AMD received a substantial coverage at VentureBeat recently, with Dean Takahashi interviewing key executives in succession. First off was an in-depth interview with Lisa Su (CEO and Chairman of AMD), followed by an interview with Raja Koduri, Head (CEO?) of Radeon Technologies Group. RTG is AMD’s spin-off which you can compare with the spin-off of manufacturing division you now know as GlobalFoundries, just without a strong sovereign wealth fund (like Mubadala Development Company)… for now.
In order to execute on a huge market opportunity in the form of Virtual Reality, RTG wants to make sure all the basics are covered. For starters, one of more painful episodes from the company history was the launch of APU – Accelerated Processing Unit. Codenamed ‘Llano’, the first APU suffered from horrible yields and mass-market unavailability, causing a significant loss in profits and ultimately, a write off. That write off caused a lot of Llano units to be destroyed – yes, perfectly functional products had to be destroyed in order to make room for new models. Now, AMD is shifting its focus back on ‘Future is Fusion’, and probably does not want to rely on a sole foundry.
Remember a few years ago when we talked about the Future is Fusion? Now we’re back to the Future is Fusion again, because of the performance requirements, the stability requirements. We organized this idea called fast pixels, deep pixels, and immersive pixels.
With the Polaris Architecture, AMD is moving all of its production silicon – APU, CPU and GPU to FinFET transistors, but 14nm is not going to be the only manufacturing process the company intends to use. AMD calls its first two demoed products ‘Polaris 10’ and ‘Polaris 11’, and according to our sources, the difference will be in manufacturing process and the foundry themselves. We are lead to believe that GlobalFoundries scored a deal to manufacture APUs and mainstream GPUs (11, 1x), while TSMC could manufacture ‘the big die’, a high-end product which should go head to head against NVIDIA’s GP100 chip based on Pascal architecture. As Koduri explains:
When we set to design this GPU, we set a completely different goal than for the usual way the PC road maps go. Those are driven by, the benchmark score this year is X. Next year we need to target 20 percent better at this cost and this power. We decided to do something exciting with this GPU. Let’s spike it so we can accomplish something we hadn’t accomplished before.
The target we set was to do console-class gaming on a thin and light notebook. What does that take for the GPU in terms of power and configuration? I’m proud to say we’ve accomplished that goal with this GPU.
While this is in domain of rumors right now, it looks like the 2016 AMD/RTG FinFET line-up will be 14nm GlobalFoundries for APU and GPU (Polaris 11) and 16nm TSMC for ‘Greenland’ (Polaris 10). Raja wasn’t shy to explain why AMD feels very confident in its lineup:
We have two versions of these FinFET GPUs. Both are extremely power efficient. This is Polaris 10 and that’s Polaris 11. In terms of what we’ve done at the high level, it’s our most revolutionary jump in performance so far. We’ve redesigned many blocks in our cores. We’ve redesigned the main processor, a new geometry processor, a completely new fourth-generation Graphics Core Next with a very high increase in performance. We have new multimedia cores, a new display engine.
AMD is expected to reveal its Polaris-based products in June 2016, most likely targeting the middle of the year with introductions at Computex Taipei (May 31- June 4) and E3 (June 14-16) in Los Angeles, CA. By that time, expect many more rumours speculating what the AMD line-up will be consist of. We know that high-end part will use HBM2 memory, while the mainstream parts will utilize GDDR5.