ByteNows. DIGISCOPE
中文 / EN Browse briefs
← All GPU comparisons

Spec comparison

GeForce RTX 4090 vs GeForce RTX 5090

GeForce RTX 4090 vs GeForce RTX 5090 — every spec row side by side; launch prices shown per market, not converted.

19 of 19 specs differ GPU highlighted rows = specs that differ

Spec table

Spec GeForce RTX 4090 NVIDIA 官方规格 GeForce RTX 5090 NVIDIA 官方规格
Released 2022-10 src 2025-01 src
Architecture Ada Lovelace src Blackwell src
Shader units 16,384 CUDA src 21,760 CUDA src
RT cores 128 src 170 src
Tensor cores 512 src 680 src
Base clock 2,235 MHz src 2,017 MHz src
Boost clock 2,520 MHz src 2,407 MHz src
VRAM 24 GB src 32 GB src leads
Memory type GDDR6X src GDDR7 src
Bus width 384-bit src 512-bit src leads
Bandwidth 1,008 GB/s src 1,792 GB/s src leads
TDP 450 W src 575 W src
Interface PCIe 4.0 src PCIe 5.0 src
FP32 82.58 TFLOPS src 104.8 TFLOPS src leads
3DMark Time Spy 36,328 src 46,923 src leads
3DMark Speed Way 10,074 src 14,614 src leads
Blender Benchmark 11,060 src 14,972 src leads
PassMark G3D Mark 38,073 src 38,949 src leads
US MSRP $1,599 src leads $1,999 src

Specs are launch values aggregated from TechAPI open data (CC BY-SA 4.0) and manufacturer spec pages. Only named benchmarks of the same version are judged; spec numbers are shown for information. Prices are US launch MSRP. Rows missing on one side are marked, never guessed.

What it means for you

GeForce RTX 4090

  • No clear spec advantage

GeForce RTX 5090

  • VRAM 32 GB vs 24 GB
  • 3DMark Time Spy 46,923 vs 36,328
  • 3DMark Speed Way 14,614 vs 10,074
  • Blender Benchmark 14,972 vs 11,060
  • PassMark G3D Mark 38,949 vs 38,073

Enter the actual prices you can get. Launch MSRPs are used until you type.

What matters most to you? Pick up to two - the matching advantages get highlighted.

Raw diff

  • Released: GeForce RTX 4090 — 2022-10 ｜ GeForce RTX 5090 — 2025-01
  • Architecture: GeForce RTX 4090 — Ada Lovelace ｜ GeForce RTX 5090 — Blackwell
  • Shader units: GeForce RTX 4090 — 16,384 CUDA ｜ GeForce RTX 5090 — 21,760 CUDA
  • RT cores: GeForce RTX 4090 — 128 ｜ GeForce RTX 5090 — 170
  • Tensor cores: GeForce RTX 4090 — 512 ｜ GeForce RTX 5090 — 680
  • Base clock: GeForce RTX 4090 — 2,235 MHz ｜ GeForce RTX 5090 — 2,017 MHz
  • Boost clock: GeForce RTX 4090 — 2,520 MHz ｜ GeForce RTX 5090 — 2,407 MHz
  • VRAM: GeForce RTX 4090 — 24 GB ｜ GeForce RTX 5090 — 32 GB
  • Memory type: GeForce RTX 4090 — GDDR6X ｜ GeForce RTX 5090 — GDDR7
  • Bus width: GeForce RTX 4090 — 384-bit ｜ GeForce RTX 5090 — 512-bit
  • Bandwidth: GeForce RTX 4090 — 1,008 GB/s ｜ GeForce RTX 5090 — 1,792 GB/s
  • TDP: GeForce RTX 4090 — 450 W ｜ GeForce RTX 5090 — 575 W
  • Interface: GeForce RTX 4090 — PCIe 4.0 ｜ GeForce RTX 5090 — PCIe 5.0
  • FP32: GeForce RTX 4090 — 82.58 TFLOPS ｜ GeForce RTX 5090 — 104.8 TFLOPS
  • 3DMark Time Spy: GeForce RTX 4090 — 36,328 ｜ GeForce RTX 5090 — 46,923
  • 3DMark Speed Way: GeForce RTX 4090 — 10,074 ｜ GeForce RTX 5090 — 14,614
  • Blender Benchmark: GeForce RTX 4090 — 11,060 ｜ GeForce RTX 5090 — 14,972
  • PassMark G3D Mark: GeForce RTX 4090 — 38,073 ｜ GeForce RTX 5090 — 38,949
  • US MSRP: GeForce RTX 4090 — $1,599 ｜ GeForce RTX 5090 — $1,999