The Opteron X2100 is the first server chip to use AMD’s APU architecture.
AMD demonstrated the world’s first server implementation of Heterogeneous System Architecture (HSA) Wednesday using its new second-generation Opteron X-Series APU, called “Berlin,” at the Red Hat Summit in San Francisco.
Like Kaveri, AMD’s HSA-compatible desktop APU, the Opteron X2100 has GPU and CPU cores on the same die linked together by what AMD calls the heterogeneous unified memory architecture (hUMA). This allows process threads to run simultaneously across GPU and CPU cores through a process called heterogeneous queueing (hQ). The performance advantage is found in the CPU no longer having to queue up the GPU with instructions, allowing the GPU’s strength in parallel processing to reach its maximum potential.
“As servers adapt to new and evolving workloads it’s critical that the software ecosystem support the requirements of these new workloads,” AMD’s server unit general manager Suresh Gopalakrishnan said in a press release. “We are actively engaged with a broad set of partners in the data center software community who are bringing to market the software infrastructure to seamlessly enable x86 APU based servers.”
On stage at the Red Hat Summit the Opteron X2100 was running Oracle’s Project Sumatra on Fedora Linux, an update to the Java Virtual Machine that allows Java programmers to write heterogeneous code across the GPU and CPU. It also allows code that’s not specifically written for the GPU to run on the GPU (with an understandable performance hit) with its parallel stream processing functions. Project Sumatra was first announced last November at AMD’s APU developer summit in San Jose.
AMD plans to ship Berlin processors within the first half of this year.