Performance Index

ID Date Classification
615781 10/10/2024 Public
Document Table of Contents

2nd Generation Intel® Xeon® Scalable Processors

Performance varies by use, configuration and other factors.

Performance results are based on testing as of dates shown in configurations and may not reflect all publicly available updates. See configuration disclosure for details. No product or component can be absolutely secure.

Intel optimizations, for Intel compilers or other products, may not optimize to the same degree for non-Intel products.

Estimates of SPECrate®2017_​int_​base and SPECrate®2017_​fp_​base based on Intel internal measurements. SPEC® , SPECrate® and SPEC CPU® are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information.

Claim Processor Family System Configuration Measurement Measurement Period
1.) Up to 33% Average Generational Gains (1.33x) on Intel® Xeon® Gold Processor Mainstream CPUs 2nd Generation Intel® Xeon® Gold processor Baseline: 1-node, 2x Intel® Xeon® Gold 5118 cpu on Wolf Pass with 384 GB (12 X 32GB 2666 (2400)) total memory, ucode 0x200004D on RHEL7.6, 3.10.0-957.el7.x86_​64, IC18u2, AVX2, HT on all (off Stream, LINPACK), Turbo on, result: SPECrate2017_​int_​base(est)=119, SPECrate2017_​fp_​base(est)=134, STREAM-Triad=148.6, LINPACK=822, server-side Java=67434, test by Intel on 11/12/2018. New configuration: 1-node, 2x Intel® Xeon® Gold 5218 cpu on Wolf Pass with 384 GB (12 X 32GB 2933 (2666)) total memory, ucode 0x4000013 on RHEL7.6, 3.10.0-957.el7.x86_​64, IC18u2, AVX2, HT on all (off Stream, LINPACK), Turbo on, result: SPECrate2017_​int_​base(est)=162, SPECrate2017_​fp_​base(est)=172, STREAM-Triad=185, LINPACK=1088, server-side java=98333, test by Intel on 12/7/2018. Geomean of SPECrate2017_​int_​base(est), SPECrate2017_​fp_​base(est), STREAM-Triad, Intel® Distribution of LINPACK, server-side Java*

Baseline: November 12, 2018

New: December 7, 2018

2.) 2x Average Generational Gains 2nd Generation Intel® Xeon® Platinum processor Baseline: 1-node, 2x Intel® Xeon® Platinum 8180 cpu on Wolf Pass with 384 GB (12 X 32GB 2666) total memory, ucode 0x200004D on RHEL7.6, 3.10.0-957.el7.x86_​64, IC19u1, AVX-512, HT on all (off Stream, LINPACK), Turbo on all (off Stream, LINPACK), result: SPECrate2017_​int_​base(est)=307, SPECrate2017_​fp_​base(est)=251, STREAM-Triad=204, LINPACK=3238, server-side Java=165724, test by Intel on 1/29/2019. New configuration: 1-node, 2x Intel® Xeon® Platinum 9282 cpu on Walker Pass with 768 GB (24x 32GB 2933) total memory, ucode 0x400000A on RHEL7.6, 3.10.0-957.el7.x86_​64, IC19u1, AVX-512, HT on all (off Stream, LINPACK), Turbo on all (off Stream, LINPACK), result: SPECrate2017_​int_​base(est)=635, SPECrate2017_​fp_​base(est)=526, STREAM-Triad=407, LINPACK=6411, server-side Java=332913, test by Intel on 2/16/2019. Geomean of SPECrate2017_​int_​base(est), SPECrate2017_​fp_​base(est), STREAM-Triad, Intel® Distribution of LINPACK, server-side Java*

Baseline: January 29, 2019

New: February 16, 2019

3.) 30x Inference Throughput Improvement on Intel® Xeon® Platinum 9282 processor with Intel® Deep Learning Boost (Intel DL Boost) 2nd Generation Intel® Xeon® Platinum processor New configuration: Tested by Intel as of 2/26/2019. Platform: Dragon Rock 2 socket Intel® Xeon® Platinum 9282 processor (56 cores per socket), HT ON, turbo ON, Total Memory 768 GB (24 slots/ 32 GB/ 2933 MHz), BIOS:SE5C620.86B.0D.01.0241.112020180249, CentOS 7 Kernel 3.10.0-957.5.1.el7.x86_​64, Deep Learning Framework: Intel® Optimization for Caffe* version: https://github.com/intel/caffe d554cbf1, ICC 2019.2.187, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a), model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/resnet50_int8_full_conv.prototxt, BS=64, No datalayer syntheticData:3x224x224, 56 instance/2 socket, Datatype: INT8 vs Baseline: Tested by Intel as of July 11th 2017: 2S Intel® Xeon® Platinum 8180 processor CPU @ 2.50GHz (28 cores), HT disabled, turbo disabled, scaling governor set to “performance” via intel_​pstate driver, 384GB DDR4-2666 ECC RAM. CentOS Linux* release 7.3.1611 (Core), Linux* kernel 3.10.0-514.10.2.el7.x86_​64. SSD: Intel® SSD Data Center S3700 Series (800GB, 2.5in SATA 6Gb/s, 25nm, MLC). Performance measured with: Environment variables: KMP_​AFFINITY='granularity=fine, compact‘, OMP_​NUM_​THREADS=56, CPU Freq set with cpupower frequency-set -d 2.5G -u 3.8G -g performance. Caffe: (http://github.com/intel/caffe/ ), revision f96b759f71b2281835f690af267158b82b150b5c. Inference measured with “caffe time --forward_​only” command, training measured with “caffe time” command. For “ConvNet” topologies, synthetic dataset was used. For other topologies, data was stored on local storage and cached in memory before training. Topology specs from https://github.com/intel/caffe/tree/master/models/intel_optimized_models (ResNet-50). Intel® C++ compiler ver. 17.0.2 20170213, Intel® Math Kernel Library (Intel® MKL) small libraries version 2018.0.20170425. Caffe run with “numactl -l“. Intel® Deep Learning Boost (Intel DL Boost)

Baseline: July 11, 2017

New: February 26, 2019

4.) Up to 14x Improvement in Inference Performance on Intel® Xeon® Platinum 8280 processor with Intel® Deep Learning Boost (Intel DL Boost) 2nd Generation Intel® Xeon® Platinum processor New configuration: Tested by Intel as of 2/20/2019. 2 socket Intel® Xeon® Platinum processor 8280, 28 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0271.120720180605 (ucode: 0x200004d), Ubuntu 18.04.1 LTS, kernel 4.15.0-45-generic, SSD 1x sda INTEL SSDSC2BA80 SSD 745.2GB, nvme1n1 INTEL SSDPE2KX040T7 SSD 3.7TB, Deep Learning Framework: Intel® Optimization for Caffe* version: 1.1.3 (commit hash: 7010334f159da247db3fe3a9d96a3116ca06b09a), ICC version 18.0.1, MKL DNN version: v0.17 (commit hash: 830a10059a018cd2634d94195140cf2d8790a75a, model: https://github.com/intel/caffe/blob/master/models/intel_optimized_models/int8/resnet50_int8_full_conv.prototxt, BS=64, synthetic Data, 4 instance/2 socket, Datatype: INT8 vs Baseline: Tested by Intel as of July 11th 2017: 2S Intel® Xeon® Platinum 8180 processor CPU @ 2.50GHz (28 cores), HT disabled, turbo disabled, scaling governor set to “performance” via intel_​pstate driver, 384GB DDR4-2666 ECC RAM. CentOS Linux* release 7.3.1611 (Core), Linux* kernel 3.10.0-514.10.2.el7.x86_​64. SSD: Intel® SSD DC S3700 Series (800GB, 2.5in SATA 6Gb/s, 25nm, MLC).Performance measured with: Environment variables: KMP_​AFFINITY='granularity=fine, compact‘, OMP_​NUM_​THREADS=56, CPU Freq set with cpupower frequency-set -d 2.5G -u 3.8G -g performance. Caffe: (http://github.com/intel/caffe/ ), revision f96b759f71b2281835f690af267158b82b150b5c. Inference measured with “caffe time --forward_​only” command, training measured with “caffe time” command. For “ConvNet” topologies, synthetic dataset was used. For other topologies, data was stored on local storage and cached in memory before training. Topology specs from https://github.com/intel/caffe/tree/master/models/intel_optimized_models (ResNet-50). Intel® C++ compiler ver. 17.0.2 20170213, Intel® Math Kernel Library (Intel® MKL) small libraries version 2018.0.20170425. Caffe run with “numactl -l“. Intel® Deep Learning Boost (Intel DL Boost)

Baseline: July 11, 2017

New: February 20, 2019

5.) Up to 3.7x avg Gain w/Intel® Xeon® Platinum 9242 Processor Vs 3-Year Old Server 2nd Generation Intel® Xeon® Platinum processor HPCG, Binary included MKL 2019u1, HT=ON, Turbo=OFF, 1 thread per corescore: 23.78. HPL 2.1, HT=ON, Turbo=OFF, 2 threads per corescore: 1204.64. WRF 3.9.1.1, conus-2.5km, HT=ON, SMT=ON, 1 thread per corescore: 4.54. OpenFOAM 6.0, 42M_​cell_​motorbike, HT=ON, Turbo=OFF, 1 thread per corescore: 3500. LS-Dyna 9.3-Explicit AVX2 binary, 3car, HT=ON, SMT=ON, 1 thread per corescore: 2814. VASP 5.4.4, CuC, HT=ON, Turbo=OFF, 1 thread per corescore: 384.99. NAMD 2.13, apoa1, HT=ON, Turbo=OFF, 2 threads per corescore: 4.4. LAMMPS version 12 Dec 2018, Water, HT=ON, Turbo=ON, 2 threads per corescore: 54.72. Black Scholes, HT=ON, Turbo=ON, 2 threads per corescore: 2573.77. Monte Carlo, HT=ON, Turbo=ON, 2 threads per corescore: 43.2.  New configuration: Intel® Xeon® 9242 processor: Intel reference platform with 2S Intel® Xeon® 9242 processors (2.2GHz, 48C), 24x16GB DDR4-2933, 1 SSD, Cluster File System: 2.12.0-1 (server) 2.11.0-14.1 (client), BIOS: PLYXCRB1.86B.0572.D02.1901180818, Microcode: 0x4000017, CentOS 7.6, Kernel: 3.10.0-957.5.1.el7.x86_​64, OFED stack: OFED OPA 10.8 on RH7.5 with Lustre v2.10.4, HBA: 100Gbps Intel® Omni-Path Architecture (Intel® OPA) 1 port PCIe x16, Switch: Intel® Omni-Path Edge Switch (Intel® OP Edge Switch) 100 Series 48 Port. STREAM OMP 5.10 , Triad, HT=ON, Turbo=OFF, 1 thread per corescore: 407. HPCG, Binary included MKL 2019u1, HT=ON, Turbo=OFF, 1 thread per corescore: 81.91. HPL 2.1, HT=ON, Turbo=OFF, 2 threads per corescore: 5314. WRF 3.9.1.1, conus-2.5km, HT=ON, SMT=ON, 1 thread per corescore: 1.44. OpenFOAM 6.0, 42M_​cell_​motorbike, HT=ON, Turbo=OFF, 1 thread per corescore: 1106. LS-Dyna 9.3-Explicit AVX2 binary, 3car, HT=ON, SMT=ON, 1 thread per corescore: 768. VASP 5.4.4, CuC, HT=ON, Turbo=OFF, 1 thread per corescore: 133.96. NAMD 2.13, apoa1, HT=ON, Turbo=OFF, 2 threads per corescore: 19.9. LAMMPS version 12 Dec 2018, Water, HT=ON, Turbo=ON, 2 threads per corescore: 276.1. Black Scholes, HT=ON, Turbo=ON, 2 threads per corescore: 9044.32. Monte Carlo, HT=ON, Turbo=ON, 2 threads per corescore: 227.62. OpenFOAM Disclaimer: This offering is not approved or endorsed by OpenCFD Limited, producer and distributor of the OpenFOAM software via www.openfoam.com, and owner of the OPENFOAM* and OpenCFD* trademark. geomean of STREAM, HPCG, HPL, WRF, OpenFOAM, LS-Dyna, VASP, NAMD, LAMMPS, Black Scholes, and Monte Carlo.

