AMD HAS ANNOUNCED that its Mantle graphics API is "entering a new phase of its development cycle" and launched a private beta programme for the Mantle SDK.
The firm supported the move by launching a new portal to give game developers the chance to experience "the benefits of low-overhead graphics APIs" for themselves.
AMD claimed there is "unprecedented demand" for the programme and said 40 development studios have already preregistered for the private beta.
"This new phase was prompted by the Mantle working group's unanimous decision that Mantle has rapidly achieved all necessary stability, performance and functionality milestones required to ready it for a broader audience in the developer community," AMD said.
"Together, AMD and this formidably large community of experienced game developers will not only shape the future of Mantle, but pioneer best practices for working with the low-overhead graphics APIs broadly emergent in our industry."
AMD is also inviting additional developers to apply via the portal to claim their spots in the next wave of invitations. However, at the moment the portal is password protected, as only a limited set of developers have access to the Mantle NDA Developer SDK and access is subject to a selection process.
Developers can register their interest to be part of the programme by emailing email@example.com with the name of their company, name and email of contact point for Mantle access, game title(s) or codenames for which they are interested in evaluating Mantle, and their reasons for requesting Mantle access.
On Tuesday, AMD launched its third generation Mullins and Beema accelerated processor units (APUs), promising "better performance at lower power".
As the successors to the Kabini and Temash APUs, which AMD released last year and were the first to deliver quad-core x86 performance, the firm's third generation Mullins and Beema APUs have four x86 cores codenamed Puma, following on from last year's Jaguar CPU cores to reduce power draw and improve performance while retaining the same chip architecture. µ