ThinkSystem SR655 Sets World Record with New One-Node SPEC ACCEL OpenACC ResultPerformance Benchmark Result

7 Aug 2019
Form Number
PDF size
5 pages, 153 KB


The Lenovo ThinkSystem SR655 has set a new 1-node 1-socket performance world record with the SPECaccel_acc_base metric from the SPEC ACCEL Benchmark.

This document summarizes the SPECaccel benchmark result for the ThinkSystem SR655 that was published on August 7, 2019. The document lists the result, summarizes the major components of the configuration under test, and provides the link to the relevant benchmark organization’s web site for details about the result.


The Lenovo ThinkSystem SR655 server has set a new 1-node 1-socket performance world record with the SPECaccel_acc_base metric from the SPEC ACCEL Benchmark. The SPEC ACCEL Benchmark suite is the industry standard to evaluate hardware-based accelerator devices and the performance of parallel computing workloads.

Lenovo ThinkSystem SR650 This new benchmark result, published in a new SPEC Report on August 7, 2019, demonstrate that the ThinkSystem SR655 continues Lenovo’s leadership with outstanding performance for the server industry.

The ThinkSystem SR655 has achieved the following score (1):

  • SPECaccel_acc_base = 13.2

This result is the best one node 1-socket performance in the industry, 8.2% faster than Lenovo’s own result (2), 5.84 time faster than the Cray’s 1-node 1-socket result (3).

Table 1. Comparison of results
System Result
Cores CPUs Memory
Lenovo (1) ThinkSystem SR655 (AMD EPYC 7742 with NVIDIA Tesla V100) 13.2 64 1 256 GB
Lenovo (2) ThinkSystem SR650 (Intel Xeon Gold 6240 with NVIDIA Tesla V100) 12.2 32 2 768 GB
Cray (3) Cray XK7 system (AMD Opteron 6274 with NVIDIA Tesla K20X) 2.26 16 1 32 GB

The SR655 was configured as follows for the benchmark audit:

  • Lenovo ThinkSystem SR655
  • 1x AMD EPYC 7742 Processor (64 cores, 2.25 GHz)
  • 256 GB memory (8 x 32GB RDIMMs at 3200 MHz)
  • 1x 480 GB SATA 2.5" SSD
  • SUSE Linux Enterprise Server SP4, Kernel 4.12.14-94.41-default
  • Accelerator: 1x NVIDIA Tesla V100 16GB

Results referenced are current as of August 7, 2019.

(1) The new Lenovo benchmark result can be found at:

(2) The previous Lenovo result can be found at:

(3) The Cray result can be found at:

About the ThinkSystem SR655

The Lenovo ThinkSystem SR655 with the next generation AMD EPYC architecture delivers outstanding TCO for virtualization, big data, analytics, and scale up software defined deployments. It offers two-socket performance in a 2U, single-socket design, featuring higher core, and storage density; increased I/O throughput with less latency; along with built-in system security.

With the enterprise class AMD EPYC 7002 Generation processor, the world’s first 7nm data center CPU, the SR655 features up to an unprecedented 64 cores with 128 PCIe Gen4 lanes in a single socket to reduce bottlenecks and increase server utilization.

Compared to the previous processor generations, ThinkSystem servers with AMD architecture deliver up to 2X performance and 4X floating point capability, providing faster data transfer and analytics without sacrificing memory capacity or I/O with PCIe Gen4 support and faster memory speeds up to 3200 MHz.

Key features of the SR655 include:

  • Multi-GPU optimized rack server, providing support for up to 6 single-wide GPUs that offer more workload acceleration in AI Inference and virtualized desktop infrastructure (VDI).
  • 64 cores to handle heavy-lift virtualization workloads; provides cutting edge application efficiency in health care applications such as EMR, PACS, and medical imaging or electronic trading platforms for financial services applications.
  • Support for up to 32 NVMe solid-state drives; when paired with high speed networking, make the system an excellent choice for workloads that need large amounts of low-latency high-bandwidth storage, including virtualized clustered SAN solutions, software-defined storage (SDS), and applications leveraging NVMe over Fabrics.

About SPECaccel

The SPEC ACCEL benchmark suite provides a comparative measure the performance of hardware accelerator devices and their supporting software tool chains using computationally-intensive parallel applications. The suite is comprised of scientific applications used in High Performance Computing (HPC) and focuses on parallel computing performance. 

The suite has been ported using several accelerator programming models each of which has been released as separate benchmark components:

  • SPEC ACCEL OpenCL -- based on the Open Computing Language (OpenCL) 1.1 framework
  • SPEC ACCEL OpenACC -- based on the Open Accelerators (OpenACC) 1.0 programming standard for parallel computing
  • SPEC ACCEL OpenMP -- based on the Open Multi-Processing (OpenMP) 4.5 application programming interface

The product consists of source code benchmarks that are developed from real user applications.

For more information and SPEC ACCEL results, see

Learn more

To learn more about solutions for parallel computing workloads, please contact your Lenovo Sales Representative.

To find out more about SPEC, visit

To learn more about the Lenovo ThinkSystem SR655 server, visit the SR655 product web page:

Related product families

Product families related to this document are the following:


Lenovo and the Lenovo logo are trademarks or registered trademarks of Lenovo in the United States, other countries, or both. A current list of Lenovo trademarks is available on the Web at

The following terms are trademarks of Lenovo in the United States, other countries, or both:

The following terms are trademarks of other companies:

Intel® and Xeon® are trademarks of Intel Corporation or its subsidiaries.

Linux® is the trademark of Linus Torvalds in the U.S. and other countries.

SPEC® and SPEC ACCEL® are trademarks of the Standard Performance Evaluation Corporation (SPEC).

Other company, product, or service names may be trademarks or service marks of others.