Intel Data Center GPU Max 1550, ARK specifications page (first-party)

VERBATIM 2026-10-04 via web_extract. This is Intel's OWN ARK page for the Max 1550; accelerators/xpu-intel-dc-gpu-max-1550 cites only two GitHub READMEs and no Intel document. ESSENTIALS: Product Collection 'Intel® Data Center GPU Max Series' | Microarchitecture 'Xe-HPC' | Code Name 'Products formerly Ponte Vecchio' | Marketing Status Launched | Launch Date "Q1'23" | 'Expected Discontinuance Jan 2026' | Warranty Period 3 yrs | Use Conditions Server/Enterprise | 'Usecase Artificial Intelligence, High Performance Computing'. GPU SPECIFICATIONS: 'Xe-cores 128' | 'Ray Tracing Units 128' | 'Intel® Xe Matrix Extensions (Intel® XMX) Engines 1024' | 'Xe Vector Engines 1024' | 'Graphics Max Dynamic Clock 1600 MHz' | 'Graphics Base Clock 900 MHz' | 'Intel® Xe Link Maximum Frequency 53 Gbps' | 'TDP 600 W' | 'PCI Express Configurations ‡ Gen 5 x16' | 'Device ID 0x0BD5'. MEMORY SPECIFICATIONS: 'Memory Size 128 GB' | 'Memory Type HBM2e' | 'Graphics Memory Interface 1024 bit' | 'Graphics Memory Bandwidth 3276.8 GB/s'. I/O SPECIFICATIONS: '# of Displays Supported‡ 0'. NO COMPUTE FIGURES ON THIS PAGE: like the Flex 170 ARK page, this page publishes no TFLOPS and no TOPS row. Verified absence, not an extraction failure: the GPU Specifications table above is complete as rendered and contains no compute-throughput row. SPARSITY CONVENTION: no sparsity footnote, no row titled 'with Structured Sparsity', no sparse/dense pair anywhere on the page. Every figure is DENSE as published and must NOT be halved. This is the documented Intel convention - sparsity is labelled only on rows explicitly titled for it - and here there is no such row.