AMD’s ready to unleash their next generation GPUs and it will happen before the end of 2018. These GPUs will based on improved Vega architecture (which is currently in use) and built on the 7nm process. AMD claims around x2 higher 7nm VEGA die density compared to the 14nm VEGA density, twice the power efficiency (for the same performance?) and x1.35 the performance (same power specifications? transistor count? price point?).
AMD has been able to compete in the performance department in most price points. However, the biggest problem in the laptop scene is its GPUs’ lack of high enough power efficiency which results in lower battery running times, making laptops with AMD gpu solutions less attractive, even for lower price. Perhaps the move to the new 7nm node will help tackle this issue, and perhaps we’ll have to wait to next generation “Navi” architecture.
With this generation, AMD will also start using its open standard technology for connecting between GPUs (and perhapsother components), the XGMI (inter-chip global memory interconnect), the direct competitor to nVidias’ NVLink which is proprietary, and it’s based on the next generation PCIe version 4 standard. This move is inline with other open-standard and open-software initiatives AMD takes, like the new Radeon ProRender that is aimed to allow implementing ray tracing in a more open and standardized way, as well as more efficient (at least on AMD’s hardware). Like many others, AMD’s investing in implementing software and hardware more suitable for the field of machine learning and it remains to be seen how the new hardware and software allow better performance and ease of use for this case.