Intel’s Response to ARM Servers: Xeon D Processors with a Twist
Intel Xeon D processors will now contain FPGAs, which directly combats the upcoming ARM-based server-grade processors.
ARM processors may not seem like the dominant processor architecture on the market today, but they exist in almost every phone or tablet sold.
They're also used in data centers, and ARM processors designed specifically for servers are expected to be right around the corner.
Intel, in an attempt to remain influential in the field of servers, have recently been updating their Xeon D processor line. In upcoming iterations, however, there will be something that hasn’t been seen before with Intel processors: a Xeon FPGA System-on-Chip (SoC).
What does an FPGA bring to the table?
An FPGA is a Field-Programmable Gate Array, which allows the user to create logic circuits by writing software. The software is compiled and then programmed onto the FPGA, which can be used in any number of ways. Currently, ARM processors are the only ones seen with this technology, as there are SoC ARM-FPGA chips.
Intel recently acquired Altera, one of the two main FPGA manufacturers, which currently has some SoCs available with FPGAs coupled with an ARM processor on the same chip. There are a few development boards available with Intel-based processors with an FPGA on a development board, but they are separate components.
The amount of flexibility an FPGA can give to the end application is immense; one chip could be used in a multitude of configurations, such as allowing for multiple PCIe lanes for data storage, more ethernet lines, or even more cores. It's even possible to create another processor on the FPGA, such as Altera’s own NIOS II processor.
FPGAs have a finite amount of space and are limited by the clock speed as well, although both limitations are usually not met in most applications. If you wish to learn more about FPGAs, click here.
Where will these new processors bring computing?
This technology is not new, but an FPGA coupled with a powerful Intel Xeon D processor could bring forth a massive amount of computing power in not only servers, but workstations and research applications. Since this SoC is reconfigurable, getting all the peripherals required for any application is as easy as just getting the code online, or writing it to program the FPGA.
These processors would be incredibly versatile. Intel provides some example scenarios, such as machine learning applications using the FPGA as a hardware-based accelerator to aid the Xeon D processor. At present, this kind of hardware acceleration would only be available when used with something like a GPU, which is usually a large, space-consuming device. However, with an FPGA, all of this is on the same chip, which is already a huge space reduction (not to mention a possible power saver, as well).
Here is a presentation outlining some more applications for how a data center could use this Xeon FPGA SoC.
Block diagram of Intel's new Xeon FPGA SoC with some specifications
The outlook for ARM-based server processors might be a little glum after this, as the RISC (Reduced Instruction Set Computing) vs. CISC (Complex Instruction Set Computing) battle continues on to this day, and now into the realm of servers. In the coming months and years, we will see how Google’s intended use of ARM processors in data centers, Intel’s new SoC, and Qualcomm’s eventual release of ARM server-grade processors play out. The future looks promising, wherever it may go.