AMD has confirmed that it will be using GlobalFoundries to manufacturer its next-generation Polaris GPUs, counter to previous rumours that said it may turn to rival silicon fabricator Samsung for the task. However it won’t be giving GlobalFoundries the entire task. Its statement suggests that the manufacturer will handle the low-to-mid-range GPUs and mobile chips, while TSMC will handle the high-end GPUs.
Although the lack of a deal with Samsung is surprising, the overall makeup of this next-generation manufacturing is quite typical for AMD. It’s also making the best of both worlds by splitting its manufacturing to exploit TSMC’s experience with GPU manufacturing, while keeping its eggs in many baskets to avoid any supply issues.
However it may not be that Samsung is completely off of the table. Since GlobalFoundries will be making use of Samsung’s FinFET design as part of its manufacturing the Zen CPUs and lower-end graphics chips, it may be that production could shift over to the Korean company if supply problems arise (according to ExtremeTech).
Polaris is expected to be a big leap forward in terms of performance and energy efficiency. One of the continuing issues with AMD hardware over the past few years has been that while its raw power could keep up with Nvidia in most respects, its cards were often far less efficient, and much hotter.
With this next generation AMD is looking to change that, promising to double performance per watt. With the potential use of HBM and HBM 2, we should see huge increases in memory bandwidth as well.
Polaris GPUs are expected to arrive around the mid-point in 2016, with Zen CPUs and its accompanying chipsets said to arrive a little later.
With Nvidia’s Pascal technology not too far away, either, which company’s hardware are you most excited by?
- Next-gen AMD and Nvidia GPUs just moved one step closer to launch
- 8 best AMD GPUs of all time
- AMD predicts 700-watt GPUs before 2025, but it has a plan
- AMD’s new tool compares its GPUs to Nvidia’s, but there’s a catch
- AMD’s 7000-series GPUs may be its most power-hungry cards yet