Baseline: February 4, 2019—February 19, 2019

New: February 28, 2019—March 4, 2019

6.) Up to 3.5x VM Density Performance 2nd Generation Intel® Xeon® Platinum processor Baseline: 1-node, 2x Intel® Xeon® processor E5-2697 v2 on Canon Pass with 256 GB (16 slots / 16GB / 1600) total memory, ucode 0x42c on RHEL7.6, 3.10.0-957.el7.x86_​64, 1x Intel 400GB SSD OS Drive, 2x P4500 4TB PCIe, 2*82599 dual port Ethernet, Virtualization Benchmark, VM kernel 4.19, HT on, Turbo on, result: VM density=21, test by Intel on 1/15/2019. New configuration: 1-node, 2x Intel® Xeon® Platinum 8280 processor on Wolf Pass with 768 GB (24 slots / 32GB / 2666) total memory, ucode 0x2000056 on RHEL7.6, 3.10.0-957.el7.x86_​64, 1x Intel 400GB SSD OS Drive, 2x P4500 4TB PCIe, 2*82599 dual port Ethernet, Virtualization Benchmark, VM kernel 4.19, HT on, Turbo on, result: VM density=74, test by Intel on 1/15/2019. VM Density

Baseline: January 15, 2019

New: January 15, 2019

7.) Up to 59% Savings with Fewer Servers (1.59x) at Similar Performance Levels when Upgrading 5 Year Old Server to 2nd Generation Intel® Xeon® Scalable processor and Reduce data center footprint by replacing twenty 5-year old servers with six servers based on 2S Intel® Xeon® Platinum 8280 processor. 2nd Generation Intel® Xeon® Platinum processor Baseline: 1-node, 2x Intel® Xeon® processor E5-2697 v2 on Canon Pass with 256 GB (16 slots / 16GB / 1600) total memory, ucode 0x42c on RHEL7.6, 3.10.0-957.el7.x86_​64, 1x Intel® SSD 400GB OS Drive, 2x P4500 4TB PCIe, 2*82599 dual port Ethernet, Virtualization Benchmark, VM kernel 4.19, HT on, Turbo on, result: VM density=21, New configuration: test by Intel on 1/15/2019. 1-node, 2x Intel® Xeon® Platinum 8280 processor on Wolf Pass with 768 GB (24 slots / 32GB / 2666) total memory, ucode 0x2000056 on RHEL7.6, 3.10.0-957.el7.x86_​64, 1x Intel® SSD 400GB OS Drive, 2x P4500 4TB PCIe*, 2*82599 dual port Ethernet, Virtualization Benchmark, VM kernel 4.19, HT on, Turbo on, result: VM density=74, test by Intel on 1/15/2019. Cost reduction scenarios described are intended as examples of how a given Intel®-based product, in the specified circumstances and configurations, may affect future costs and provide cost savings. Circumstances will vary. Intel does not guarantee any costs or cost reduction. Example based on estimates as of March 2019 of equivalent rack performance over 4-year operation on virtualization workload running VMware* vSphere Enterprise Plus on Red Hat Enterprise Linux* Server and comparing 20 installed 2-socket servers with Intel® Xeon® processor E5-2697 v2 (formerly "IvyBridge") at a total cost of $796,563 [Per server cost $39.8K: acquisition=13.7K, infrastructure, and utility=4.2K, os & software=12.2K, maintenance=9.7K ] vs. 6 new Intel® Xeon® Platinum 8280 (costs based on Platinum 8180 assumptions) at a total cost of $325,805 [Per server cost $54.3K: acquisition=28.9K, infrastructure, and utility=3.5K, os & software=12.2K, maintenance=9.7K]. Assumptions based on https://xeonprocessoradvisor.intel.com, assumptions as of Feb 13, 2019. Infrastructure Cost Savings

Baseline: January 15, 2019

New: January 15, 2019

8.) Up to 24.8x Performance Gains on Data Warehousing Queries on the new 2nd Generation Intel® Xeon® Platinum 8280 processor with Windows Server 2016 vs. 4-year old legacy server with old hardware and software. 2nd Generation Intel® Xeon® Platinum processor Baseline: 1-node, 2x Intel® Xeon® processor E5-2699 v3 on Wildcat Pass with 768 GB (24 slots / 32GB / 2666) total memory (workload uses 691GB), ucode 0x3D on Windows Server 2008 R2, 1 x S710 (200GB), 1 x S3500 (1.6TB), 2 x P4608 (6.4TB), SQL Server 2008 R2 SP1 (Enterprise Edition), HT on, Turbo on, result: queries per hour at 1TB =33681, test by Intel on 12/21/2018. New configuration: 1-node, 2x Intel® Xeon® Platinum 8280 processor on Wolf Pass with 1536 GB (24 slots / 64GB / 2666 (1866)) total memory (workload uses 691GB), ucode 0xA on Windows Server 2016 (RS1 14393), 1 x S710 (200GB), 1 x S3500 (1.6TB), 4 x P4610 (7.6TB), SQL Server 2017 RTM - CU13 (Enterprise Edition), HT on, Turbo on, result: queries per hour at 1TB =836261, test by Intel on 3/13/2019. Analytics Data Wearhousing Queries

Baseline: December 21, 2018

New: March 13, 2019

9.) Up to 4.3X Big Data SPARK* Performance on 2nd Generation Intel® Xeon® Scalable processor vs 5 year old 2-socket server. 2nd Generation Intel® Xeon® Gold processor Baseline: 1+4-node, 2x Intel® Xeon® processor E5-2697 v2 on S2600JF with 128 GB (8 slots / 16GB / 1866 ) total memory, ucode 0x42d on CentOS-7.6.1810, 4.20.0-1.el7.x86_​64, 1x 180GB SATA3 SSD, 3 x Seagate ST4000NM0033 (4TB), 1 x Intel I350, HiBench v7.1 / bigdata, Mllib, OpenJDK-1.8.0_​191, python-2.7.5, Apache Hadoop-2.9.1, Apache Spark-2.2.2, , HT on, Turbo on, result: SparkKmeans=119.5M, SparkSort=121.4M, SparkTerasort=107.4M, New configuration: test by Intel on 1/23/2019. 1+4-node, 2x Intel® Xeon® Gold 6248 processor on S2600WF with 768 GB (384 GB used) (12 slots* / 64 GB / 2400 (384GB used)) total memory, ucode 0x400000A on CentOS-7.6.1810, 4.20.0-1.el7.x86_​64, Intel SSD DC S3710, 6 x Seagate ST2000NX0253 (2TB), 1 x Intel X722, HiBench v7.1 / bigdata, Mllib, OpenJDK-1.8.0_​191, python-2.7.5, Apache Hadoop-2.9.1, Apache Spark-2.2.2, HT on, Turbo on, result: SparkKmeans=1235.8M, SparkSort=518.4M, SparkTerasort=589.3M, test by Intel on 1/23/2019. Geomean of SparkKmeans, SparkSort, SparkTerasort

Baseline: January 23, 2019

New: January 23, 2019

10-a.) 36% more VMS per node (1.36x) for multi-tenant virtualized OLTP databases with Intel® Optane™ DC Persistent Memory Module (DCPMM) 2nd Generation Intel® Xeon® Platinum processor Baseline: 1-node, 2x 26-core 2nd Generation Intel® Xeon® Scalable Processor, HT on, Turbo on, 768GB, 0(24 slots / 32GB / 2666 DDR)1x Samsung PM963 M.2 960GB, 7 x Samsung PM963 M.2 960GB, 4x Intel SSDs S4600 (1.92TB), 1x Intel X520 SR2 (10Gb), Windows Server 2019 RS5-17763, OLTP Cloud Benchmark, New configuration: test by Intel as of 1/31/2019. 1-node, 2x 26-core 2nd Generation Intel® Xeon® Scalable Processor, HT on, Turbo on, 192GB, 1TB(12 slots / 16 GB / 2666 DDR + 8 slots /128GB / 2666 Intel® Optane™ DCPMM), 1x Samsung PM963 M.2 960GB, 7x Samsung PM963 M.2 960GB, 4x Intel SSDs S4600 (1.92TB), 1x Intel X520 SR2 (10Gb), Windows Server 2019 RS5-17763, OLTP Cloud Benchmark, test by Intel as of 1/31/2019. multi-tenant virtualized OLTP database

Baseline: January 31, 2019

New: January 31, 2019

10-b.) 30% lower cost per VM (1.30x) for multi-tenant virtualized OLTP databases with Intel® Optane™ DC Persistent Memory Module (DCPMM) 2nd Generation Intel® Xeon® Platinum processor Baseline cost: Total System Cost: $29,408 [cpu cost=$7310, Memory subsystem cost @ Capacity: 24x 64GB=$16,998, Storage cost=$2,100, Chassis, PSUs, Bootdrive, etc.=$3,000] vs. New configuration cost: Total System Cost: $22,024 [cpu cost=$7310, Memory subsystem cost @ full capacity: =$9,614 ($2,690 for DDR4 + $6,924 for Intel® Optane™ DC PMEM), Storage cost=$2,100, Chassis, PSUs, Bootdrive, etc.=$3,000]. Cost Savings for multi-tenant virtualized OLTP databases

Baseline: January 31, 2019

New: January 31, 2019

11.) Up to 4X More VMs When Quadrupling Memory Capacity with Intel® Optane™ DC Persistent Memory Module (DCPMM) Running Redis+Memtier 2nd Generation Intel® Xeon® Platinum processor Baseline: 1-node, 2x Intel® Xeon® Platinum 8280L cpu on Intel reference platform with 768 GB (12 slots / 32GB / 2666) total memory, ucode 0x400000A on Fedora-27, 4.20.4-200.fc29.x86_​64, 2x40GB , Redis 4.0.11, memtier_​benchmark-1.2.12, KVM, 1 45GB instance/VM, centos-7.0, ww06'19 BKC, HT on, Turbo on, score VM=14, test by Intel on 2/21/2019. New configuration: 1-node, 2x Intel® Xeon® Platinum 8280L cpu on Intel reference platform with 192GB DDR, 3072GB Intel Optane DCPMM (12 slots / 16 GB / 2666 DDR + 12 slots / 256GB/ 2666) total memory, ucode 0x400000A on Fedora-27, 4.20.4-200.fc29.x86_​64, 2x40GB , Redis 4.0.11, memtier_​benchmark-1.2.12, KVM, 1 45GB instance /VM, centos-7.0, ww06'19 BKC, AEP firmware 5346, HT on, Turbo on, score VM=56, test by Intel on 2/21/2019. Redis+Memtier

