Moor Insights & Strategy provides or has provided paid (wish services to technology companies, like all tech industry research and analyst firms. That, in turn, is good for all of us, from developers to data scientists to anyone impacted by AI being everywhere. This competition has helped sharpen Intel and accelerated its efforts to restart its innovation engine. AMD continues to deliver on generation after generation of processors with EPYC, and Ampere has come on very strong with Altra and AmpereOne. Intel’s competitors aren’t going anywhere. (Note the company's $100 billion investment in manufacturing globally, including in Oregon, Arizona, New Mexico and Ohio.) There's plenty of room for everybody.Īs Intel Innovation came and went, I walked away with a renewed appreciation for how the company has reestablished itself in the datacenter space in both design and manufacturing. ![]() The siliconomy is huge, and it's growing exponentially. And while Intel, AMD, Nvidia and Arm players such as Ampere each fight for market share, I believe there also must be an eye on interoperability and a commitment to embracing open standards that allow customers to choose without compromise. Regardless, this is the world we live in. It may be a training environment completely different from the inferencing environment. Maybe it's a CPU and accelerator from other vendors. While Intel has a solution that maps to each of the unique requirements in an AI stack, chances are that many organizations will use different technology blocks to deliver a performance and cost profile that best meets their needs. When you think about setting up an AI environment, there is a lot of specialized hardware supporting training (machine learning) and inferencing environments. This makes what Intel has been doing in bringing AI everywhere all the more impressive.įoes need to be friends in the “siliconomy” Understanding the upstream (or up-the-stack) requirements of platforms, tools, software and data and how these influence design is even more overwhelming. Even as an analyst immersed in the semiconductor space, it is still a bit overwhelming to fully comprehend everything that gets done at the silicon level to enable complex workloads such as AI. In addition to Developer Cloud, Intel demonstrated the full breadth of its AI everywhere enablement-from the CPU to manufacturing to UCIe (which enables interconnectivity of silicon providers on the same die) to glass-based wafers. Intel brings all of the elements to help customers play in the "siliconomy." Moor Insights & Strategy It’s a fully functioning environment that allows developers to build, test and tweak their datacentric applications on the full range of Intel technologies-CPUs, GPUs and FPGAs married with Intel's full suite of software libraries and tools. Finally, it enables architects and developers to experiment, build and optimize AI for their specific needs.Īt this year’s Innovation event, Intel announced the general availability of Intel Developer Cloud, a virtual laboratory enabling the scenario I described above. It's about all of this silicon and infrastructure supporting the software ecosystem-languages, models, frameworks and all of the elements of a deployment. It's not CPUs, GPUs and other custom silicon designed for advanced workloads. ![]() It takes more than chips to drive the “siliconomy”ĭuring his keynote, Gelsinger coined the phrase “siliconomy” in reference to how the modern economy is being driven by technology-a shift away from the forces that drove the global economy over the past several decades.īut this siliconomy isn't just about silicon. This was a smart strategy that puts Xeon in a stronger competitive position as Intel closes the core count/core performance gap, something it will achieve with Granite Rapids. Rather than trying to get into a “cores race” with its competition, it focused on driving workload acceleration through discrete silicon engines that were part of the compute complex. ![]() While this CPU was delayed by manufacturing snafus, Intel took a novel approach in its design. However, I think the company more subtly showed this with the launch of its 4th Gen Xeon processor, codenamed Sapphire Rapids. The Sierra Forest announcement exemplifies how I see Intel being back in its innovation groove. The company has gone from trailing in the core-count race by a significant margin to taking the top spot in just a single generation. While there are a lot of questions around performance, power, I/O etc., this is still quite an impressive accomplishment. This doesn't just put Sierra Forest ahead of AMD and Ampere-it beats them by a country mile. That Sierra Forest chip? Yeah, that's going to support up to 288 cores, not 144. ![]() Fast forward three weeks or so, and Intel flipped the script at Innovation 2023 with a surprise reveal.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |