Intel announced a set of new enterprise products today aimed at furthering its strengths in the TOP500 supercomputing market. As of today, the Chinese Tiahne-2 supercomputer (aka Milky Way 2) is now the fastest supercomputer on the planet at roughly ~54PFLOPs. That's double the speed of the old leader, the AMD
powered Titan at Oak Ridge National Laboratory. The new system draws significantly more power than the Titan, (17.8MW vs. 8.2MW) but overall power efficiency is fairly similar. The Titan's GFLOP/W ratio is 2.143, while the new Tianhe hits 1.935.
Intel is putting its own major push behind heterogeneous computing with the Tianhe-2. Each node contains two Ivy Bridge sockets and three Xeon Phi cards. Each node, therefore, contains 422.4GFLOP/s in Ivy Bridge performance -- but 3.43TFLOPs/s worth of Xeon Phi
. The size of that gap illustrates why Intel felt building Xeon Phi was a worthwhile investment -- it gives the company a chance to keep its own x86 architectures at the forefront of super computing technology, rather than being reduced to the server hardware that advanced supercomputing platforms rely on.
Speaking of Ivy Bridge, we'll see new Xeons based on this technology later this year, in the 22nm E5-2600 V2 family, with up to 12 cores. The new chips will be built on Ivy Bridge technology (the current E5-2600 family is Sandy-Bridge based) and will offer up to 12 cores / 24 threads. That's significantly wider than the SNB-based hardware, which tops out at eight cores / 16 threads. The new Xeons, however, aren't really the interesting part of the story. Today, Intel is adding cards to the current Xeon Phi lineup.
New Xeon Phi's Flesh out Card Family
For the past year, Intel's Xeon Phi, like the original Model T, has shipped in just one flavor. The standard Xeon Phi 5110P is a 60-core chip running at 1053MHz with 8GB of RAM, 16 memory channels, and 320GB/s of aggregate bandwidth. Single precision performance is 2TFLOPs, double precision tops 1TFLOP. Today, Intel is introducing several new versions of the card -- the 7120P, 3120P, 3120A, and 5120D.
The 3120P and 3120A are the same card -- the "P" is passively cooled, while the "A" integrates a fan. Both of these solutions have 57 CPUs and 6GB of RAM. Intel states that they offer ~1TFLOP of performance, which puts them on par with the 5110P that launched last year, but with slightly less memory (and presumably a lower price point). At the top of the line, Intel is introducing the 7120P and 7120X -- the 7120P comes with an integrated heat spreader, the 7120X doesn't. Clock speeds are higher on this card, it has 61 cores instead of 60, 16GB of GDDR5, and 352GBps of memory bandwidth. This type of expansion will give Intel more flexibility in meeting customer workload demands. The 16GB of RAM for the 7120 family makes far more memory available to each core, though we expect the price to jump commensurately with the additional memory.
If the first Xeon Phi cards were a foot in the door, these new models are clearly aimed at improving the use-case in specific scenarios. Customers who need lots of cores and not much RAM can opt for one of the cheaper 3100 cards, while the 7100 family allows for much greater data sets. Intel hasn't taken the overall efficiency lead, but the Tianhe-2 is close on the heels of the Oak Ridge Titan in that regard.It's a strong showing for Intel and for Xeon Phi, particularly given the fact that Intel has only been playing the many-core game for 12 months.