⋮  

AMD x86 Zen Processors Rumored To Launch in Q4 2016 – R&D and Yield Rates Affect 14nm FinFET Node

Hassan Mujtaba
Posted Sep 11, 2015
161Shares
Share Tweet Submit

Earlier this year, AMD announced their latest x86 Zen core architecture which will power their next generation processors. The AMD Zen processors will be featured in the client and server markets with focus on overall CPU performance increases in the IPC and efficiency sector. A new rumor has been brought forward by Digitimes that mentions that AMD will be pushing the launch of Zen processors to late 2016.

AMD x86 Zen Processors Might Launch in Q4 2016 – Poor Yield Rates of 14nm FinFET Node Will Affect Production

While it was revealed a few months ago that AMD has taped out two FinFET products which include Zen CPU and Arctic Islands based GPUs, the overall production has yet to commence and will be affected by poor yield and R&D issues over at Global Foundries which are said to produce Zen chips for AMD. The Zen processors are officially planned for launch in 2016 but no exact time frame has been set in the ground. Digitimes has reports straight from the motherboard industry that AMD has planned Zen processors for a launch in Q4 2016 and that decision will put AMD in a rather difficult position.

Seeing PC demand remaining weak and its server, embedded and tablet product shipments failing to achieve major growth, AMD reported losses of US$360 million for the first half of 2015.

The sources believe Zen architecture’s late release is due to Globalfoundries’ issues with R&D and yield rates for its 14nm FinFET manufacturing process. via Digitimes

Now it is not known whether AMD is actually using Global Foundries to manufacturer their Zen CPUs but they did use their 28nm Bulk High Density process to manufacture Carrizo and 28nm Super High Performance node to manufacture Kaveri before it. While GloFo has already slipped way past their roadmap to deliver FinFET 14 nm node, AMD is once again expected to be using them for preparation of their latest processors. The problem that most motherboard vendors worry about is that AMD will be directly competing against a revised version of the Skylake Core, codenamed Kaby Lake that launches in Q3 2016. Now it is possible that AMD might actually kick start the AM4 platform with 7th generation APUs that are still based on the same design as Carrizo and are suggested to be codenamed as the Bristol Ridge series while the Zen powered lineup is part of the Summit Ridge family.

NVIDIA Gains Graphics Market Share in Q3 2016, Discrete GPU Shipments Up 39.8%, AMD Gains 34.7% - Intel and AMD Witness Slight Market Share Loss

From previous details, we know that the next update to the FX CPUs will be based around the x86 Zen core architecture and will be feature on the AM4 platform which being a totally revised design, will adopt several key technologies such as Dual Channel DDR4 memory support and a decent feature set that reflects a true update on AMD’s product portfolio. Some details regarding the new FX CPUs will include high core count with simultaneous multi-threading support which Intel has been using on almost the last 7 generations of their core family. The new Zen CPUs will deliver 40% IPC improvement over excavator with an additional 5-10% IPC that will come later in the form of Zen+. The Zen processors will also ship with a high-bandwidth and low latency cache system which will be similar to Intel’s eDRAM implementation which is a LLC (Last Level Cache) featured on the die to improve system bandwidth and offer lower latency.

All of these improvements will be true for both consumer and server/workstation class Opteron CPUs. The next generation Zen based Opteron APUs will feature high core count with multi-threading support along with disruptive memory bandwidth with higher native I/O capacity. These Zen cores will also be featured on the high-performance server APUs with multi-teraflops performance which will be added through both Zen and Arctic Islands powered discrete GCN graphics core while transformation memory architecture will be using both HBM and DDR4 high-density memory designs for higher bandwidth, delivering exponentially faster compute performance in a true HPC class APU design.

Share Tweet Submit