PrimeGrid — Mapping Prime Numbers with Computers Worldwide

PrimeGrid is not one project searching for one enormous prime. Through BOINC it operates searches with different mathematical forms and purposes: Cullen, Woodall, generalized Fermat, Proth, factorial, Sierpiński and Riesel problems, and arithmetic progressions of primes. Participants select a subproject that fits their device, supported application, and acceptable task length. Whereas GIMPS focuses mainly on Mersenne numbers of the form 2^p−1, PrimeGrid hosts a portfolio of prime families and conjecture searches.

At 00:53 UTC on 29 August 2026 the official homepage displayed cumulative totals of 357,988 registered users, 893,527 hosts, 101,804 primes found, 38,124 reports to The Prime Pages, and 3,738 mega primes of at least one million digits. Estimated throughput was 3,071.389 TFLOPS. These are not counts of simultaneously connected people, nor do all discoveries carry the same mathematical weight. Because cumulative and estimated values move, this article timestamps the observation and leaves the latest values to the official dashboard.

PrimeGrid offers the possibility that a personal computer will be the first to test an unknown number. A discovery is not an isolated winning ticket, however. People who sieved candidates, implemented and optimized applications, independently checked the computation, and operated the servers and databases all support one public record.

The story began on 12 June 2005. Rytis Slatkevičius opened Message@Home to 50 users from a home laptop while testing a Perl implementation of a BOINC server. Its first Message7 experiment brute-forced a message encoded with MD5. The name PrimeGrid was chosen on 1 September 2005. The present mathematics project did not arrive fully formed; it grew from a small experiment in making distributed-computing infrastructure work.

Early work included the RSA-640 and then RSA-768 factoring challenges. In March 2006 the project moved away from factorization toward prime generation and search. Collaboration with Riesel Sieve and Twin Prime Search and the introduction of the LLR application that year shaped the number-theory project that followed. Its former experiment with a cryptographic challenge and its present prime searches should not be treated as one objective.

Challenge Series, PRPNet, generalized Fermat searches, and GPU-enabled Genefer were added over time. A peer-reviewed 2015 paper describes the transformation from brute-force factorization of RSA numbers into a large volunteer-computing platform for number theory. BOINC supplies distribution and accounting; PrimeGrid builds search plans, specialized applications, validation, and discovery reporting for each family of numbers on top.

Restricting the form—to Cullen n×2^n+1, Woodall n×2^n−1, or generalized Fermat b^(2^m)+1 numbers, for example—does more than make unknown primes easier. It enables specialized tests and faster big-integer arithmetic, permits observations to be compared with theoretical distributions, and can eliminate candidates in open conjectures one at a time. A result may matter because it is the first of a form, excludes a k, or establishes a new arrangement even when it is not the largest by digits.

The first AP27, found on 23 September 2019 after more than three years of searching, embodies that distinction. Twenty-seven numbers are all prime and every adjacent pair has the same difference. PrimeGrid is not merely a longest-number contest; it is a portfolio of investigations that survey the terrain of primes in different ways.

Candidate design comes first. A mathematical form and unexplored range are selected, and known factors and prior results are removed. Sieving then cheaply eliminates candidates divisible by many smaller primes. A deeper sieve leaves fewer survivors, but eventually the sieve itself costs more than testing the remaining candidates directly. That trade-off is why PrimeGrid documentation discusses an optimal sieve depth.

Survivors receive a probable-prime (PRP) test in software such as LLR, PRST, or Genefer. Nearly all composites fail here. A passing result is exceptionally promising, but one host displaying “prime” is not yet a record. Some long main computations produce data that a short proof task can verify; other work is independently repeated on a second host. Arithmetic-progression searches and sieves do not all use the fast-proof path.

Where required, another program or a method such as N−1 then supplies a mathematical primality proof, and The Prime Pages applies its own checks when registering a result. “Every task is always computed twice” and “one proof task proves everything mathematically” are therefore both inaccurate. Error detection, reproducibility of the computation, and proof of primality are combined differently for different work.

