AMD had its first keynote at Computex this week and this is really a coming out party for a new, revitalized, AMD, one that is increasingly out from under Intel’s shadow. This change has as much to do with Intel’s now decade-long execution problems (it starts with the board) and AMD CEO Lisa Su’s near laser-like focus on strengthening AMD. Su and her team have made several critical changes to the firm and now it seems to be able to execute like clockwork. At Computex, Su showcased the progress the firm has made while announcing a number of new products. This couldn’t have been timelier as Intel again had security issues that it again didn’t disclose in a timely manner, again flirting with a potential recall and leaving its OEM partners and end customers exposed.
Pushing Technology to the Limits
Su opened with a trip down memory lane, reminding the audience that AMD released the first Gigahertz CPU, and the first Gigahertz PPU (Physics Processing Unit), the first 32-core high-performance desktop processor (Threadripper), and recently the first 7nm GPU. She uses this as a foundation to point to what AMD will be doing in the future.
Su believes that there are two huge drivers underneath the technology market. One is the demand for a better experience through ever more intelligent products. Kind of like Amazon Echo on steroids. Second is the fact that these devices are producing massive amounts of data that can be used to analyze people and the environment, redefining scale. The data will be used to improve living conditions and quality, make better decisions, and improve the world.
AMD has one goal in mind, and that is to lead the industry in high-performance computing, whether on the desktop, the laptop, under the TV (game console), or in the data center. To do this, you have to make big bets because these bets redefine the competitive landscape. AMD’s big bets are on high-performance leadership. AMD chose to be at the leading edge of process technology, and is proud to work with TSMC to make that leadership stick.
The announcements at the show are to showcase this leadership and show that the company can’t just build parts but must help create solutions. It is going to talk about its data center products ROME, its next-generation graphics platform Navi, its third-generation Ryzen CPUs, and the High-Performance Ecosystem it is driving into the market.
AMD EPYC “Rome”
Su starts with the EPYC platform and the release, code-named “Rome.” The data center helps solve some of the world’s biggest problems. From data analysis to machine and deep learning, AMD’s EPYC platform has been making inroads since its introduction two years ago. (May I just say I love this brand.) Adoption has been strong, with 60 platforms based on this product, which has deeply penetrated the world’s cloud environments (Su called out Azure, AWS, Baidu, Tencent). There are over 50 EPYC cloud instances in production. AMD was also selected by Oak Ridge National Laboratory and Cray to produce what is expected to be the world’s fastest supercomputer at 1.5 exaFLOPs, which will use a future version of EPYC and a future version of Radeon.
They then shared a film starring Forrest Norrod, who runs AMD’s data center business, along with leaders from Cray and Oak Ridge National Laboratory (among others, like Dr. Alexandra Weisberg from the USDA and Mark Papermaster, CTO of AMD). They started with an example from Oregon state, which is using supercomputers to look at environmental impacts and to explore the future of artificial intelligence, among other things, and then moved to medical research, implying that some of the great medical problem solutions are just around the technology corner.
Azure Cloud Supercomputer
This brings us to the first major AMD announcement. Apparently, the Azure Cloud will have access to over 11,500 EPYC CPUs, and apparently Microsoft reports that the result is a 7500x speed increase over what it had been able to offer previously.
Back to “ROME”
The new Rome EPYC processor has some impressive performance gains with a 2x performance per socket improvement and a 4x floating point performance increase per socket, as well. Norrod returned to stage and showcased a molecular dynamic protein simulation demonstration where the latest EPYC was outperforming the latest Intel “Cascade Lake,” a $10K processor, by around 2x. This processor will launch next quarter.
Graphics Radeon “Navi”
Su now moves to graphics and the latest Radeon graphics platform, code-named “Navi.” Apparently, 4M gamers are using this architecture for gaming across PCs, notebooks, vendors (including Apple), game consoles (Microsoft and Sony), and even the new Google Stadia platform (cloud gaming) is based on Radeon. Radeon VII was the first use of a 7nm process used in GPU production. There are over 2B gamers in the world, according to AMD, and it is both demanding and very social.
The goal for the gaming market is to create a platform where you can game on whatever hardware you want, and Su argues that Navi was designed for this end use, platform-independent technology. The PlayStation will use the next generation of Navi and Ryzen in a custom package. Navi’s improvements include a brand-new 7nm Radeon DNA (RDNA) architecture, faster clock speeds using less power, and the world’s first PCIe 4.0 enabled GPUs. RDNA is a brand-new compute unit design, which increases the number of instructions that can be performed per clock cycle. It has a new multi-level cache hierarchy for higher bandwidth at lower power, and a streamlined graphics pipeline optimized for performance. This results in a 25 percent improvement in performance and a 50 percent improvement in performance per watt.
