GeForce RTX 3090 vs GeForce RTX 4070 SUPER: specs and benchmark comparison
ANVIDIA · Ampere
NVIDIA GeForce RTX 3090
- 24 GB GDDR6X
- 350 W
- PSU 750 W+
September 24, 2020 · $1,499 at launch
BNVIDIA · Ada Lovelace
NVIDIA GeForce RTX 4070 SUPER
- 12 GB GDDR6X
- 220 W
- PSU 650 W+
January 17, 2024 · $599 at launch
Short verdict
NVIDIA GeForce RTX 4070 SUPER 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 +18.2%NVIDIA GeForce RTX 4070 SUPER leads in this scenario (up to 18.2% 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 3090 is better2
Video memory 12 GB more24 GB vs 12 GB
Memory bus 192 bit more384 bit vs 192 bit
Where GeForce RTX 4070 SUPER is better2
Blender Benchmark 4.5.0 15.6% faster6,268 samples/min vs 5,424 samples/min
Launch price $900 lower$599 vs $1,499
Benchmarks
Results are compared within one benchmark, one version and the same conditions.
- B ahead by 15.6%
Blender Benchmark 4.5.0
GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) · samples/min, higher is better
GeForce RTX 30905,423.6 samples/minGeForce RTX 4070 SUPER6,267.9 samples/minPosition among tested models (67, OptiX)
A · 5,424B · 6,26805,00010,00015,000Each dot is one model’s median in this benchmark. The scale starts at zero, samples/min. NVIDIA GeForce RTX 4070 SUPER is ahead by 15.6% in this benchmark.
About the measurement and source
- A · NVIDIA GeForce RTX 3090
5,423.6 samples/min · community measurement, sample 596 · November 14, 2026
Blender Open Data (Blender Foundation)
Median of 596 submissions (34 outliers excluded), interquartile range 5240.15–5558.47; submissions 2025-07-17 to 2026-10-19; OPTIX backend. Snapshot 2026-11-14.
- B · NVIDIA GeForce RTX 4070 SUPER
6,267.9 samples/min · community measurement, sample 448 · November 14, 2026
Blender Open Data (Blender Foundation)
Median of 448 submissions (21 outliers excluded), interquartile range 6009.46–6439.99; submissions 2025-07-17 to 2026-11-11; 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 (9)
| Benchmark | A | B | Difference |
|---|---|---|---|
| Blender Benchmark 4.2.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,354.8 | 6,067.5 | B +13.3% |
| Blender Benchmark 5.1.1GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,253.3 | 6,209.5 | B +18.2% |
| Blender Benchmark 5.2.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,093.8 | 5,970.4 | B +17.2% |
| Blender Benchmark 3.6.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 6,305.5 | 6,923.4 | B +9.8% |
| Blender Benchmark 4.0.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,536.1 | 6,370.1 | B +15.1% |
| Blender Benchmark 4.1.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,536.4 | 6,352.9 | B +14.7% |
| Blender Benchmark 4.3.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,347.6 | 6,078.3 | B +13.7% |
| Blender Benchmark 4.4.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,333 | 6,169.7 | B +15.7% |
| Blender Benchmark 5.1.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,266.1 | 6,113.7 | B +16.1% |
Each benchmark version has its own scale: numbers are compared only within a row.
