Skip to content
RU

Graphics cards

235 models. Pick two and compare.

Benchmark rankings

AMD · RDNA 2

Radeon RX 6900 XT

  • 16 GB GDDR6
  • 300 W
  • PSU 850 W+
Render (HIP) · 6 of 1360%
At launch
$999

NVIDIA · Ada Lovelace

GeForce RTX 4050 Laptop GPU

  • 6 GB GDDR6
  • 35–115 W
Render (OptiX) · 40 of 6717%

AMD · RDNA 3

Radeon RX 7800 XT

  • 16 GB GDDR6
  • 263 W
  • PSU 700 W+
Render (HIP) · 7 of 1357%
At launch
$499

AMD · RDNA 2

Radeon RX 6800 XT

  • 16 GB GDDR6
  • 300 W
  • PSU 750 W+
Render (HIP) · 8 of 1356%
At launch
$649

NVIDIA · Turing

GeForce RTX 2080 SUPER

  • 8 GB GDDR6
  • 250 W
Render (OptiX) · 41 of 6715%
At launch
$699

NVIDIA · Turing

GeForce RTX 2080

  • 8 GB GDDR6
  • 215 W
Render (OptiX) · 42 of 6715%
At launch
$699

NVIDIA · Turing

GeForce RTX 2070 SUPER

  • 8 GB GDDR6
  • 215 W
Render (OptiX) · 43 of 6715%
At launch
$499

NVIDIA · Ampere

GeForce RTX 3060

  • 12 GB GDDR6
  • 170 W
  • PSU 550 W+
Render (OptiX) · 45 of 6714%
At launch
$329

NVIDIA · Turing

GeForce RTX 2070

  • 8 GB GDDR6
  • 175 W
Render (OptiX) · 46 of 6714%
At launch
$499

AMD · RDNA 3

Radeon RX 7700 XT

  • 12 GB GDDR6
  • 245 W
  • PSU 700 W+
Render (HIP) · 9 of 1348%
At launch
$449

NVIDIA · Turing

GeForce RTX 2060 SUPER

  • 8 GB GDDR6
  • 175 W
Render (OptiX) · 47 of 6714%
At launch
$399

AMD · RDNA 2

Radeon RX 6800

  • 16 GB GDDR6
  • 250 W
  • PSU 650 W+
Render (HIP) · 10 of 1343%
At launch
$579

Apple · Apple Silicon

Apple M5 (GPU, 10 cores)

  • 10 GPU cores
Render (Metal) · 9 of 1833%

NVIDIA · Turing

GeForce RTX 2060

  • 6 GB GDDR6
  • 160 W
  • PSU 500 W+
Render (OptiX) · 48 of 6711%

AMD · RDNA 2

Radeon RX 6700 XT

  • 12 GB GDDR6
  • 230 W
  • PSU 650 W+
Render (HIP) · 11 of 1334%
At launch
$479

NVIDIA · Ampere

GeForce RTX 3050

  • 8 GB GDDR6
  • 130 W
  • PSU 550 W+
Render (OptiX) · 49 of 679%

AMD · RDNA 3

Radeon RX 7600

  • 8 GB GDDR6
  • 165 W
  • PSU 550 W+
Render (HIP) · 12 of 1327%
At launch
$269

The bar on a card is the result in Blender Benchmark 4.5.0 relative to the leader on the same render backend (OptiX, HIP, oneAPI or Metal; rendering, not games); “N of M” is the position among M tested models. Manufacturer's suggested launch price (US), rounded to whole dollars.