CPU Benchmark - Group S

Available Options

Choose the benchmark to execute (replace BENCHMARK with the name of one of the benchmarks down below):

.\benchmarker.exe --benchmark={BENCHMARK}

Automatically detect the number of logical processors and execute the benchmark multithreaded (number of threads = number of logical processors)

.\benchmarker.exe --multithreaded

Manually set the amount of threads to use. Overwrites --multithreaded

.\benchmarker.exe --threads=4

Run the benchmark X times and average out the results. By default the benchmark will be run 3 times

\.benchmarker.exe --runs=5

Run the benchmarks in a memory efficient manner (on low-memory devices for example)

\.benchmarker.exe --memory-efficient

Run a stress test for the selected benchmark (currently only available for all extension benchmarks). Stop by pressing any key.

\.benchmarker.exe --stress

List all available benchmarks

\.benchmarker.exe --list-benchmrks

To combine the above

.\benchmarker.exe --benchmark=zip --multithreaded --runs=64

Available Benchmarks

General

• ALL (runs all of the below benchmarks)
• COMPRESSION (runs all compression benchmarks)
• DECOMPRESSION (runs all decompression benchmarks)
• ARITHMETIC (runs all arithmetic benchmarks)
• EXTENSION (runs all instruction extension benchmarks)
• CRYPTOGRAPHY (runs all cryptography benchmarks)
• INT (runs all integer-related benchmarks)
• FLOAT (runs all float-related benchmarks)
• AVX (runs all AVX benchmarks)
• SSE (runs all SSE benchmarks)
• Latency (runs all latency-related benchmarks)
• Experimental (runs all experimental benchmarks)

Compression

Note: All compression algorithms are run with the highest compression level (if applicable).

• ZIP
• GZip
• BZip2
• Deflate

Decompression

Note: All compression/decompression algorithms are run with the highest compression level (if applicable).

• ZIP
• GZip
• BZip2
• Deflate/Inflate

Arithmetic

• Arithmetic_Int (Simple Integer Benchmark for byte, short, int and long performance)
• Arithmetic_Float (Simple Float Benchmark)
• Arithmetic_Double (Simple Double Benchmark)
• Arithmetic_FP16 (Simple FP16 (half-precision 16-bit floating point) Benchmark)

Instruction extensions

• AVX
• SSE
• SSE2 (128-bit integer)
• AVX2Int (256-bit integer)
• FMA (fused multiply-add of 128-bit float)
• SSE4 (SSE4.1 Benchmark)
• SSE4-CRC32C (SSE4.2 CRC32C Benchmark)

Cryptography

• Encryption
• Decryption
• CSPRNG (cryptographically secure pseudo-random number generator)
• Hashing (Hash with SHA-256)

Latency

• L2CacheLatency (Tests L2 Cache Latency and outputs it into the file ./nanoseconds.txt, name is currently l2cachelatency-experimental)

References

The benchmark will print reference points for a 3900x (non-overclocked) system. Detailed system specs:

• R9 3900x
• 32 GB 3200 MHz CL16 RAM
• ASUS Hero VIII

Authors

• Siddharth Sharma

Acknowledgements

• CommandLine for the command line options
• ShellProgressBar for the progress bar
• HTMLAgilityPack used in the HTML benchmark
• Newtonsoft.Json used in the JSON benchmark and for saves
• SharpZipLib used for the (de)compression benchmarks