CHIP DESIGNER AMD said its collaboration with Oracle to bring improved GPGPU support in OpenJDK is a preparatory step prior to the launch of an Opteron accelerated processor unit (APU).
AMD has bet the farm on its APUs, the concept of mating a CPU with a GPGPU on the same die, and it has seen favourable uptake in the consumer market since Llano launched over a year ago but has yet to launch an Opteron APU. Now the firm has admitted that its collaboration with Oracle on improving GPGPU support in OpenJDK is one of many steps it needs to take prior to the launch of an Opteron APU.
As AMD works on improving OpenJDK performance with its APUs, the real benefit comes in the server market where Java is widely used for presentation layer applications and middleware. Should AMD be able to engineer GPGPU support in the JVM, it could make an Opteron APU a very attractive processor for the many companies that rely on Java.
Margaret Lewis, director of software at AMD's server business told The INQUIRER, "If you had a server based APU you have to make sure you had a rich software ecosystem for it, because if you brought that type of hardware out you would want to have software that takes advantage of it from the get go. [...] There are a lot of our server customers that are starting to play with our client APUs to get ready for the concept of a server APU and we also have discrete graphics cards in servers for the same purpose."
Lewis added, "It would make sense that we would start working on the ecosystem in preparation for that server APU type of timeframe and so this would be one of many steps."
AMD has said an Opteron APU is in the pipeline, an admission that should come as no surprise as it is about the only unique selling point AMD has against Intel's all conquering Xeon chips. AMD's problem is that its work with Oracle is not likely to yield production results until sometime between the release of Java 8 and Java 9, the latter of which has yet to receive a launch date. µ
Sign up for INQbot – a weekly roundup of the best from the INQ