Although Advanced Micro Devices wants to cease using special process technologies for its products in order to cut-down their manufacturing and development costs, it looks like next year the company will have to use GlobalFoundries’ 28nm super high-performance (SHP) technology for its Radeon graphics processors in order to ensure their higher performance.
At present Advanced Micro Devices uses GlobalFoundries’ 28nm super high-performance (SHP) fabrication process to make code-named Kaveri accelerated processing units (APUs), but going forward the company plans to use the technology for other products as well. BitsAndChips web-site reports that in 2015 the company intends to use the 28SHP technology to make graphics processing units.
The 28nm SHP manufacturing technology is designed to enable high clock-rates at relatively low voltage levels. Therefore, it can allow AMD to boost frequencies without increasing power consumption. Moreover, since 28nm SHP also sports higher transistor density, it also helps to increase the amount of execution units compared to GPUs made using TSMC’s 28nm process technology without increasing sizes of the chips and thus manufacturing costs.
Previously it was reported that AMD intends to make code-named Bermuda and Fiji graphics processing units at GlobalFoundries using 20nm process technology. While the company does plan to use 20nm manufacturing processes, it looks like it may use 28nm SHP technology as well. Unfortunately, it is now not clear which AMD GPUs will be made where and using which process technology.
Separately, the web-site reports that AMD plans to use TSMC’s 16nm FinFET manufacturing technology in the first half of 2016 for its upcoming K12 and Zen central processing units.
AMD and GlobalFoundries did not comment on the news-story.
Discuss on our Facebook page, HERE.
KitGuru Says: Given that the information comes from an unofficial source, take it with a pinch of salt. AMD has publicly said that it wants to unify fabrication processes it uses in order to lower development costs by re-using IP across different product lines. As a result, it makes absolutely no sense to use TSMC’s 20nm SoC and GlobalFoundries’ 28nm SHP at the same time. It makes a lot more sense to use only one process technology for different kinds of products. At least, from AMD’s perspective it outlined for a number of times in the past 18 months.