Reference results (84)
| Benchmark | A | B | Difference |
|---|---|---|---|
| Blender Benchmark 2.90GPU, bmw27 (CUDA) | 21.5 | — | no pair |
| Blender Benchmark 2.90.1GPU, bmw27 (CUDA) | 20.5 | — | no pair |
| Blender Benchmark 2.91.2GPU, bmw27 (CUDA) | 18.5 | — | no pair |
| Blender Benchmark 2.92GPU, bmw27 (CUDA) | 18.5 | — | no pair |
| Blender Benchmark 2.93.1GPU, bmw27 (CUDA) | 16.9 | — | no pair |
| Blender Benchmark 2.90GPU, bmw27 (OPTIX) | 8.9 | — | no pair |
| Blender Benchmark 2.90.1GPU, bmw27 (OPTIX) | 8.9 | — | no pair |
| Blender Benchmark 2.91.2GPU, bmw27 (OPTIX) | 9.2 | — | no pair |
| Blender Benchmark 2.92GPU, bmw27 (OPTIX) | 9.1 | — | no pair |
| Blender Benchmark 2.93.1GPU, bmw27 (OPTIX) | 9.3 | — | no pair |
| Blender Benchmark 2.90GPU, classroom (CUDA) | 70.6 | — | no pair |
| Blender Benchmark 2.90.1GPU, classroom (CUDA) | 67.4 | — | no pair |
| Blender Benchmark 2.91.2GPU, classroom (CUDA) | 58.8 | — | no pair |
| Blender Benchmark 2.92GPU, classroom (CUDA) | 55.5 | — | no pair |
| Blender Benchmark 2.93.1GPU, classroom (CUDA) | 79.5 | — | no pair |
| Blender Benchmark 2.90GPU, classroom (OPTIX) | 35.2 | — | no pair |
| Blender Benchmark 2.90.1GPU, classroom (OPTIX) | 38.9 | — | no pair |
| Blender Benchmark 2.91.2GPU, classroom (OPTIX) | 37.1 | — | no pair |
| Blender Benchmark 2.92GPU, classroom (OPTIX) | 34.4 | — | no pair |
| Blender Benchmark 2.93.1GPU, classroom (OPTIX) | 33.5 | — | no pair |
| Blender Benchmark 2.90GPU, fishy_cat (CUDA) | 39 | — | no pair |
| Blender Benchmark 2.90.1GPU, fishy_cat (CUDA) | 37.3 | — | no pair |
| Blender Benchmark 2.91.2GPU, fishy_cat (CUDA) | 33.9 | — | no pair |
| Blender Benchmark 2.92GPU, fishy_cat (CUDA) | 34.8 | — | no pair |
| Blender Benchmark 2.93.1GPU, fishy_cat (CUDA) | 34.4 | — | no pair |
| Blender Benchmark 2.90GPU, fishy_cat (OPTIX) | 16.5 | — | no pair |
| Blender Benchmark 2.90.1GPU, fishy_cat (OPTIX) | 16.8 | — | no pair |
| Blender Benchmark 2.91.2GPU, fishy_cat (OPTIX) | 16.7 | — | no pair |
| Blender Benchmark 2.92GPU, fishy_cat (OPTIX) | 18.4 | — | no pair |
| Blender Benchmark 2.93.1GPU, fishy_cat (OPTIX) | 17.9 | — | no pair |
| Blender Benchmark 2.90GPU, koro (CUDA) | 58 | — | no pair |
| Blender Benchmark 2.90.1GPU, koro (CUDA) | 57.5 | — | no pair |
| Blender Benchmark 2.91.2GPU, koro (CUDA) | 57.5 | — | no pair |
| Blender Benchmark 2.92GPU, koro (CUDA) | 58 | — | no pair |
| Blender Benchmark 2.93.1GPU, koro (CUDA) | 54.6 | — | no pair |
| Blender Benchmark 2.90GPU, koro (OPTIX) | 36.5 | — | no pair |
| Blender Benchmark 2.90.1GPU, koro (OPTIX) | 37.1 | — | no pair |
| Blender Benchmark 2.91.2GPU, koro (OPTIX) | 36.3 | — | no pair |
| Blender Benchmark 2.92GPU, koro (OPTIX) | 57.1 | — | no pair |
| Blender Benchmark 2.93.1GPU, koro (OPTIX) | 37.9 | — | no pair |
| Blender Benchmark 2.90GPU, pavillon_barcelona (CUDA) | 119.6 | — | no pair |
| Blender Benchmark 2.90.1GPU, pavillon_barcelona (CUDA) | 108.8 | — | no pair |
| Blender Benchmark 2.91.2GPU, pavillon_barcelona (CUDA) | 108.5 | — | no pair |
| Blender Benchmark 2.92GPU, pavillon_barcelona (CUDA) | 109.3 | — | no pair |
| Blender Benchmark 2.93.1GPU, pavillon_barcelona (CUDA) | 99.3 | — | no pair |
| Blender Benchmark 2.90GPU, pavillon_barcelona (OPTIX) | 45.4 | — | no pair |
| Blender Benchmark 2.90.1GPU, pavillon_barcelona (OPTIX) | 46.6 | — | no pair |
| Blender Benchmark 2.91.2GPU, pavillon_barcelona (OPTIX) | 46.9 | — | no pair |
