GeForce RTX 4080 vs GeForce RTX 5090: specs and benchmark comparison
ANVIDIA · Ada Lovelace
NVIDIA GeForce RTX 4080
- 16 GB GDDR6X
- 320 W
- PSU 750 W+
November 16, 2022 · $1,199 at launch
BNVIDIA · Blackwell
NVIDIA GeForce RTX 5090
- 32 GB GDDR7
- 575 W
- PSU 1000 W+
January 30, 2025 · $1,999 at launch
Short verdict
NVIDIA GeForce RTX 5090 leads in every scenario with comparable benchmarks: rendering and compute. This does not imply an advantage in tasks without benchmarks.
Compare with AIRendering and compute
B +95.4%NVIDIA GeForce RTX 5090 leads in this scenario (up to 95.4% in comparable benchmarks).
No comparable benchmarks yet: gaming: rasterization, gaming: ray tracing, power efficiency.
Advantages
Only measured and rated values: a noticeable benchmark gap and specs where more is better.
Where GeForce RTX 4080 is better2
Board power 255 W lower320 W (TGP) vs 575 W (TGP)
Launch price $800 lower$1,199 vs $1,999
Where GeForce RTX 5090 is better3
Blender Benchmark 4.5.0 78.1% faster14,978 samples/min vs 8,409 samples/min
Video memory 16 GB more32 GB vs 16 GB
Memory bus 256 bit more512 bit vs 256 bit
Benchmarks
Results are compared within one benchmark, one version and the same conditions.
- B ahead by 78.1%
Blender Benchmark 4.5.0
GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) · samples/min, higher is better
GeForce RTX 40808,409.3 samples/minGeForce RTX 509014,978.4 samples/minPosition among tested models (67, OptiX)
A · 8,409B · 14,97805,00010,00015,000Each dot is one model’s median in this benchmark. The scale starts at zero, samples/min. NVIDIA GeForce RTX 5090 is ahead by 78.1% in this benchmark.
About the measurement and source
- A · NVIDIA GeForce RTX 4080
8,409.3 samples/min · community measurement, sample 199 · November 14, 2026
Blender Open Data (Blender Foundation)
Median of 199 submissions (20 outliers excluded), interquartile range 8230.5–8528.15; submissions 2025-07-18 to 2026-08-19; OPTIX backend. Snapshot 2026-11-14.
- B · NVIDIA GeForce RTX 5090
14,978.4 samples/min · community measurement, sample 1334 · November 14, 2026
Blender Open Data (Blender Foundation)
Median of 1334 submissions (221 outliers excluded), interquartile range 14677.66–15255.56; submissions 2025-07-17 to 2026-10-15; OPTIX backend. Snapshot 2026-11-14.
difference = |A − B| ÷ the smaller value × 100%. The 5% noticeable threshold is a UX heuristic, not statistical significance.
Cycles render in Blender Benchmark. Median of the sum of samples per minute over three scenes after removing outliers.
Other comparable benchmarks (5)
| Benchmark | A | B | Difference |
|---|---|---|---|
| Blender Benchmark 5.1.1GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 8,486.5 | 16,330.8 | B +92.4% |
| Blender Benchmark 5.2.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 8,284.8 | 16,189.8 | B +95.4% |
| Blender Benchmark 4.3.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 8,293 | 14,904.4 | B +79.7% |
| Blender Benchmark 4.4.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 8,456.2 | 15,097.9 | B +78.5% |
| Blender Benchmark 5.1.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 8,492.3 | 16,312.9 | B +92.1% |
Each benchmark version has its own scale: numbers are compared only within a row.
Reference results (14)
| Benchmark | A | B | Difference |
|---|---|---|---|
| Blender Benchmark 3.6.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 6,156.7 | — | no pair |
| Blender Benchmark 4.0.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 5,359.2 | — | no pair |
| Blender Benchmark 4.3.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 5,208.7 | — | no pair |
| Blender Benchmark 4.2.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 8,287 | 14,263.3 | reference |
| Blender Benchmark 3.2.1GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 8,005.5 | — | no pair |
| Blender Benchmark 3.3.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 9,238.6 | — | no pair |
| Blender Benchmark 3.4.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 9,347.7 | — | no pair |
| Blender Benchmark 3.5.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 9,732.9 | — | no pair |
| Blender Benchmark 3.6.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 9,661.3 | 17,822.2 | reference |
| Blender Benchmark 4.0.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 8,618.1 | — | no pair |
| Blender Benchmark 4.1.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 8,621.9 | — | no pair |
| Blender Benchmark 5.0.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 7,946.5 | 14,411.8 | reference |
| Blender Benchmark 5.3.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 10,165.2 | 19,417.5 | reference |
| Blender Benchmark 4.4.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 10,016.4 | no pair |
Each benchmark version has its own scale: numbers are compared only within a row.
