At first glance, the announced acquisition of Cray’s interconnect technology by Intel comes as a surprise to the HPC community, simply because Cray still stands for innovation, and with this deal they are selling one of their last proprietary assets.
At second glance, however, you find a couple of good reasons, from the perspective of both parties, which makes it a win-win situation, at least in the short-term. And Wall Street agreed. It sent Cray shares up 24 percent after-hours.
The reasons for this deal, in my opinion, are as follows:
-
The general trend to commodity components continues. For small companies like Cray (with about 800 employees), it is simply too expensive to innovate and develop sophisticated hardware such as an interconnect for exascale computing. And Intel is certainly able to take this on, especially now with all the expertise gained from the previous QLogic acquisition and now from the 74 interconnect experts moving from Cray to Intel.
-
It is certainly not easy to achieve exascale (and beyond) just with standard InfiniBand technology from QLogic. With the acquisition, Intel will own two different technologies, one for the high-end and one for the mainstream HPC market, including cloud computing. By the way, Intel is a strong promoter of the current ‘waking up the missing middle’ trend, so the mainstream HPC market might grow faster in the near future.
-
Strengthening their HPC business comes in waves with Intel. We have seen this in the past, several times, in the early 80s when the company came up with the first parallel systems (iPSC), or 10 years later with Paragon, or with leading supercomputing systems, like the Pentium-based ASCI Red. With Knights Corner, its upcoming Many Integrated Core (MIC) product, Intel will deliver a chip with the same performance range as ASCI Red, which led the TOP500 list from June 1997 to June 2000. Intel knows that HPC technology is one of the drivers for its mainstream leadership.
-
Especially in HPC, Intel has strong competitors, like AMD (which recently acquired SeaMicro and their 3D ‘Freedom’ torus interconnect), or the HPC newcomers NVIDIA and potentially ARM.
-
With Cray’s interconnect technology, Intel is able to integrate and optimize hardware components to speed-up data transfer on a more systematic basis. For example, with the PCI Express bus to link devices at the board level and the Cray interconnect for inter-processor communication, Intel could gain a competitive edge.
-
And, finally, good and bad news for the consumers. Intel’s own technology and expertise together with the QLogic and Cray assets form a great basis now for further innovation in interconnect technology. Potentially, that could ease the memory bandwidth bottleneck for future HPC systems. On the other hand, consumers will be somewhat suspicious about such a strong dominance in this field.
-
Beside technology, there are certainly economic drivers as well. Cray can reduce its workforce, without losing access to important interconnect technology, thus strengthening the company’s balance sheet. And Cray employees won’t lose their jobs.
However, a final critical remark. I can imagine that selling these interconnect assets could make Cray more vulnerable against their competitors, in the mid-term, because the company is losing a great differentiating asset (and patents), and that would be a pity. Cray now has to compete on the integration and services levels, which is more difficult to do, because every competitor tries this. Most of the HPC vendors here look alike, because all depend on the same commodity hardware and software technology providers — processor and interconnect makers, and middleware, tools, and ISV vendors.
The air is getting thinner towards exascale.