Performance Random Performance

Random Number Generator Performance Benchmarks

Auto-generated via RandomPerformanceTests.Benchmark. Run the test to refresh these summary and detail tables.

State repair for a generator restored from JSON or protobuf happens in its constructor or the shared after-deserialization callback. This includes each generator's state, the shared bit/byte reservoirs and GUID scratch buffer. Repair is not repeated in a draw method, so malformed/default serialized state is repaired before the first draw without adding a guard to every later draw. PhotonSpin's one-time warmup priming also happens there, leaving only its block-boundary check in the draw path.

For statistical batteries, the repository's Generator~/WallstopStudios.UnityHelpers.RandomQuality host emits a reproducible little-endian byte stream from an explicit generator, GUID seed and byte count. Long PractRand/TestU01 runs belong in scheduled reporting with pinned tools and expected weak-generator failures; they are not suitable as nondeterministic pull-request gates.

Summary (fastest first)

Random	NextUint (ops/s)	Speed	Quality	Notes
LinearCongruentialGenerator	1,323,700,000	Fastest	Poor	Numerical Recipes 'quick and dirty' LCG (a=1664525, c=1013904223, m=2^32) returning the raw state, so bit k has period only 2^(k+1) -- measured linear complexity of bit k is 2^k+1, and bit 0 simply alternates. Cosmetic use only.
WaveSplatRandom	1,283,400,000	Fastest	Experimental	Single-word chaotic generator; author notes period 2^64 but provides no formal test results—treat as experimental.
BlastCircuitRandom	1,054,900,000	Very Fast	Good	Chaotic ARX mixer rather than a proven statistically optimal generator. Verified here: PractRand 0.95 clean through 8GB, the depth at which SystemRandom fails.
SplitMix64	1,052,300,000	Very Fast	Very Good	Well-known SplitMix64 mixer; passes TestU01 BigCrush and PractRand up to large data sizes in literature. Vigna 2014
FlurryBurstRandom	923,200,000	Fast	Excellent	Six-word ARX-style generator tuned for all-around use. Verified here: PractRand 0.95 clean through 8GB, the depth at which SystemRandom fails. The author reports TestU01 BigCrush passes; that run cannot be checked -- the upstream repository is offline.
PcgRandom	897,900,000	Fast	Excellent	PCG XSH RR 64/32 variant; passes TestU01 BigCrush and PractRand in published results. O'Neill 2014
XoroShiroRandom	754,700,000	Fast	Fair	xoroshiro128+, returning the low 32 bits -- the half its authors document as linear. Measured: bit 0 has linear complexity exactly 128, so NextBool is predictable from 128 draws. Prefer Xoshiro128StarStar or PcgRandom where single bits matter. Blackman & Vigna 2018
IllusionFlow	754,500,000	Fast	Excellent	Five-word rotate/xor/add generator driven by a 32-bit Weyl counter, and the generator PRNG.Instance returns. Verified here: PractRand 0.95 clean through 8GB, the depth at which SystemRandom fails. The author reports 64GB; that run cannot be checked -- the upstream repository is offline.
RomuDuo	750,100,000	Fast	Good	ROMU multiplier with a modified duo update; returns the low 32 bits of the 64-bit product. Not the published romuDuo or romuDuoJr, so their measured results do not transfer. Overton 2020
StormDropRandom	705,200,000	Moderate	Excellent	Large-state ARX generator over a 1024-word (4 KB) ring buffer with two 32-bit control words. Verified here: PractRand 0.95 clean through 8GB, the depth at which SystemRandom fails. The author's own results cannot be checked -- the upstream repository is offline.
XorShiftRandom	602,100,000	Moderate	Fair	Classic 32-bit xorshift; known to fail portions of TestU01 and PractRand, acceptable for lightweight effects only. Marsaglia 2003
WyRandom	440,500,000	Slow	Very Good	Wyhash-based generator; published testing shows it clears BigCrush/PractRand with wide seed coverage. Wang Yi 2019
SquirrelRandom	414,000,000	Slow	Fair	Hash-based generator built on Squirrel3. Measured: fails PractRand 0.95 FPF-14+6/16 at 1GB, reproducibly across four seeds. Good equidistribution for the table lookups it was designed for; not a general-purpose stream. Squirrel Eiserloh
PhotonSpinRandom	261,100,000	Very Slow	Excellent	SHISHUA-inspired generator. Verified here: PractRand 0.95 clean through 8GB, the depth at which SystemRandom fails. The author reports 128GB; that run cannot be checked -- the upstream repository is offline.
UnityRandom	87,600,000	Very Slow	Fair	Mirrors UnityEngine.Random, documented by Unity as Xorshift 128; suitable for legacy compatibility but not high-stakes simulation. UnityEngine.Random
SystemRandom	64,700,000	Very Slow	Poor	Reimplements the classic .NET System.Random algorithm (Knuth subtractive lagged-Fibonacci, mod 2^31-1) so its sequence is serializable and stable across runtimes. Fails modern statistical batteries. System.Random considered harmful
DotNetRandom	55,900,000	Very Slow	Poor	Wraps System.Random, which on Mono and for seeded .NET Core is Knuth's subtractive lagged-Fibonacci generator (mod 2^31-1), not an LCG. The sequence is runtime-dependent, so do not rely on it for cross-platform determinism. System.Random considered harmful