| Blender Benchmark 2.92GPU, pavillon_barcelona (OPTIX) | 45.1 | — | no pair |
| Blender Benchmark 2.93.1GPU, pavillon_barcelona (OPTIX) | 45.1 | — | no pair |
| Blender Benchmark 2.90GPU, victor (CUDA) | 159.1 | — | no pair |
| Blender Benchmark 2.90.1GPU, victor (CUDA) | 148.7 | — | no pair |
| Blender Benchmark 2.91.2GPU, victor (CUDA) | 124.3 | — | no pair |
| Blender Benchmark 2.92GPU, victor (CUDA) | 178.3 | — | no pair |
| Blender Benchmark 2.93.1GPU, victor (CUDA) | 138.7 | — | no pair |
| Blender Benchmark 2.90GPU, victor (OPTIX) | 52.8 | — | no pair |
| Blender Benchmark 2.90.1GPU, victor (OPTIX) | 57.7 | — | no pair |
| Blender Benchmark 2.91.2GPU, victor (OPTIX) | 53.9 | — | no pair |
| Blender Benchmark 2.92GPU, victor (OPTIX) | 62.5 | — | no pair |
| Blender Benchmark 2.93.1GPU, victor (OPTIX) | 60.4 | — | no pair |
| Blender Benchmark 3.0.1GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,696.5 | — | no pair |
| Blender Benchmark 3.1.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,418.3 | — | no pair |
| Blender Benchmark 3.2.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,138.2 | — | no pair |
| Blender Benchmark 3.2.1GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,448.2 | — | no pair |
| Blender Benchmark 3.3.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,729.7 | — | no pair |
| Blender Benchmark 3.4.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,797.9 | — | no pair |
| Blender Benchmark 3.5.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,864 | — | no pair |
| Blender Benchmark 3.6.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 4,060.8 | — | no pair |
| Blender Benchmark 4.0.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,724.8 | — | no pair |
| Blender Benchmark 4.1.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,735.6 | 4,070.7 | reference |
| Blender Benchmark 4.2.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,699.3 | 3,969.7 | reference |
| Blender Benchmark 4.3.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,730.8 | 3,989.7 | reference |
| Blender Benchmark 4.5.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,798 | 4,088 | reference |
| Blender Benchmark 3.0.1GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,841.8 | — | no pair |
| Blender Benchmark 3.1.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 6,560.4 | — | no pair |
| Blender Benchmark 3.1.2GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,826.5 | — | no pair |
| Blender Benchmark 3.2.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 6,566.2 | — | no pair |
| Blender Benchmark 3.2.1GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,625.7 | 6,140.3 | reference |
| Blender Benchmark 3.3.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 6,018.2 | 6,920.2 | reference |
| Blender Benchmark 3.4.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 6,014.6 | 6,990.3 | reference |
| Blender Benchmark 3.5.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 6,296.9 | 6,903.7 | reference |
| Blender Benchmark 5.0.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 4,875.8 | 5,931.9 | reference |
| Blender Benchmark 5.3.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,143.8 | 7,092.6 | reference |
| Blender Benchmark 4.4.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 4,090.2 | no pair |
Each benchmark version has its own scale: numbers are compared only within a row.
