Published in Processors

AMD announces glorious five year plan

by on30 March 2015


GPU and APUs for everyone

AMD shared its glorious five year long GPU and APU roadmap with the world for the first time.

AMD's Junji Hayashi revealed AMD's cunning plan at a special event in Japan, according to WCCFtech.

As we expected AMD's upcoming ARM K12 and its sister x86 CPU core Zen got a mention AMD wants to develop and introduce both x86 and ARM powered SOCs to the market in a pin for pin compatible platform code named SkyBridge.

According to the latest reports, AMD's upcoming K12 and Zen will support multiple threads. Basically this means that AMD will move away from clustered multi-threading used on Bulldozer-based processors and move to simultaneous multi-threading instead. This is good news on more than one front.

Simultaneous Multi-Threading 

Simultaneous Multi-Threading (SMT) will allow AMD to boost utilisation and efficiency. With SMT, underutilized parts of the CPU core can be put to use via a secondary execution thread. The end result is superior resource utilization.

This is a contrast to clustered multi-threading which shares resources between two different CPU cores, instead of doing it inside a single CPU core.

AMD never really commented about the how much multi-threading support its upcoming cores would have. Now it seems that AMD's K12 ARM core will support "many threads" instead of just supporting one additional thread as is the case with Intel's high performance CPUs.

When it comes to GPUs, Hayashi said AMD will be employing a two year cadence to updating its GPU architecture inside APUs.

AMD GPU Roadmap 2015-2020 AMD will introduce Accelerated Processing Units with updated GPU architectures once every two years.

AMD plans to introduce what it described as a High Performance Computing APU or HPC for short. This APU will carry a sizable TDP between 200 and 300 watts.

APU loves HPC when it is stacked

This sort of APU loves HPC. So far powerful APUs were not attempted because the amount of memory bandwidth required to keep such a powerful APU fed was too much.

Stacked HBM (High Bandwidth Memory ) will make such designs extremely effective. The second generation of HBM is nine times faster than GDDR5 memory and 128 times faster than DDR3.

Code names for future GPU architectures unfortunately were not revealed. But it is pretty likely that AMD's upcoming GPU architecture to debut on 16nm FinFET will be code named Arctic Islands. We'll know more details in May during the company's scheduled Financial Analyst Day event.

Last modified on 01 April 2015
Rate this item
(21 votes)

Read more about: