With a share price that has rocketed in the last decade off the back of aggressive launch schedules and best-selling products, AMD (NASDAQ:AMD) has gone from strength to strength. I recently caught up with its CTO, Mark Papermaster, to discuss how he helped turn the company around, the creation of the Zen architecture and Ryzen CPUs up to the current range and the exiting products we can look forward to in 2022.
Antony: So mark you joined AMD in 2011, But before that I know you worked for several other high profile companies. Could you give us a brief overview of what those companies were and any key projects or products that you worked on?
Mark: Sure well I’ve been very fortunate as I’m coming on to 40 years in the industry and when you think about it, when I started it was like the birth of the PC. The IBM PC had just come out and the whole way in which we work with computers was about to change. Before that it used to be these big behemoths that were behind walls. The phrase personal computer was just that.
MORE FOR YOU
It was also a time of great technological change with new semiconductor approaches that were much lower power and could scale well plus many of the underpinnings of chip technology that we still use today. I was able to grow the use of technology and apply that across PCs, mainframes and super computers. It was a great way to start a career.
Steve Jobs then recruited me to run iPod and iPhone where I worked on the iPhone 3 and 4 series and just like at IBM, it was all about bringing the hardware and the software together. Personal computing was then in your pocket. I then had a stint at Cisco working on routing and switching technologies and then 10 years I joined AMD as CTO and our products span PCs to intense graphics and gaming to super computers. We’re also building what we think will be the worlds biggest super computer with over one Exoflop of computing capabilities.
Antony: I’m very envious you got in there a few years before me to see all that! I think it’s no secret to anybody that AMD wasn’t particularly competitive when you joined it in 2011 and it was a full six years before we’d see the launch of the Zen architecture and ryzen CPUs.
I think everybody is interested in how things were turned around both in terms of amd’s goals and its internal culture to go from that time in 2011 to the launch of the Zen architecture in 2017. Can you tell us a bit more about you achieved that?
Mark: Well Antony, what attracted me to joing AMD it was really the opportunity of a lifetime. The talent at AMD was unparalleled and the company is 50 years old. It’s brought competition and innovation to the table with with its acquisition of ATI in 2006, it brought on a strong graphics capability. It was a very smart move. The challenge was to bring all of that together. To bring the CPU roadmap back to leadership computing capability and really make sure we revamp the engineering processes so AMD could be there each and every product cycle and be dependable, on time and with the quality and innovation it’s been famous for.
What a huge opportunity and the key thing was bringing the leadership together and aligning with them on this vision of how we could reset that engineering approach. I really can’t say enough on the talent at AMD, but I mean not everyone stayed. The surgery you’re doing in meeting this goals – it wasn’t for everyone, but so many of the key leaders and innovators stayed and they’re the real story of AMD and this turn around.
You did mention it took a few years to turn things around and , well, that’s what differentiates us from a software startup. It just takes longer. We started right away – putting all the building blocks together we called it the Infinity Architecture – we started that all the way back Q1 in 2012. That was huge – we had to get the underpinnings right and late 2012 we started working on the Zen processor and it takes 5 years to get a new CPU design- a new microarchitecture out there.
Antony: It’s amazing you started working on Zen all the way back in 2012 for a 2017 launch and I’m sure plenty of people didn’t know that. I think most enthusiast’s are familiar with the Zen architecture and the shift towards a modular CPU design and the more tech savvy might even be familiar with core complexes and chiplets.
What were the reasons for opting for that very different design for the Zen architecture – I mean it obviously lent itself two boosting multi-threaded performance which had languished for a number of years with very little movement in core and thread counts on desktop CPUs so was that the main reason or was it for a whole bunch of reasons And I guess the question is also was it maybe a risky move?
Mark: Well change is a risky move, when you’re in the position we were in, change is required. To change the trajectory of AMD, as a technology company you need to do that with technological innovation. In a way the position we were in did us a favor. We were in crisis mode and the team just galvanised not only the innovative spirit at AMD but the fighting spirit. Being that underdog and to find a way to come back – it was an overpowering culture across the company and with collaboration.
You talk about modularity, well we really looked at where the industry was going. Moore’s Law was slowing down. The technology trend that all of us had followed where every new semi conductor technology node allowed you to scale performance and you could raise the frequency and it would still be at the same cost. You could double the number of transistors but the product would still operate at the same power requirements.