Specifications
| Specification | A · GeForce RTX 3090 | B · GeForce RTX 4070 SUPER | Difference |
|---|---|---|---|
| Chip | |||
| Architecture | Ampere | Ada Lovelace | differs |
| Shader unitsShader unit counts are comparable only within one architecture. | 10,496 (CUDA cores) | 7,168 (CUDA cores) | not directly comparable |
| Clocks | |||
| Base clock | 1,400 MHz | 1,980 MHz | B: 580 MHz more |
| Boost clockClocks of different architectures do not reflect performance directly. For laptop chips, the clock depends on the laptop’s power limit. | 1,700 MHz | 2,480 MHz | B: 780 MHz more |
| Memory | |||
| Video memory | 24 GB | 12 GB | A: 12 GB more |
| Memory type | GDDR6X | GDDR6X | equal |
| Memory bus | 384 bit | 192 bit | A: 192 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. | 350 W (board power) | 220 W (TGP) | not directly comparable |
| Recommended PSU | 750 W | 650 W | A: 100 W higher |
| Interface | no data | PCIe 4.0 | no data for A |
| Length (reference / Founders Edition)Partner cards have different dimensions; check the specific SKU. | 313 mm | 244 mm | A: 69 mm higher |
| Thickness (slots) | 3 | 2 | A: 1 higher |
| Power connectors (reference) | 2x 8-pin (adapter to 1x 12-pin included) — FE | 1x 16-pin (adapter to 2x 8-pin included) — Founders Edition | differs |
Chip
Architecture
AAmpere
BAda Lovelace
differsShader units
A10,496 (CUDA cores)
B7,168 (CUDA cores)
not directly comparableShader unit counts are comparable only within one architecture.
Clocks
Base clock
A1,400 MHz
B1,980 MHz
B: 580 MHz moreBoost clock
A1,700 MHz
B2,480 MHz
B: 780 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
A24 GB
B12 GB
A: 12 GB moreMemory type
AGDDR6X
BGDDR6X
equalMemory bus
A384 bit
B192 bit
A: 192 bit more
Power and dimensions
Board power
A350 W (board power)
B220 W (TGP)
not directly comparableTGP (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
B650 W
A: 100 W higherInterface
Ano data
BPCIe 4.0
no data for ALength (reference / Founders Edition)
A313 mm
B244 mm
A: 69 mm higherPartner cards have different dimensions; check the specific SKU.
Thickness (slots)
A3
B2
A: 1 higherPower connectors (reference)
A2x 8-pin (adapter to 1x 12-pin included) — FE
B1x 16-pin (adapter to 2x 8-pin included) — Founders Edition
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 3090 and NVIDIA GeForce RTX 4070 SUPER have?
NVIDIA GeForce RTX 3090: 24 GB GDDR6X, 384-bit bus. NVIDIA GeForce RTX 4070 SUPER: 12 GB GDDR6X, 192-bit bus.
What power supply does the manufacturer recommend?
For NVIDIA GeForce RTX 3090, 750 W or more; for NVIDIA GeForce RTX 4070 SUPER, 650 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 3090 or NVIDIA GeForce RTX 4070 SUPER?
NVIDIA GeForce RTX 4070 SUPER 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 3090, specifications of NVIDIA GeForce RTX 4070 SUPERchecked November 7, 2026
- Blender Open Data (Blender Foundation) — benchmark Blender Benchmark 2.90, benchmark Blender Benchmark 2.90.1, benchmark Blender Benchmark 2.91.2, benchmark Blender Benchmark 2.92, benchmark Blender Benchmark 2.93.1, benchmark Blender Benchmark 4.1.0, benchmark Blender Benchmark 4.2.0, benchmark Blender Benchmark 4.3.0, benchmark Blender Benchmark 4.4.0, benchmark Blender Benchmark 4.5.0, benchmark Blender Benchmark 3.0.1, benchmark Blender Benchmark 3.1.0, benchmark Blender Benchmark 3.2.0, 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 3.6.0, benchmark Blender Benchmark 4.0.0, 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 3.1.2checked November 14, 2026
- Wikipedia (English) — release date and launch pricechecked November 14, 2026
How we compare: methodology. Where the data comes from: sources.