Baseline: February 21, 2019

New: February 21, 2019

12-a.) Network Specialized 2nd Generation Intel® Xeon® Scalable Processors Offer 1.25x to 1.58x Gains on Various Network Workloads 2nd Generation Intel® Xeon® Gold processor Baseline: Tested by Intel on 1/17/2019 1-Node, 2x Intel® Xeon® Gold 6130 processor on Neon City platform with 12x 16GB DDR4 2666MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 6x Intel XXV710-DA2, Bios: PLYDCRB1.86B.0155.R08.1806130538, ucode: 0x200004d (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.15.0-42-generic, Benchmark: VPP IPSec w/AESNI (AES-GCM-128) (Max Gaits/s (1420B)), Workload version: VPP v17.10, Compiler: gcc7.3.0, Results: 179. Tested by Intel on 1/17/2019 New configuration: 1-Node, 2x Intel® Xeon® Gold 6230N processor on Neon City platform with 12x 16GB DDR4 2999MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 6x Intel XXV710-DA2, Bios: PLYXCRB1.PFT.0569.D08.1901141837, ucode: 0x4000019 (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.20.0-042000rc6-generic, Benchmark: VPP IPSec w/AESNI (AES-GCM-128) (Max Gbits/s (1420B)), Workload version: VPP v17.10, Compiler: gcc7.3.0, Results: 225. 12. VPP IP Security

Baseline: January 17, 2019

New: January, 17 2019

12-b.) Network Specialized 2nd Generation Intel® Xeon® Scalable Processor SKUs Offer 1.25x to 1.58x Gains on Various Network Workloads 2nd Generation Intel® Xeon® Gold processor Baseline: Tested by Intel on 1/17/2019 1-Node, 2x Intel® Xeon® Gold 6130 processor on Neon City platform with 12x 16GB DDR4 2666MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 6x Intel XXV710-DA2, Bios: PLYDCRB1.86B.0155.R08.1806130538, ucode: 0x200004d (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.15.0-42-generic, Benchmark: VPP FIB (Max Mpackets/s (64B)), Workload version: VPP v17.10 in ipv4fib configuration, Compiler: gcc7.3.0, Results: 160. New configuration: Tested by Intel on 1/17/2019 1-Node, 2x Intel® Xeon® Gold 6230N processor on Neon City platform with 12x 16GB DDR4 2999MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 6x Intel XXV710-DA2, Bios: PLYXCRB1.PFT.0569.D08.1901141837, ucode: 0x4000019 (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.20.0-042000rc6-generic, Benchmark: VPP FIB (Max Mpackets/s (64B)), Workload version: VPP v17.10 in ipv4fib configuration, Compiler: gcc7.3.0, Results: 212.9. VPP FIB

Baseline: January 17, 2019

New: February 17, 2019

12-c.) Network Specialized 2nd Generation Intel® Xeon® Scalable Processor SKUs Offer 1.25x to 1.58x Gains on Various Network Workloads 2nd Generation Intel® Xeon® Gold processor Baseline: Tested by Intel on 10/26/2018 1-Node, 2x Intel® Xeon® Gold 6130 processor on Neon City platform with 12x 16GB DDR4 2666MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 4x Intel X710-DA4, Bios: PLYDCRB1.86B.0155.R08.1806130538, ucode: 0x200004d (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.15.0-42-generic, Benchmark: Virtual Firewall (64B Mpps), Workload version: opnfv 6.2.0, Compiler: gcc7.3.0, Results: 38.9. Tested by Intel on 2/04/2019 New configuration: 1-Node, 2x Intel® Xeon® Gold 6230N processor on Neon City platform with 12x 16GB DDR4 2999MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 6x Intel XXV710-DA2, Bios: PLYXCRB1.PFT.0569.D08.1901141837, ucode: 0x4000019 (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.20.0-042000rc6-generic, Benchmark: Virtual Firewall (64B Mpps), Workload version: opnfv 6.2.0, Compiler: gcc7.3.0, Results: 52.3. Virtual Firewall

Baseline: October 26, 2018

New: February 04, 2019

12-d.) Network Specialized 2nd Generation Intel® Xeon® Scalable Processor SKUs Offer 1.25x to 1.58x Gains on Various Network Workloads 2nd Generation Intel® Xeon® Gold processor Baseline: Tested by Intel on 11/06/2018 1-Node, 2x Intel® Xeon® Gold 6130 processor on Neon City platform with 12x 16GB DDR4 2666MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 6x Intel XXV710-DA2, Bios: PLYDCRB1.86B.0155.R08.1806130538, ucode: 0x200004d (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.15.0-42-generic, Benchmark: Virtual Broadband Network Gateway (88B Mpps), Workload version: DPDK v18.08 ip_​pipeline application, Compiler: gcc7.3.0, Results: 56.5. New configuration: Tested by Intel on 1/2/2019 1-Node, 2x Intel® Xeon® Gold 6230N processor on Neon City platform with 12x 16GB DDR4 2999MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 6x Intel XXV710-DA2, Bios: PLYXCRB1.PFT.0569.D08.1901141837, ucode: 0x4000019 (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.20.0-042000rc6-generic, Benchmark: Virtual Broadband Network Gateway (88B Mpps), Workload version: DPDK v18.08 ip_​pipeline application, Compiler: gcc7.3.0, Results: 78.7 Virtual Broadband Network Gateway

Baseline: November 06, 2018

New: January 02, 2019

12-e.) Network Specialized 2nd Generation Intel® Xeon® Scalable Processor SKUs Offer 1.25x to 1.58x Gains on Various Network Workloads 2nd Generation Intel® Xeon® Gold processor Baseline: Tested by Intel on 1/22/2019 1-Node, 2x Intel® Xeon® Gold 6130 processor on Supermicro*-X11DPH-Tq platform with 12x 16GB DDR4 2666MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 4x Intel XXV710-DA2, Bios: American Megatrends Inc.* version: '2.1', ucode: 0x200004d (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.20.0-042000rc6-generic, Benchmark: Virtual Converged Cable Access Platform (iMIX Gbps), Workload version: vcmts 18.10, Compiler: gcc7.3.0 , Other software: Kubernetes* 1.11, Docker* 18.06, DPDK 18.11, Results: 54.8. New configuration: Tested by Intel on 1/22/2019 1-Node, 2x Intel® Xeon® Gold 6230N processor on Neon City platform with 12x 16GB DDR4 2999MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 6x Intel XXV710-DA2, Bios: PLYXCRB1.PFT.0569.D08.1901141837, ucode: 0x4000019 (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.20.0-042000rc6-generic, Benchmark: Virtual Converged Cable Access Platform (iMIX Gbps), Workload version: vcmts 18.10, Compiler: gcc7.3.0, Other software: Kubernetes* 1.11, Docker* 18.06, DPDK 18.11, Results: 83.7. Virtual Cable Modem Termination System (vCMTS)

Baseline: January 22, 2019

New: January 22, 2019

12-f.) Network Specialized 2nd Generation Intel® Xeon® Scalable Processor SKUs Offer 1.25x to 1.58x Gains on Various Network Workloads 2nd Generation Intel® Xeon® Gold processor Baseline: Tested by Intel on 1/21/2019. Baseline: 1-Node, 2x Intel® Xeon® Gold 6130 processor on Neon City platform with 12x 16GB DDR4 2666MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 4x Intel XXV710-DA2, Bios: PLYXCRB1.86B.0568.D10.1901032132, ucode: 0x200004d (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.15.0-42-generic, Benchmark: Open Virtual Switch (on 4C/4P/8T 64B Mpacket/s), Workload version: OVS 2.10.1, DPDK-17.11.4, Compiler: gcc7.3.0, Other software: QEMU-2.12.1, VPP v18.10, Results: 9.6. Tested by Intel on 1/18/2019 New configuration: 1-Node, 2x Intel® Xeon® Gold 6230N processor on Neon City platform with 12x 16GB DDR4 2999MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 6x Intel XXV710-DA2, Bios: PLYXCRB1.86B.0568.D10.1901032132, ucode: 0x4000019 (HT= ON, Turbo= OFF), OS: Ubuntu* 18.04 with kernel: 4.20.0-042000rc6-generic, Benchmark: Open Virtual Switch (on 6P/6C/12T 64B Mpacket/s), Workload version: OVS 2.10.1, DPDK-17.11.4, Compiler: gcc7.3.0, Other software: QEMU-2.12.1, VPP v18.10, Results: 15.2. Tested by Intel on 1/18/2019 1-Node, 2x Intel® Xeon® Gold 6230N processor with SST-BF enabled on Neon City platform with 12x 16GB DDR4 2999MHz (384GB total memory), Storage: 1x Intel® 240GB SSD, Network: 6x Intel XXV710-DA2, Bios: PLYXCRB1.86B.0568.D10.1901032132, ucode: 0x4000019 (HT= ON, Turbo= ON (SST-BF)), OS: Ubuntu* 18.04 with kernel: 4.20.0-042000rc6-generic, Benchmark: Open Virtual Switch (on 6P/6C/12T 64B Mpacket/s), Workload version: OVS 2.10.1, DPDK-17.11.4, Compiler: gcc7.3.0, Other software: QEMU-2.12.1, VPP v18.10, Results: 16.9. Open Vitual Switch (OVS) DPDK

Baseline: January 21, 2019

New: January 18, 2019

13.) 50% faster runtimes (1.5x) for Twitter Hadoop Animation. Intel® Xeon® processor E5-26xx v4 Baseline: Dual-socket Intel® Xeon® processor E5-2630 v4 @ 2.2 GHz (10 cores/20 threads per socket); 128 GB RAM; 12x 6 TB 7200 RPM SATA HDD; 1x SATA SSD boot disk; 25 GbE Ethernet; 102 nodes spread across 6 racks. Workload: Gridmix* and Terasort*. Gridmix Score: 3309 seconds; Terasort Score: 5504 seconds. New configuration: Dual-socket Intel® Xeon® processor E5-2630 v4 @ 2.2 GHz (10 cores/20 threads per socket); 128 GB RAM; 12x 6 TB 7200 RPM SATA HDD; 1x SATA SSD boot disk; 1x 750 GB Intel® Optane™ DC P4800X NVMe*-based SSD; 25 GbE Ethernet; 102 nodes spread across 6 racks. Workload: Gridmix and Terasort. Gridmix Score: 2396 seconds; Terasort Score: 2640 seconds. OS: Twitter CentOS* 6 Derivative, Kernel Version 2.6.74-t1.el6.x86_​64 (based on upstream 4.14.12 Kernel), BIOS Version: D3WWM11. Microcode Version: 0xb000021. Note that the test cluster used a higher core count than Twitter's production Hadoop* clusters, which provided only 4 cores/8 threads per HDD. Twitter Hadoop Animation (Gridmix* and Terasort*)

Baseline: February 1, 2019

New: February 1, 2019

14.) 30% lower TCO (1.30x) for Twitter Hadoop Animation. 2nd Generation Intel® Xeon® Gold processor Baseline: Single-socket Intel® Xeon® processor E3-1230 v6 (4 cores); 32 to 64 GB RAM; 1x 1 TB or 2 TB HDDs; Intel® S4500 240 GB boot disk; 1 GbE to 10 GbE Ethernet; no caching. New configuration: Single-socket Intel® Xeon® Gold 6262 processor (24 cores); 192 GB RAM; Intel S4500 240 GB boot disk; 8x 6 TB HDDs; 1x Intel® SSD DC P4610 6.4TB; 25 GbE Ethernet; caching using Intel® Cache Acceleration Software (Intel® CAS). OS: Twitter CentOS* 6 Derivative, Kernel Version 2.6.74-t1.el6.x86_​64 (based on upstream 4.14.12 Kernel), BIOS Version: D3WWM11, Microcode Version: 0xb000021. Lower TCO using Twitter Hadoop Animation

