GPU Series
The generation and tier the card belongs to. It tells you which feature set you are buying into — the upscaling, frame generation and ray tracing support a card gets is decided by its architecture, not by its score.
Pick two cards to compare performance, VRAM, power and value. Below: RTX 5080 vs RTX 4090, our most-compared matchup.
Scores are modelled from published benchmark averages, not measured on your PC — see the methodology. Wondering which processor keeps either card fed? Run the pair through the bottleneck calculator.
| Spec | GeForce RTX 5080 | GeForce RTX 4090 |
|---|---|---|
| 3DMark Time Spy (graphics) | 32,264 | 35,519 |
| Estimated FPS · 1080p | 180 | 190 |
| Estimated FPS · 1440p | 149 | 160 |
| Estimated FPS · 4K | 99 | 108 |
| Video memory | 16 GB | 24 GB |
| Board power | 360 W | 450 W |
| Launch price (MSRP) | $999 | $1,599 |
| PCI Express | 5.0 ×16 | 4.0 ×16 |
| Release year | 2025 | 2022 |
| Series | NVIDIA RTX 50 Series | NVIDIA RTX 40 Series |
Verdict: The RTX 4090 is about 9% faster than the RTX 5080 at 4K and 7% faster at 1440p. the RTX 4090 is the faster card. The RTX 5080 is only the smarter buy if you can get it for at least 7% less than the RTX 4090; any smaller discount and you are paying more per frame.
This is the matchup people ask about most: last generation's flagship against this generation's second card. The 4090 still has the bigger chip and 24GB of memory, while the 5080 brings DLSS 4 Multi Frame Generation at a much lower launch price and noticeably lower power draw.
Read the full RTX 5080 vs RTX 4090 comparison, with per-game FPS estimates and power supply guidance.
The matchups people search for most, each with its own page: a full spec table, an FPS chart for every resolution, per-game estimates and a written verdict.
Every card in the comparison has two kinds of performance figure. The first is a 3DMark Time Spy graphics score. Time Spy is a DirectX 12 benchmark that both NVIDIA and AMD cards run the same way, which makes it a fair single number for lining up cards from different brands and generations. Our figure is modelled from published results, so it represents a typical card at stock settings rather than one particular overclocked sample.
The second is an estimated average frame rate at 1080p, 1440p and 4K. It is based on the average frame rates of a sample of recent AAA games at a high preset, with no ray tracing and no upscaling, normalised to a common test scene. Those figures are calibrated against published benchmark averages for a set of reference cards and then extended across the database from each card's architecture, memory and position in its range. The frame rates assume a processor fast enough not to hold the card back; if yours is not, the bottleneck calculator shows what you will actually get.
The winner badge in the tool is decided at 1440p, the resolution that separates two cards most cleanly: at 1080p the processor starts to flatten the differences, and at 4K most mid-range cards fall below playable frame rates. The full formula, the sources and what the model leaves out are on the methodology page.
The fastest current-generation card in each price band by its US launch price, with the best alternative from another brand. Street prices move every week, so treat the bands as a guide and compare the two cards you can actually buy.
| Budget (launch price) | Best pick | 1440p FPS | VRAM | Alternative |
|---|---|---|---|---|
| Under $260 | Intel Arc B580 | 69 | 12 GB | GeForce RTX 5050 |
| $260 – $360 | Radeon RX 9060 XT 16GB | 90 | 16 GB | GeForce RTX 5060 |
| $360 – $450 | GeForce RTX 5060 Ti 16GB | 91 | 16 GB | — |
| $450 – $650 | Radeon RX 9070 XT | 125 | 16 GB | GeForce RTX 5070 |
| $650 – $1,000 | GeForce RTX 5080 | 149 | 16 GB | — |
| Over $1,000 | GeForce RTX 5090 | 204 | 32 GB | — |
Below about $250, an 8GB card limits you to 1080p in recent games, so the extra memory on cards like the Arc B580 is worth having. From $450 upward, every pick has 16GB, which is enough for high textures at 1440p and 4K for the life of the card.
Each current generation from fastest to slowest, with the modelled 3DMark score and 1440p frame rate. Within a generation the tiers are evenly spaced; across generations, the same tier number usually moves up by about one step, which is why an RTX 5070 lands close to an RTX 4070 Super rather than a 4070.
