Performance Index

ID Date Classification
615781 11/27/2024 Public
Document Table of Contents

Vision 2024

Keynote

Number Speaker Session Claim Claim Details/Citation Testing Date
1 Pat Gelsinger Day 2 Keynote End to End RAG deployment on Intel® Gaudi® 2 and Intel® Xeon® based platforms provide up to ~1.5x better TCO compared to NVIDIA H100 SXM based platforms, when deployed with industry standard software. Gaudi2: 1-node, HLS-Gaudi2 with 8x Gaudi® 2 HL-225H and 2x Intel® Xeon® Platinum 8380 CPU, Total Memory 1TB, Ubuntu 22.04.03, Kernel 5.15.0, Test by Intel as of 04/02/24 April 2, 2024
2 Demo Presenter or Pat Gelsinger Day 2 Keynote Next token latency reduction of up to 6.5x on Next Generation Intel Xeon using MXFP4 over 4th Gen Xeon using FP16 H100: 1x Lambda labs instance gpu_​8x_​h100_​sxm5 with 8xH100 SXM and 2x Intel® Xeon® Platinum 8480 CPU, Total Memory 1.8TB, Ubuntu 20.04.6 LTS, Kernel 5.15.0, Test by Intel as of 04/02/24 March 27, 2024

Demos

Number Session Speaker Claim Claim Details/Citation Testing Date
1 Consolidation for Cloud-Scale Workloads

Vision Demo Showcase

Travis Vanderzanden Xeon 6 delivers a 1.45x performance per watt over 5th Gen Intel Xeon on media transcode workloads Baseline: 1-node, 2x INTEL(R) XEON(R) PLATINUM 8592+, 64 cores, HT On, Turbo On, NUMA 4, Integrated Accelerators Available [used]: DLB 2 [0], DSA 2 [0], IAA 2 [0], QAT 2 [0], Total Memory 512GB (16x32GB DDR5 5600 MT/s [5600 MT/s]), BIOS EGSDCRB1.SYS.0107.D20.2310211932, microcode 0xa10001d1, 1x Ethernet Controller I225-LM, 1x 1TB INTEL SSDSC2KW010T8, CentOS Stream 9, 6.2.0-emr.bkc.6.2.18.4.51.x86_​64, ffmpeg n4.4, x264 tag 5db6aa6cab1b146e07b60cc1736a01f21da01154, x265 3.1, SVT_​AV1 v0.9.1, SVT_​HEVC v1.5.1, GCC8.1, score=Frames per second (FPS). Tested by Intel on 12/11/2023. New: 1-node, 2x SRF CPUs, 144 cores, No HT, Turbo On, NUMA 2, Integrated Accelerators Available [used]: DLB 2 [0], DSA 2 [0], IAA 2 [0], QAT 2 [0], Total Memory 1024GB (16x64GB DDR5 6400 MT/s [6400 MT/s]), BIOS BHSDCRB1.IPC.0030.D85.2403122316, microcode 0x13000131, 1x I210 Gigabit Network Connection, 1x Micron_​7450_​MTFDKBG1T9TFR, CentOS Stream 9, 6.5.0-2023-08-28-intel-next-02186-g67af28a5c039, ffmpeg n4.4, x264 tag 5db6aa6cab1b146e07b60cc1736a01f21da01154, x265 3.1, SVT_​AV1 v0.9.1, SVT_​HEVC v1.5.1, GCC8.1, score=Frames per second (FPS). Tested by Intel on 03/29/2024 December 11, 2023 and March 29, 2024
2 Enhancing Generative AI - Business Relevant Results with RAG

Booth 310

Devin Ryles Gen on Gen Increased Vector DB throughput of 2.35x New: 1-node, pre-production platform with 2x (1x used) Next Generation Intel Xeon, HT On, Turbo On, Total Memory 3072GB 3072GB (DDR5 8800 MT/s [6400 MT/s]), microcode 0x21000200, Ubuntu 22.04.2 LTS, 5.15.0-101-generic, Test by Intel as of 03/2/24. Baseline: 1-node, 2x Intel(R) Xeon(R) Platinum 8592+, HT On, Turbo On, Total Memory 1024GB (DDR5 4800 MT/s), microcode 0xa10001c0, Ubuntu 22.04.2 LTS, 5.15.0-101-generic, Test by Intel as of 03/27/24. Software: Redis 7.2 100m vector similarity search, HNSW ef512, 1 Redis instance, 240 shards per instance, 48 users per instance March 27, 2024
3 Consolidation for Cloud-Scale Workloads

Vision Demo Showcase

Travis Vanderzanden Up to 2.7X More Performance per Rack New: 1-node, pre-production platform with 1x Next Generation Intel Xeon, 144 cores, HT Off, Turbo On, NUMA 2, Total Memory 512GB 5600MT/s, microcode 0x81000183, Ubuntu 22.04 LTS, 5.15.0-27-generic, 9x Ethernet Controller E810-C for QSFP, 1x I210 Gigabit Network Connection, 1x 223.6G INTEL SSDSC2KB240GZ, BIOS BHSDCRB1.86B.0026.D88.2310041923, Integrated Accelerators Available [used]: DLB 0 [0], DSA 4 [0], IAA 4 [0], QAT 0 [0], Test by Intel as of 02/02/24. Baseline: 1-node, 2x Intel(R) Xeon(R) Gold 6252N CPU, 24 cores, HT On, Turbo Off, Total Memory 96GB DDR 2934 MT/s, microcode 0x5003003, Ubuntu 18.04.3 LTS, 4.15.0-132-generic, 2x Intel E810-CQDA2 (CVL, Total - 4x100G ports), 1x Intel 240GB SSD, BIOS 3.3, Integrated Accerators: none, Test by Intel as of 02/02/24. Tools Versions: GCC 7.5.0, DPDK 20.11 February 2, 2024