Baseline: February 1, 2019

New: February 1, 2019

15.) Approximately 75% lower power consumption (1.75x) for Twitter Hadoop Animation. 2nd Generation Intel® Xeon® Gold processor Source: Twitter estimate, based on 4 racks (10KW each) consolidating into 1 rack (10KW). Lower Power consumption using Twitter Hadoop Animation

Baseline: February 1, 2019

New: February 1, 2019

16.) 25% more data (1.25x) to be available in the database main store and saves 10% in costs (1.10x) for Intel® Optane™ DC persistent memory use case 2nd Generation Intel® Xeon® Platinum processor Baseline: 1-node, 4x Intel® Xeon® Platinum 8280M on Lightning Ridge with 48x 128GB DDR4 2666 MHz GB total memory, ucode TBD on SUSE 15, 60x Intel® SSD DC S4600 SATA 480GB TB, SAP HANA* analytic workload operating on 5.83TB database, HT off, Turbo off, test by Intel on 3/15/2019. New configuration: 1-node, 4x Intel® Xeon® Platinum 8280L processor on Lightning Ridge with total memory - (24x 128GB DRAM and 24x 256GB Intel® Optane™ DC PMEM), ucode TBD on SUSE 15, 75x Intel SSD DC S4600 SATA 480GB, SAP HANA* analytic workload operating on 7.3TB database, HT off, Turbo off, test by Intel on 3/15/2019. Cost savings using SAP HANA* analytic workload

Baseline: March 15, 2019

New: March 15, 2019

17.) Up to 39% lower total system cost (1.39x) per database for Intel® Optane™ DC persistent memory use case 2nd Generation Intel® Xeon® Platinum processor Baseline: 5-node, 4x Intel® Xeon® Platinum 8280L on Lightning Ridge with 6TB total memory (48 slots / 128 GB / 2666), ucode TBD on SUSE 15, 60x Intel SSD DC S4600 SATA 480GB, SAP HANA* analytic workload operating on 3TB database, HT off, Turbo off, test by Intel on 3/15/2019. New configuration: 5-node, 4x Intel® Xeon® Platinum 8280M on Lightning Ridge with 9TB total memory (24x 256GB Intel Optane DCPMM + 24x 128GB DDR4 2666), ucode TBD on SUSE 15, 90x Intel SSD DC S4600 SATA 480GB, SAP HANA* analytic workload operating on 6TB, HT off, Turbo off, test by Intel on 3/15/2019. Cost savings using SAP HANA* analytic workload

Baseline: March 15, 2019

New: March 15, 2019

18.) Up to 40% lower memory cost (1.40x) for content delivery of high-quality video 2nd Generation Intel® Xeon® Gold processor Baseline: 1-node, 2x Intel ® Xeon ® Gold 6252 CPU at 2.10 GHz on S2600WFT platform with 1.5 TB total memory (24x 64 GB @ 2666 MT/s), ucode 0x04000010 on CentOS Linux release 7.5.1804 (Core) 4.19.0-rc3+ (Host), 4.19.0-rc3, Intel® SSD DC P4510 1TB, 2x Dual port Intel Corporation Ethernet Controller XXV710 for 25GbE SFP28 (rev 02) NUMA Aligned 100Gbps LAG, Apache Traffic server 7.1.4, NGINX 1.12.2, HT on, Turbo on, Dataset = 512 X 103, 1 MB Randomized Web Page Content, test by Intel on 1/15/2019. New configuration: 1-node, 2x Intel ® Xeon ® Gold 6252 CPU at 2.10 GHz S2600WFT platform with 192 GB + 1.5 TB total memory (12x 128 GB Intel Optane DCPMM + 12x 16 GB @ 2666 MT/s DDR4), ucode 0x04000010 on CentOS Linux release 7.5.1804 (Core) 4.19.0-rc3+ (Host), 4.19.0-rc3, Intel SSD DC P4510 1TB, 2x Dual port Intel Corporation Ethernet Controller XXV710 for 25GbE SFP28 (rev 02) NUMA Aligned 100Gbps LAG, Apache Traffic server 7.1.4, NGINX 1.12.2, HT on, Turbo on, Dataset = 512 X 103, 1 MB Randomized Web Page Content, test by Intel on 1/15/2019. Cost savings for content dilivery of high-quality video

Baseline: January 15, 2019

New: January 15, 2019

19.) Up to 43% lower memory cost (1.43x) running SAS machine learning workload for Intel® Optane™ DC persistent memory use case 2nd Generation Intel® Xeon® Platinum processor Baseline: 1-node, 2x Intel® Xeon® Platinum 8280 processor on Purley Wolfpass (2S) with 24x 64GB DDR4 1536GB total memory, ucode 0x4000013 on CentOS 7.6, 4.19.8, 1x 1.5TB Intel® SSD DC P4610 NVMe Drive, SAS Machine learning workload running 3 concurrent logistic regression tasks on 400GB of data each, HT on, Turbo on, elapsed time=15:39min: test by Intel on 2/14/2019. New configuration: 1-node, 2x Intel® Xeon® 8280 CPU on Purley Wolfpass (2S) with 1536GB total memory (12x 128GB Intel® Optane™ DCPMM + 12x 16GB DDR4 GB), ucode 0x4000013 on CentOS 7.6, 4.19.8, 1x 1.5TB Intel SSD DC P4610 NVMe Drive, SAS Machine learning workload running 3 concurrent logistic regression tasks on 400GB of data each, HT on, Turbo on, result=16:06min: test by Intel on 2/15/2019. Cost savings for SAS Machine Learning

Baseline: February 14, 2019

New: February 15, 2019

20.) Up to 35% more VMs (1.35x) and 27% lower costs (1.27x) for hyper-converged infrastructures for Intel® Optane™ DC persistent memory use case 2nd Generation Intel® Xeon® Gold processor Baseline: 4-node, 2x Intel® Xeon® Gold 6230 processors on S2600WFD platform with 384GB total memory (24x 16GB DDR4 @ 2933 MT/s), ucode 0x04000013, running on Windows Server 2019, 10.0.17763, 2x Intel® Optane SSD P4800X 375GB, 1x Chelsio 25G NIC (iWARP), workload: vmfleet and diskspd result=41 VMs (settings: Benchmark Setup: Vmfleet Test: Each VM with 1 Core,8 GB Memory, 40 GB VHDX, Test setup: Threads=2 , Buffer Size= 4KiB ,Pattern: Random , Duration = 300 Seconds, Queue Depth=16, 30% write, OS: Windows Server 2019 Standard (Desktop) with updated patch), HT On, Turbo On, test by Microsoft on 2/15/2019. New configuration: 4-node, 2x Intel® Xeon® Gold 6230 CPU on S2600WFD platform with 512GB total memory (12x 16GB DDR4 + 4x 128GB Intel® Optane™ DCPMM), ucode 0x04000013, running on Windows Server 2019, 10.0.17763, 2x Intel® Optane SSD P4800X 375GB, 1x Chelsio 25G NIC (iWARP), workload: vmfleet and diskspd result=56 VMs (settings: Benchmark Setup: Vmfleet Test: Each VM with 1 Core,8 GB Memory, 40 GB VHDX, Test setup: Threads=2 , Buffer Size= 4KiB ,Pattern: Random , Duration = 300 Seconds, Queue Depth=16, 30% write, OS: Windows Server 2019 Standard (Desktop) with updated patch), HT On, Turbo On, test by Microsoft on 2/15/2019. Hyper-converged Infrastructure (HCI)

Baseline: February 15, 2019

New: February 15, 2019

21.) SPEC CPU2017 Floating Point Rate World Record on Intel® Xeon® Platinum Processor 9282 2nd Generation Intel® Xeon® Platinum processor 1 node, 2x Intel® Xeon® Platinum processor 9282 with 768GB (24 x 32GB 2933Mhz) total memory. CentOS 7.6.1810 with kernel 4.20.0+, Version 19.0.1.144 of Intel® C/C++ Compiler. IMC Interleaving set to 1-way Interleave, Sub_​NUMA Cluster set to Enabled. Source: https://spec.org/cpu2017/results/res2019q2/cpu2017-20190513-13797.pdf, SPECrate2017_​fp_​base score: 522. Tested by Intel as of 3/2019. SPECrate2017_​fp_​base Baseline: March 11, 2019
22.) Up to 2.41x Performance Advantage Over Nvidia* V100 GPUs 2nd Generation Intel® Xeon® Platinum processor New Configuration: 2 socket Intel® Xeon® Platinum 8268 processor, 24 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0286.011120190816 (ucode: 0x4000013), CentOS 7.6, Kernel 4.19.5-1.el7.elrepo.x86_​64, SSD 1x INTEL SSDSC2KG96 960GB, Deep Learning Framework: MXNet https://github.com/apache/incubator-mxnet.git commit f1de8e51999ce3acaa95538d21a91fe43a0286ec applying https://github.com/intel/optimized-models/blob/v1.0.2/mxnet/wide_deep_criteo/patch/patch.diff, Compiler: gcc 6.3.1, MKL DNN version: commit: 08bd90cca77683dd5d1c98068cea8b92ed05784, Wide & Deep: https://github.com/intel/optimized-models/tree/v1.0.2/mxnet/wide_deep_criteo commit: c3e7cbde4209c3657ecb6c9a142f71c3672654a5, Dataset: Criteo Display Advertisement Challenge, Batch Size=512, 2 instance/2 socket, Datatype: FP32; with recommendation results: 678,000 records /seconds. vs. Baseline: host system: 2 socket Intel® Xeon® Platinum 8180 processor (28 cores), HT ON, Total memory 128 GB (16 slots/8 GB/ 2666 MHz), Ubuntu 18.04.2 LTS Accelerator: Nvidia* Turing V100 GPU accelerator, 32GB HBM2, 32GB/sec Interconnect BW, System interface x16 PCIe Gen3, Driver Version 410.78, CUDA Version 10.0.130, CUDNN Version 7.5, CUDA CUBLAS 10.0.130 Deep learning workload: MxNet 1.4.0 https://pypi.org/project/mxnet-cu92/, DatatType:FP32, Batch Size= 512, Running 2 instances Model: Wide & Deep: https://github.com/intel/optimized-models/blob/master/mxnet/wide_deep_criteo/model.py Commit ID for the current state is c3e7cbde4209c3657ecb6c9a142f71c3672654a5 Training dataset (8,000,000 samples): wget https://storage.googleapis.com/dataset-uploader/criteo-kaggle/large_version/train.csv Evaluation dataset (2,000,000 samples): wget https://storage.googleapis.com/dataset-uploader/criteo-kaggle/large_version/eval.csv python3 inference.py --batch-size $bs --num-batches 10000 >> $outdir/bs$bs- $runid.2xbgout 2>&1 & python3 inference.py --batch-size $bs --num-batches 10000 >> $outdir/bs$bs-$runid.2xfgout 2>&1. Recommendation results: 281,211 records/second. Tested by Intel as of 3/26/2019. Deep Learning Framework

