Moore's Law, Ryzen and PCIe 4.0, Laptop Market
AMD rose from the brink of bankruptcy to one of the best performers in the S&P 500 in just three short years. The surprising turnaround comes against all odds as the company competes with both Intel and Nvidia, two of the largest semiconductor vendors in the world, yet is a fraction of their size.
AMD CEO Dr. Lisa Su, who holds a Ph.D. in Electrical Engineering, has been largely credited for the company’s resurgence. Su has led the company back to profitability and competitiveness, earning multiple plaudits for her work along the way, including being recognized by Forbes 50 for being one of the top 50 world leaders. We had a chance to sit down with AMD CEO Lisa Su to discuss the company’s successes over the last year and its plans for the future.
Tom’s Hardware: It’s been another solid year of execution for AMD for being the top performer in the S&P 500 for 2018 and the recent addition to the Nasdaq 100. For you, what were the biggest highlights of last year, and how do you plan to build on those successes as you move into 2019?
Lisa Su: Thanks for those comments. You know we are always focused on our products. That’s sort of our mantra to make sure that we’re bringing something to the market that adds true differentiation and value for the industry, so we’re pretty pleased with the performance in 2018. If you look across our CPU roadmaps and the reception of our second generation of Ryzen, as well as the reception of our second-generation EPYC processors, I think we’re really pleased with how they’ve done in the market. And as we go into 2019, I think it’s going to be a significant opportunity for us to expand that portfolio. And so we’re looking forward to really continuing this mantra of bringing out great products and helping the industry drive high performance computing.
Tom’s Hardware: We often hear that Moore's law is dead, and even some of AMD's competitors have embraced that message recently. How does AMD plan to deliver similar improvements to what we enjoyed during the height of the Moore's Law era, and do you think Moore's Law is ending, or just changing?
Lisa Su: I don’t personally use the terminology that Moore’s Law is dead. Moore’s Law is certainly one of the things that that has governed our industry for the last almost 50 years, and it is changing, it is definitely changing. And it has been changing, frankly, for the past decade, so it didn’t just start changing. The slow down in just pure manufacturing technology and process improvements has actually opened up new opportunities for innovation around architecture and design.
Our focus is to double down where we think we can add value, and that’s in architecture and design. So, for us, it is aggressive use of technology. A big theme for us in 2019 is to be first to market in 7nm process technology for high performance computing, but it’s also doing quite a bit of innovation on the architectural side, and that’s both on the CPU front and the GPU front. We’ll have new architectures that will be introduced in 2019. And then also doing a different look at system partitioning and how we partition the processor elements and the memory elements and the I/O elements. We like to say that we’re leading the way on multi-chip modules and chiplets in high performance computing.
Tom’s Hardware: It’s worth noting that even some of your competitors publicly questioned the merits of the Zen design when it first came out. Now that other semiconductor vendors and the industry as a whole are embracing the benefits of multi-chip designs, do you feel that validates your approach?
Lisa Su: We do. As you know, these decisions were made literally four or five years ago in terms of how we think about not just our Zen design, but about Zen 2 and Zen3 and how we’re putting together those pieces. I like to say that we were early to the idea that there is value in multi-chip designs and we’re going to keep on that path. It’s natural that other vendors would also use similar approaches, but I think we’re quite aggressive with the use of multi-chip designs to get optimal system performance.
Tom’s Hardware: The adoption of a multi-chip architecture was important largely because it allowed the use of the same basic building blocks for both enterprise and consumer products, but part of that was a necessity borne of limited engineering resources. Now that AMD has considerably more money to invest in R&D, can we expect a similar approach in the future of using the same basic building blocks, or could we see microarchitectures targeted more for specific market segments?
Lisa Su: We’ve always thought about what the right technical solution is. Put aside the thought process of reuse. Reuse is an important secondary reason to do something, but it's not the primary reason. From our standpoint, the primary reason is that there are different optimization points that you would choose for processors versus some of the high-speed I/O that doesn’t scale as readily. As we’ve seen, we have separate notebook and separate server designs and we are looking at how we serve this very large market with the right set of products. Again, back to the point that the multi-chip approach was extremely valuable to get the right products in the right technology.
