NVIDIA GeForce RTX 5070 vs NVIDIA GeForce RTX 5090 graphics comparison: performance score, cores and clock speeds, VRAM, TGP, cache, and RT and AI units. The better value in each row is highlighted in green.
Summary
NVIDIA GeForce RTX 5070 offers a higher boost clock (2347 MHz vs 2160 MHz). NVIDIA GeForce RTX 5090 offers more compute units (CUDA cores: 10496 vs 4608), more VRAM (24 GB vs 8 GB), a wider memory bus (256 bit vs 128 bit), higher memory bandwidth (896 GB/s vs 384 GB/s), and more L2 cache (64 MB vs 32 MB).
Performance score
0
Coming soon
0
Coming soon
Key specs
CUDA cores4608
CUDA cores10496
Boost2347 MHz
Boost2160 MHz
VRAM8 GB
VRAM24 GB
Memory bus128 bit
Memory bus256 bit
TGP50–115 W
TGP95–175 W
MemoryGDDR7
MemoryGDDR7
General
SeriesRTX 50 Mobile
SeriesRTX 50 Mobile
GPU architectureBlackwell
GPU architectureBlackwell
GPU chipGB206
GPU chipGB203
Chip variantGN22-X6-A1
Chip variantGN22-X11-A1
FoundryTSMC
FoundryTSMC
Process node5 nm
Process node5 nm
Die size181 mm²
Die size378 mm²
Transistors21.9 billion
Transistors45.6 billion
InterfacePCIe 5.0 x8
InterfacePCIe 5.0 x16
Release dateJanuary 6, 2025
Release dateJanuary 6, 2025
Cores and clocks
CUDA cores4608
CUDA cores10496
36× SM
82× SM
Boost clock (max)2347 MHz
Boost clock (max)2160 MHz
Boost clock (min)1425 MHz
Boost clock (min)1597 MHz
Memory
VRAM8 GB
VRAM24 GB
MemoryGDDR7
MemoryGDDR7
Memory bus128 bit
Memory bus256 bit
Memory speed24 Gb/s
Memory speed28 Gb/s
Bandwidth384 GB/s
Bandwidth896 GB/s
TGP
TGP range50 - 100 W
TGP range95 - 150 W
Dynamic Boost15 W
Dynamic Boost25 W
Cache
L2 cache32 MB
L2 cache64 MB
L1 cache128 KB
L1 cache128 KB
Units
TMUs144
TMUs328
ROPs48
ROPs112
RT cores36
RT cores82
Tensor cores144
Tensor cores328
SM blocks36
SM blocks82
AI
AI performance798 TOPS
AI performance1824 TOPS