by Anton Shilov
02/02/2012 | 11:44 PM
Advanced Micro Devices said at its event for financial analysts that it would ensure performance leadership of its graphics processing units (GPUs) as the technology is critical for the company's further growth. But at the same time, the company is slowing down the progress of its high-performance x86 processors for enthusiast-class desktops as well as servers.
"Graphics technology ends up being the central piece of our entire roadmap. Graphics is a basic building block in terms of parallel processing capability that allows us to accelerate many-many applications. [...] This is really a secret sauce that goes into our APU line," said Lisa Su, general manager of global business units at AMD.
Since highly-parallel graphics processing units can accelerate different types of workloads dramatically, it is crucial for AMD to ensure development of the best graphics architecture possible so that its accelerated processing units (APUs) could provide better overall user experience than Intel Corp.'s microprocessors. AMD calls processing on different types of cores as heterogeneous system architecture (HSA) and is working hard to improve efficiency of this architecture.
At present AMD has the highest-performing Radeon HD 7900-series "Tahiti" discrete graphics adapters as well as highest-performing integrated graphics adapters with Fusion A-series "Llano" APUs. 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.
"End-to-end leadership in discrete graphics is something that we are very committed to and something we will continue to drive forward as this is a key market and a key way for us to leverage our IP," stressed Ms. Su.
Boosting efficiency of heterogeneous computing in general as well as heterogeneous multi-core "Fusion" chips is one of the main tasks for AMD nowadays. As a result, 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-related enhancements in addition to better-performing graphics. In particular, "Southern Islands" graphics processor of Kaveri will be able to access to CPU memory, whereas a combination of a Sea Island graphics processor with an appropriate microprocessor will allow CPU and GPU to have unified memory.
But while AMD is quickly improving performance and efficiency of APUs and GPUs, its high-end central processing units seem to stagnate. The AMD FX-series "Zambezi" processor failed to impress and it will look even more pale once Intel releases its Core i-series "Ivy Bridge" chips this spring.
AMD will launch its next-gen FX-series chip code-named Vishera with Piledriver x86 cores (that will increase instructions per clock count and thus will be faster than existing ones) sometimes in the second half of the year, some project November-December timeframe. But unfortunately for AMD, it will have to sell those chips throughout the whole 2013 and compete against Intel's code-named Ivy Bridge-E and Haswell offerings as next year the company does not plan to refresh its high-end lineup at all. Similar situation occurs with AMD's server lineup: in 2013 the company currently has no plans to introduce new server processors.
Without tangible improvements of "discrete" central processing units, AMD has to count only on performance advantages brought by its graphics processors whether in standalone or integrated variations.