SAN JOSE, Calif. – Altera's deal in June to use Intel's 14 nm FinFET process took an interesting twist when the FPGA designer announced today it will pack ARM's 64-bit cores into its chips. That means Intel will fabricate starting in 2014 high-end Altera Stratix 10 parts that use four ARM Cortex-A53 cores.
Altera says the Stratix 10 devices -- which will also have embedded DSPs and other logic -- will be its highest performance parts to date. They target a broad range of apps from search engine accelerators to communications data plane processors and radar guidance and security processing.
The deal may give Intel pause about its outsourcing strategy. To date the chip giant has experimented with offering its leading-edge fab processes as foundry services to a handful of chip designers, Altera being one of its largest planned customers to date.
Intel is also experimenting with licensing its x86 core. It made a brief effort several years ago to license its mobile Atom through TSMC, but it failed to gain traction. More recently it announced it will license its even lower end Quark x86 though it has not disclosed terms or any initial customers.
Industry watchers will be interested to kick the tires on the combination of Intel's leading-edge process and ARM's 64-bit cores. Altera claims the Stratix 10 parts in Intel's 14nm process will provide FPGA "performance over one gigahertz, two times the core performance of current high-end 28-nm FPGAs."
The ARMv8 is aimed primarily at driving the architecture into servers and communications, markets where the x86 has dominant and rising positions, respectively, and ARM is starting from nearly zero. To be a credible player in these markets, ARM added hardware virtualization and ECC protection on L1 and L2 cache designs, features the x86 has long supported.
Altera noted the Cortex-A53 core can run in 32-mode, which will run Cortex-A9 operating systems and code unmodified. The backward compatibility will ease the transition to ARMv8 for Altera’s 28 and 20nm SoC customers, it said.
Both Altera and ARM are focusing on the OpenCL programming environment for parallel programming jobs. They see OpenCL as their best shot as a de facto standard in a fragmented sector of options from Microsoft (DirectCompute), Google (Renderscript), and Nvidia (Cuda).
In terms of fab processes, Intel is seen as a node ahead of the competition. It is now ramping up 14nm FinFET process. Foundry rival TSMC is ramping up its 20nm process now but does not expect to have a version with FinFETs ready for commercial use until late next year.