X

Everything you want and need to know about Oracle SPARC systems performance

SHA Digest Encryption: SPARC S7 Performance, Beats Intel E5-2699 v4 Per Core Under Load

Brian Whitney
Principal Software Engineer

Oracle's cryptography benchmark measures security performance on important Secure Hash Algorithm (SHA) functions. Oracle's SPARC S7 processor with its security software in silicon is faster than current and recent x86 servers. In this test, the performance of on-processor digest operations is measured for three sizes of plaintext inputs (64, 1024 and 8192 bytes) using three SHA2 digests (SHA512, SHA384, SHA256) and the older, weaker SHA1 digest. Multiple parallel threads are used to measure each processor's maximum throughput. Oracle's SPARC S7-2 server shows dramatically faster digest computation compared to current x86 two processor servers.

  • SPARC S7 processors ran 7.5 times faster per core computing multiple parallel SHA512 digests of 8 KB inputs (in cache) than Cryptography for Intel Integrated Performance Primitives for Linux (library) on Intel Xeon Processor E5-2699 v4.

  • SPARC S7 processors ran 7.1 times faster per core computing multiple parallel SHA256 digests of 8 KB inputs (in cache) than Cryptography for Intel Integrated Performance Primitives for Linux (library) on Intel Xeon Processor E5-2699 v4.

  • SPARC S7 processors ran 2.6 times faster per core computing multiple parallel SHA1 digests of 8 KB inputs (in cache) than Cryptography for Intel Integrated Performance Primitives for Linux (library) on Intel Xeon Processor E5-2699 v4.

SHA1 and SHA2 operations are an integral part of Oracle Solaris, while on Linux they are performed using the add-on Cryptography for Intel Integrated Performance Primitives for Linux (library).

Oracle has also measured AES (CFB, GCM, CCM, CBC) cryptographic performance on the SPARC S7 processor.

Performance Landscape

Presented below are results for computing SHA1, SHA256, SHA384 and SHA512 digests for input plaintext sizes of 64, 1024 and 8192 bytes. Results are presented as MB/sec (10**6). All SPARC S7 processor results were run as part of this benchmark effort. All other results were run during previous benchmark efforts.

Digest Performance – SHA512

Performance is presented for SHA512 digest. The digest was computed for 64, 1024 and 8192 bytes of pseudo-random input data (same data for each run).

Processors Total
Cores
Performance (MB/sec) Perf/Core (MB/sec/core)
64B 1024B 8192B 64B 1024B 8192B
2 x SPARC M7, 4.13 GHz 64 39,201 167,072 184,944 613 2,611 2,890
2 x SPARC T5, 3.6 GHz 32 18,717 73,810 78,997 585 2,307 2,469
2 x SPARC S7, 4.26 GHz 16 10,231 43,099 47,820 639 2,694 2,989
2 x Intel E5-2699 v4, 2.2 GHz 44 6,973 15,412 17,616 158 350 400
2 x Intel E5-2699 v3, 2.3 GHz 36 3,949 9,214 10,681 110 256 297
2 x Intel E5-2697 v2, 2.7 GHz 24 2,681 6,631 7,701 112 276 321

Digest Performance – SHA384

Performance is presented for SHA384 digest. The digest was computed for 64, 1024 and 8192 bytes of pseudo-random input data (same data for each run).

Processors Total
Cores
Performance (MB/sec) Perf/Core (MB/sec/core)
64B 1024B 8192B 64B 1024B 8192B
2 x SPARC M7, 4.13 GHz 64 39,697 166,898 185,194 620 2,608 2,894
2 x SPARC T5, 3.6 GHz 32 18,814 73,770 78,997 588 2,305 2,469
2 x SPARC S7, 4.26 GHz 16 10,315 43,158 47,763 645 2,697 2,985
2 x Intel E5-2699 v4, 2.2 GHz 44 6,909 15,353 17,618 157 349 400
2 x Intel E5-2699 v3, 2.3 GHz 36 4,061 9,263 10,678 113 257 297
2 x Intel E5-2697 v2, 2.7 GHz 24 2,774 6,669 7,706 116 278 321

Digest Performance – SHA256

Performance is presented for SHA256 digest. The digest was computed for 64, 1024 and 8192 bytes of pseudo-random input data (same data for each run).

