Methodology

How every number is made

No black boxes. Every calculator on this site is built on one of ten published models, each with named inputs, a stated error band, and a dated changelog. This page is the method behind all of them.

Our principles

Every number on this site carries a stated error band and a dated source list. If a claim can't be sourced, it doesn't go in. If a model can't be reproduced, it doesn't ship.

  • Estimates, labelled as estimates. Where Rockstar has not published a number — PC specs, exact PC performance — the tool says so on the same screen as the estimate.
  • Error bands, not false precision. Frame rates are ranges, VRAM is a band, tiers are letters. A single-point FPS number is a claim; a band is a model.
  • Public sources. Every external number traces to a named source you can check yourself, listed per model below.
  • Reproduction before correction. A reported mismatch is re-run on the reference rig before a fix ships. The fix is then dated in the changelog.

Reference rig

The rig every model is validated against. Where a claim needs a different rig, the model page says so.

Reference rig components
ComponentSpec
CPUIntel Core i9-14900K (24 cores / 32 threads)
GPUNVIDIA GeForce RTX 4090, 24 GB
Memory64 GB DDR5-6400 CL32, dual channel
StorageSamsung 990 Pro 2 TB NVMe
Operating systemWindows 11 24H2

Ten models, one method

Every calculator and planner on the site is built on one of ten models below. Each lists its inputs, its output shape, its anchoring dataset, and its error band.

Bottleneck model

Calculates CPU vs GPU headroom as a function of draw-call throughput, thread saturation and raster limits; validated against GTA V and RDR2 at 4K and 1080p. Full detail in the model changelog.

  • Inputs: GPU tier, CPU tier, resolution, target frame rate
  • Output shape: Bottleneck score 0-100 and a single call — CPU-bound, GPU-bound, or balanced.
  • Anchor: GTA V and RDR2 benchmarks at 4K and 1080p
  • Error band: ±8 points on the score

VRAM model

Maps texture preset and resolution to allocated memory, plus a fixed headroom for ray tracing and upscalers. Calibrated against public VRAM captures on RDR2.

  • Inputs: GPU tier, texture quality, resolution, ray tracing state, upscaler
  • Output shape: Required VRAM in GB, plus the next tier up if within 1 GB.
  • Anchor: RDR2 VRAM allocation captures
  • Error band: ±1 GB

FPS estimator

Estimates average and 1% low frame rates per GPU tier and resolution, with a CPU scaling factor. Derived from RDR2, Cyberpunk 2077 and GTA V at equivalent presets.

  • Inputs: GPU tier, CPU tier, resolution, preset, target frame rate
  • Output shape: Average FPS and 1% low FPS as a range, not a point.
  • Anchor: RDR2, Cyberpunk 2077 and GTA V at equivalent presets
  • Error band: ±15% on the mean

Readiness score

Combines bottleneck, VRAM, storage and internet results into a single 0-100 score and letter grade. The weakest component dominates the score, so the result is actionable rather than cosmetic.

  • Inputs: Bottleneck score, VRAM margin, storage headroom, internet grade
  • Output shape: A 0-100 score, a letter grade F to S, and the named weakest link.
  • Anchor: Aggregates the bottleneck, VRAM, storage and internet models
  • Error band: Scores are diagnostic; no single band applies

GPU tier ranking

Ranks cards into S, A, B, C, D and F bands from measured raster performance, VRAM capacity and price band. Re-runs on every driver release with known GTA V or RDR2 impact.

  • Inputs: Raster performance, VRAM, memory bandwidth, launch price band
  • Output shape: Tier letter per card, with deltas per resolution.
  • Anchor: TechPowerUp and Hardware Unboxed datasets
  • Error band: ±1 tier

Network model

Grades latency, jitter, packet loss and throughput against a realistic target for your region and device. Latency and jitter dominate the grade because the launch-day risk is server load, not raw speed.

  • Inputs: Ping, jitter, packet loss, download speed
  • Output shape: A latency-led grade (Good / Edge / Poor) plus a preload-time estimate.
  • Anchor: Public GTA Online latency measurements
  • Error band: ±10 ms

Economy model

Ranks money methods by net $/hour using published payout tables and spawn rates. Each method lists its assumptions so a rank change can be checked against the source.

  • Inputs: Activity payout, spawn rate, cooldown, mission time
  • Output shape: Net $/hour per method, ranked, with named assumptions.
  • Anchor: Published GTA Online payout tables
  • Error band: Directional only; not a guarantee

Pricing model

Converts the announced US price to local currency using a bank-rate lookup. Tax treatment is explicit per market so the local price is not inflated by hidden tax.

  • Inputs: Base price, country, local tax rate, bank rate date
  • Output shape: Local price, USD equivalent, and the tax split.
  • Anchor: Bank rate at the stated date
  • Error band: Exact at the stated rate date

Storage model

Computes total footprint as install size plus patch working-set plus user-selected add-ons. The patch working-set term is the part other planners omit.

  • Inputs: Install size, patch size, DLC add-ons, console or PC
  • Output shape: Total GB required, plus headroom guidance.
  • Anchor: GTA V and RDR2 install and patch sizes
  • Error band: ±15 GB

Pre-order decision desk

Scores six buying channels on cancellability, price protection, preload timing, bonuses and resale value. Scoring weights are published, not hidden.

