Last month I sent out a message on Twitter announcing that the latest PowerVR Wizard ray tracing GPU test silicon had taped out, based on Imagination’s reference design and the IC Compiler™ II place and route solution from Synopsys.



Given the enthusiastic response from developers, I thought I should write a short article about how we got here.

Firstly, I’d also like to take this opportunity to congratulate everyone in the engineering team; they have worked really hard to take the Wizard architecture from a theoretical concept to a mature, real-world implementation that can now be found in silicon.

For those who haven’t kept track of our ray tracing solution, the PowerVR GR6500 GPU is the culmination of eight years of effort from a global team of designers. The technology was originally conceived at Caustic Graphics, a startup with a vision for interactive ray tracing, going against twenty years of industry momentum.

PowerVR GR6500 GPU - PowerVR Wizard GPUs

The PowerVR GR6500 ray tracing GPU

I recently had a chat with Luke Peterson, Imagination’s director of research for PowerVR Ray Tracing, who revealed some details about the team’s beginnings in the semiconductor industry.

We approached the problem differently. While others in the industry were focused on solving ray tracing using GPU compute, we came up with a new approach leveraging on our prior expertise in rasterized graphics. This allowed us to challenge assumptions that the rest of the industry took for granted and cleared the way for some major breakthroughs.

Their incredible work was soon recognized by several high-profile peers working in graphics. Academy Award Honoree Matt Pharr, co-author of Physically Based Rendering, from Theory to Implementation, had this to say about our technology:

The [PowerVR] Wizard architecture offers the possibility of a quantum leap forward in image quality in mobile interactive graphics applications.

In addition, John Carmack, co-founder of id Software and current CTO of Oculus, declared the following about PowerVR Ray Tracing:

Overcoming the initial challenges

One of the first technical challenges the ray tracing team faced was that the memory bandwidth requirements were orders of magnitude higher than what available technologies of the time could offer. This prompted the engineers to cleverly invent an efficient scheduling and sorting algorithm to amortize data reads across many simultaneous queries into a spatial database.

A big business challenge came in 2009; during the global financial crisis, Caustic’s investors could no longer fund the company. The team however continued development, based on the strong belief in the potential of the technology that they were building.

Soon after, Imagination acquired Caustic and integrated the ray tracing group within the PowerVR Multimedia division, leading up to the launch of the PowerVR Wizard ray tracing architecture.

The road to tape out

One of the many advantages of PowerVR Wizard ray tracing GPUs lies in the power efficiency gains delivered by the architecture. Thanks to the innovative nature of the design, the majority of data can remain in a subset of the local cache, eliminating the need for any high bandwidth random access path between processing elements and internal cache memory.

This coherence gathering results in huge savings in memory bandwidth and power consumption, making Wizard ray tracing GPUs 100x more efficient compared to using GPU compute or other software-only ray tracing approaches on traditional graphics architectures. The approach has the added benefit of recollecting SIMD coherence, enabling graphics processors to tackle previously unapproachable challenges.

PowerVR Ray Tracing - cascaded vs ray traced-1A screenshot from an upcoming article comparing ray traced and cascaded shadows (click on the image for the hi-res version)

The success of the Wizard design team is based on a number of factors, including commitment to a long-term vision, ambition and tenacity.

Along the way, the PowerVR Ray Tracing team collected multiple nominations for high-profile awards; recent examples include Best Technology Innovation at The Inquirer 2015 Tech Hero Awards and Design Team of the Year at the 2015 UBM ACE Awards.

Having delivered the technology that will change the state of the art in mobile and embedded graphics, we can’t wait to put it in the hands of developers!

If you want to be among the first to receive the latest news about our PowerVR Ray Tracing technology, subscribe to this blog and make sure to follow us on Twitter (@ImaginationTech, @PowerVRInsider, @PowerVR_RT), LinkedIn, Facebook and Google+.

Comments

  • Ken Gray

    So, how many more years til we are likely to see it in product Alex ? You’ve previously suggested ‘not years’ but there seems so little industry chatter about real SOC development – that can be used in product that this seems unlikely to be this year or next year.

    So are we looking 2017 or as far away as 2018 or beyond ?

    I know you cannot speak for customers’ schedules but you’ve got thoughts, and you know normal silicon schedules from license to product so when would you feel confident about predicting ?

    • We can disclose certain activities (e.g. taping out a test chip) but we really can’t talk about customer engagements. We’ve also said that we are going to work closely with developers to ensure that they have everything (hardware, SDK, tools etc.) needed to write ray tracinng shaders.

      This is a new technology – and with all new technologies, you can’t rely on the tried and tested models of the past. I hope you can therefore understand why it is very difficult to make statements about product availability; most of that is out of our control anyway.

      Regards,
      Alex.

      • Ken Gray

        Hi Alex, thanks for your reply.

        I totally understand you can’t say exactly who will be releasing what and when but you’ve already told someone it ‘won’t be years’ – I was only after something 1% less vague. eg less than 3,5 years – or more if that’s the case.

        Also as there seems to be a decent amount of developer activity I can’t imagine they’re doing anything from purely altruistic motives – through IMG they must have a view on when their work may actually generate some revenue. And this won’t be until there is hardware supporting RT in the market. They aren’t charities after all.

        • When I said “it won’t be years”, I meant we’re getting close to a test chip (i.e. implementing in silicon). This can be used for dev kits and/or other projects. I really can’t anticipate when you’ll see it in consumer electronics products because this is a brand new technology.

          Therefore it requires investment and careful development – the last thing you want is to rush technology to market.

          Regards,
          Alex.

          • rogue_warriorwiz

            Surely with the amount of investment put in we would be much closer to market. What has gone wrong!?!

          • Ken Gray

            Come off it Alex ! You were asked by the person on Twitter in relation to him upgrading his next phone to which you replied ‘it won’t be years’, it was blatantly obvious to everyone he wasn’t talking about test silicon so let’s not go to that place !

            Waiting 3 years and still having no licensees is hardly ‘rushing’ and you don’t go from no licensees to product in under another 3 years (according to our CEO at countless AGMs and in countless presentations).

            IMG are an incredible company with amazingly talented people but when you start teasing product PR fluff too soon with too little end result it just seems embarrassing.

  • LDM

    Hi Alex,

    In terms of pure performance, how do you compare this GPU? Is closer to G6430 or GX6450?

    Also do you guys expect to see any further implementation inside the future Api such as Vulkan etc?

    Regards

    L

    • Since it has that brand new Series6XT DNA, it is equivalent in performance to a GX6450 in terms of graphics and compute. In addition, it features dedicated ray tracing hardware which delivers 300m rays per second; the beauty is this ray tracing performance does not take away from the graphics and compute numbers, it adds on top.

      Vulkan is a low level API whereas OpenGL ES is more high-level – they aren’t mutually exclusive. The market wants both for different use cases and we will continue development on both – and others.

      Regards,
      Alex.

  • Wizardinsilicon

    So this is ray tracing week – yawn

    • The in-depth articles are coming tomorrow and on Thursday.

      • roninja

        Look forward to reading these. As for getting Wizard in silicon and out to developers is 2017 a more realistic timeframe. There seems like much work and convincing to be done. Not sure what can accelerate this disruptive tech. I want IMG to be first movers and not see ARM, Qualcomm etc catch up!

  • Nemesis

    Well, we’re waiting to see some video in realtime, Alexandru, how much we have to wait for some sugar?

  • LDM

    Hi Alex,

    In the next generation of iphones/Ipads what gpu will be implemented?
    Series 6 GR, or Series 7?

    L