Processors Total
Cores
Performance (MB/sec) Perf/Core (MB/sec/core)
64B 1024B 8192B 64B 1024B 8192B
2 x SPARC M7, 4.13 GHz 64 45,148 113,648 119,929 705 1,776 1,874
2 x SPARC T5, 3.6 GHz 32 21,140 49,483 51,114 661 1,546 1,597
2 x SPARC S7, 4.26 GHz 16 11,872 29,371 30,961 742 1,836 1,935
2 x Intel E5-2699 v4, 2.2 GHz 44 5,103 11,174 12,037 116 254 274
2 x Intel E5-2699 v3, 2.3 GHz 36 3,446 7,785 8,463 96 216 235
2 x Intel E5-2697 v2, 2.7 GHz 24 2,404 5,570 6,037 100 232 252

Digest Performance – SHA1

Performance is presented for SHA1 digest. The digest was computed for 64, 1024 and 8192 bytes of pseudo-random input data (same data for each run).

Processors Total
Cores
Performance (MB/sec) Perf/Core (MB/sec/core)
64B 1024B 8192B 64B 1024B 8192B
2 x SPARC M7, 4.13 GHz 64 47,640 92,515 97,545 744 1,446 1,524
2 x SPARC T5, 3.6 GHz 32 21,052 40,107 41,584 658 1,253 1,300
2 x SPARC S7, 4.26 GHz 16 12,665 23,899 25,209 792 1,494 1,576
2 x Intel E5-2699 v4, 2.2 GHz 44 8,566 23,901 26,752 195 543 608
2 x Intel E5-2699 v3, 2.3 GHz 36 6,677 18,165 20,405 185 505 567
2 x Intel E5-2697 v2, 2.7 GHz 24 4,649 13,245 14,842 194 552 618

Configuration Summary

SPARC S7-2 server
2 x SPARC S7 processor, 4.26 GHz
1 TB memory
Oracle Solaris 11.3

SPARC T7-2 server
2 x SPARC M7 processor, 4.13 GHz
1 TB memory
Oracle Solaris 11.3

SPARC T5-2 server
2 x SPARC T5 processor, 3.60 GHz
512 GB memory
Oracle Solaris 11.2

Oracle Server X6-2L system
2 x Intel Xeon Processor E5-2699 v4, 2.20 GHz
256 GB memory
Oracle Linux 7.2
Intel Integrated Performance Primitives for Linux, Version 9.0 (Update 2) 17 Feb 2016

Oracle Server X5-2 system
2 x Intel Xeon Processor E5-2699 v3, 2.30 GHz
256 GB memory
Oracle Linux 6.5
Intel Integrated Performance Primitives for Linux, Version 8.2 (Update 1) 07 Nov 2014

Sun Server X4-2 system
2 x Intel Xeon Processor E5-2697 v2, 2.70 GHz
256 GB memory
Oracle Linux 6.5
Intel Integrated Performance Primitives for Linux, Version 8.2 (Update 1) 07 Nov 2014

Benchmark Description

The benchmark measures cryptographic capabilities in terms of general low-level encryption, in-cache and on-chip using various digests, including SHA1 and SHA2 (SHA256, SHA384, SHA512).

The benchmark results were obtained using tests created by Oracle which use various application interfaces to perform the various digests. They were run using optimized libraries for each platform to obtain the best possible performance. The encryption tests were run with pseudo-random data of sizes 64 bytes, 1024 bytes and 8192 bytes. The benchmark tests were designed to run out of cache, so memory bandwidth and latency are not the limitations.

See Also

Disclosure Statement

Copyright 2016, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 6/29/2016.

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.