CHIP DESIGNER AMD has confirmed that its ARM based Opteron accelerated processing units (APUs) will use AMD's own GPU designs rather than ARM's Mali GPU.
AMD's decision to design ARM based Opteron processors will eventually lead to APUs that include graphics cores for vector computation. The company told The INQUIRER that the graphics core will be based on its own technology rather than licensing ARM's Mali GPU but said its first ARM products will be CPU chips.
ARM has been touting the full profile OpenCL capabilities of its Mali GPU for a few months but despite AMD buying a CPU licence for its upcoming ARM based Opteron chips, the firm said it will stick with its own GPU when it eventually designs an ARM based Opteron APU. However the firm said workloads that are expected to be deployed on ARM servers are not presently making use of vector processors and that will delay the launch of an ARM based Opteron APU.
John Williams, VP of server marketing and business development at AMD told The INQUIRER, "The initial ARM product will be a CPU, we will have future products including X86 that will be APUs. The products of that will be APUs using graphics IP that will be AMD's graphics IP, because it is really focused on the vector compute aspect rather just graphical imaging."
"The workloads that can accurately utilise APU capabilities those are as far along as the workloads that can adequately utilise ARM CPUs. You'll see the ecosystem developing equally as rapidly for APUs but lag a little bit behind that of ARM CPUs," said Williams.
AMD's decision to use its own GPU designs is not surprising but the firm wouldn't confirm this during its initial announcement. The move will see it effectively do what Nvidia has been doing with its Tegra chips, meaning that the two long time rivals in the graphics market might well go up against each other in the far more lucrative server market within the next three years. µ
Sign up for INQbot – a weekly roundup of the best from the INQ