Baseline: March 26, 2019

New: March 26, 2019

23.) Up to 90% Average Generational gains (1.90x) on Intel® Xeon® Gold Processor Mainstream CPUs 2nd Generation Intel® Xeon® Gold processor New configuration: 2 socket Intel® Xeon® Platinum 8268 processor, 24 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0286.011120190816 (ucode:0x4000013), CentOS 7.6, Kernel 4.19.5-1.el7.elrepo.x86_​64, SSD 1x INTEL SSDSC2KG96 960GB, Deep Learning Framework: MXNet https://github.com/apache/incubator-mxnet.git commit f1de8e51999ce3acaa95538d21a91fe43a0286ec applying https://github.com/intel/optimized-models/blob/v1.0.2/mxnet/wide_deep_criteo/patch/patch.diff, Compiler: gcc 6.3.1, MKL DNN version: commit: 08bd90cca77683dd5d1c98068cea8b92ed05784, Wide & Deep: https://github.com/intel/optimized-models/tree/v1.0.2/mxnet/wide_deep_criteo commit: c3e7cbde4209c3657ecb6c9a142f71c3672654a5, Dataset: Criteo Display Advertisement Challenge, Batch Size=512, 2 instance/2 socket, Datatype: Int8, with 1,299,000 records/ second vs. Baseline: processor, 24 cores HT On Turbo ON Total Memory 384 GB (12 slots/ 32GB/ 2933 MHz), BIOS: SE5C620.86B.0D.01.0286.011120190816 (ucode: 0x4000013), CentOS 7.6, Kernel 4.19.5-1.el7.elrepo.x86_​64, SSD 1x INTEL SSDSC2KG96 960GB, Deep Learning Framework: MXNet https://github.com/apache/incubator-mxnet.git commit f1de8e51999ce3acaa95538d21a91fe43a0286ec applying https://github.com/intel/optimized-models/blob/v1.0.2/mxnet/wide_deep_criteo/patch/patch.diff, Compiler: gcc 6.3.1, MKL DNN version: commit: 08bd90cca77683dd5d1c98068cea8b92ed05784, Wide & Deep: https://github.com/intel/optimized-models/tree/v1.0.2/mxnet/wide_deep_criteo commit: c3e7cbde4209c3657ecb6c9a142f71c3672654a5, Dataset: Criteo Display Advertisement Challenge, Batch Size=512, 2 instance/2 socket, Datatype: FP32, with 678,000 records/ second. Tested by Intel as of 3/26/2019. 30.) Deep Learning Framework

Baseline: March 26, 2019

New: March 26, 2019

24.) Up to 76% Average Gains (1.76x) on Intel® Xeon® Gold Processor 6230N CPU Over Intel® Xeon® Gold Processor 6130 CPU 2nd Generation Intel® Xeon® Gold processor Baseline: 1-node, 1x Intel® Xeon® Gold processor 6130 cpu on Wolf Pass with 192 GB (12 X 16GB 2666) total memory, 2x Intel Corporation Ethernet Controller X710-DA2, Ubuntu 18.04, 4.15.0-33, Bios PLYXCRB1.86B.0532.D14.1804240330, 64B MPackets, 4P/4C/8T 500000 flows per port w/ 2VM's, result: 9.6 Mpps. New configuration: 1-node, 1x Intel® Xeon® Gold processor 6230N cpu on Wolf Pass with 192 GB (12 X 16GB 2666) total memory, 3x Intel Corporation Ethernet Controller X710-DA2, Ubuntu 18.04, 4.20.0-042000rc6, Bios PLYXCRB1.86B.0568.D10.1901032132, 64B MPackets, 6P/6C/12T 500,000 flows per port w/ 3 VM's, SST-BF result: 16.9 Mpps. Open Virtual Switch (OVS) DPDK
25.) Delivering additional memory at the Edge 2nd Generation Intel® Xeon® Gold processor 2x 2nd Generation Intel® Xeon® Gold 6252 processors @ 24 core, 2.10GHz. System A: 1536GB (6x256GB AEP + 6x16GB DRAM, 2-2-2, Memory Mode, 16:1) System B: 768GB (6x128GB) 2666 DRAM. Configured CDN software: Qwilt* Instance software with cache on ramdisk. Testing performed March 2019. CDN software: Qwilt* Instance March 2019
26.) Up to 5.5x Average Generational Gains on Intel® Xeon® Gold Processor Mainstream CPUs 2nd Generation Intel® Xeon® Gold processor Baseline: Intel® Xeon® Scalable processor Platinum 8280, configured with 192 GB of memory, Intel® Solid State Drive Data Center 480 GB, and CentOS Linux* 7.4.1708; Intel® Math Kernel Library for Deep Neural Networks (Intel® MKL-DNN), OpenVINO™ toolkit (R5 Release), Siemens Healthineers custom topology, and dataset, Datatype: Int8, inference result: 201/seconds vs. New configuration: Intel® Xeon® Scalable processor Platinum 8180, configured with 192 GB of memory, Intel® Solid State Drive Data Center 480 GB, and CentOS Linux* 7.4.1708; Intel® Math Kernel Library for Deep Neural Networks (Intel® MKL-DNN), OpenVINO™ toolkit (R5 Release), Siemens Healthineers custom topology, and dataset, Datatype: FP32, inference result: 37/seconds. Tested by Siemens in March 2019. Intel® Math Kernel Library for Deep Neural Networks (Intel® MKL-DNN), OpenVINO™ toolkit (R5 Release), Siemens Healthineers custom topology, and dataset, March 2019
27.) Up to 8x speedup improvement in IO intensive queries with Intel® Optane™ DC persistent memory + HDD vs. DRAM + HDDs 2nd Generation Intel® Xeon® Platinum processor Baseline: 1-node, 2x Intel® Xeon® Platinum 8280L CPU @ 2.70GHz processor on S2600WF (Wolf Pass) with 768GB DDRGB (DDR Mem: 24 slots / 32GB / 2666 MT/s) total memory, ucode 0x0400000a running Fedora release 29 kernel Linux-4.18.8-100.fc27.x86_​64-x86_​64-with-fedora-27,and 9 decision support I/O intensive queries, storage is 8x HDD (ST1000NX0313), 10-Gigabit SFI/SFP+ network connection. Source: score: geomean baseline 1. Tested by Intel, on 24 Feb 2019, vs. New configuration: 1-node, 2x Intel® Xeon® Platinum 8280L CPU @ 2.70GHz processor on S2600WF (Wolf Pass) with 192GB DDR + 1TB Intel® Optane™ DC persistent memory (DDR Mem: 12 slots / 16GB / 2666 MT/s + 8 slots / 128GB / 2666 MT/s) total memory, ucode 0x0400000a running Fedora release 29 kernel Linux-4.18.8-100.fc27.x86_​64-x86_​64-with-fedora-27,and 9 decision support I/O intensive queries, storage is 8x HDD (ST1000NX0313), 10-Gigabit SFI/SFP+ network connection. Score: Geomean speedup 8x. Tested by Intel, on 24 Feb 2019. IO intensive queries

Baseline: February 24, 2019

New: February 24, 2019

28.) Up to 96% efficiency on Redis* Memtier benchmark running inside Intel® Software Guard Extensions (Intel® SGX) 1st Generation Intel® Xeon® Platinum processor Baseline: 1 node, 2x Intel® Xeon® Platinum 8180 scalable processor with 128GB total memory running Centos 7.6 3.10.0-514.el7.centos.2.1.13.VCA.x86_​64, Redis Memtier executing with 50,000 keys over 50 clients & 64 byte objects. Baseline: Unmodified Redis 5.0-rc6 without SGX enclave. Score: 55062.81 ops/sec, vs. New configuration: One Intel® SGX Card with three Intel® Xeon® E3-1585L v5 processors, 16GB memory per node, EEPROM 2.3.26, BIOS 2.3.26, SGX enabled with GPU-APERTURE set to 256. All three nodes running Ubuntu_​16.04.3_​2.3.26 kernel 4.14.20-1.2.3.26.vca, modified Redis SGX 5.0-rc6 with Linux SGX driver 2.0 inside SGX enclave. Score: 53239.08 ops/sec. Redis* Memtier

Baseline: March 4, 2019

New: March 4, 2019

29.) LAMMPS protein models runs 51% faster on 2-socket Intel® Xeon® Platinum 8260 processors 2nd Generation Intel® Xeon® Platinum processor Baseline: AMD* EPYC* 7601 Processor: Supermicro* AS -1023US-TR4, 2S AMD EPYC 7601 (2.2GHz, 32C), 16x16GB DDR4-2666, 1 SSD, BIOS ver: 1.1c (10/04/2018), Microcode ver: 0x8001227, Oracle* Linux Server release 7.6 (compatible with Red Hat Enterprise Linux* (RHEL) 7.6) on a 7.5 kernel using ksplice for security fixes, Kernel: 3.10.0-957.5.1.el7.crt1.x86_​64, Cluster File System: Panasas (124 TB storage) Firmware v6.3.3.a & EDR based IEEL Lustre*, HBA: 100Gbps Mellanox* EDR MT27700, 36 Port Mellanox EDR IB Switch, OFED stack: OFED MLNX mlnx-4.3-3.0.2.0, LAMMPS version 12 Dec 2018, Protein workload, Intel® Compiler 2019u2, Intel® MPI 2019u2, SMT=ON, Turbo=ON. Score: 1 node= 15.813 timestamp/sec, 16 node= 95.72 timestamp/sec, higher is better, tested by Intel, March 7, 2019. New configuration: Intel Reference Platform with 2x Intel Xeon Platinum 8260 processor (2.4GHz, 24C), 12x16GB DDR4-2933, 1 SSD, Cluster File System: Panasas (124 TB storage) Firmware v6.3.3.a & Intel® Omni-Path Architecture (Intel® OPA) based IEEL Lustre*, BIOS: SE5C620.86B.0D.01.0286.011120190816, Microcode: 0x4000013, Oracle* Linux Server release 7.6 (compatible with Red Hat Enterprise Linux* (RHEL) Server 7.6) on a 7.5 kernel using ksplice for security fixes, Kernel: 3.10.0-957.5.1.el7.crt1.x86_​64, OFED stack: OFED Intel OPA 10.9 on Oracle Linux 7.6 (Compatible w/RHEL 7.6) w/Lustre v2.10.6, HBA: 100Gbps Intel OPA 1 port PCIe x16, Switch: Intel OPA Edge Switch 100 Series 48 Port, LAMMPS version 12 Dec 2018, Protein workload, Intel® Compiler 2019u2, Intel® MPI 2019u2, HT=ON, Turbo=ON. Score: 1 node=24.015 timestamp/sec, 16 node=226.691 timestamp/sec, higher is better, tested by Intel, March 5, 2019. HPC applications (LAMMPS workloads)

