AMD latest processor architecture may be trademarked Zen, but that’s not the laid-back approach the chip designer is taking with its Silicon Valley competitors. Rather, it’s pushing out new silicon and graphics engines for computing and gaming.
Is it a half-century of punching above its price point? Or, a jab at the market leaders?
But when AMD’s chief executive Lisa Su delivered the company’s first-ever keynote addressOpens a new window at the long-running Computex expo in Taiwan last week, she took shots at rivals Intel and Nvidia as she detailed design innovations for chips and chipsets that reach from data center to the desktop.
Among them are the new EPYC and Ryzen central processing units (CPUs) baked into forthcoming graphics cards built on a motherboard that AMD says makes better use of its size and shape as well as device real estate.
To press the point, head-to-head tests in Su’s presentation showed a pre-production version of AMD’s second-generation EPYC server platform getting the better of machines running Intel’s Xeon 8280 silicon in performance benchmarking. In a game-play demonstration side-by-side with Nvidia’s RTX 2070 chipset, the AMD graphics cards built on the RDNA architecture delivered 1.25 times better clock speed by wresting 1.5 times more per-watt performance from the silicon.
Size and speed also figured into keynote speeches from Intel’s Gregory Bryant whose company dominates the original equipment manufacture (OEM) market for CPUs, and Nvidia’s Jeff Fisher whose company is the graphics processing unit (GPU) sales leader. Bryant brought long-awaited chips built on 10nm coresOpens a new window to the OEM market, while Fisher presented a software stack for the tracing of light rays in virtual environments that the company’s RTX GPUsOpens a new window are designed to accelerate.
Both Bryant and Fisher cited an emerging class of creators – designers for whom enhanced graphics engines are less toy and more tool – as driving the market for advanced chips and chipsets. While neither mentioned AMD, Intel did issue a releaseOpens a new window in the wake of Su’s event-opening address that took issue with some of her claims.
Those companies share a history that dates to second-source licensing agreements signed in 1982 that AMD broke when it chose to compete with Intel in the CPU market nine years later. Since then, AMD has aspired to be more than a producer of lower cost silicon designed on Intel’s x86 architecture.
Like the move to 7nm cores, AMD’s 2006 acquisition of GPU designer API acted as a means of diversification. Since then, AMD’s focus on marrying graphics engines with its processors has led it to a gaming market largely dominated by Nvidia.
Su told her audience at the 30-year-old trade fair that a twin focus on computing and user experience provides a wellspring for industry innovations that reaches back to the company’s founding, as Advanced Micro Devices, in 1959. The RDNA architecture built into its line of NAVI graphics cards melds PC and console gaming with cloud connections and computing to enhance processing power and lower energy consumption.
Fisher, who heads Nvidia’s PC business, said the same attributes underlie the Turing architecture now being delivered in OEM laptops that facilitate the ray tracing tech used in games and films. Turing’s 18 billion transistors let users push more instructions through a larger bus, he said, with artificial intelligence in Nvidia’s newly introduced Studio Stack boosting times for rendering and framing.
Introduced last year, Turing is helping Nvidia’s OEM partners access what Fisher said is a $15 billion market for gaming laptops. AMD is aiming for a piece of that market with third-generation Ryzen chips and Radeon graphics cards aimed at PC users.
Intel is targeting the emerging creator segment by marrying its 10nm Ice Lake architecture with Iris Plus graphics engines built on the same die, Bryant told his Computex audience. Intel has worked for years to bring Ice Lake to market, with a production shortage for the 14nm Coffee Lake silicon being blamed for a recent slowdown in PC sales.
Repeated delaysOpens a new window have ended with the launch of Intel’s 10th generation Core processors, but not before letting AMD make inroads with its Zen chips. In an effort to claw back some of that share, Bryant released details of Intel’s Project Athena innovation program aimed at helping OEMs bring laptops bearing the 10nm architecture.
AMD also is giving Intel a run for its money in the data center arena. In addition to EPYC chips providing the processing power for the world’s fastest supercomputer that Seattle-based CrayOpens a new window is building for the Oak Ridge Laboratory, Microsoft is achieving high-performance computing with AMD on its Azure cloud.
Microsoft executives joined Su on the Computex stage to praise the value of the first-generation EPYC platform in knitting together 11,500 processer cores for a fluid-dynamics computing simulation. The operating systems giant — which released virtual machines for high performance applications that run Intel x86 architectures in September 2018 — says that EPYC offers a performance advantageOpens a new window that the form factor has yet to match.