GPU model hub
GeForce RTX 4070 Ti
Specs and their source, the 6 listings on this site with a price range, bottleneck pairings and common problems for this GPU.
"Add to a BTO build" starts a new build with this model and opens it in the BTO configurator.
Specifications
| GPU chip | AD104 |
|---|---|
| Memory | 12 GB · GDDR6X · 192-bit |
| Shaders | 7,680 |
| Clock (base / boost) | – / 2,610 MHz |
| Board power (TDP) | 285 W |
| Suggested PSU (TechPowerUp) | 600 W |
| Bus | PCIe 4.0 x16 |
| Released | Jan 5th, 2023 |
| PassMark G3D Mark | 31,509 (PassMark public score, fetched 2026/10/10) |
Source: TechPowerUp product page · fetched 2026/10/10
Listings for this model (6)
Price range¥102,364 〜 ¥289,8595 store prices, fetched 2026/5/27–2026/9/2
1 listing(s) with a price under review are left out of the range.
- MSI GeForce RTX 4070 Ti GAMING X SLIM 12G VD8659 — High-performance slim gaming graphics card¥102,364as of 2026/7/19
- ZOTAC GAMING GeForce RTX 4070 Ti Trinity — High-performance gaming graphics card with DLSS 3¥192,089as of 2026/5/27
- Palit RTX 4070 Ti White OC 12GB Gaming¥248,800as of 2026/5/29
- ASUS TUF Gaming NVIDIA GeForce RTX™ 4070 Ti OC Edition — Durable enthusiast-grade gaming GPU¥260,517as of 2026/9/2
- Gigabyte GeForce RTX 4070 Ti AERO OC V2 12G — White-themed GPU with 12GB GDDR6X memory¥289,859as of 2026/9/1
- ASUS ProArt GeForce RTX 4070 Ti OC Edition 12GB GDDR6X — High-performance graphics card for creative workChecking price
Prices are as fetched for each product page, not live. Check the store buttons on each page for the current price.
Bottleneck guide
Same scores and verdicts as the compare tool's bottleneck analysis. Partner CPUs are taken from this site's model hubs at the lower, middle and top of the range. This is not an FPS estimate.
| CPU | 1080p | 1440p | 4K | Compare tool |
|---|---|---|---|---|
| EntryCore i5-12500 | CPU-Constrained (GPU-heavy build) | GPU Has Headroom (better at higher res) | GPU-Limited (Normal Gaming) | Open in compare |
| Mid-rangeRyzen 5 8400F | CPU-Constrained (GPU-heavy build) | GPU Has Headroom (better at higher res) | GPU-Limited (Normal Gaming) | Open in compare |
| High-endCore Ultra 9 285K | GPU Has Headroom (better at higher res) | GPU-Limited (Normal Gaming) | GPU-Limited (Normal Gaming) | Open in compare |
How this verdict is calculated
The formula
ratio = (CPU score × 0.45 × resolution factor) ÷ GPU score
0.45 is a calibration constant that puts the CPU and GPU scores on the same scale. The resolution factor is the actual pixel ratio: 1080p = 1, 1440p = 1.78, 4K = 4.
This combination, step by step
| CPU score | 61.5 / 100（Ryzen 5 8400F） |
|---|---|
| GPU score | 58.5 / 100（GeForce RTX 4070 Ti） |
| Resolution factor | 1.78（1440p） |
| ratio | 61.5 × 0.45 × 1.78 ÷ 58.5 = 0.84 |
Reading the ratio
| ratio | Verdict |
|---|---|
< 0.55 | CPU-constrained (GPU-heavy build) |
0.55 – 0.85 | GPU has headroom (better at higher resolution) |
0.85 – 1.15 | Balanced |
1.15 – 3.5 | GPU-limited (normal for gaming) |
> 3.5 | GPU significantly underpowered |
"GPU-limited" is not a fault. In games the GPU hitting its limit first is normal, and generally what you want.
Where the scores come from
- GPU score: compute (shaders × clock) blended with PassMark (G3D Mark), normalised so the fastest gaming GPU is 100. As of 2026-09-01 that reference is the GeForce RTX 5090
- The reference pool excludes datacenter and professional cards (Instinct, Tesla, RTX PRO, Rubin, …) and anything above 32GB VRAM. Including them flattens every consumer GPU — an RTX 5070 was scoring 4.0/100 before this was fixed
- Only GPUs with an actual PassMark result are used as the reference. Rumoured or unreleased models (an RTX 6090, say) have no measurement and cannot set the scale
- CPU score: clock × generation (inferred from process node) × performance-core count (capped at 8) × 3D V-Cache adjustment, normalised the same way
- PassMark single-thread figures are deliberately not used for the gaming score: a parsing issue in the source table over-rates older CPUs (e.g. an i7-7700K scoring above the plausible ceiling)
What this cannot tell you
- It cannot give you actual frame rates. Nothing here is measured in-game; it is an estimate built from specifications
- It does not account for individual games. CPU-heavy and GPU-heavy titles behave differently
- Memory speed and capacity, storage, power supply and cooling are not part of this verdict
- VRAM is reported only as "likely enough or not" for the chosen resolution. How much performance is lost when it runs short depends on the title and settings, so no figure is given (rough guides: 12GB at 4K, 8GB at 1440p, 6GB at 1080p)
- The generation term in the CPU score is inferred from the process node, not measured IPC
Common problems
Symptoms commonly reported with GPUs and how to narrow them down (not a list of faults specific to this model).
- Visual artifacts on screenTroubleshooting (Graphics Card)
- No display outputTroubleshooting (Graphics Card)
- Screen goes black in games, then recoversTroubleshooting (Graphics Card)
- Lower frame rates than expectedTroubleshooting (Graphics Card)
- Sudden shutdown under heavy loadTroubleshooting (Power Supply)