| Card | 3DMark | 1440p FPS | 4K FPS | VRAM |
|---|---|---|---|---|
| GeForce RTX 5090 | 49,256 | 204 | 139 | 32 GB |
| GeForce RTX 5080 | 32,264 | 149 | 99 | 16 GB |
| GeForce RTX 5070 Ti | 25,559 | 125 | 79 | 16 GB |
| GeForce RTX 5070 | 20,485 | 106 | 65 | 12 GB |
| GeForce RTX 5060 Ti 16GB | 16,870 | 91 | 54 | 16 GB |
| GeForce RTX 5060 Ti 8GB | 16,090 | 88 | 52 | 8 GB |
| GeForce RTX 5060 | 13,436 | 77 | 44 | 8 GB |
| GeForce RTX 5050 | 9,179 | 58 | 32 | 8 GB |
| Card | 3DMark | 1440p FPS | 4K FPS | VRAM |
|---|---|---|---|---|
| Radeon RX 9070 XT | 25,559 | 125 | 79 | 16 GB |
| Radeon RX 9070 | 22,146 | 112 | 70 | 16 GB |
| Radeon RX 9060 XT 16GB | 16,479 | 90 | 53 | 16 GB |
| Radeon RX 9060 XT 8GB | 15,704 | 86 | 51 | 8 GB |
| Card | 3DMark | 1440p FPS | 4K FPS | VRAM |
|---|---|---|---|---|
| Intel Arc B580 | 11,614 | 69 | 39 | 12 GB |
| Intel Arc B570 | 9,179 | 58 | 32 | 10 GB |
| Card | 3DMark | 1440p FPS | 4K FPS | VRAM |
|---|---|---|---|---|
| GeForce RTX 4090 | 35,519 | 160 | 108 | 24 GB |
| GeForce RTX 4080 Super | 28,640 | 136 | 88 | 16 GB |
| GeForce RTX 4080 | 27,751 | 133 | 86 | 16 GB |
| GeForce RTX 4070 Ti Super | 24,265 | 120 | 76 | 16 GB |
| GeForce RTX 4070 Ti | 22,146 | 112 | 70 | 12 GB |
| GeForce RTX 4070 Super | 20,485 | 106 | 64 | 12 GB |
| GeForce RTX 4070 | 16,870 | 91 | 54 | 12 GB |
| GeForce RTX 4060 Ti 16GB | 14,182 | 80 | 46 | 16 GB |
| GeForce RTX 4060 Ti | 14,182 | 80 | 46 | 8 GB |
| GeForce RTX 4060 | 10,904 | 66 | 37 | 8 GB |
| Card | 3DMark | 1440p FPS | 4K FPS | VRAM |
|---|---|---|---|---|
| Radeon RX 7900 XTX | 28,195 | 134 | 86 | 24 GB |
| Radeon RX 7900 XT | 24,694 | 122 | 77 | 20 GB |
| Radeon RX 7900 GRE | 20,485 | 106 | 65 | 16 GB |
| Radeon RX 7800 XT | 19,059 | 100 | 60 | 16 GB |
| Radeon RX 7700 XT | 16,090 | 88 | 52 | 12 GB |
| Radeon RX 7600 XT | 10,553 | 64 | 36 | 16 GB |
| Radeon RX 7600 | 10,205 | 63 | 35 | 8 GB |
Older generations are on the graphics card list, grouped by series.
Five rows, five different questions. Only one of them is about raw speed, and the others decide whether that speed is any use to you.
The generation and tier the card belongs to. It tells you which feature set you are buying into — the upscaling, frame generation and ray tracing support a card gets is decided by its architecture, not by its score.
How old the design is. Newer usually means more performance per watt and a longer window of driver support, though a previous-generation flagship still beats plenty of newer mid-range cards.
The closest thing to a single answer on this page. It is a modelled Time Spy graphics score, so it lines up cards from either brand on the same scale and tracks real gaming performance closely.
The buffer for textures, shadow maps and frame data. It sets the settings and resolution a card can hold without stuttering — it is a capacity limit, not a speed rating.
The link between card and motherboard, shown as generation and lane count. Almost never a problem on a full ×16 slot, and worth a second look on cut-down ×8 cards in older boards.
The winner badge is a starting point, not a verdict. Work down these six checks and you will know whether the gap is worth paying for.
Look at the 3DMark row first and read the percentage under the winner. Under about 10% is close enough that price, cooling and availability should decide it. Past 25%, you are looking at a genuine tier difference you will feel in demanding games.
8GB still works at 1080p, but modern titles at 1440p with high textures are comfortable at 12GB, and 4K with ray tracing wants 16GB. If the faster card has noticeably less memory, weigh that against how long you plan to keep it.
Divide the price you would actually pay by the card’s 1440p frame rate. A card that is 10% faster but 30% more expensive is worse value, and the gap only earns its price if you have a monitor fast enough to show it.
A newer series brings newer upscaling and frame generation support, which can matter more day to day than a small raw performance gap. If the older card wins the score but loses the feature set, it is a closer call than the numbers suggest.
Board power decides the power supply you need and how much heat the card puts into your case. A 300W-plus card in a small case with a 650W supply is a problem no benchmark will show you.
A card can only stretch its legs if the CPU can feed it. Once you have a favourite, put it and your processor through the bottleneck calculator and see whether the upgrade actually lands.
Yes, modestly. In our model the RTX 5080 is about 10% faster at 1440p and 13% faster at 4K. Both have 16GB and both launched at $999, so the 5080 is the better buy at the same price, mainly for DLSS 4 Multi Frame Generation. If you already own a 4080 Super, the upgrade is not worth it.
For high-refresh 1440p, the RTX 5070 Ti (125 FPS in our model) and RX 9070 XT (125 FPS) are the sweet spot, with 16GB each. For a steady 60 to 100 FPS on a tighter budget, the RTX 5060 Ti 16GB and RX 9060 XT 16GB both average around 91 FPS. Avoid 8GB cards if 1440p is the goal.
No, but it sets a floor. Performance decides your frame rate; VRAM decides whether you can use high textures without stutter. Between two cards with enough memory for your resolution (8GB at 1080p, 12GB at 1440p, 16GB at 4K), buy the faster one. If the faster card falls below that line, the slower card with more memory is often the better long-term choice.
Yes. In our model the RTX 4090 is about 9% faster at 4K and 7% faster at 1440p, and it has 24GB of memory against 16GB. The RTX 5080 launched at $999 against $1,599, draws less power, and supports DLSS 4 Multi Frame Generation, which can more than close the gap in games that support it.
Neither across the board. At most price points AMD gives you more raw performance and more memory for the money, while NVIDIA leads in ray tracing and in how many games support its upscaling and frame generation. Compare the specific cards you can afford: the gap between two models usually matters more than the badge.
They are modelled estimates, calibrated against published benchmark averages, with an error margin of roughly ±15% against a real benchmark run. They are reliable for ranking two cards against each other and for sizing the gap between them, and less reliable for predicting an exact frame rate in one specific game. The methodology page explains the model.
Data last reviewed October 28, 2026. Launch prices are US MSRPs and board power is the reference design's rating. Comparing processors instead? Use the CPU vs CPU comparison.