Eric Demers - Farewell Interview



Author: James Prior
Editor: Charles Oliver
Date: April 16th, 2012

So long, and thanks for all the fish

PC enthusiast websites exploded on Valentine's day this year as, on the eve of the Cape Verde GPU launch, Semiaccurate's Charlie Demerjian reported that Eric Demers had left AMD. This was quickly confirmed by AMD and widely acknowledged as a Bad Thing. Speculation abounded on where he might have landed - where do you go from Chief Technology Officer of AMD's Graphics Product Group (and Vice President, to boot)?

Why, Vice President of Engineering, of course, for none other than Qualcomm; home of the former ATI acquisition 'Bitboys Oy' team whom Qualcomm acquired with the January 2009 purchase of the Imageon IP, as part of AMD's mobile business. Qualcomm's current graphics technology, featured in the Snapdragon processors, is named Adreno - an curious anagram of Radeon. As the mobile marketplaces offer more and more capabilities in the hands of consumers, demands will increase - creating more and more challenges for delivering power efficient, feature rich, great performance technology to power it. Lots of work for a guy who helped make the hardware in so many consumer gaming systems over the last 15 years.

Eric was kind enough to answer a few questions for us, as we bid him farewell from the current PC desktop and notebook graphics realm and wish him well on his now mobility-focused journey:

Eric Demers, covertly photographed plotting his world domination

Rage3D - Your final role at AMD titled you as Graphics CTO and VP of AMD, did the executive level position help or hinder you in your engineering roles?

Eric Demers - It's hard to really compare. From a pure technical standpoint, the deep dives afforded by a lower level engineering position are hard to beat; you get to deep dive into technical issues and get to contribute at a low level in the problem solving. At a CTO level, the abstraction level is much higher. You have to have a much wider span of view, but it tends to stick to a more modest depth. The CTO role is still much more technical than any other executive role (most of which are concerned with execution and management), but less vertical than an individual contributor role. On the other hand, you also gain more overall control and get to set long term direction to the efforts. In fact, that's the value of a CTO - to help set the long term technical direction for the company or group you represent.

R3D - What are you most proud of doing for ATI/AMD?

Eric - It's hard to say. Projects like the Nintendo GameCube/Wii and R300 projects certainly stand out. But at the same time, some of the engineers and teams I've helped to grow and develop into amazing contributors are also great sources of pride. Creating processes and roadmaps that would deliver for years are also potential favorites. Overall, though, it's hard for me to pinpoint a single one. Ask me again in a year, and I'll see which one pass the filter of time :-)

R3D - What is your favorite product that you worked on so far (if it's released)

Eric - Of my career, probably the work at ArtX (both the t&l 3d chipset and the GameCube). That was a special time, with a small group of friends, where we worked hard and did great things. Not so much the products, but the atmosphere which we designed in. Always will be a special time for me.

Tahiti, heart of the AMD Radeon HD 7900 series

R3D - What design or engineering decision do you regret?

Eric - There are so many! The reality is that you make a bunch of decisions, and after the fact, you can always go back and say 'well, that could of been better here' or 'we should of done that.' On any relatively complex project like designing a GPU, you will always find things you could have done much better. The worst ones were decisions that delayed projects. Like picking processes that didn't pan out on time, or selecting vendors that didnt have their libraries ready, or many other decisions of the sort. Any place where you can really point to a single decision that could of made a difference :-)

R3D - What's a standout engineering/design moment in your time with ATI/AMD, is there a 'heck yeah!' point you think back on and relish?

Eric - As an individual contributor, there were certainly cases. For example, designing some clever circuit to do this or that. Examples in the last 10 years include designing the pixel shader for r300 with Mark Leather, where we did many clever things; or r600 shader core sequencer. Many of those actually even ended up in patents. I'd have to check, but at least 20-30 of those. At a CTO level, it's hard to pin point select items, because most of the time, it's group efforts and oversight; that's harder to find 'moments' in.

My new work phone is THIS big! (Just kidding, Eric - Ed.)

R3D - What drives you to innovate in your engineering, is it a reactive process driven by problem solving inside a framework towards a set of goals or working towards executing a grander, bigger solution thats a superset of the current goals, or something else?

Eric - A bit of both. Obviously, on the road to a goal, you will encounter many specifics problems that need problem solving; finding elegance and simplicity in those solutions is always a pleasure. But at the same time, making progress on long terms goals can be really satisfying, though it's generally after you are done that you look back and find those moments. So both cases occur, and both bring satisfaction, just at different times. Overall, though, accomplishing the delivery of a program and it's goals over a long period is the most satisfying. It's the culmination of a vision, and represents the effort of many over long periods - its quite amazing.

R3D - How challenging are business unit requirement constraints when trying to bring an awesome new design into play?

Eric - They represent the boundary conditions for your work. While having some flexibility, they represent the challenge of the program - its actually not that fun to design without constraints! You need to work closely with the product teams, to see where any flexibility might exist, to push the boundaries and to deliver a product on schedule and meeting the needs of the business. And a really successful product from a business standpoint is really satisfying :-)

ATI R300

R3D - When you were involved in R300, did you ever envision that Radeon GPUs would be as powerful as they are now?

Eric - Nah! We all know that doubling transistors every 18 months can lead to great things. But you rarely really think about it. And you never really think in concrete ways about what will happen 10 years down the line. It really would not of been that difficult - over 10 years, you would have expected a 100x increase, and I think we got there at least on the shader side. But obviously, the system did not scale equally accross all parts - so bottlenecks shifted. But overall, it's now amazing to see small handheld devices deliver what took 100s of watts in a larger computer just a few years ago. And looking forward, it's even more amazing to think of what can be delivered in the individual consumer product space! A truly personal computing experience driven by amazing graphics and parallel processing, all in your pocket!

To round things out we have a final comment from another long-time Rage3D member - Terry CatalystMaker Makedon, now in charge of the AMD Fusion product management:

Terry Makedon - Long before there was Facebook or Twitter there were only forums on websites to interact with other enthusiasts. 10 years ago Eric and I used to post like mad on the Rage3D forums, but we never met. He was working at ATI's Silicon Valley office and I was working in the Toronto office. We both spent countless hours on those forums reading every single post that was put up there. Ahh the good ole days!

And so it is with heavy hearts but grateful thanks we bid Eric Demers farewell from AMD. It was always a pleasure listening and conversing with him at ATI's and AMD's technical press events, and we can't help but wonder if in five or eight years time we'll be back listening to him again as Personal Computing moves to using quite different devices from the big boxes we use today to deliver the great gaming experiences we so enjoy.