The market for server processors is changing, creating openings for new vendors. With the emergence of mega data centers and cloud computing, server economics no longer focuses on capital expenses alone. Demand for ultimate performance from a single processor has been replaced by a balanced view of capital and operating costs. Performance per watt and performance per watt per dollar are the new metrics driving purchasing decisions in large data centers. Physical density is also growing in importance, driving greater scalability and new form factors such as microservers that pack more nodes into precious rack space.
The market is moving to a new era where backward compatibility is less important than before and innovation takes the front seat. Intel and AMD—the incumbent vendors—continue to innovate and advance their Xeon and Opteron designs, respectively. Integration, microarchitecture advances, and process technology are the primary factors when evolving these x86 processors. But new entrants are eyeing cloud-computing environments as an opening for radically different architectures and more-power-efficient CPU designs. With the merchant server-processor market exceeding $7 billion, success requires taking only a few percentage points of share from Intel.
Having reached practical power limits, server-processor designers are increasing performance primarily by adding cores rather than increasing clock speeds. Mainstream x86 processors now offer 16 cores per chip, while startup Tilera is already shipping 36-core processors. Mainstream server processors are currently using 32nm technology, although Intel is shipping its first processors using the 22nm node. By moving to a finer geometry process, vendors get more transistors in the same die area and power envelope. This additional transistor count can be used to add CPUs or to increase cache sizes. Larger caches increase performance by absorbing DRAM latency, which is not decreasing as rapidly as processor performance is growing.
Following the earlier integration of memory controllers, processors are now integrating PCI Express (PCIe) controllers. This step eliminates one system-logic component, the “north bridge,” reducing the chipset to a single “south bridge” chip. For desktop and workstation applications, processors are also integrating the graphics controller. New entrants like Calxeda, Marvell, and Tilera offer system-on-a-chip (SoC) designs that integrate Ethernet controllers in addition to memory and PCIe controllers. For high-density server form factors, this additional level of integration provides differentiation.
Intel offers the broadest line of server processors, which are built using its industry-leading process technology. The company’s current two-socket design is the Xeon E5-2600 platform (Romley) based on the 32nm Sandy Bridge architecture. New to Intel’s Xeon lineup is a four-socket version of this platform, the E5-4600. This new platform offers a lower-cost alternative to the Westmere-based Xeon E7 line, which is designed for scale-up servers with four or more sockets (4P and above). By adding reliability, availability, and serviceability (RAS) features to the E7 line, Xeon processors now serve mission-critical designs that formerly required Itanium (IA-64) processors.
Using its tick-tock development approach, Intel shrank Sandy Bridge to produce the 22nm Ivy Bridge architecture. The first server processors to use Ivy Bridge comprise the Xeon E3-1200 v2 line, which serves uniprocessor designs. By the end of 2012, Intel plans to introduce a server processor based on its low-power Atom architecture. At six watts, this chip will compete with new entrants fielding ARM-based designs.
AMD’s Opteron line of server processors offers a good alternative to Intel for high-volume two- and four-socket platforms. With its 16-core Opteron 6200 (Interlagos), AMD continues to lead in core count for x86 processors. The 32nm Opteron 6200 processors are based on a novel microarchitecture called Bulldozer that improves density. The company also offers Opteron 4200 and 3200 processors based on this design. In 4Q12, AMD introduced the Opteron 6300 Series (Abu Dhabi), which uses an update to Bulldozer called Piledriver. AMD has also licensed ARM’s Cortex-A57 CPU and plans to introduce processors using this core in 2014.
New entrants fielding non-x86 server processors are targeting cloud computing, where compatibility requirements are more manageable compared with traditional enterprise applications. Most of these new vendors implement the ARM instruction set using either licensed or custom CPU designs. Marvell was first to market with a four-core ARM processor suitable for servers. Startup Calxeda added unique server features to its quad-core ARM processor. Other vendors are skipping 32-bit ARM designs and instead using the new 64-bit ARMv8 instruction set. AppliedMicro and Cavium are designing custom 64-bit CPUs for their respective server processors, whereas ARM is licensing its new Cortex-A57 (Atlas) core to vendors that prefer an off-the-shelf CPU design.
One vendor, Tilera, relies solely on open-source Linux distributions and open-source or customer-ported applications. The company is shipping 64-bit server processors with up to 36 proprietary CPUs connected using a unique on-chip mesh network. The startup’s chips deliver a several-fold improvement in performance per watt compared with x86 processors.