Household hosts are exposed to overclocking, memory errors, drivers, power instability, and software faults. PrimeGrid spends resources on computational proofs and double checks not simply because volunteers are untrusted, but because honest hardware can err during a long big-integer calculation. A speed ranking and reproducibility are separate goals.

“Probable prime” does not mean fake. Appropriate tests are designed so that a composite passing them is extraordinarily unlikely, and PRP tests are indispensable for discovery. World records and mathematical claims nevertheless record the probable or proven status, program, hardware, checker, and date so readers can distinguish levels of evidence.

PrimeGrid’s official application list includes implementations for x86 and Arm CPUs, NVIDIA CUDA, AMD and Intel OpenCL, and Apple platforms. Not every subproject can use every device. Candidate form, integer size, FFT or multiplication method, memory, double precision, drivers, checkpointing, and validation determine which hardware fits. There is no single table in which having a GPU makes every task faster than a CPU.

LLR and LLR2, PRST, Genefer, OpenPFGW, AP-specific code, and sievers such as sr2sieve and AthGFNSieve perform different roles. A sieve rejects many candidates with light operations; a PRP test subjects one survivor to long big-integer arithmetic; proof needs different data and algorithms. Even on one host, changing the application or number size moves the bottleneck among arithmetic units, memory, communication, and heat.

A February 2025 generalized Fermat prime, 13520762^524288+1, was found as a PRP with Genefer on a GeForce RTX 3060 Ti, then confirmed with LLR on a Ryzen 9 7950X3D in about 20 hours 40 minutes. It is a tangible case of GPU search and CPU confirmation using different methods within one discovery. The hardware names are only part of a chain that also includes software authors, sievers, server operators, and checkers.

On 12 October 2025 PrimeGrid found 2524190^2097152+1. It is the first GFN-21 prime and has 13,426,224 digits. In August 2026 The Prime Pages recorded it as proven and sixth among all known primes. Its N−1 verification there took 78.12 days. The gap between completing the discovery computation and recording a proven result exposes the long verification stage.

A different kind of value appears in the 2016 result 10223×2^31172165+1. Beyond its 9,383,761 digits, it eliminated k=10223 from Seventeen or Bust. It did not merely add a giant number; it removed one candidate from the finite open set concerning the smallest Sierpiński number.

The 2019 AP27 found 27 primes beginning near an 18-digit value and separated by an equal interval. Each number is far smaller than a million-digit record, yet their 27-term arithmetic progression was a world first. Digit count, special form, elimination of a conjecture candidate, and a first arrangement should not be collapsed into one league table.

BOINC credit is participation accounting for returned computation. PrimeGrid ranks participants, teams, countries, computers, and prime finders. Alongside credit badges for individual subprojects are discovery badges for mega primes, elimination of conjecture candidates, AP26, AP27 and AP28, and world firsts. Challenge Series and Tour de Primes add timed individual and team contests, giving the search a season and a story.

That system does not contradict contributing without financial compensation. Credit is neither fiat money nor a cryptoasset, and it grants no ownership of a research result. Rankings can sustain motivation and belonging, while validators and proofs separately determine correctness. A high-credit participant cannot make a number prime by vote.

The prime-reporting policy sets a contact procedure for the initial finder and, if no response arrives, the double checker and eventually anonymous registration. Behind a finder’s name are also people who prepared the range, ran the sieve, and checked the result. Preserving public results and program names turns a leaderboard into a research history that others can inspect.

Primes can look irregular while still obeying known structure: density described by the prime number theorem, congruence restrictions, and possible positions within special formulas. The Cullen, Proth, generalized Fermat, Sierpiński, and Riesel numbers studied by PrimeGrid are all numbers with form. Knowing the form enables specialized algorithms; it does not produce a formula in which every value is prime or reveal the secret factors of arbitrary cryptographic keys.

The 2002 AKS result showed that deciding whether an integer N is prime can be done deterministically in polynomial time. That was a major theoretical breakthrough, but it is not an algorithm for efficiently recovering secret p and q from an RSA public value N=p×q. “Is this number prime?” and “Which primes made this product?” have different inputs and request different answers.