Baseline: March 7, 2019

New: March 5, 2019

30.) When running molecular dynamics HPC applications (NAMD workloads), Intel's 2-socket Intel® Xeon® Platinum 9282 processor delivers up to 23% higher performance than a 2-socket AMD* Rome CPU 2nd Generation Intel® Xeon® Platinum processor New configuration: 2S Intel® Xeon® Platinum 9282 processor configuration: Intel Reference Platform with 2x Intel Xeon Platinum 9282 processor (2.6GHz, 56C), 24x32GB DDR4-2933, 1 SSD, BIOS: SE5C620.86B.0D.01.0541.052120190651, Microcode: 0x4000024, Red Hat Enterprise Linux* (RHEL) Server release 7.6, Kernel: 3.10.0-957.12.2.el7.x86_​64, Intel® Compiler 2019, Intel® MPI 2019, HT=ON, Turbo=ON, 1 thread per core, NAMD ver 2.13 Intel Optimized Build, apoa1 workload, FFTW 3.3.8, Charm++ 6.8.2, Tcl 8.6.8. Score: 24.16ns/day, tested by Intel on May 29, 2019. 2S Intel® Xeon® Platinum 9242 processor configuration: Intel Reference Platform with 2x Intel Xeon Platinum 9242 processor (2.3GHz, 48C), 24x16GB DDR4-2933, 1 SSD, BIOS: PLYXCRB1.86B.0572.D02.1901180818, Microcode: 0x4000017, CentOS 7.6, Kernel: 3.10.0-957.5.1.el7.x86_​64, Intel® Compiler 2019, Intel® MPI 2019, HT=ON, Turbo=OFF, 2 threads per core, NAMD ver 2.13 Intel Optimized Build, apoa1 workload, FFTW 3.3.8, Charm++ 6.8.2, Tcl 8.6.8. Score: 19.9ns/day, tested by Intel on February 28, 2019. 2S Intel® Xeon® Platinum 8280 processor configuration: Intel Reference Platform with 2x Intel Xeon Platinum 8280 processors (2.7GHz, 48C), 12x16GB DDR4-2933, 1 SSD, BIOS: SE5C620.86B.0D.01.0286.011120190816, Microcode: 0x4000013, Oracle* Linux Server release 7.6 (compatible with RHEL 7.6) on a 7.5 kernel using ksplice for security fixes, Kernel: 3.10.0-957.5.1.el7.crt1.x86_​64, Intel® Compiler 2019, Intel® MPI 2019, HT=ON, Turbo=ON, 2 threads per core, NAMD ver 2.13 Intel Optimized Build , apoa1 workload, FFTW 3.3.8, Charm++ 6.8.2, Tcl 8.6.8. Score: 12.65ns/day, tested by Intel on May 29, 2019. Intel build notes for config #30: FLOATOPTS = -xCORE-AVX-512 -qopt-zmm-usage=high -O3 -g -fp-model fast=2 -no-prec-div -qoverride-limits -DNAMD_​DISABLE_​SSE; CXX = icpc -std=c++11 -DNAMD_​KNL; CXXOPTS = -static-intel -O2 $(FLOATOPTS); CXXNOALIASOPTS = -O3 -fno-alias $(FLOATOPTS) -qopt-report-phase=loop,vec -qopt-report=4; CXXCOLVAROPTS = -O2 -ip; CC = icc; COPTS = -static-intel -O2 $(FLOATOPTS); ./config Linux-KNL-icc --charm-base $base_​charm --charm-arch mpi-linux-x86_​64-ifort-smp-mpicxx --with-fftw3 --fftw-prefix $base/fftw3_​icc19.1_​SKX --tcl-prefix $base/tcl --charm-opts -verbose; No CPU performance changes between NAMD v2.13 versus v2.12. NAMD v2.13 added changes for GPU; Baseline: Configurations for AMD* Computex claims unknown. Score: 19.6ns/day (Source: Lisa Su Computex* keynote May 27, 2019). HPC applications (NAMD workloads)

Baseline: May 27, 2019

New: May 29, 2019

31.) 31% Higher Performance with 2S Intel Xeon-AP vs 2S AMD* EPYC* "Rome" 7742 2nd Generation Intel® Xeon® Platinum processor 2S Intel Xeon-AP vs 2S AMD* EPYC* "Rome" 7742: Intel measured as of October 8, 2019 using geomean of STREAM Triad, HPCG, HPL, WRF (2 workloads), OpenFOAM 42M_​cell_​motorbike, ANSYS® (14 workloads), LS-DYNA (3 workloads), VASP (4 workloads), NAMD (2 workloads), GROMACS (9 workloads), LAMMPS (9 workloads), FSI Kernels (3 workloads). Intel® Xeon® Platinum 9282 processor configuration: Intel "Walker Pass" S9200WKL platform with 2-socket Intel® Xeon® Platinum 9282 processors (2.6GHz, 56C), 24x16GB DDR4-2933, 1 SSD, BIOS: SE5C620.86B.2X.01.0053, Microcode: 0x5000029, Red Hat Enterprise Linux* 7.7, kernel 3.10.0-1062.1.1. AMD EPYC™ 7742 processor configuration: Supermicro AS-2023-TR4 (HD11DSU-iN) with 2-socket AMD EPYC™ 7742 "Rome" processors (2.25GHz, 64C), 16x32GB DDR4-3200, 1 SSD, BIOS: 2.0 CPLD 02.B1.01, Microcode: 830101C, CentOS* Linux release 7.7.1908, kernel 3.10.0-1062.1.1.el7.crt1.x86_​64. STREAM OMP 5.1 Triad: Intel® Xeon® Platinum 9282 processor: Intel® Compiler 2019u5, BIOS: HT ON, Turbo ON, SNC ON, 1 thread/core; AMD EPYC™ 7242: Intel® Compiler 2019u5, BIOS: SMT ON, Boost ON, NPS 4, 1 thread/core. HPCG Intel optimized version: Intel® Xeon® Platinum 9282 Processor: Intel® Compiler 2019u4, Intel® Math Kernel Library (Intel® MKL) 2019u4, Intel MPI 2019u4, BIOS: HT ON, Turbo OFF, SNC OFF, 1 thread/core; AMD EPYC™ 7742: Intel® Compiler 2019u4, Intel® MKL 2019u4, Intel MPI 2019u4, BIOS: SMT ON, Boost ON OFF, NPS 4, 1 thread/core. HPL v2.3: Intel® Xeon® Platinum 9282 processor: Intel Optimized LINPACK Benchmark, Intel® Distribution for LINPACK* Benchmark, Compiler: Intel MPI 2018u1N=80000, NB=384, P=2, Q=1, BIOS: HT ON, Turbo ON, SNC OFF, 1 thread/core; AMD EPYC™ 7742: AMD official HPL binary https://developer.amd.com/amd-aocl/blas-library/, Compiler: Netlib HPL + BLIS, OpenMPI3 N=16000, NB=192, P=2, Q=4; BIOS: SMT ON, Boost ON, NPS 4, 1 thread/core. WRF 3.9.1.1: Geomean (2 workloads: conus-12km, conus-2.5km): Intel® Xeon® Platinum 9282 processor: Intel® Compiler 2018u3, Intel MPI 2018u3, AVX2 build, BIOS: HT ON, Turbo ON, SNC OFF, 1 thread/core; AMD EPYC™ 7742: Intel® Compiler 2018u3, Intel MPI 2018u3, AVX2 build, BIOS: SMT ON, Boost ON, NPS 4, 1 thread/core. OpenFOAM v6.0 42M_​cell_​motorbike: Intel® Xeon® Platinum 9282 processor: Intel® Compiler 2019u3, Intel MPI 2019u3, BIOS: HT ON, Turbo ON, SNC OFF, 1 thread/core; AMD EPYC™ 7742: Intel® Compiler 2019u3, Intel MPI 2019u3, BIOS: SMT ON, Boost ON, NPS 4, 1 thread/core. ANSYS® Fluent® 2019R1: Geomean (14 workloads: aircraft_​wing_​14m, aircraft_​wing_​2m, combustor_​12m, combustor_​16m, combustor_​71m, exhaust_​system_​33m, f1_​racecar_​140m, fluidized_​bed_​2m, ice_​2m, landing_​gear_​15m, oil_​rig_​7m, pump_​2m, rotor_​3m, sedan_​4m): Intel® Xeon® Platinum 9282 Processor: Intel® Compiler 2017u3, Intel MPI 2018u3, BIOS: HT ON, Turbo ON, SNC ON, 1 thread/core; AMD EPYC™ 7742: Intel® Compiler 2017u3, Intel MPI 2018u3, BIOS: SMT ON, Boost ON, NPS 4, 1 thread/core. LS-DYNA v9.3: Geomean (3 workloads: 3cars/150ms, car2car/120ms, ODB_​10M/30ms): Intel® Xeon® Platinum 9282 processor: Intel® Compiler 2016u3, Intel MPI 2018u3, AVX2 build, BIOS: HT OFF, Turbo ON, SNC ON, 1 thread per core; AMD EPYC™ 7742: Intel® Compiler 2016u3, Intel MPI 2018u3, AVX2 build, BIOS: SMT OFF, Boost ON, NPS 4, 1 thread/core. VASP, developer branch based on v5.4.4: Geomean (4 workloads: CuC, PdO4, PdO4_​K221, Si): Intel® Xeon® Platinum 9282 processor: Intel® Compiler 2019u4, Intel® Math Kernel Library (Intel® MKL) 2019u4, Intel MPI 2019u4, BIOS: HT ON, Turbo OFF, SNC OFF, 1 thread per core; AMD EPYC™ 7742: Intel® Compiler 2019u4, Intel® MKL 2019u4, Intel MPI 2019u4, BIOS: SMT ON, Boost ON, NPS 4, 1 thread per core. NAMD v2.13: Geomean (2 workloads: Apoa1, STMV): Intel® Xeon® Platinum 9282 processor: Intel® Compiler 2019u4, Intel MPI 2019u4, BIOS: HT ON, Turbo ON, SNC OFF, 2 threads per core; AMD EPYC™ 7742: Compiler: AOCC 2.0, Intel MPI 2019u4, BIOS: SMT ON, Boost ON, NPS 4, 2 threads/core. GROMACS 2019.4: Geomean (5 workloads: archer2_​small, ion_​channel_​pme, lignocellulose_​rf, water_​pme, water_​rf): Intel® Xeon® Platinum 9282 processor: Intel® Compiler 2019u4, Intel® Math Kernel Library (Intel® MKL) 2019u4, Intel MPI 2019u4, AVX-512 build, BIOS: HT ON, Turbo OFF, SNC OFF, 2 threads per core for: ion_​channel_​pme, lignocellulose_​rf, water_​rf. 1 thread per core for: water_​pme, archer2_​small; AMD EPYC™ 7742: Intel® Compiler 2019u4, Intel® MKL 2019u4, Intel MPI 2019u4, AVX2_​256 build, BIOS: SMT ON, Boost ON, NPS 4, 2 threads per core. LAMMPS v2019: Geomean (9 workloads: Atomic Fluid, Copper, DPD, Liquid Crystal, Polyethylene, Protein, Stillinger-Weber, Tersoff, Water): Intel® Xeon® Platinum 9282 processor: Intel® Compiler 2019u5, BIOS: HT ON, Turbo ON, SNC ON, 2 threads/core; AMD EPYC™ 7742: Compiler: AOCC 2.0, Intel MPI 2019u5, BIOS: SMT ON, Boost ON, NPS 4, 2 threads/core. FSI Kernels v2.0: Geomean (3 workloads: Binomial Options, Black Scholes, Monte Carlo): Intel® Xeon® Platinum 9282 processor: Intel® Compiler 2019u5, Intel® Math Kernel Library (Intel® MKL) 2019u5, BIOS: HT ON, Turbo ON, SNC OFF, 2 threads/core, HT OFF, Turbo ON, SNC OFF, 1 threads/core, HT ON, Turbo ON, SNC OFF, 2 threads/core; AMD EPYC™ 7742: Intel® Compiler 2019u5, Intel® MKL 2019u5, BIOS: SMT ON, Boost ON, NPS 4, 2 threads/core, SMT OFF, Boost ON, NPS 4, 1 thread/core, SMT ON, Boost ON, NPS 4, 2 threads/core. HPC applications workloads

