As GPUs get smaller and faster, there’s always room for error, and new technologies don’t always impress as much as they should. Word on the street is that as AMD prepares its next series of GPUs, the smaller 20 nano-meter production process they’d planned on using may not be panning out. Of course if that were the case, AMD would want to keep it on the down low, so for now it’s all speculation, but there are a few cues that 20nm dies may be passed over entirely for 14nm.
Users are always expecting each new set of GPUs to be a significant improvement over the previous generation, expect it to pull less power, and fit in a smaller form factor. Rumors coming out of the manufacturers are pointing to 20nm chips not operating stably, or at all, when faced with the power draw and stress they’ll face when used to produce a GPU.
Meanwhile, silicon die manufacturer TSMC has stepped up production and research of 16nm FinFET chips, with a launch planned for later in the year. That means we could be seeing a step up for AMD’s Arctic Circle chips in late 2015 or early 2016, depending on demand for the new chips.
That being said, AMD may decide to stick with the 28nm die for the time being, since there’s already infrastructure for it in place. An updated series of the same size chip wouldn’t be a too much of a backstep for the brand, and would allow more development time for the smaller chips in 2015. Its prime competitor in the video card space, Nvidia, has already committed to 28nm, as its most recent GTX 900 series uses that production process – just like the previous 700/800 GPUs.
Either way, AMD won’t keep us waiting too long to hear about the plan for its next series of GPUs. They have a number of events scheduled for the next couple of weeks, with the annual stockholder’s meeting on April 29th, and a Financial Analyst day shortly thereafter.