# passmark

> Passmark benchmark workloads: PassMark: CPU Compression Test, PassMark: CPU Encryption Test, PassMark: CPU Extended Instructions Test and 13 more.

Passmark

## PassMark: CPU Compression Test {#passmark-cpu_compression_test}

Using Gzip (Crypto++ 8.6) to compress blocks of data with a 4MB memory buffer size per core.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:cpu_compression_test</code></dd><dt>Unit of measurement</dt><dd>Kilobytes per second (kB/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>2.1K – 7.1M <small>kB/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: CPU Compression Test](/benchmarks/histograms/passmark-cpu_compression_test.svg)

## PassMark: CPU Encryption Test {#passmark-cpu_encryption_test}

Encrypting blocks of random data using AES, SHA256 and ECDSA with any available specialized CPU instruction sets and memory buffers totaling about 1MB per core.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:cpu_encryption_test</code></dd><dt>Unit of measurement</dt><dd>Megabytes per second (MB/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>73.1 – 423.2K <small>MB/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: CPU Encryption Test](/benchmarks/histograms/passmark-cpu_encryption_test.svg)

## PassMark: CPU Extended Instructions Test {#passmark-cpu_extended_instructions_test}

Testing how fast the CPU can perform mathematical operations using extended instructions, such as SSE, FMA, AVX, AVX512 and NEON.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:cpu_extended_instructions_test</code></dd><dt>Unit of measurement</dt><dd>Millions of matrices per second (Mmat/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>138.6 – 485.3K <small>Mmat/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: CPU Extended Instructions Test](/benchmarks/histograms/passmark-cpu_extended_instructions_test.svg)

## PassMark: CPU Floating Point Maths Test {#passmark-cpu_floating_point_maths_test}

Testing how fast the CPU can perform mathematical floating point operations, using large sets of an equal number of random 32-bit and 64-bit floating point numbers for addition (30% of the time), subtraction (30% of the time), multiplication (30% of the time) and division (10% of the time), with floating point buffers totaling about 240kb per core.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:cpu_floating_point_maths_test</code></dd><dt>Unit of measurement</dt><dd>Millions of operations per second (Mops/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>355.1 – 1.5M <small>Mops/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: CPU Floating Point Maths Test](/benchmarks/histograms/passmark-cpu_floating_point_maths_test.svg)

## PassMark: CPU Integer Maths Test {#passmark-cpu_integer_maths_test}

Testing how fast the CPU can perform mathematical integer operations, using large sets of an equal number of random 32-bit and 64-bit integers for addition, subtraction, multiplication and division, with integer buffers totaling about 240kb per core.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:cpu_integer_maths_test</code></dd><dt>Unit of measurement</dt><dd>Millions of operations per second (Mops/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>567.8 – 2M <small>Mops/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: CPU Integer Maths Test](/benchmarks/histograms/passmark-cpu_integer_maths_test.svg)

## PassMark: CPU Mark {#passmark-cpu_mark}

A composite average of the Integer, Floating point, Prime and String Sorting test scores, which can be used to compare CPUs from different platforms (even e.g. desktop vs mobile).

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:cpu_mark</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>0 – 192.9K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: CPU Mark](/benchmarks/histograms/passmark-cpu_mark.svg)

## PassMark: CPU Physics Test {#passmark-cpu_physics_test}

Simulating the same physics interactions as many times as possible within a timeframe, using the Bullet Physics Engine (version 2.88 for x86, 3.07 for ARM).

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:cpu_physics_test</code></dd><dt>Unit of measurement</dt><dd>Frames per second (fps)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>0 – 64.5K <small>fps</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: CPU Physics Test](/benchmarks/histograms/passmark-cpu_physics_test.svg)

## PassMark: CPU Prime Numbers Test {#passmark-cpu_prime_numbers_test}

Finding prime numbers using the Sieve of Atkin formula (with a limit of 32 million) on 64-bit integers with 4MB of memory per thread.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:cpu_prime_numbers_test</code></dd><dt>Unit of measurement</dt><dd>Million prime numbers per second (Mnums/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>1.1 – 4.6K <small>Mnums/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: CPU Prime Numbers Test](/benchmarks/histograms/passmark-cpu_prime_numbers_test.svg)

## PassMark: CPU Single Threaded Test {#passmark-cpu_single_threaded_test}

Using a single logical core for a mixture of floating point, string sorting and data compression tests.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:cpu_single_threaded_test</code></dd><dt>Unit of measurement</dt><dd>Millions of operations per second (Mops/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>161.6 – 4.6K <small>Mops/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

:::warning

This benchmark measures single-threaded CPU performance only, so scores do not increase with vCPU count beyond one core.

