Skip to content
RU

NVIDIA · Ampere

NVIDIA GeForce RTX 3090: specs, benchmarks and comparison

  • 24 GB GDDR6X
  • 350 W
  • PSU 750 W+
Manufacturer site
Illustration, not a product photo
Render rank (OptiX)
17 of 67
among models tested in Blender Benchmark 4.5.0 · 36% of the leader
Launch price
$1,499
MSRP, US
Release
September 24, 2020
Power supply
750 W+
manufacturer recommendation

Performance

Rank among models tested in the same benchmark on the same render backend.

Blender Benchmark 2.93.1

GPU, bmw27 (CUDA)

16.92 s

lower is better

Rank 1 of 31 tested on CUDA100% of the leader
17
0100200300
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 279 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.9218.49 s1 of 19
Blender Benchmark 2.91.218.5 s1 of 9
Blender Benchmark 2.90.120.55 s · small sample—
Blender Benchmark 2.9021.54 s1 of 27

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 2.93.1

GPU, bmw27 (OPTIX)

9.31 s

lower is better

Rank 1 of 20 tested on OptiX100% of the leader
9
050100150
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 244 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.929.07 s2 of 12
Blender Benchmark 2.91.29.22 s1 of 4
Blender Benchmark 2.90.18.88 s · small sample—
Blender Benchmark 2.908.95 s1 of 17

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 2.93.1

GPU, classroom (CUDA)

79.51 s

lower is better

Rank 1 of 26 tested on CUDA100% of the leader
80
0200400600800
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 260 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.9255.51 s1 of 16
Blender Benchmark 2.91.258.78 s1 of 5
Blender Benchmark 2.90.167.39 s · small sample—
Blender Benchmark 2.9070.6 s1 of 25

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 2.93.1

GPU, classroom (OPTIX)

33.49 s

lower is better

Rank 1 of 14 tested on OptiX100% of the leader
33
050100150200
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 182 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.9234.4 s1 of 10
Blender Benchmark 2.91.237.12 s1 of 3
Blender Benchmark 2.90.138.93 s · small sample—
Blender Benchmark 2.9035.23 s1 of 15

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 2.93.1

GPU, fishy_cat (CUDA)

34.43 s

lower is better

Rank 1 of 21 tested on CUDA100% of the leader
34
0100200300
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 213 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.9234.77 s1 of 13
Blender Benchmark 2.91.233.93 s1 of 5
Blender Benchmark 2.90.137.3 s · small sample—
Blender Benchmark 2.9039.03 s1 of 22

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 2.93.1

GPU, fishy_cat (OPTIX)

17.93 s

lower is better

Rank 1 of 13 tested on OptiX100% of the leader
18
020406080
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 141 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.9218.35 s1 of 9
Blender Benchmark 2.91.216.68 s1 of 3
Blender Benchmark 2.90.116.79 s · small sample—
Blender Benchmark 2.9016.52 s1 of 13

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 2.93.1

GPU, koro (CUDA)

54.57 s

lower is better

Rank 1 of 20 tested on CUDA100% of the leader
55
0200400600
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 216 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.9258.04 s1 of 12
Blender Benchmark 2.91.257.49 s1 of 5
Blender Benchmark 2.90.157.54 s · small sample—
Blender Benchmark 2.9057.95 s1 of 22

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 2.93.1

GPU, koro (OPTIX)

37.95 s

lower is better

Rank 1 of 13 tested on OptiX100% of the leader
38
050100150200
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 143 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.9257.13 s2 of 10
Blender Benchmark 2.91.236.35 s1 of 3
Blender Benchmark 2.90.137.08 s · small sample—
Blender Benchmark 2.9036.48 s1 of 13

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 2.93.1

GPU, pavillon_barcelona (CUDA)

99.33 s

lower is better

Rank 1 of 21 tested on CUDA100% of the leader
99
0200400600800
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 215 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.92109.26 s1 of 12
Blender Benchmark 2.91.2108.52 s1 of 4
Blender Benchmark 2.90.1108.8 s · small sample—
Blender Benchmark 2.90119.61 s1 of 23

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 2.93.1

GPU, pavillon_barcelona (OPTIX)

45.09 s

lower is better

Rank 1 of 13 tested on OptiX100% of the leader
45
0100200300
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 145 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.9245.05 s1 of 9
Blender Benchmark 2.91.246.94 s1 of 3
Blender Benchmark 2.90.146.55 s · small sample—
Blender Benchmark 2.9045.35 s1 of 13

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 2.93.1

GPU, victor (CUDA)

