AMD and Intel Trade Blows in AI Chip Battle -- Virtualization Review

AMD and Intel Trade Blows in AI Chip Battle

By David Ramel
10/10/2024

Prominent chipmakers AMD and Intel are duking it out as they chase NVIDIA for AI hardware supremacy, this week trading announcements about their latest/greatest offerings while holding dueling AI events.

Intel, fresh on the heels of announcing new Xeon 6 processors and Gaudi 3 AI accelerators, revealed it was changing the name of its Intel Tiber Developer Cloud to Intel Tiber AI Cloud with the mission to "Speed up AI development using Intel-optimized software on the latest Intel Core Ultra processor, Intel Xeon processor, Intel Gaudi AI Accelerator, and GPU compute."

Then AMD launched its own AI processors and accelerators two days later (today Oct. 10) at the same time Intel delivered a follow-up blow, announcing the first AI PC Intel Core Ultra desktop processors.

And Intel this week kicked off its fourth-annual AI Global Impact Festival, while AMD held its Advancing AI 2024 conference.

And so it goes as they both chase NVIDIA's coattails.

AMD announced a slew of developments around new chips, CPUs and accelerators, along with showcasing partner solutions, the growth of its open-source AI software and more.

"The data center and AI represent significant growth opportunities for AMD, and we are building strong momentum for our EPYC and AMD Instinct processors across a growing set of customers," said AMD Chair and CEO Dr. Lisa Su. "With our new EPYC CPUs, AMD Instinct GPUs and Pensando DPUs we are delivering leadership compute to power our customers' most important and demanding workloads. Looking ahead, we see the data center AI accelerator market growing to $500 billion by 2028. We are committed to delivering open innovation at scale through our expanded silicon, software, network and cluster-level solutions."

The company's Advancing AI site shows press releases for AI PRO 300 Series processors, AMD Instinct MI325X accelerators, 5th Gen AMD EPYC CPUs and much more.

Intel's Sept. 24 announcement, meanwhile, unveiled two highlights:

The launch of Xeon 6 with Performance-cores (P-cores), doubling the performance for AI and HPC workloads.
New Gaudi 3 AI accelerators that offer up to 20 percent more throughput and 2x price/performance vs H100 for inference of LLaMa 2 70B.

"Demand for AI is leading to a massive transformation in the data center, and the industry is asking for choice in hardware, software and developer tools," said Justin Hotard, Intel executive vice president and general manager of the Data Center and Artificial Intelligence Group. "With our launch of Xeon 6 with P-cores and Gaudi 3 AI accelerators, Intel is enabling an open ecosystem that allows our customers to implement all of their workloads with greater performance, efficiency and security."

The company also this week explained how its newly named Intel Tiber AI Cloud works, detailing its building blocks of silicon, systems, software, and services:

Silicon:
- Intel Tiber AI Cloud is first-in-line and Customer Zero for the latest Intel CPUs, GPUs, and AI accelerators, with direct support channels into Intel engineering.
- As a result, Intel's AI cloud can offer early access to the latest Intel hardware and software at a low cost, which may not always be generally available yet.
Systems:
- The demands of compute infrastructure rapidly evolve due to increasing AI workload scale and performance requirements.
- Intel Tiber AI Cloud offers various computing options to deploy AI workloads, including containers, virtual machines, and dedicated systems. In addition, we recently introduced small and super-compute scale Intel Gaudi clusters with high-speed Ethernet networking for foundation model training.
Software:
- Intel supports the growing open software ecosystem for AI and actively contributes to open source projects such as Pytorch and UXL Foundation.
- Dedicated Intel cloud computing is allocated to support open source projects, enabling users to access the latest AI frameworks and models in Intel Tiber AI Cloud.
Services:
- Enterprise users and developers consume computing AI services differently, depending on technical expertise, implementation requirements, and use cases.
- Intel Tiber AI Cloud offers server CLI access, curated interactive access via notebooks, and serverless API access to meet these diverse access requirements.