:::

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: CPU Single Threaded Test](/benchmarks/histograms/passmark-cpu_single_threaded_test.svg)

## PassMark: CPU String Sorting Test {#passmark-cpu_string_sorting_test}

Sorting strings using the Quicksort algorithm with memory buffers totaling about 25MB per core.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:cpu_string_sorting_test</code></dd><dt>Unit of measurement</dt><dd>Thousands of strings per second (Kstrings/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>276.9 – 963.3K <small>Kstrings/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: CPU String Sorting Test](/benchmarks/histograms/passmark-cpu_string_sorting_test.svg)

## PassMark: Database Operations {#passmark-database_operations}

Single threaded and multi-threaded CRUD operations, such as INSERT (40%), SELECT (26%), UPDATE (24%), and DELETE (10%) on a relational database with 4 tables and 1k rows per table.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:database_operations</code></dd><dt>Unit of measurement</dt><dd>Thousands of operations per second (Kops/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>0 – 62.6K <small>Kops/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

:::warning

This benchmark does not scale well on instances with more than 32 vCPUs. Empirical data shows measured scores plateau and may even degrade as vCPU count increases, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: Database Operations](/benchmarks/histograms/passmark-database_operations.svg)

## PassMark: Memory Latency {#passmark-memory_latency}

Measuring the time it takes for a single byte of memory to be transferred to the CPU for processing. A 512 MB buffer is allocated and then filled with pointers to other locations in the buffer, looping through a linked list.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:memory_latency</code></dd><dt>Unit of measurement</dt><dd>Nanoseconds (ns)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>0 – 405.3 <small>ns</small></dd><dt>Better</dt><dd>↓ Lower is better</dd></dl>

:::warning

This benchmark score is largely independent of vCPU count within an instance family and mainly reflects hardware generation and memory characteristics, not instance size.

:::

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: Memory Latency](/benchmarks/histograms/passmark-memory_latency.svg)

## PassMark: Memory Mark {#passmark-memory_mark}

A composite score of PassMark's Database and Memory test cases

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:memory_mark</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>0 – 4.7K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

:::warning

This benchmark does not scale well on instances with more than 16 vCPUs. Empirical data shows measured scores plateau and may even degrade as vCPU count increases, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: Memory Mark](/benchmarks/histograms/passmark-memory_mark.svg)

## PassMark: Memory Read Cached {#passmark-memory_read_cached}

Read a combination of 32-bit and 64-bit data from memory.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:memory_read_cached</code></dd><dt>Unit of measurement</dt><dd>Megabytes per second (MB/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>0 – 42K <small>MB/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

:::warning

This benchmark score is largely independent of vCPU count within an instance family and mainly reflects hardware generation and memory characteristics, not instance size.

:::

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: Memory Read Cached](/benchmarks/histograms/passmark-memory_read_cached.svg)

## PassMark: Memory Read Uncached {#passmark-memory_read_uncached}

Read a combination of 32-bit and 64-bit data from memory using a 512 MB block size.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:memory_read_uncached</code></dd><dt>Unit of measurement</dt><dd>Megabytes per second (MB/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>0 – 40.3K <small>MB/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

:::warning

This benchmark score is largely independent of vCPU count within an instance family and mainly reflects hardware generation and memory characteristics, not instance size.

:::

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: Memory Read Uncached](/benchmarks/histograms/passmark-memory_read_uncached.svg)

## PassMark: Memory Write {#passmark-memory_write}

Write a combination of 32-bit and 64-bit data to the memory using a 512 MB block size.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>passmark:memory_write</code></dd><dt>Unit of measurement</dt><dd>Megabytes per second (MB/s)</dd><dt>Server types evaluated</dt><dd>3,681</dd><dt>Total measurements</dt><dd>3,681</dd><dt>Range of measurements</dt><dd>0 – 35.9K <small>MB/s</small></dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

:::warning

This benchmark score is largely independent of vCPU count within an instance family and mainly reflects hardware generation and memory characteristics, not instance size.

:::

Score distribution <small>(3,681 measurements)</small>

![Score distribution of PassMark: Memory Write](/benchmarks/histograms/passmark-memory_write.svg)