Baseline: October 8, 2019

New: October 8, 2019

32.) Recommender Engines Up to 2.2X Faster on 2-Socket Intel® Xeon® Platinum 8280 Processors with Intel® Deep Learning Boost (Intel® DL Boost) 2nd Generation Intel® Xeon® Platinum processor Baseline: Supermicro AS-2023-TR4 (HD11DSU-iN) with 2-socket AMD EPYC™ 7742 "Rome" processors (2.25GHz, 64C), 32x 32GB DDR4-3200, BIOS: 2.0 CPLD 02.B1.01, Microcode: 830101C, running Ubuntu Linux release 19.10, kernel 5.3.0-rc3-custom, gcc version 9.2.0. Software: OOB TF build with eigen (pip install tensorflow), Wide & Deep: https://github.com/IntelAI/models/tree/master/benchmarks/recommendation/tensorflow/wide_deep_large_ds, commit id: 4ead44aa254a84109ac8019f5d386e3adb75ac26, Model: https://storage.googleapis.com/intel-optimized-tensorflow/models/wide_deep_int8_pretrained_model.pb, https://storage.googleapis.com/intel-optimized-tensorflow/models/wide_deep_fp32_pretrained_model.pb, Dataset: Criteo Display Advertisement Challenge, Batch Size=512, 2instance/2socket, Datatype: FP32, INT8. New configuration: Intel reference platform "WolfPass" with 2x Intel® Xeon® Platinum 8280L processor (2.7GHz, 28C), 12x 32GB DDR4-2933, 1 SSD, BIOS: SE5C620.86B.02.01.0008.031920191559, Microcode: 0x5000029, running Ubuntu* Linux* release 19.10, kernel 5.3.0-rc3-custom, gcc version 9.2.0. Software: TensorFlow public docker: docker.io/intelaipg/intel-optimized-tensorflow:nightly-latestprs-bdw (https://github.com/tensorflow/tensorflow.git A3262818d9d8f9f630f04df23033032d39a7a413 + Pull Request PR26169 + Pull Request PR26261 + Pull Request PR26271), MKL DNN version: v0.18, Wide & Deep: https://github.com/IntelAI/models/tree/master/benchmarks/recommendation/tensorflow/wide_deep_large_ds, commit id: 4ead44aa254a84109ac8019f5d386e3adb75ac26, Model: https://storage.googleapis.com/intel-optimized-tensorflow/models/wide_deep_int8_pretrained_model.pb, https://storage.googleapis.com/intel-optimized-tensorflow/models/wide_deep_fp32_pretrained_model.pb, Dataset: Criteo Display Advertisement Challenge, Batch Size=512, 2instance/2socket, Datatype: FP32, INT8. Intel measured as of November 14, 2019 Intel® Deep Learning Boost (Intel® DL Boost)

Baseline: November 14, 2019

New: November 14, 2019

33.) 9x higher inference performance with Intel® Xeon® Platinum 9282 processor. 2nd Generation Intel® Xeon® Platinum processor New Configuration: Tested by Intel as of 11/13/2019. 2 socket Intel® Xeon® Platinum 9282 processors (56C), HT ON, Turbo ON, Total Memory 384 GB (24 slots, 16GB, 2934Mhz), BIOS: SE5C620.86B.2X.01.0053.081920190637, Microcode: 0x500002c, Ubuntu 19.10, Kernel 5.3.0-22-generic, SSD 1x Micron_​5100_​MTFDDAV480TBY 447G, Intel® Deep Learning Framework: PyTorch (master + PR for MLPerf)*git fetch origin pull/25235/head:mlperf; git checkout mlperf, Compiler GCC 9.2.1.20191008, MobileNetV1, Batch Size=64, Iterations: 1000, Datatype: INT8 vs Baseline: AMD EPYC™ 7742 processor configuration: Tested by Intel as of 11/13/2019. 2-socket AMD EPYC™ 7742 "Rome" processors (64C), HT ON, Turbo ON, Total Memory 512 GB (16 slots, 32GB, 3200Mhz), BIOS: 2.0, Microcode 0x830101C, Ubuntu 19.10, Kernel 5.3.0-22-generic, SSD 1x INTEL® SSD D3-S4610 1.8T, Deep Learning Framework: PyTorch (master + PR for MLPerf) *git fetch origin pull/25235/head:mlperf; git checkout mlperf, GCC 9.2.1.20191008, MobileNetV1, Batch Size=64, Iterations: 1000, Datatype: FP32. Deep Learning Framework: PyTorch (master + PR for MLPerf)

Baseline: November 13, 2019

New: November 13, 2019

34.) Supports Up to 24 Streams in Parallel Using Visual Compute Accelerator Card for Analytics 2nd Generation Intel® Xeon® Gold processor Intel Reference Platform “WolfPass” with 2x Intel® Xeon® Gold 6252 processor (2.3GHz, 24C), 12x 16GB DDR4-2666, 2x 480 GB Intel® SSD SATA (for OS and primary data), 1x Intel® C627 chipset with Intel® QuickAssist Technology (Intel® QAT), 1x Dual Port 25GbE Intel® Ethernet Network Adapter XXV710 SFP28, running CentOS 7.3, kernel 5.1.3-1.el7.elrepo.x86_​64, Bios: SE5C620.86B.0D.01.0438. 1x high density Intel® Visual Compute Accelerator (Intel® VCA) for Analytics running Ubuntu18.01.1, kernel 3.10.0-693.17.1.el7.2.5.20.VCA.x86_​64. Software workloads: Host system (Libvirt-4.10.0, QEMU-4.0.0, SST-3.0.1054, CollectD-5.8.1.git-master-090afcd, DPDK-19.05, i40evf-3.2.3-k, ixgbe-5.1.0-k, ixgbevf-4.1.0-k, Docker-19.03.1 build 74b1e89, IPR-clx_​Media_​Analytics_​R2_​CI_​84, Gstreamer-GST 1.16 Package clx_​1.2, FFMPEG-FFMPEG4.1.0), Intel VCA-A card (Docker-19.08.1 build 75c2e88, IPR-vca_​disk48_​reference_​k4.19_​ubuntu16.04_​1.0.51_​00_​ISS_​release, Gstreamer-GST 1.16 Package vcaa_​1.2.1, FFMPEG-FFMPEG4.2). BIOS Options enabled: Intel® Virtualization Technology (Intel® VT) with Intel CPU VMX Support and Intel IO Virtualization; Intel Boot Guard; Intel® Trusted Execution Technology (Intel® TXT). Input video: AVC 1080p30 at 6 Mbps. All data points are using the IPR software framework except for Object Detection on the Plus configuration. Inference Models are available from the Open Model Zoo project at https://github.com/opencv/open_model_zoo.git, checkout 2019_​R2; Object Detection: Mobilenet-SSD; Face Recognition: vehicle-detection-adas-0002 and vehicle-attributes-recognition-barrier-0039; Car Classification: face-detection-adas-0001 and face-reidentification-retail-0095. Media analytics with visual accelerator

November 15, 2019

35.) 36% More Estimated Performance and 42% More Estimated Performance/Dollar: Geomean of SPECrate®2017_​int_​base(est), SPECrate®2017_​fp_​base(est), STREAM Triad, and Intel® Distribution for LINPACK* Benchmark Across Ten New 2-Socket 2nd Gen Intel® Xeon® Gold Processors Vs. First Generation. 2nd Generation Intel® Xeon® Gold processor 2nd Gen Intel® Xeon® Gold R processors: 1-node, 2x 2nd Gen Intel® Xeon® Gold processor (62xxR/$$) on Intel Reference platform with 384GB (12 slots / 32 GB / 62xx@2933,52xx@2666) total memory, ucode 0x500002c, HT on for all except off for STREAM (GB/s), LINPACK (GFLOPS/s), Turbo on, with Ubuntu19.10, 5.3.0-24-generic, 6258R/$3950: SPECrate®2017_​int_​base(est)=323, SPECrate®2017_​fp_​base(est)=262, STREAM=224, LINPACK =3305; 6248R/$2700: SPECrate®2017_​int_​base(est)=299, SPECrate®2017_​fp_​base(est)=248, STREAM=224, LINPACK =3010; 6246R/$3286: SPECrate®2017_​int_​base(est)=238, SPECrate®2017_​fp_​base(est)=217, STREAM=225, LINPACK =2394; 6242R/$2529: SPECrate®2017_​int_​base(est)=265, SPECrate®2017_​fp_​base(est)=231, STREAM=227, LINPACK =2698; 6240R/$2200: SPECrate®2017_​int_​base(est)=268, SPECrate®2017_​fp_​base(est)=228, STREAM=223, LINPACK=2438; 6238R/$2612: SPECrate®2017_​int_​base(est)=287, SPECrate®2017_​fp_​base(est)=240, STREAM=222, LINPACK =2545; 6230R/$1894: SPECrate®2017_​int_​base(est)=266, SPECrate®2017_​fp_​base(est)=227, STREAM=222, LINPACK =2219; 6226R/$1300: SPECrate®2017_​int_​base(est)=208, SPECrate®2017_​fp_​base(est)=192, STREAM=200, LINPACK =2073; 5220R/$1555: SPECrate®2017_​int_​base(est)=257, SPECrate®2017_​fp_​base(est)=220, STREAM=210, LINPACK =1610; 5218R/$1273: SPECrate®2017_​int_​base(est)=210, SPECrate®2017_​fp_​base(est)=188, STREAM=199, LINPACK =1290, test by Intel on 12/25/2019. First Gen Intel® Xeon® Gold processors: 1-node, 2x Intel® Xeon® Gold processor (61xx/$$) on Intel Reference platform with 384GB (12 slots / 32 GB / 61xx@2666,51xx@2400) total memory, ucode 0x500002c, HT on for all except off for STREAM (GB/s), LINPACK (GFLOPS/s), Turbo on, with Ubuntu19.10, 5.3.0-24-generic, 6152/$3655: SPECrate®2017_​int_​base(est)=224, SPECrate®2017_​fp_​base(est)=198, STREAM=200, LINPACK =1988; 6148/$3072: SPECrate®2017_​int_​base(est)=225, SPECrate®2017_​fp_​base(est)=198, STREAM=197, LINPACK =2162; 6146/$3286: SPECrate®2017_​int_​base(est)=161, SPECrate®2017_​fp_​base(est)=175, STREAM=185, LINPACK =1896; 6142/$2946: SPECrate®2017_​int_​base(est)=193, SPECrate®2017_​fp_​base(est)=176, STREAM=185, LINPACK =1895; 6140/$2445: SPECrate®2017_​int_​base(est)=202, SPECrate®2017_​fp_​base(est)=183, STREAM=188, LINPACK =1877; 6138/$2612: SPECrate®2017_​int_​base(est)=189, SPECrate®2017_​fp_​base(est)=195, STREAM=189, LINPACK =1976; 6130/$1894: SPECrate®2017_​int_​base(est)=172, SPECrate®2017_​fp_​base(est)=165, STREAM=185, LINPACK =1645; 6126(proj)/$1776: SPECrate®2017_​int_​base(est)=141, SPECrate®2017_​fp_​base(est)=157, STREAM=170, LINPACK =1605; 5120(proj)/$1555: SPECrate®2017_​int_​base(est)=148, SPECrate®2017_​fp_​base(est)=148, STREAM=159, LINPACK =924, 5118/$1273: SPECrate®2017_​int_​base(est)=134, SPECrate®2017_​fp_​base(est)=132, STREAM=149, LINPACK =818, test by Intel on 2/18/2020. Geomean of SPECrate2017_​int_​base(est), SPECrate2017_​fp_​base(est), STREAM-Triad, Intel® Distribution of LINPACK

