Renowned chip designer Imagination Technologies has received massive funding to launch its operations and develop native technologies to meet the demand for AI.
Imagination is responsible for the PowerVR GPU architecture, the company’s technology is still present in 12 billion devices and is now planning a major resurgence
It won’t be wrong to categorize Imagination Technologies as the “NVIDIA” of the late 20th century, as the company is known for its PowerVR GPUs, which saw massive adoption in both the desktop and mobile markets, with companies like Apple being a key partner for the GPU maker. However, things started going south for the company in 2017 when Apple decided to ditch it in favor of its own custom GPU solution. Imagination Technologies was then acquired by Chinese-funded firm Canyon Bridge, and after that, the company saw decent popularity in China.
Imagination Technologies is now poised to make strides in the high-end AI segment as the company has seen a major change in leadership and has worked its way from the ground up to evolve its GPU IP technologies. Additionally, they have received $100 million from Fortress Investment Group in a bid to revive their operations, focusing primarily on AI this time around.
To gain an edge on the AI hype, Imagination believes it needs to focus on the software side, as its GPU IPs are already well-refined when it comes to power efficiency and data movement.
There are two ways to approach [AI]“Either take something like an NPU and make it more general purpose, sacrificing some PPA, or start with a GPU, which is inherently flexible, and then add to your PPA by adding dense, low-precision compute capability.”
– Tim Mamtora, Head of Innovation and Engineering via EE Times
Mamtora says that GPUs have reached their peak when it comes to the capabilities that manufacturers can integrate into them, and the only way to be competitive in this segment is by creating a robust AI software ecosystem, and the area of innovation here is to achieve some kind of engine that Imagination can use to compete with NVIDIA’s CUDA software stack, likely by providing a platform that allows developers to run their hardware on any existing software stack, suggesting something like ZLUDA.
The main problem we are trying to solve is to develop something to address [Nvidia’s] CUDA. Developers want [their applications] running on a platform, they just want to translate it and for it to work — forget about performance to begin with, it just needs to work. The fundamental problem we had with our AI accelerator was getting things to work in the first place.
Imagination Technologies is known for having a decent arsenal of AI products, the most notable being the PowerVR Series 9XE and the company’s Neural Network Accelerator, which is specifically designed for neural network computations. The company plans to launch next-gen products as well, but for now, they are primarily focused on improving their software stacks and waiting to see how demand evolves in the markets.
While it’s too early to say whether Imagination can make a comeback here, a shake-up in both planning and leadership shows that the company is committed to taking a step forward in the markets. We’ll have to wait and see what they have in store for us.
News source: EE Times