Detailed Metrics

Random	NextBool	Next	NextUint	NextFloat	NextDouble	NextUint (Range)	NextInt (Range)
LinearCongruentialGenerator	785,200,000	538,800,000	1,323,700,000	184,300,000	296,300,000	582,900,000	498,700,000
WaveSplatRandom	787,300,000	528,500,000	1,283,400,000	184,300,000	297,900,000	529,000,000	458,200,000
BlastCircuitRandom	788,200,000	537,400,000	1,054,900,000	183,800,000	293,200,000	479,700,000	421,200,000
SplitMix64	795,900,000	537,200,000	1,052,300,000	184,400,000	297,700,000	482,500,000	439,100,000
FlurryBurstRandom	767,800,000	526,500,000	923,200,000	182,300,000	293,100,000	449,800,000	404,100,000
PcgRandom	249,800,000	527,300,000	897,900,000	181,000,000	274,700,000	447,700,000	405,000,000
XoroShiroRandom	761,000,000	359,000,000	754,700,000	157,900,000	192,400,000	422,100,000	376,900,000
IllusionFlow	779,300,000	529,200,000	754,500,000	183,000,000	281,600,000	439,200,000	390,000,000
RomuDuo	784,600,000	359,400,000	750,100,000	156,100,000	188,700,000	437,700,000	384,600,000
StormDropRandom	776,600,000	523,600,000	705,200,000	180,600,000	252,400,000	393,500,000	356,700,000
XorShiftRandom	783,300,000	534,100,000	602,100,000	184,100,000	283,200,000	474,500,000	383,500,000
WyRandom	749,500,000	363,000,000	440,500,000	159,000,000	185,300,000	290,000,000	277,800,000
SquirrelRandom	756,300,000	382,700,000	414,000,000	158,800,000	197,800,000	355,500,000	309,500,000
PhotonSpinRandom	713,300,000	220,800,000	261,100,000	119,900,000	115,800,000	217,600,000	214,200,000
UnityRandom	628,900,000	76,700,000	87,600,000	59,700,000	38,800,000	81,800,000	81,900,000
SystemRandom	146,200,000	145,700,000	64,700,000	131,300,000	138,600,000	58,700,000	57,800,000
DotNetRandom	544,600,000	53,100,000	55,900,000	44,700,000	26,700,000	53,400,000	51,700,000

Generators added since the last benchmark run

The tables above are rewritten only by a benchmark run, so a generator added since the last one is absent until the .github/workflows/unity-benchmarks.yml workflow next runs. Absence here says nothing about quality: statistical standing is measured separately, by the bit-plane linearity gate on every pull request and by the scheduled PractRand battery. See Random Generators for the current ratings.