The QCT DevCloud, in collaboration with Intel^®, is designed to help developers install, develop, test, and optimize their HPC/AI applications on QCT’s latest hardware platforms. This platform provides accessibility to QCT’s computing, storage, and networking supporting the latest Intel generation of CPU innovations such as the Intel^® Xeon^® CPU Max and Intel’s cutting-edge technologies and toolkits. The Intel^® Xeon^® CPU Max series is Intel’s first and only x86 processor that features high-bandwidth memory (HBM), allowing for accelerated performance of HPC/AI workloads. Leveraging the full potential of Intel’s latest hardware empowers the QCT DevCloud systems with a comprehensive HPC/AI environment and drives advancements in HPC and AI technology.

In recent years, with the rapid development of HPC/AI technologies, a demand for processing large amounts of data to perform complex calculations in parallel has been generated. HBM has emerged as a solution that can provide extremely high bandwidth and low latency, accelerating data access and processing speeds, and enhancing the training and inferencing efficiency of AI models. The HBM stack in these processors can be used inline with traditional DRAM, as tiered memory, or even standalone without any DIMM modules at all (see Fig. 1).

Fig. 1 First and only x86 CPU with HBM, the Intel^® Xeon^® CPU Max series

With that said, the Intel^® Xeon^® CPU Max series offers 56 P-cores across 4 tiles and incorporates 64GB of on-package HBM2e memory divided into 4x clusters of 16GB for a net total of 1 TB/s memory bandwidth. Its power target is also in the 350-watt range, and there is support for PCIe 5.0 and Compute Express Link (CXL) 1.1 (see Fig. 2 below). These processors are also HPC optimized and placed in a higher tier over the 4th Gen Intel^® Xeon^® Scalable Bronze, Silver, Gold and Platinum processors, which is why it is in the “Intel^® Max series” product family.

Fig. 2 Intel^® Xeon^® CPU Max series features 

In QCT’s testing scenario, we conducted a comparison test between 4th Gen Intel^® Xeon^® Scalable processors and Intel^® Xeon^® CPU Max, the memory performance of Intel^® HBM’s on-board memory was found to be 1.2x to 1.5x better than 4th Gen Intel^® Xeon^® Scalable processors in a memory-bound HPC workload scenario. This demonstrates the improvement HBM brings to memory-intensive tasks, accelerating data access and reducing bottlenecks in real-world memory-bound workloads.

Developing and testing HPC/AI applications can be difficult due to the fact that developers must install, test, and troubleshoot using a wide variety of hardware, frameworks, and software toolkits. QCT DevCloud provides developers with remote access to precompiled workloads, development toolkits, and multiple QCT hardware platforms with the latest technology such as Intel^® Xeon^® CPU Max series. Therefore, developers can seamlessly streamline their processes in QCT’s comprehensive cluster environment, enhancing time and cost efficiency.

Fig. 3 QCT DevCloud Benefits for Developers
Join the QCT DevCloud Program and bring your ideas to life.