  • Inputs: Platform, region, retailer policy, edition
  • Output shape: A ranked recommendation per retailer, with the weights shown.
  • Anchor: Retailer policy pages
  • Error band: Ranking only; not a price forecast

Confidence: what each number is, and how far it can move

A number without an error band is a promise. Ours are not promises, so each one carries its band, its inputs, and the dataset it is anchored to.

Confidence bands for every model
ModelNatureBandDepends onAnchored to
Download timeEstimate±3%Network speed, patch sizeMeasured sustained throughput at the same time of day.
Storage footprintEstimate±15 GBInstall size, patch working-set, add-onsGTA V and RDR2 installs plus patch history.
VRAM requirementEstimate±1 GBTexture preset, resolution, RT, upscalerRDR2 VRAM allocation captures.
Bottleneck scoreEstimate±8 pointsGPU tier, CPU tier, resolutionGTA V and RDR2 benchmarks at 4K and 1080p.
GPU tier rankingEstimate±1 tierRaster performance, VRAM, price bandTechPowerUp and Hardware Unboxed datasets.
FPS estimateEstimate±15%GPU tier, CPU tier, resolution, presetRDR2, Cyberpunk 2077 and GTA V at equivalent presets.
Price by countryCalculatedExact at the stated rate dateBase price, country, local tax, rate dateBank rate at the stated date.
Internet gradeEstimate±10 msPing, jitter, packet loss, speedPublic GTA Online latency measurements.
Pre-order rankingRankingNot a numberRetailer policy, platform, regionRetailer policy pages.
Money method rankingEstimateDirectional onlyPayout, spawn rate, cooldown, mission timePublished GTA Online payout tables.

What we do not claim

  • We do not claim to know GTA 6's PC system requirements. Rockstar has not published them.
  • We do not claim a single-point FPS figure for any card. Frame rates are ranges with a stated band.
  • We do not claim price forecasts. Price pages show current conversion at a stated rate date.
  • We do not claim insider access. Unverified material is discussed and labelled, never modelled.
  • We do not claim a tier is permanent. Tiers re-run on every driver cycle with a known impact.
  • We do not claim a recommendation is a purchase order. The decision desk ranks channels; you buy where you choose.

How we verify every claim

A short checklist, applied before anything ships.

  • Every external number names its source in the model card above.
  • Every model has a reviewer sign-off from an independent benchmark reviewer before it ships.
  • Every claim is reproduced on the reference rig before a correction is published.
  • Every model re-runs on the trigger list below, and the change is dated in the tool's changelog.
  • Every published change keeps the old value visible, so a move can be checked.
  • Every page that shows an estimate labels it as an estimate on the same screen.

Retest triggers

The changelog events that force a re-run of the affected models.

  • 2026-01-15 — Every GPU vendor driver release with a known GTA V or RDR2 impact triggers a tier re-run.
  • 2026-03-02 — Rockstar's March system update changed the console target; all tiers were re-anchored.
  • 2026-05-20 — The May engine patch changed draw-call throughput assumptions; bottleneck deltas re-checked.
  • 2026-11-19 — Full re-validation against real GTA 6 performance when the game ships.

Correction process

  1. A reader reports a mismatch to corrections@toolwools.com with the tool, exact inputs, the value we gave, the value they measured, and how they measured it.
  2. We reproduce the case on the reference rig. Reproduction is the gate.
  3. We ship a dated fix in the tool's changelog, keeping the old value visible.
  4. We credit the reporter in the changelog entry unless they ask us not to.

Baseline sources for the anchoring datasets: Digital Foundry, TechPowerUp, Hardware Unboxed, and the Rockstar GTA VI page.

Launch reference: GTA 6 launches November 19, 2026 at midnight Eastern. All launch-relative dates on this site use that instant.

Last verified .

Methodology, answered

Error bands, retesting, the reference rig and corrections in short form.

  • Why do some numbers have error bands?

    Because they are estimates, not measurements of GTA 6. We estimate against the announced console targets and measured behaviour in comparable RAGE-engine titles. The band tells you how far the real value could land from the estimate, and it is computed from the variance in the anchoring datasets, not picked.

  • What do you do when Rockstar announces a change?

    Every model re-runs on a fixed trigger list: new driver releases with a known impact, Rockstar system updates, and launch day itself. The change is dated in the tool's changelog and the old value is kept visible, so you can see what moved and when.

  • What is the reference rig?

    The rig we validate every model against is listed in full on this page: a Core i9-14900K, RTX 4090, 64 GB DDR5-6400 and a Samsung 990 Pro NVMe drive, on Windows 11 24H2. Where a model's claim needs a different rig, the model page says so.

  • How often do you re-test?

    On every driver release with a known GTA V or RDR2 impact, on every Rockstar system update, and on launch day. Between triggers, a model is considered stable unless a correction report reproduces a mismatch.

  • How do I report a mismatch with my own rig?

    Use the corrections address on the About page: tool, exact inputs, the value we gave, the value you measured, and how you measured it. Reproduction is the gate: we re-run the case on the reference rig, then ship a dated fix in the tool's changelog.

  • Do you model leaked specs?

    No. Leaked material may be discussed and labelled as unconfirmed, but it never changes a calculation. Only Rockstar announcements and measured behaviour in released games move a model.