects to make a big dent in November's list should give cause for concern at Intel for its Xeon Phi board. µ
Sign up for INQbot – a weekly roundup of the best from the INQ