You an still improve the density with each of our new technology nodes today but it’s even more expensive and you cannot keep on the same frequency phase. We saw that coming over 10 years ago so we realised that we needed a modular approach to piece this together and you see this today with our high end CPUs and GPUs and we’re leveraging that modularity.
And Antony your question on the CPU, I really give the credit to Mike Clark our chief engineer. We knew we could never let go of single thread performance – how does a single x86 CPU perform. We knew we had to have clear leadership and you have to combine it with scalability so not only is a single CPU providing leadership capability but as you add more and more cores, you could scale too.
We basically brought up the amount of memory and interconnect and I/O capability and that was the breaktrough. The x86 industry had stagnated and our goal was to bring significant performance gains with each generation of Zen processor and bring competition back. It was really exciting.
Antony: I’ve enjoyed seeing the progress AMD has made over the four generations of Ryzen desktop CPUs and because it was such an aggressive road map I set out my own requirements for what I considered to be a successful range of CPUs. For example, with the first generation it was about establishing Ryzen as force to be reckoned with, boosting multi-threaded performance that had languished for years and starting to claw back market share. With the second generation, I wanted to see that performance envelope extended and further market share gains and with the Ryzen 5 2600 being the best selling desktop CPU after its launch, I think you were pretty successful at that. What did AMD want to achieve with each of the four generations of Ryzen desktop CPUs and do you think it succeeded?
Mark: When you think about what we were trying to achieve with each generation – it was to bring a better experience. With the PC market – it wasd being under-served. With notebooks, the battery life wasn’t improving, the graphics technology wasn’t being exploited that fast and we set out to change that, better CPU and GPU performance, better performance per watt and it turns out it also helped desktop, especially gamers – they benefit from having a great CPU and a great GPU and how they communicate between each other. That’s what gives PC users a great experience and that’s what we focussed on.
We’re so excited about the coming CES in January – it will be quieter than usual with a big virtual presence, but we could not be more thrilled than with the next generation of products. We can’t reveal too much now, but Antony, stay tuned. We are not letting the foot off the preverbial gas peddle. We’re full speed ahead.
Antony: So moving on from the initial launch of Zen in 2017, what were the main learnings over the last four years? I mean obviously CPUs aren’t designed and made in six months–but what’s your take on AMD’s progress since then and are there any specific challenges that come to mind that you had to overcome?
Mark: Absolutely – the one thing I love about AMD’s culture, is it’s not only innovbative, but we’re good at listening to our customers. By no means was our first generation of products perfect. It got us back in the game, but we didn’t have single thread performance leadership. That improved with Zen 2 and really with Zen 3 – the Ryzen 5000-series. It really has been just an outstanding leader in the industry in both single and multi-thread.
So we learned to innovate and grow that performance gain with each generation, but also we learn along the way, so what we found with the first generation with gaming for example, the latency – the delay of our CPU and memory communication – was not the best. We improved it on generation too and game performance went up and our market share also went up significantly – gamers noticed immediately when we made those changes.
The quality across the board was a factor too – with the first generation we learned we needed to improve drivers and we worked on fixing those issues and fixing them quickly. We now carry out testing in exactly the same way our customers use our products so we have no surprises when our products are released. I’m very proud of the team both in terms of performance and quality the team has achieved.
Antony: Is there anything you might have done slightly differently either with a specific CPU, range of CPUs or more generally, if you could go back to 2011 again?
Mark: Well we did the changes along the way. You’ve got to be quick on your feet. We’d done the first generation of Zen 1 and as we did Zen 2, the original design we had, there were good metrics, but we were going to be late to market. I mentioned this to Mike Clark our chief architect and we flagged the problem. He came in and said look, we can modify how we were thinking about doing Zen 2 and hold our schedule and make sure we’re on time. As you mentioned earlier, the later generations of Zen were really big hits, but people didn’t see what was under the covers.
If we weren’t reacting in real time and watching our progress, we would have been late. So, rather than saying what we’d do differently, we just do different things along the way to make sure we are seamless in our execution and don’t drop a bit.
