Intel introduces the MAX series of CPUs and GPUs that includes Sapphire Rapids-HBM and Ponte Vecchio

Intel Max Series delivers impressive memory bandwidth and performance to HPC and AI

The new range of products delivers a high-bandwidth memory CPU, 4.8 times faster than competitors, and Intel’s highest-density GPU, to solve the world’s biggest challenges.

What’s new: Prior to Supercomputing 22 in Dallas, Intel introduced the Intel Max Series family of products with two flagship products for High Performance Computing (HPC) and artificial intelligence (AI): Intel® Xeon® CPU Max Series (codenamed Sapphire Rapids HBM) and Intel® Data Center GPU Max Series (codenamed Ponte Vecchio). The new products will power the upcoming Aurora supercomputer at Argonne National Laboratory, with updates sharing its post today.

The Xeon Max CPU is the first and only x86-based processor with high-bandwidth memory, which speeds up many HPC workloads without having to change code. The Max Series GPU is Intel’s highest density processor, packing more than 100 billion transistors into a 47-segment package with up to 128 gigabytes (GB) of high-bandwidth memory. The oneAPI open software ecosystem provides a single programming environment for each of the new processors. Intel 2023 oneAPI and AI tools will provide capabilities to enable advanced features of Intel Max Series products.

“To ensure that the HPC workload is not left behind, we need a solution that maximizes bandwidth, increases compute, increases developer productivity, and ultimately increases impact. The Intel Max Series family of products brings high-bandwidth memory to the broader market, Combined with oneAPI, it makes it easier to share code between CPUs and GPUs and solve the world’s biggest challenges faster.”

— Jeff McPhee, Corporate Vice President and General Manager, Super Compute Group at Intel

why does it matter: High Performance Computing (HPC) is the cutting edge of technology, using the latest innovations at scale to solve the biggest challenges of science and society, from mitigating the effects of climate change to treating the world’s deadliest diseases.

Max Series products meet the needs of this community with scalable and balanced CPUs and GPUs, which include breakthroughs in memory bandwidth, and unified by oneAPI, an open, standards-based, multi-architecture programming framework. Researchers and companies will solve problems faster and more sustainably with Max Series products.

Upon arrival: Max Series products are scheduled to launch in January 2023. In fulfillment of its commitments to customers, Intel is shipping blades with Max Series GPUs to the Argonne National Laboratory for Aurora Supercomputers, and will deliver Xeon Max CPUs to Los Alamos National Laboratory, Kyoto University and other locations. other supercomputing.

What the Intel Xeon Max CPU offers: The Xeon Max CPU delivers up to 56 performance cores built from four squares connected using Intel’s Embedded Multiple Interconnect Bridge (EMIB) technology, in a 350W envelope. Xeon Max CPUs feature 64 GB of high-bandwidth packet memory, as well as PCI Express 5.0 and CXL1.1 I/O. Xeon Max CPUs will provide more than 1 GB of high-bandwidth memory (HBM) capacity per Center, enough to fit most common HPC workloads. Max Series CPU delivers up to 4.8 times better performance compared to real HPC workloads.¹

• Uses 68% less power than the AMD Milan-X range for the same HPCG performance.
• AMX Extensions boost AI performance and deliver 8x higher throughput compared to AVX-512 for INT8 with INT32 stacks.²
• Provides flexibility to run in different HBM and DDR memory configurations.
• Workload Standards:
  • Climate modeling: 2.4x faster than AMD Milan-X on MPAS-A using only HBM.
  • Molecular Dynamics: At DeePMD, performance is improved 2.8 times over competing products with DDR memory.

What Intel Max Series GPU Offers: Max Series GPU offers up to 128X^eHPC cores, the new foundational architecture that targets the most demanding computing workloads. In addition, the Max Series GPU features:

• 408MB L2 cache – the highest in the industry – and 64MB L1 cache for increased productivity and performance.
• The only HPC/AI GPU with native ray tracing acceleration, designed to accelerate scientific visualization and animation.
• Workload Standards:
  • Funding: 2.4 times the performance compared to NVIDIA’s A100 on Riskfuel credit option pricing.
  • Physics: 1.5x improvement over the A100 to simulate the NekRS virtual reactor.

Max Series GPUs will be available in several variants to meet the needs of different customers:

• Max Series 1100 GPU: 300W dual-width PCIe card with 56 X^e cores and 48 GB of HBM2e memory. Multiple cards can be connected via Intel Xe Link bridges.
• Max Series 1350 GPU: 450W OAM with 112 X^e cores and 96 GB of HBM.
• Max Series 1550 GPU: Intel’s Maximum Performance 600W CPU (OAM) with 128X^e cores and 128 GB of HBM.

In addition to individual cards and modules, Intel will offer the Intel Data Center GPU Max Series subsystem with an x4 GPU OAM bus board and Intel Xe Link to enable high-performance communication between multiple GPUs within the subsystem.

Intel Max GPU Series, Source: Phoronix

What Max Series Products Enable: In 2023, the Aurora supercomputer, currently under construction at Argonne National Laboratory, is expected to become the first supercomputer to exceed 2 exaflops of peak dual-precision computing performance.³. Aurora will also be the first to showcase the power of pairing Max Series GPUs and CPUs into a single system, with more than 10,000 blades, each containing six Max Series GPUs and two Xeon Max CPUs.

Prior to SC22, Argonne and Intel unveiled Sunspot, the Aurora test development system consisting of 128 production codes. Researchers from Aurora Early Science will have access to the system starting in late 2022.

Max Series products will power many other HPC systems critical to national security and basic research, including Crossroads at Los Alamos National Laboratory, CTS-2 systems at Lawrence Livermore National Laboratory and Sandia National Laboratory, and Camphor 3 at Kyoto University.

What’s Next: At Supercomputing ’22, Intel and its customers will showcase more than 40 upcoming system designs from 12 OEMs using Max Series products. Attendees can explore demos showcasing the performance and capabilities of Max Series products for a range of AI and HPC applications, as well as hear from Intel architects, customers, and end users about the power of Intel’s platform solutions at Intel Booth #2428. More information about Intel’s activities is available in SC22.

The Intel Data Center Max Series GPU, codenamed Rialto Bridge, is the successor to the Max Series GPU and aims to arrive in 2024 with improved performance and a smooth upgrade path. Intel then plans to launch a major new architecture innovation to enable the future of HPC. The company’s upcoming XPU, codenamed Falcon Shores, will combine the X^e and x86 on one package. This groundbreaking new architecture will also have the flexibility to integrate new IP addresses from Intel and customers, manufactured using our IDM 2.0 model.