“A pattern in primes would bring down the cryptographic fortress” is therefore not factual without a condition. What would change cryptography is a practical method for factoring arbitrary deployed keys, a practical discrete-log algorithm, a way to predict secret values from biased key generation, or an attack exploiting the structure of particular parameters. A beautiful distribution law or another enormous prime does not itself supply that capability.

Large-prime searches advance number theory, big-integer arithmetic, error detection, distributed verification, and hardware optimization. Their contribution to cryptography is indirect in this sense. Discovering a public record prime with a special form does not change the security level of deployed RSA keys or Bitcoin keys. Key strength must account for randomness, parameters, protocol, implementation, and operations—not bit length alone.

Prime research and cryptography are not unrelated, either. An algorithm that makes factorization or discrete logarithms practical, bias in the primes selected by a weak key generator, or precomputation against reused parameters can change security. The decisive question is not whether both fields use the word “prime,” but whether a discovery gives an attacker a practical route to the secret value.

FIPS 186-5 specifies that RSA signing-key p and q be generated as random probable or provable primes satisfying required sizes and conditions. Public million-digit record primes are not used as secret factors. An RSA-2048 modulus is about 617 decimal digits and each factor is roughly half that size, but a digit comparison alone does not determine security strength. PrimeGrid’s huge primes are public research results, not secret-key material.

Real failures have occurred when insufficient randomness made different RSA keys share a prime factor. Heninger and colleagues compared large populations of public keys and showed cases where taking the greatest common divisor of two moduli recovered the common factor and therefore the private keys. This was not a mysterious pattern in all primes; it was a failure of implementation and entropy to select independent, unpredictable p and q.

In finite-field Diffie–Hellman, short and widely reused groups can also attract precomputation. The lesson of Logjam research and RFC 7919 concerns group size, structure, provenance, reuse, and implementation—not the mere fact that the prime is public. Separating public parameters from secrets and identifying the actual hard problem is more accurate than the slogan “a larger prime is safe.”

Bitcoin’s ECDSA and BIP 340 Schnorr signatures use the secp256k1 elliptic curve. The field prime p, curve equation, and generator G are public parameters. The secret is x and the public key is Q=xG. Security rests principally on the difficulty of the elliptic-curve discrete logarithm, not factoring an RSA modulus. A new large prime found by PrimeGrid does not automatically strengthen the secp256k1 parameters or existing keys.

Bitcoin mining is not a prime search either. Miners vary the nonce and other parts of a block header, trying inputs until the double-SHA-256 hash is below the network target. A PrimeGrid PRP test asks whether N is prime; mining asks whether the hash of an input meets a threshold. The large space of independent trials may look similar, but the mathematical problem, meaning of a result, validation, and reward are different.

Shor’s algorithm on a sufficiently large, fault-tolerant quantum computer offers polynomial-time routes to both integer factorization and discrete logarithms, threatening RSA, DH, and ECC. That is an algorithmic breakthrough, not “finding a pattern in primes.” No published device has recovered Bitcoin private keys in practical time, so research should not be turned into an invented deadline. Yet migration takes years, as NIST’s post-quantum standardization demonstrates, which is why early design work remains valuable.

Participation begins by installing the official BOINC client, attaching PrimeGrid, and selecting subprojects and resource limits. Check the official application list for operating-system, CPU or GPU, and driver support. Start conservatively with CPU share, schedules, battery, temperature, fans, and networking. Some tasks run for days or longer, so inspect deadlines and checkpoints, and never use a device without its owner’s permission.

Spare computation still consumes non-zero power and produces heat. Marginal impact changes with whether hardware was purchased for the work, whether its heat is useful, and the local electricity mix. Overclocking that creates errors can increase recomputation while looking fast. Treat valid results, electricity, hardware life, and noise as one operating system rather than optimizing only credit.

There is no guarantee of finding a prime. A composite result still advances the searched range, sieving saves later work, and a proof checks someone else’s result. PrimeGrid’s romance does not belong only to the lucky host. It lies in ordinary computers around the world filling a verifiable map one failed candidate at a time.