This week we traveled to Santa Fe for the IDC User Forum to get a better grasp on upcoming trends for the high performance computing market in 2014 and beyond.
While the emphasis at this particular meeting was on the value of industrial partnerships, there were a number of talks around scientific computing initiatives, developments toward exascale, and the role of large-scale analytics in research and enterprise settings.
As we reported not long ago, IDC shared news of a decline in the overall HPC server market, although these numbers are far more reflective of unexpected growth that caused last year’s numbers to soar. While the numbers this year reflect a slight dip for these reasons, the supercomputing market is still strong—right around the $10 billion mark.
To put that market in better context on the chip side, consider that over 30 million processors were shipped in 2013 across the HPC spectrum. As seen in the chart below, it’s clear that x86 has continued to dominate the market but the chatter around the future of ARM, especially as more vendors look to the 64-bit future, will shake up the market. IDC’s Steve Conway said that they’ve counted over 30 vendors working now to move the ecosystem forward around enhancements to ARM and via movements around OpenPower.
In terms of where those cores are going, the chart below shows some interesting momentum among the vendors, with HP maintaining share. This won’t change, says Conway, especially in the wake of IBM selling off its x86 business to Lenovo.
While there were plenty of conversations around how long that deal might take to close (it is expected by the end of the year, but there are complications we’re following for a future article), if and when it does there could be a more pared-down vendor market with HP gaining more of a lead and others grabbing around a 15% share.
Momentum for these vendors will continue to build to meet the demands of pre-exascale systems in the 10-100 petaflop range. Conway says they’re still expecting “exasize” systems to emerge in the 2020 (or slightly before) timeframe, although what the dominant architecture and power profile for these systems will look like remains to be seen. As we’ve noted here, there are numerous efforts with homegrown and commodity parts in Japan in particular that are promising candidates in the exascale race, but China and the U.S. are certainly still contenders.
IDC laid out a number of other predictions for the coming year, a few of which we’ve touched on in earlier articles. In the list below, the most prominent seem to the need for centers and enterprises to make stronger arguments about the real return on investment for handing over millions for HPC systems. One potential trigger for new investment that could be easier to justify, especially in commercial segments, is how the increased need for advanced analytics requires hardware and software tools that go beyond the basic offerings from mainstream big data vendors.
We’ll be bringing more stories this week that highlight the above trends, particular in terms of experiences in public/private industrial partnerships, the growth around ARM and low power server approaches for HPC and high performance data analysis, as well as the ability of the market to latch firmly onto the growth around data analysis to yield new market opportunities.