by Anton Shilov
04/16/2012 | 04:54 PM
Even though in the recent years central processing units (CPUs) from Advanced Micro Devices were competitive on the mainstream markets, the company virtually ceased to compete for high-end desktop systems. Although we will see higher-performance AMD's microprocessors with Piledriver cores this year, the company will only truly introduce something radically better with Steamroller generation of chips, which are due in 2013 - 2014 timeframe.
According to a report from VR-Zone web-site, which cites sources with knowledge of AMD's plans, "there will be substantial changes in both cores and system architecture from Steamroller onwards, that should help make AMD competitive closer to the top". Some executives from AMD have reportedly acknowledged low instructions-per-clock (IPC) performance of the Bulldozer micro-architecture, something that clearly affected overall performance of AMD's latest chips.
While no exact details about the improvements of the Steamroller family of microprocessors are known, it is expected that both mainstream and high-performance platforms featuring Steamroller-class processing engines will support new technologies, such as PCI Express 3.0, more DDR3 memory channels, better integration between CPU and GPU chips and other improvements.
What should be kept in mind is that AMD's very first code-named Kaveri accelerated processing units with Steamroller x86 cores that will become available in 2013 will not feature fully-fledged Steamroller cores, but something better than Piledriver, but less advanced than the Steamroller cores found in the next-gen server and desktop platforms. AMD's current plans to not include Steamroller-class high-performance CPUs for next year, hence, the company will only be able to get back to high-end CPU-centric systems earlier than in 2014.
Another important factor that should be considered is that the Steamroller-based central processing units for high-performance servers and desktops [due in very late 2013 or in 2014] will compete not against Intel's current-generation or next-generation processors, but against server offerings based on Haswell and desktop offerings powered by Rockwell/Broadwell, which will be made using 22nm and 14nm process technologies, respectively.
In the meantime, boosting efficiency of heterogeneous computing in general as well as heterogeneous multi-core "Fusion" chips in particular will be the two main tasks for AMD. The company's next-generation discrete Radeon HD "Sea Islands" family of chips as well as future-generations of APUs code-named Kaveri and Kabini will feature numerous HSA [heterogeneous systems architecture]-related enhancements in addition to better-performing graphics and x86 cores.
Right now there around 200 apps that can be accelerated by stream processing units of GPUs, but several years from now that number will increase and the compute performance of GPU cores will be critical for success on the mass market. Still, x86 performance clearly remains an important factor, but the breakthrough in that direction may occur only in two years down the road.