The worlds largest chip supplier has dealt another hand of power with the launch of their new Xeon Phi coprocessors, the new wave of chips are based on the many-core MIC micro-architecture for the high-performance computing market and launched on Monday of this week.
Now many of you will no doubt be thinking "what are these?", well to put it simply, this generation of super powered CPUs are already featured in 7 of the worlds super computers and offer a solution to the high processing demand of today's biggest technical problem solvers and scientific applications. With Xeon Phi being able to handle highly parallel processing to calculate weather pasterns, develop cures, industrial simulations and extensive physics research are a few things that just scratches the surface of applications of this technology
Intel says these new chips will lay the foundation for "Exascale computing", marking a 1000x the computational capabilities over today's Petascale systems, which is quite simply a near unimaginable amount of processing power for most mortals to even comprehend in their life time and a serious achievement in computer sciences.
Integration of systems should be fairly straightforward also, relatively speaking of course. This is thanks to their abilitiy to use many familiar programming languages, models, techniques and dev tools that already exists for the x86 architecture. Meaning that users can take advantage of Intels software tools which will help scientists, engineers and more optimize their coding to take full advantage of these modern coprocessors.
The Intel Xeon Phi 3100 can serve up a whopping 1TFLOPS double-precision performance, with support for up to 6GB memory at 240GB/s bandwidth all withing a 300w TDP. The 5110p will feature 60 cores with 4-way simultaneous multi=threading and 512KB of L2 cache (per core) @ 1.01TFLOPS double-precision performance, support for 8GB GDDR5 memory at 320 GB/s and only 225w TDP.