Tom’s Hardware: Expectations are high because everyone knows that AMD is beating Intel to a smaller process. Given AMD's disruptive 52% leap in IPC, lower pricing, and increased core counts with the first-gen Ryzen models, how does it plan to manage customer expectations for future processors?
Lisa Su: We are flattered that there is so much anticipation as to what comes next. When we thought about what we wanted to do in computing, we really thought about it as a three-generation family. So, it wasn’t let’s do Zen, then come back and let’s do Zen 2, and then let’s come back and think about Zen 3. It was a three-generation family of where did we think the innovation was needed, and what was the right way to roll it out.
We’re very happy with our initial Zen and Zen+. I think it’s actually quite satisfying to see, especially among the enthusiast community, the amount of attention that first and second-generation Ryzen has gotten. But I also believe that you are only as good as your last product, and so there is no question that we are working hard to meet and exceed expectations as we go forward into our Zen 2 and Zen 3 designs. That’s expected in our industry, so we know that.
Tom’s Hardware: Speaking of the enthusiast community, backward-compatibility is a big selling point for Ryzen processors. How does AMD balance the tradeoff between maintaining compatibility with your current platforms, which does restrict options, but also continue to deliver new and exciting technologies, like support for PCIe 4.0?
Lisa Su: A very deliberate decision to have this backwards compatibility for both our server product line and our desktop product line. And the reason is for exactly what you said. We wanted enthusiasts to be able to join the Ryzen family, and then continue to upgrade as Ryzen and Epyc upgraded. It is a balance, but I think it is the right balance that we’ve chosen. We will maintain backward compatibility, but if you want to use all the new features and functions there will be some modest modifications for the new functions, like PCIe Gen 4, for example.
Tom’s Hardware: AMD is announcing a new series of 2nd Gen Ryzen Mobile chips at CES, building upon the goal that you stated last year to expand penetration with OEMs. How has that progressed, particularly with laptop OEMs?
Lisa Su: We are really excited about the progress that we’ve made in the notebook market. If you take a look at even the second half of 2018 compared to the first half of 2018 there are many many more Ryzen-based notebooks out there for this holiday season. I think as we go into 2019, we are launching early in the year such that we can get that breadth of product offerings. The notebook market is really important, it is 2/3 of the PC market, clearly there are lots of opportunities in both consumer and commercial notebooks. As we look at ultrathins, I think our products just get better as we go into second-generation Ryzen and 12nm technology. We’re enthusiastic about what second-generation Ryzen will look like in the notebook form factor.
Tom’s Hardware: What do you think are the things that have been holding back AMD from getting a bigger share in notebooks, and what are you doing to address them?
Lisa Su: When you look at, for example, comparing the desktop and the notebook market, particularly the do-it-yourself market or the builders, they’re really up on the specs and the capabilities of the processors and so we’ve really made a lot of progress in terms of market share in the enthusiast Ryzen community, and we’ll continue to focus there.
On the notebook side I think you have a much broader set of platforms and the platforms go through different OEMs and different retailers so there is just a longer life cycle as you bring out the notebook platforms. The way we look at it is a multi-generation opportunity for us. The first generation of Ryzen notebooks were great, the second generation are even going to be better. There are going to be more of them, they incorporate some newer features, they’re still excellent for gaming, improvements in battery life, and other features.
I think the time scale is just a little bit longer on the notebook side because you do go through so many OEM platforms and so many retailers, and there’s a process of just expanding the portfolio.
So... AMD competes against itself?
Go easy on them. Writing and publishing this during CES is a tremendous achievement. Great job Paul!
Edit: and it's fixed :D Thanks for the eagle eyes shrapnel!
Reality is, I hope everyone competes against themselves anyway... It keeps them from from being too complacent and challenges them to grow.
I was hoping to hear a question about if AMD is working with Microsoft to fix its obvious kernel issue. That fix should be huge for the upcoming high core count threadripper and EPYC CPU's.