Specifications
| Specification | A · GeForce RTX 4080 | B · GeForce RTX 5090 | Difference |
|---|---|---|---|
| Chip | |||
| Architecture | Ada Lovelace | Blackwell | differs |
| Shader unitsShader unit counts are comparable only within one architecture. | 9,728 (CUDA cores) | 21,760 (CUDA cores) | not directly comparable |
| Clocks | |||
| Base clock | 2,210 MHz | 2,010 MHz | A: 200 MHz more |
| Boost clockClocks of different architectures do not reflect performance directly. For laptop chips, the clock depends on the laptop’s power limit. | 2,510 MHz | 2,410 MHz | A: 100 MHz more |
| Memory | |||
| Video memory | 16 GB | 32 GB | B: 16 GB more |
| Memory type | GDDR6X | GDDR7 | differs |
| Memory bus | 256 bit | 512 bit | B: 256 bit more |
| Power and dimensions | |||
| Board powerTGP (NVIDIA) and TBP (AMD/Intel) are calculated differently; higher is not better. For laptops, the chip maker sets the range and the laptop maker picks the value, so the same chip behaves differently in different laptops. | 320 W (TGP) | 575 W (TGP) | B: 255 W higher |
| Recommended PSU | 750 W | 1,000 W | B: 250 W higher |
| Interface | no data | PCIe 5.0 | no data for A |
| Length (reference / Founders Edition)Partner cards have different dimensions; check the specific SKU. | 304 mm | 304 mm | equal |
| Thickness (slots) | 3 | 2 | A: 1 higher |
| Power connectors (reference) | 3x 8-pin (adapter) or 450+ W PCIe Gen 5 — FE | 4x 8-pin (adapter) or 1x 600 W PCIe Gen 5 — FE | differs |
Chip
Architecture
AAda Lovelace
BBlackwell
differsShader units
A9,728 (CUDA cores)
B21,760 (CUDA cores)
not directly comparableShader unit counts are comparable only within one architecture.
Clocks
Base clock
A2,210 MHz
B2,010 MHz
A: 200 MHz moreBoost clock
A2,510 MHz
B2,410 MHz
A: 100 MHz moreClocks of different architectures do not reflect performance directly. For laptop chips, the clock depends on the laptop’s power limit.
Memory
Video memory
A16 GB
B32 GB
B: 16 GB moreMemory type
AGDDR6X
BGDDR7
differsMemory bus
A256 bit
B512 bit
B: 256 bit more
Power and dimensions
Board power
A320 W (TGP)
B575 W (TGP)
B: 255 W higherTGP (NVIDIA) and TBP (AMD/Intel) are calculated differently; higher is not better. For laptops, the chip maker sets the range and the laptop maker picks the value, so the same chip behaves differently in different laptops.
Recommended PSU
A750 W
B1,000 W
B: 250 W higherInterface
Ano data
BPCIe 5.0
no data for ALength (reference / Founders Edition)
A304 mm
B304 mm
equalPartner cards have different dimensions; check the specific SKU.
Thickness (slots)
A3
B2
A: 1 higherPower connectors (reference)
A3x 8-pin (adapter) or 450+ W PCIe Gen 5 — FE
B4x 8-pin (adapter) or 1x 600 W PCIe Gen 5 — FE
differs
“More” does not always mean “better”: higher power draw or size is a system requirement, not an advantage (such bars are dimmed).
Compare with AI
An AI model reviews the specs and benchmarks of this pair and gives a plain-language verdict on who each model suits.
Questions and answers
How much video memory do NVIDIA GeForce RTX 4080 and NVIDIA GeForce RTX 5090 have?
NVIDIA GeForce RTX 4080: 16 GB GDDR6X, 256-bit bus. NVIDIA GeForce RTX 5090: 32 GB GDDR7, 512-bit bus.
What power supply does the manufacturer recommend?
For NVIDIA GeForce RTX 4080, 750 W or more; for NVIDIA GeForce RTX 5090, 1000 W or more. This is a recommendation for a typical system; also check the power connectors of the specific card model.
Which is faster: NVIDIA GeForce RTX 4080 or NVIDIA GeForce RTX 5090?
NVIDIA GeForce RTX 5090 leads in every scenario with comparable benchmarks: rendering and compute. This does not imply an advantage in tasks without benchmarks.
See also
Data sources (3)
- NVIDIA — specifications of NVIDIA GeForce RTX 4080, specifications of NVIDIA GeForce RTX 5090checked November 7, 2026
- Blender Open Data (Blender Foundation) — benchmark Blender Benchmark 4.4.0, benchmark Blender Benchmark 3.6.0, benchmark Blender Benchmark 4.0.0, benchmark Blender Benchmark 4.3.0, benchmark Blender Benchmark 4.2.0, benchmark Blender Benchmark 4.5.0, benchmark Blender Benchmark 5.1.1, benchmark Blender Benchmark 5.2.0, benchmark Blender Benchmark 5.0.0, benchmark Blender Benchmark 5.1.0, benchmark Blender Benchmark 5.3.0, benchmark Blender Benchmark 3.2.1, benchmark Blender Benchmark 3.3.0, benchmark Blender Benchmark 3.4.0, benchmark Blender Benchmark 3.5.0, benchmark Blender Benchmark 4.1.0checked November 14, 2026
- Wikipedia (English) — release date and launch pricechecked November 14, 2026
How we compare: methodology. Where the data comes from: sources.