138.75 s

lower is better

Rank 1 of 19 tested on CUDA100% of the leader
139
02505007501,000
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 217 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.92178.26 s1 of 12
Blender Benchmark 2.91.2124.25 s1 of 4
Blender Benchmark 2.90.1148.72 s · small sample—
Blender Benchmark 2.90159.13 s1 of 20

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 2.93.1

GPU, victor (OPTIX)

60.44 s

lower is better

Rank 1 of 13 tested on OptiX100% of the leader
60
0100200300
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 144 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (4)
BenchmarkResultRank
Blender Benchmark 2.9262.54 s1 of 8
Blender Benchmark 2.91.253.95 s1 of 3
Blender Benchmark 2.90.157.67 s · small sample—
Blender Benchmark 2.9052.76 s1 of 13

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 3.4.0

GPU, sum of scenes monster + junkshop + classroom (CUDA)

3,798 samples/min

higher is better

Median of 22 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (12)
BenchmarkResultRank
Blender Benchmark 4.5.03,798 samples/min · small sample—
Blender Benchmark 4.3.03,731 samples/min · small sample—
Blender Benchmark 4.2.03,699 samples/min · small sample—
Blender Benchmark 4.1.03,736 samples/min · small sample—
Blender Benchmark 4.0.03,725 samples/min · small sample—
Blender Benchmark 3.6.04,061 samples/min · small sample—
Blender Benchmark 3.5.03,864 samples/min · small sample—
Blender Benchmark 3.3.03,730 samples/min · small sample—
Blender Benchmark 3.2.13,448 samples/min · small sample—
Blender Benchmark 3.2.03,138 samples/min · small sample—
Blender Benchmark 3.1.03,418 samples/min · small sample—
Blender Benchmark 3.0.13,697 samples/min · small sample—

Each benchmark version has its own scale; results from different versions are not compared.

Blender Benchmark 4.5.0

GPU, sum of scenes monster + junkshop + classroom (OptiX)

5,424 samples/min

higher is better

Rank 17 of 67 tested on OptiX36% of the leader
5,424
05,00010,00015,000
Each dot is one catalog model in this benchmark; GeForce RTX 3090 is highlighted.

Median of 596 runs · Blender Open Data (Blender Foundation) · November 14, 2026

Other benchmark versions (19)
BenchmarkResultRank
Blender Benchmark 5.3.05,144 samples/min · small sample—
Blender Benchmark 5.2.05,094 samples/min15 of 48
Blender Benchmark 5.1.15,253 samples/min14 of 46
Blender Benchmark 5.1.05,266 samples/min11 of 25
Blender Benchmark 5.0.04,876 samples/min8 of 11
Blender Benchmark 4.4.05,333 samples/min16 of 59
Blender Benchmark 4.3.05,348 samples/min13 of 50
Blender Benchmark 4.2.05,355 samples/min10 of 47
Blender Benchmark 4.1.05,536 samples/min9 of 45
Blender Benchmark 4.0.05,536 samples/min9 of 47
Blender Benchmark 3.6.06,306 samples/min10 of 49
Blender Benchmark 3.5.06,297 samples/min7 of 41
Blender Benchmark 3.4.06,015 samples/min7 of 43
Blender Benchmark 3.3.06,018 samples/min5 of 37
Blender Benchmark 3.2.15,626 samples/min3 of 30
Blender Benchmark 3.2.06,566 samples/min1 of 9
Blender Benchmark 3.1.25,827 samples/min · small sample—
Blender Benchmark 3.1.06,560 samples/min1 of 38
Blender Benchmark 3.0.15,842 samples/min1 of 9

Each benchmark version has its own scale; results from different versions are not compared.

Specifications

Reference values for the chip. Clocks, length and connectors depend on the card model.

Chip

Architecture
Ampere
Shader units
10,496 (CUDA cores)

Clocks

Base clock
1,400 MHz
Boost clock
1,700 MHz

Memory

Video memory
24 GB
Memory type
GDDR6X
Memory bus
384 bit

Power and dimensions

Board power
350 W (board power)
Recommended PSU
750 W
Length (reference / Founders Edition)
313 mm
Thickness (slots)
3
Power connectors (reference)
2x 8-pin (adapter to 1x 12-pin included) — FE

Checked against the manufacturer’s official page on 2026-11-08. Fields missing from the page are left empty (unknown).

Specifications: NVIDIA, checked November 8, 2026. Release date and launch price: Wikipedia (CC BY-SA 4.0).

Will it fit your system

Checks the power supply, case length and PCIe slot.

  • Needs checkingPower supply. Manufacturer recommendation: 750 W or more. Enter your PSU wattage to check.
  • Needs checkingCard length and case. Enter the maximum graphics card length for your case.
  • Needs checkingPCIe interface. The card’s PCIe version is unknown.