Antony: Do you have any stand-out CPUs that you were particularly proud of that you maybe felt nailed their intended market or were particularly competitive? I mean I’ve always had a bit of a soft spot for the early six-core CPUs such as the Ryzen 5 1600 and the 16-core flagships like the Ryzen 9 5950X with had with the last couple of generations.
Mark: Well, we are at AMD are proud of all of our generations and I can honestly say that the teams responsible for the CPUs think of them as their babies. You talk to the engineers and you can see the pride as they bring their products to bear. For me, I do want to call out our current generation of Zen 3 CPUs and what we’ve done with the whole Ryzen 5000 line here is pretty amazing. They brought absolute leadership and it allowed us to extend the product range in terms of cores. We can now offer great performance at value, but also if you want high performance, we can scale you all the way up, not only to the 16-core CPUs, but remember Threadripper too – what a beast. We scale all the way to workstations with 64 cores. Just a phenomenal beast of computing. So that’s what we’ve been focused on.
With our innovation, we announced at Computex we announced 3D stacking technique called V-Cache, which we will introduce on future Zen 3 CPUs, giving up to 96MB L3 cache. As we enter 2022, we’ll be taking a product line that we’re already proud of and put it on steroids.
Antony:The launch of Windows 12 threw up a few performance issues for Ryzen processors. Do you consider these fixed or is it an ongoing process?
Mark: We have a deep partnership with Microsoft and we do extensive testing on any release and it was prior to the release of Windows 11, but too late to that we caught that we had some issues in working in partnership with them. That’s why you saw the patch go up so quickly to resolve the issue. We’ve also modified our way in which we work with Microsoft where we do this kind of testing much earlier.
Antony:Looking to 2022 and we’ve obviously very excited about the launch of Zen 4, what are you hopes for the new architecture, what do you think PC enthusiasts like me will be most impressed with and has it presented any specific challenges?
Mark: I can’t go into too many details but it will be a completely enhanced experience. At the start we had to get back to the forefront of consumers’ minds and if you want get value, go AMD. Well we wanted to go well beyond value. We wanted to offer the best performance and the best experience, so once we’d done that, we got a seat at the table with our gaming partners, our PC partners and all the way up to our server partners thanks to our leadership. We can deeply optimize for the end experience, so what customers can expect soon is ever increasing battery life, quality of service, enhanced application performance. We expect our customers to have a high bar of expectation for AMD’s next launch and we’re working hard to meet that with Zen 4.
Antony:Has the pandemic had any significant impacts in Zen 4’s launch and can you confirm a rough eta for the launch?
Mark: Firstly the pandemic has been such a global challenge and certainly all of us in the industry have had to double down because as people went to work from home and to play from home with gaming for example, they’re really banking on the whole tech industry to respond. That’s what we focused on at AMD and we worked hard with our supply chain – I can’t say enough about them and our partners. They’ve had to deal with huge growth and also boost our capacity for 2022 and beyond.
Antony, with regards to the upcoming generation – I point to CES in January. We’re excited to be revealing some additional details on our new product launches that will deliver phenomenal experiences and as we’ve said, later in the year as it progresses we’ll share more detail on Zen 4 with some mentioned at CES and more announcements on it over the course of 2022. It will be a very exciting year for AMD.
Antony: Finally, looking beyond Zen 4, what are AMDs goals here in terms of areas of innovation and can you elaborate on your Moore’s Law Plus philosophy I’ve heard a low about?
Mark: I called it Moore’s Law Plus because Moore’s law is gone, but yet we still need to improve performance and capability at the same exponential pace the industry has always been on. What we do at AMD in going beyond Moore’s law is continuing to build and architecting solutions that started with the acquisition of ATI to focus on heterogeneous computing.
It can’t be all about the CPU, it can’t be all about the GPU. It’s both of those things, it’s other accelerators you put in for AI workloads, video processing – we have all of that in our chips. That’s Moore’s Law Plus and it’s how you put it all together. Its not on one piece of silicon anymore. IT’s on chiplets and that goes back to to how we started with Zen 10 years ago. We want to stay on an exponential pace of improving computing capabilities going forwards.
I’d like to thank Mark for taking the time to talk to me and I’ll be back soon with all the announcements from AMD and others from CES in 2022 so follow me here on Forbes for the latest PC hardware news and reviews and don’t forget to check me out on YouTube, Instagram and Facebook as well.