Intel Announces Max Series ‘Sapphire-Rapids’ CPUs & ‘Ponte Vecchio’ GPUs

Just recently, Intel announced their new Max Series targeted for data centers. This includes Intel’s upcoming Sapphire-Rapids HEDT server CPU and the new Xe-HPC based Max GPUs. Team blue’s new CPUs are tipped to offer 4.8x more bandwidth than the current competition. 

Xeon Max CPU

Intel’s new Max CPU is the world’s first x86 CPU to ship with high-bandwidth memory. The Xeon Max CPU features 56 Performance cores. These cores span across 4 tiles which are connected through Intel’s embedded multi-die interconnect bridge in a 350W package.  The Xeon Max CPU contains 64GB of high bandwidth memory along with support for PCI Express Gen 5 and CXL1.1 Input/Output.

Intel also provided some benchmarks for their upcoming Max Series CPU;

• 68% less power usage than an AMD Milan-X cluster for the same HPCG performance.
• AMX extensions boost AI performance and deliver 8x peak throughput over AVX-512 for INT8 with INT32 accumulation operations.
• Provides flexibility to run in different HBM and DDR memory configurations.
• Workload benchmarks:
  • Climate modeling: 2.4x faster than AMD Milan-X on MPAS-A using only HBM.
  • Molecular dynamics: On DeePMD, 2.8x performance improvement against competing products with DDR memory.

Data Center GPU Max

After that, we have Intel’s new Data Center Max GPUs. Based on the Xe-HPC architecture, they offer up to 128 Xe Cores specifically designed for handling high-speed computing. In addition, we see 408MB of L2 cache which is termed the highest in the industry along with 64MB of L1 cache. What’s better is that these GPUs support ray tracing which can help accelerate scientific visualizations. 

The Max GPUs will arrive in different form factors to suit the customers’ needs. 

• Max Series 1100 GPU: A 300-watt double-wide PCIe card with 56 X^e cores and 48GB of HBM2e memory. Multiple cards can be connected via Intel Xe Link bridges.
• Max Series 1350 GPU: A 450-watt OAM module with 112 X^e cores and 96GB of HBM.
• Max Series 1550 GPU: Intel’s maximum performance 600-watt OAM module with 128 X^e cores and 128GB of HBM.

Performance-wise, these GPUs offer significant uplifts in performance as compared to their competition;

• Climate modeling: 2.4x faster than AMD Milan-X on MPAS-A using only HBM.
• Molecular dynamics: On DeePMD, 2.8x performance improvement against competing products with DDR memory.

Apart from the individual products, Intel also introduced an all-encompassing subsystem featuring 4 OAM modules. 

Release Date

The Intel Max series is slated to launch in January 2023. In the same year, the Aurora supercomputer which is currently under production will make use of Intel’s new Max lineup. Boasting the performance numbers, Intel claims that this supercomputer will surpass 2 exaflops of peak double-precision performance. Similarly, the successor to the Max Series GPU codenamed ‘Rialto Bridge‘ is expected to arrive in 2024. 

Source : Intel