February 18, 2020

36.) 6.0x Inference Performance for Object Detection SSD-MobileNet vs. AMD Top Bin CPU 2nd Generation Intel® Xeon® Gold processor New configuration: 1-node, 2x Intel® Xeon® Gold 6258R processor on Intel Reference platform with 384GB (12 slots / 32 GB / 2933) total memory, ucode 0x500002c, HT on, Turbo on, with Ubuntu 18.04, 5.3.0-24-generic, AIXPRT v1.01, Intel® distribution of OpenVINO™ toolkit 2019 R3, Object Detection SSD-MobileNet using INT8 with Intel® DL Boost, BS=8, no. of instances=56, test by Intel on 1/23/2020 vs. Baseline: 1-node, 2x AMD EPYC 7742 on Supermicro* AS-2023US-TR4 with 512GB (16 slots / 32 GB / 3200) total memory, ucode 0x8301025, HT on, Turbo on, with Ubuntu 18.04 , 5.3.0-24-generic, AIXPRT v1.01 TensorFlow 1.15, Object Detection SSD-MobileNet using FP32, BS=2, no. of instances=128, test by Intel on 2/12/2020. Inference performance for Object Detection using AIXPRT

February 12, 2020

37.) 19% IOPS increase, and a 15% latency decrease in VMware vSAN™ 2nd Generation Intel® Xeon® Platinum and Intel® Xeon® Gold processors Baseline: 4x Intel R2224WFTZS servers with 2x Intel® Xeon® Platinum 8168 processors each (2.7GHz, 24cores), ucode version: 0x0200004d, BIOS version: SE5C620.86B.00.01.0014.070920180847, HT ON (in BIOS), Turbo ON, total memory: 256GB (8x 32GB DDR4-2666, running at 2666), Storage: OS drive: 1x 128 GB SATA SSD, Cache device: 2x Intel® Optane™ DC P4800X (375GB) NVMe 2.5", Capacity devices: 6x Intel® SSD DC P4510 (4TB) NVMe 2.5", VMware vSAN configured in 2 groups (1 cache drive + 3 capacity drives in RAID 10), Network: 1x Intel® 82599 10 GbE Controller SFI/SFP+ for management, 2x Intel® X722 10Gigabit BASE-T for Vmotion, Clients, and VMs. Hypervisor: VMware ESXi, 6.7.0, 10302608, w/L1TF mitigation which disables HT by ESXi, Data Center Management: VMware vCenter 6.7.0 10244857, HCIBench v1.6.8.7. Tested by Evaluator Group on April 2, 2019. New Configuration: Intel Xeon Gold 6248R processor: Performance is estimated by Intel to be equal in performance to Intel Xeon Platinum 8268 based on core count (24C) and similar frequency (8268: 2.9GHz, 6248R: 3.0GHz). 4x Intel S2600WFT servers with 2x Intel Xeon Platinum 8268 processors each (2.9GHz, 24 cores), ucode version: 0x5000021, BIOS version: SE5C620.86B.02.01.0008.031920191559, HT ON (in BIOS), Turbo ON, total memory: 768GB (12x 64GB DDR4-2993), Storage: OS drive: 1x SSD, Cache device: 2x Intel Optane DC P4800X (375GB) NVMe 2.5", Capacity devices: 6x Intel P4510 (4TB) NVMe 2.5", VMWare vSAN configured in 2 groups (1 cache drive + 3 capacity drives in RAID 10), Network: 1x Intel XL710 40GbE QSFP+ for management, 1x XL710 40GbE QSFP+ for vMotion, Clients, and VMs. Hypervisor: VMware ESXi, 6.7.0, 14320388, All available security patches enabled for Hypervisor and Guest OS. Data Center Management: VMware vCenter 6.7.0 14351034, HCIBench v2.3.1. Tested by Evaluator Group March 9, 2020. VMware HCIBench

March 9, 2020, April 2, 2020

38.) Up to 8.5x better performance on L3 Packet Forwarding from over a decade of software optimizations and Intel® Xeon® Scalable processor platforms from 2010 to 2020 2nd Generation Intel® Xeon® Gold processor Baseline: 1-node, 1x Intel® Xeon® processor E5645, 2.40GHz, 12MB L3 cache on Green City Platform (CRB), Intel 5520 chipset with 6GB (DDR3 1066MHz, 3x dual rank 2GB, Triple-Channel Configuration) total memory running CentOS* 5.3 x86_​64, kernel 2.6.32.12, and DPDK v1.2.3r0, IxNetwork 5.60.301.28, Intel® ixgbe-3.0.12 10 GbE NIC driver with Multi-queue, using 1x Intel® 82599 Dual Port PCI-Express x8. Tested by Intel on 8 Mar 2010. New Configuration: 1-node, 2x Intel® Xeon® Gold 6252N CPU @ 2.30GHz processor on S2600WFD with 192GB (12 slots / 16 GB / 2933 + ) total memory, ucode 0x5000026 running Ubuntu* 19.04 kernel 5.0.0-23-generic, BIOS version:SE5C620.86B.0X.02.0040.060420190144, and DPDK v19.05, using 1x Intel XXV710-DA4, 3x Intel XXV710-DA2. Tested by Intel on 28 Oct 2019. DPDK L3fwd packet processing

Baseline: March 8, 2010

New: October 28, 2019

39.) Intel® Xeon® processor outperforms Graviton2 by up to 1.76x in database computations (M5.16xlarge instances outperform Graviton2 M6g.16xlarge instances) 2nd Generation Intel® Xeon® Platinum processor

Baseline: M6g.16xlarge 64 vCPU (Graviton2). New configuration: M5.16xlarge 64vCPU (Xeon); both with 256GB memory capacity/instance.

1.65x higher MongoDB v4.4.1 Enterprise, Storage/Instance: EBS io1 2x 800GB, Network BW/Instance 20Gbps for M5, 25Gbps for M6g, Ubuntu 20.04, Kernel 5.4.0-1025-aws.

1.76x higher MySQL 8.0.20, HammerDB v3.2, Storage/Instance: EBS io1 1x 800GB, Network BW/Instance 20Gbps for M5, 25Gbps for M6g, Ubuntu 20.04, Kernel 5.4.0-1029-aws.

1.19x higher PostgreSQL 13.1, HammerDB v4.0, Storage: SSD, Network BW/Instance 20Gbps for M5 and AWS default for M6g, Ubuntu 20.04, Kernel 5.4.0.

MySQL, MongoDB: New Orders per/Minute

PostgreSQL: Maximum Throughput / Under Latency SLA

MySQL, HammerDB: Dec 2020

PostgreSQL, HammerDB: Dec 2020 - Jan 2021

MongoDB: Nov 2020

40.) Intel® Xeon® processor outperforms Graviton2 by up to 1.67x in web/Java computations (M5.16xlarge instances outperform Graviton2 M6g.16xlarge instances) 2nd Generation Intel® Xeon® Platinum processor

Baseline: M6g.16xlarge 64 vCPU (Graviton2). New configuration: M5.16xlarge 64vCPU (Xeon); both with 256GB memory capacity/instance.

1.20x higher Server-side Java, openjdk version "11.0.2" 2019-01-15 for M5 and AdoptOpenJDK version "11.0.7" for M6g, Storage/Instance: EBS, Ubuntu 20.04, Kernel 5.4.0-1009-aws.

1.04x higher WordPress 5.2.0, Storage/Instance: SSD, Complier: PHP 7.3.26-1+ubuntu20.04.1+deb, Other SW: Nginx 1.18.0, MariaDB 10.3, Siege 2.78, Kernel 5.4.0, Ubuntu 20.04 LTS.

1.67x higher ASP.NET Core SDK 5.0.101, Storage/Instance: SSD, Network BW/Instance: 20Gbps for M5 and 25Gbps for M6g, Ubuntu 20.04 LTS, Other SW: TFB ASP.NET KPIs, Kernel 5.4.0.

Server-side Java: Operations per/Second

WordPress: Transactions per/Second

ASP.NET: Geomean RPS (Request per/Second) across ASP.NET KPIs https://github.com/aspnet/Benchmarks/

Server-side Java: Oct 2020

WordPress: Jan 2021

ASP.NET: Dec 2020

41.) Intel® Xeon® processor outperforms Graviton2 by up to 1.30x in compute throughput (M5.16xlarge instances outperform Graviton2 M6g.16xlarge instances) 2nd Generation Intel® Xeon® Platinum processor

Baseline: M6g.16xlarge 64 vCPU (Graviton2). New configuration: M5.16xlarge 64vCPU (Xeon); both with 256GB memory capacity/instance.

1.30x higher SPEC CPU®2017 v1.1.0 and SPEC CPU®2017 v1.1.0 (1-Copy) estimate, Storage/Instance: EBS 500GB, Network BW/Instance: 20Gbps for M5 and 25Gbps for M6g, Ubuntu 20.04.2 LTS for M5 and Ubuntu 20.04.1 LTS for M6g, Other SW: -xCORE-AVX512 -O3 for M5 and -march=armv8.2-a+lse-mtune=neoverse-n1 aarch64-linux-gnu-gcc-10 -O3 for M6g, Kernel 5.4.0-1037-aws for M5 and 5.4.0-1035-aws for M6g, Compiler: ICC 2021.1 for M5 and GCC 10.2.0 for M6g.

1.26x higher SPEC CPU®2017 v1.1.0 and SPEC CPU®2017 v1.1.0 (1-Copy) estimate, Storage/Instance: EBS 500GB, Network BW/Instance: 20Gbps for M5 and 25Gbps for M6g, Ubuntu 20.04.2 LTS for M5 and Ubuntu 20.04.1 LTS for M6g, Other SW: -xCORE-AVX512 -O3 for M5 and -march=armv8.2-a+lse-mtune=neoverse-n1 aarch64-linux-gnu-gcc-10 -O3 for M6g, Kernel 5.4.0-1037-aws for M5 and 5.4.0-1035-aws for M6g, Compiler: ICC 2021.1 for M5 and GCC 10.2.0 for M6g.

Throughput Feb 2021
42.) Intel® Xeon® processor outperforms Graviton2 by up to 1.80x in single-threaded computations (M5.16xlarge instances outperform Graviton2 M6g.16xlarge instances) 2nd Generation Intel® Xeon® Platinum processor

Baseline: M6g.16xlarge 64 vCPU (Graviton2). New configuration: M5.16xlarge 64vCPU (Xeon); both with 256GB memory capacity/instance.

1.80x higher SPEC CPU®2017 v1.1.0 and SPEC CPU®2017 v1.1.0 (1-Copy) estimate, Storage/Instance: EBS 500GB, Network BW/Instance: 20Gbps for M5 and 25Gbps for M6g, Ubuntu 20.04.2 LTS for M5 and Ubuntu 20.04.1 LTS for M6g, Other SW: -xCORE-AVX512 -O3 for M5 and -march=armv8.2-a+lse-mtune=neoverse-n1 aarch64-linux-gnu-gcc-10 -O3 for M6g, Kernel 5.4.0-1037-aws for M5 and 5.4.0-1035-aws for M6g, Compiler: ICC 2021.1 for M5 and GCC 10.2.0 for M6g.

1.56x higher SPEC CPU®2017 v1.1.0 and SPEC CPU®2017 v1.1.0 (1-Copy) estimate, Storage/Instance: EBS 500GB, Network BW/Instance: 20Gbps for M5 and 25Gbps for M6g, Ubuntu 20.04.2 LTS for M5 and Ubuntu 20.04.1 LTS for M6g, Other SW: -xCORE-AVX512 -O3 for M5 and -march=armv8.2-a+lse-mtune=neoverse-n1 aarch64-linux-gnu-gcc-10 -O3 for M6g, Kernel 5.4.0-1037-aws for M5 and 5.4.0-1035-aws for M6g, Compiler: ICC 2021.1 for M5 and GCC 10.2.0 for M6g.

Throughput Feb 2021