Su then announces the 5000RX family (the 50 in 5000 is to commemorate AMD’s 50th anniversary). They showcased a game demo and availability for the Navi Radeon cards in July 2019.
Su talks about the need to add performance and energy efficiency to this class of product, as well as providing an upgrade path for those who want to create content, or use that content, on the device or while streaming. Unlike its larger competitor, which seemed to abandon this segment, AMD continued to invest in it. The PC market requires a huge ecosystem to bring products to market. Su calls out Microsoft as the core partner (heart of the PC ecosystem) for creating the future of PCs. She brought Roanne Sones, CVP of OS Platforms, up on stage to chat about the future of PCs.
Sones spoke about the future of connected machines and how they will increase in capability significantly as the market shifts from endpoint computing to the cloud. She confirmed that Ryzen is making a huge move on this segment with massive increases in design wins and she implied that AMD is working far closer with Microsoft than others (that’s new). Sones implies that the best PCs coming to market would be based on Ryzen and Radeon (Navi). Typically, Microsoft plays Switzerland. Not this time. This time, it is presented as a huge AMD fan. (As a side note, Microsoft and Intel have been having “issues” for some time and they are again likely pissed over the lack of timely disclosure of Intel’s unique security problems.) Acer, Asus, Dell, HP and Lenovo are all announcing top laptops based on this technology. Acer and Asus were brought on stage to praise the platform and AMD’s CEO (this is also very unusual for AMD to get this kind of OEM support).
It is particularly interesting that Acer’s co-COO indicated that his investor friends made a ton of money investing in AMD but that he thinks their technology is revolutionary (Acer’s laptop offering, Nitro5, was one of the strongest at the event and it is all AMD). He really went off on the “better together” message that AMD often underplays, but it remains AMD’s strongest competitive advantage. Acer is all in with FreeSync AMD based monitors and the Predator desktop computers are heavy users of AMD as well.
This moved us to Ryzen and the announcement of the next third-generation CPU platform. There are some significant improvements to this generation of product. Typically, the third generation of a product is the first to really consider what was learned by the first generation. This generation makes significant improvements to performance by doubling cache size, providing a 15 percent IPC uplift. She talks about the Ryzen 7 3700X with 8 cores, 16 threads, 26MB cache, 65W TDP, and up to 4.4 GHz with Boost (3.6 GHz base). They then showed on-stage comparisons, indicating that this new Ryzen part outperforms Intel’s shipping comparable offering by 30 percent. They match Intel at single-threaded performance and exceed Intel in multi-threaded performance by 30 percent. She moves to the 3800X, which moves base performance to 3.9GHz with a 105W TDP (and it has a 36 MB cache). The competitive demo with Intel was unique in that it showcased that the Intel part bottlenecked with the GPU while the Ryzen 3800X did not (and the implication is it worked better with any GPU including those not built by AMD). They then compared an Intel/NVIDIA system against an AMD/AMD system for rendering and, as you’d expect, there was a huge performance difference (nearly 40 percent).
She then showcases the massive number of partners from system makers to motherboard makers that have committed to the AM4 platform.
One More Thing, Ryzen 9
This is a 12-core, 24-thread mainstream processor. This is also a 105 TDP part and it gives up a little (3.8 vs. 3.9 MHz base performance) for the four extra cores and eight extra threads. They ran this against the Intel Core i9 9920X Intel low-volume high-performance part, which is wicked expensive using a Bender benchmark (content creation). They beat the Intel part (also 12 cores) by 18 percent even though the AMD part uses 50 less watts of power.
She discloses prices for the products. The 3700X will retail for $329, 3800X for $399, and the 3900X for $499, or less than half Intel’s lower performing offering. These will hit retail shelves on 7/07/2019.
Wrapping Up: AMD’s Opportunity Against Intel Is Now
AMD has been outperforming Intel for some time. But Intel is entrenched, making it very hard for AMD to displace Intel on shelves. Intel’s lack of focus on PCs, coupled with its near continuous and untimely disclosures of major security problems has significantly soured the market on that company, though, allowing AMD to make serious inroads across the ecosystem.
For me, though, it is on trust. Intel has become untrustworthy, apparently more interested in areas like drone swarms and its lagging automotive effort than in maintaining its core business. This has allowed AMD to slip into ever greater design wins, and the firm is so much easier to work with than Intel, the OEMs are finding the grass on AMD’s side of the fence far greener.