Samsung Has Designed a New Mobile GPU Architecture That Could be Used in Smartphones and Supercomputers, Says Analyst
Earlier this month, Samsung was reportedly making in-house GPU solutions, but they would be featured in affordable devices to create higher margins for the company while being sold at competitive prices. However, the latest development of Samsung’s quest to make in-house silicon has landed at a very interesting stage, because the company is not just focusing on affordable smartphones with their own GPUs, but future flagships branded with Samsung’s name will also be getting their own solutions, and they will be packing quite the punch.
Samsung’s Work on Its New Graphics Processing Unit Is the First New Design Seen in a Decade, According to a Person Familiar With the Matter
According to an analyst briefed on the work being carried out, Samsung’s GPU will feature a novel architecture that could allow it to work for a wide range of applications, ranging from smartphones to supercomputers. This news is coming at a time where there is growing competition in the smartphone graphics processor segment. Apple unveiled its first ever custom GPU with the launch of the iPhone 8, the iPhone 8 Plus and the iPhone X.
It was revealed earlier that the Exynos 9820 would feature its own GPU, but it looks like Samsung is going to be relying on ARM’s Mali-G76 GPU for the time being. According to Jon Peddie, who first broke the news that Samsung was working on this custom GPU praises the Korean giant’s efforts as he states the following:
“Put it on par with Apple. The only question is where and when Samsung’s GPU will show up. This design is so good they could deploy it in every platform — it’s a function of their ambition. If I owned it, it would be in everything including cockpits and supercomputers.”
This new GPU design is expected to be present in the next Samsung custom chipset, which will obviously mean the Exynos 9820, but we have been hearing opposing rumors so far, so remember to take this information with a pinch of salt for now. It is also unclear if Samsung is going to be keeping this technology exclusively for its own chipsets, or whether it will license it out to other companies.
To achieve a new echelon of performance per watt, EETimes reports that the architecture bundles multiple instructions into a group that can be executed in a single cycle. However, it does not use the sorts of VLIW techniques that have fallen out of fashion because they generate other processing overheads.
The GPU design team is being led by Chien-Ping Lu, a graphics veteran who started his career at NVIDIA. It is also reported that Samsung could find itself a new avenue of resource generation if it ends up licensing this technology to phone OEMs, creating lots of trouble for Qualcomm and its SoCs.
Are you excited for the future launch of the Samsung custom-made GPU? Tell us down in the comments.