Compared to general purpose versions, processors custom-designed processors at the foundry level are synonymous with substantial performance gains for a few privileged clients eligible because of their size or pocket depth. It’s also very trendy and in any case quite logical considering the rise of Big Data, Cloud computing and the size of data centers.

Officially announced at ISC’14, the combination of an FPGA (almost certainly from Altera) with a Xeon on the same chip should also allow Intel to slow the progression of ARM and POWER, which are now considered as credible alternatives to x86 for compute-or data-intensive processes and become, therefore, of particular interest to large operators.

By hardwiring a certain number of low and higher level operations, FPGAs can boost performance in the range of 10x to 30x, depending on the nature of the code, while maintaining binary compatibly with Xeon cores. The downside is of course the complexity of development because this kind of platform requires both HDL (for designing the logic and programming the component) and application level skills (to integrate these specialized pieces of hardware into codes). But the major operators can afford this highly sought-after expertise. Moreover, as Intel officials have indicated, such hybrid Xeons will allow for the testing of new, specialized operations which, if proven relevant and effective, could be integrated into future iterations of Xeon targeting mere mortals.

The actual date of availability for this new breed of CPUs hasn’t been specified, but it will probably be next year. As so often happened with the blue team lately, we are a bit frustrated regarding technical details but what’s important to note is the confirmed trend towards integrated hybridation. As everyone is striving towards scalability, generic and specialized cores tend to multiply on the same die, with the repeatedly expressed goal to increase performance and energy efficiency – if possible simultaneously.