APU13: Silicon? Check. HSA? Check. All done? Wrong!

Share on linkedin
Share on twitter
Share on facebook
Share on reddit
Share on digg
Share on email
This year’s AMD Developer Summit, called APU13, has been bigger and better than ever before. With an expected audience of 1000+ from around the globe, APU13 is definitely one of the must-attend conferences for developers, technology leaders, academics and startup investors. This year Tony King-Smith, EVP of marketing for Imagination, delivered a keynote on the first day titled “Silicon? Check. HSA? Check. All done? Wrong!” and I had the chance to be in the audience when it happened.
If you’ve met Tony in person, you already know how passionate he is about the need for SoC technology platforms to address the needs of the application developer, not the benchmark chaser.

“The reality is that today’s SoCs are heterogeneous, with powerful CPUs and GPUs whose power cannot be necessarily be harnessed by today’s software APIs. That’s why we are a founder member of the HSA Foundation: to ensure relevant APIs are defined as open standards, and to further ensure their relevance to all markets, not just the high performance end of the spectrum”.

 Tony Keynote 1 small APU13: Tony King-Smith explains how everything, from mobile to the data center, starts with the SoC

According to Tony, we must ensure that HSA addresses as broad a community of developers as possible – and that means ensuring HSA works on not only tomorrow’s coherent architectures, but today’s SoCs.

“If HSA waits until fully compliant SoC platforms are deployed in high volumes to attract developers, it will simply take too long for developers to adopt, and risks becoming marginalized. HSA needs to run on as many of today’s platforms as possible as well as tomorrows, so that developers embrace HSA at the earliest opportunity”.

Tony also explained how tomorrow’s applications would be heterogeneous in a different and complementary way: sharing processing between the cloud and the local device based on factors such as real-time response and access to data. This does not mean that mobile platforms should demand desktop capabilities; rather that apps should minimize their use of high end capabilities such as high precision except when really needed, allowing them to freely migrate between cloud and local processor resources.

“The reality is that desktop or HPC apps don’t migrate to mobile platforms – why should they?” said King-Smith. “Mobile apps are designed to be highly interactive, touch-based and very responsive, They also need to take advantage of cloud-based compute and data power, but must also cater for the fact the link from mobile to cloud is often unreliable and at times non-existent”.

King-Smith concluded that HSA must better embrace the needs of the hundreds of millions of mobile platforms already deployed and those being shipped over the next few years, to ensure its relevance to developers. “We need a consistent heterogeneous API across low end to high end – HSA has the potential to be that API, but only if it truly embraces the needs of all developers”.

Tony Keynote 2 smallAPU13: Tony King-Smith explains how why the HSA standard must address everything from mobile to workstation processors

In case you missed it, here’s what @AMDHSA live tweeted during Tony’s presentation












You can also check out a livestream of the keynote from Tom’s Hardware, including photos and short snippets of what was said, here. If you want to get all the news about our initiatives with the @HSAFoundation, GPU compute and other open APIs, follow us on Twitter (@ImaginationTech, @PowerVRInsider and @GPUCompute) and bookmark our blog.

Jen Bernier

Jen Bernier

Jen Bernier-Santarini is director of technology communications, part of Imagination’s global marketing communications organization.

1 thought on “APU13: Silicon? Check. HSA? Check. All done? Wrong!”

Please leave a comment below

Comment policy: We love comments and appreciate the time that readers spend to share ideas and give feedback. However, all comments are manually moderated and those deemed to be spam or solely promotional will be deleted. We respect your privacy and will not publish your personal details.

Blog Contact

If you have any enquiries regarding any of our blog posts, please contact:

United Kingdom

Tel: +44 (0)1923 260 511

Search by Tag

Search by Author

Related blog articles

shutterstock 1175807410 1

Vulkan synchronisation and graphics-compute-graphics hazards: Part I

How do you mix and match rasterisation and compute in a modern GPU? In modern rendering environments, there are a lot of cases where a compute workload is used during a frame. Compute is generic (non-fixed function) parallel programming on the GPU, commonly used for techniques that are either challenging, outright impossible, or simply inefficient to implement with the standard graphics pipeline (vertex/geometry/tessellation/raster/fragment).

Read More »
self drive car

Unveiling XS: the ultimate GPU family for automotive

Introducing the XS family of GPU for automotive. XS is the first GPU IP compliant with ISO 26262, offering a new methodology and architecture for functional safety with specific safety-optimised mechanisms designed to protect against transient and permanent faults. Tuned on real-world ADAS applications, it offers unrivalled performance for graphics and compute in next-generation cars.

Read More »


Sign up to receive the latest news and product updates from Imagination straight to your inbox.