AMD has made a variety of bulletins associated to its analysis and improvement efforts, and has publicly disclosed model new roadmap updates at its Financial Analyst Day occasion. While gentle on product-specific information, we now know lots about how the following few generations of CPUs and GPUs will form up, and the way AMD plans to realize market share within the essential enterprise and information centre markets. AMD has had an excellent run over the previous three years, principally because of the model new Zen CPU structure which got here after practically a decade of struggling to be aggressive. The firm introduced rising income, elevated profitability, and important market share features.
Describing AMD’s “multi-generational” graphics and compute roadmaps, CEO Dr Lisa Su outlined the corporate’s company technique and recapped successes throughout the shopper, cloud, gaming, and supercomputing areas lately. AMD has to this point shipped over 260 million Zen cores throughout its Ryzen, Threadripper, and Epyc processors since they debuted in 2017. For desktop customers, new CPUs codenamed ‘Vermeer’ based mostly on the Zen three structure, will launch this yr and in 2021, relying on market phase, as anticipated.
The subsequent technology of Epyc processors, based mostly on the Zen three structure and codenamed ‘Milan’, are anticipated to be launched in late 2020. AMD additionally introduced that work on the following Epyc ‘Genoa’ technology based mostly on Zen Four cores is progressing, and these CPUs are anticipated to be manufactured on a brand new 5nm course of, to hit the market by 2022.
AMD’s scalable Infinity Fabric interconnect, which has allowed for an environment friendly modular chiplet-based design, is now generally known as Infinity Architecture. The third technology will optimise coherent reminiscence sharing between a CPU and GPU for improved efficiency in addition to simplified software program improvement. Future CPUs shall be manufactured utilizing a die stacking course of that AMD is looking ‘X3D’ which might end in as much as 10x will increase in bandwidth density.
As for graphics, we are able to count on new high-performance flagship-grade graphics playing cards in the direction of the tip of 2020, based mostly on the second-gen RDNA (Radeon DNA) structure. AMD guarantees hardware-accelerated ray tracing in addition to variable price shading, in addition to a 50 % enchancment in energy effectivity.
This additionally sheds some gentle on what we are able to count on from Sony and Microsoft’s next-gen sport consoles, each based mostly on semi-custom AMD CPUs and CPUs, and each anticipated to launch this yr. With a standard {hardware} and software program ecosystem, sport builders will have the ability to goal PCs and consoles with much less effort. AMD additionally disclosed that third-generation RDNA will succeed it in or close to 2022.
Alongside the RDNA structure for client gaming graphics playing cards, AMD will introduce a brand new CDNA (Compute DNA) product line for information centre compute workloads. Products will launch this yr and goal machine studying purposes within the close to future, with exascale-class supercomputing purposes on the roadmap for the longer term. This will permit the corporate to develop targeted merchandise for this essential market, which could be deployed along side Epyc server CPUs.
That ties in with a current announcement by AMD that it has a received a contract to provide a brand new supercomputer known as El Capitan for the Lawrence Livermore National Laboratory which shall be extra highly effective than the world’s present prime 200 supercomputers mixed. As the world’s strongest supercomputer, it will likely be the primary to interrupt the two Exaflop barrier. It shall be based mostly on future Zen 4-based Epyc ‘Genoa’ CPUs, Radeon Instinct GPUs, and the Infinity Architecture interconnect. It shall be accomplished in early 2023.
This post was last modified on March 6, 2020 9:56 pm