Comparigon Logo
GPU • NVIDIA • ADA LOVELACE
GeForce RTX 4070

GeForce RTX 4070

AD104 Released: April 12, 2023
Compare
Memory
12
GB GDDR6X
Boost Clock
2475
MHz
TGP
200
W

Technical Highlights

Architecture : Ada Lovelace
Generation : GeForce 40
Shading Units : 5888
RT Cores : 46

About this GPU Card

GeForce RTX 4070 is a discrete graphics card from NVIDIA released on April 12, 2023 based on the Ada Lovelace architecture with 12 GB GDDR6X memory. It runs at 1920 MHz base clock, reaches up to 2475 MHz boost clock, and has a 200 W power rating. It supports DirectX 12.2, Vulkan 1.4, and OpenGL 4.6.

Last updated: March 28, 2026
Jump to
Overall Scores General Info Graphics Processor Raw Performance Memory Board Design APIs Benchmarks Performance Charts FAQ

Overall Score Of GeForce RTX 4070

This summarizes how GeForce RTX 4070 scores across key areas like gaming, rendering, AI, and Energy Efficiency.

71
Overall
Gaming Performance
55
AI & Machine Learning
58
Hardware Architecture
79
Workstation
49
Energy Efficiency
100

General Info of GeForce RTX 4070

Vendor

NVIDIA

Release Date

April 12, 2023

Foundry

TSMC

Architecture

Ada Lovelace

Process Size

Excellent

5 nm

80% of the current best Process Size score

Transistors

35.8 billion

Launch Price

599 USD

Graphics Processor of GeForce RTX 4070

Base Clock

Good

1920 MHz

72% of the current best Base Clock score

Boost Clock

Excellent

2475 MHz

79% of the current best Boost Clock score

Shading Units

5888

27% of the current best Shading Units score

Compute Units

46

27% of the current best Compute Units score

Ray Tracing Cores (RTCs)

46

27% of the current best Ray Tracing Cores (RTCs) score

Tensor Cores

184

27% of the current best Tensor Cores score

Textured Mapping Units (TMUs)

184

27% of the current best Textured Mapping Units (TMUs) score

Render Output Units (ROPs)

64

25% of the current best Render Output Units (ROPs) score

Die Size

294 mm²

Density

121.8 MTr/mm²

L1 Cache

128 KB per SM

50% of the current best L1 Cache score

L2 Cache

Good

36 MB

38% of the current best L2 Cache score

Raw Performance of GeForce RTX 4070

Pixel Fill Rate

158.4 GPixel/s

33% of the current best Pixel Fill Rate score

Texture Fill Rate

455.4 GTexel/s

28% of the current best Texture Fill Rate score

FP32 Performance

29.15 TFLOPs

28% of the current best FP32 Performance score

Memory of GeForce RTX 4070

Memory Size

Excellent

12 GB

38% of the current best Memory Size score

Memory Type

Excellent

GDDR6X

80% of the current best Memory Type score

Bus Width

192-bit

Memory Clock

1313 MHz

50% of the current best Memory Clock score

Effective Speed

Good

21000 Mbps

70% of the current best Effective Speed score

Bandwidth

504.2 GB/s

28% of the current best Bandwidth score

Board Design of GeForce RTX 4070

Bus Interface

Excellent

PCIe 4.0 x16

80% of the current best Bus Interface score

PCIe Version

Excellent

4.0

80% of the current best PCIe Version score

PCIe Lanes

x16

TGP

200 W

Length

240 mm

Width

40 mm

Height

110 mm

APIs of GeForce RTX 4070

DirectX

Top tier

12.2

100% of the current best DirectX score

OpenGL

Top tier

4.6

100% of the current best OpenGL score

OpenCL

Top tier

3.0

100% of the current best OpenCL score

Vulkan

Top tier

1.4

100% of the current best Vulkan score

HDMI

Top tier

2.1

100% of the current best HDMI score

DisplayPort

Excellent

1.4a

80% of the current best DisplayPort score

CUDA

8.9

Ray Tracing

Yes

DLSS

DLSS 3

Benchmarks of GeForce RTX 4070

3D Mark

A standardized suite that approximates real-world gaming workloads to compare graphics performance.

Steel Nomad Lite:
18093
34% of max score
Time Spy:
17811 (25th)
38% of max score
Solar Bay:
82720
35% of max score
Port Royal:
11154
29% of max score
Fire Strike:
44071
51% of max score
Wild Life Extreme:
35487
33% of max score
Night Raid:
153448
74% of max score
Good

GeekBench 6 OpenCL

GPU compute benchmark focused on workloads like image processing and parallel compute tasks.

Compute Score:
167522 (21st)
40% of max score
Background Blur:
233 img/sec
55% of max score
Face Detection:
141.7 img/sec
47% of max score
Horizon Detection:
6.62 Gpixels/sec
31% of max score
Edge Detection:
9.19 Gpixels/sec
27% of max score
Gaussian Blur:
9.2 Gpixels/sec
27% of max score
Feature Matching:
1.78 Gpixels/sec
70% of max score
Good
Stereo Matching:
845.4 Gpixels/sec
30% of max score
Particle Physics:
25043.9 FPS
42% of max score

Passmark Graphics

A multi-test benchmark that evaluates raster, memory bandwidth, and compute performance.

G3D Mark Score:
26914 (16th)
69% of max score
Good
G2D Mark Score:
1168
83% of max score
Excellent
DirectX 12:
103
58% of max score
GPU Compute:
14633 Ops/s
56% of max score

GFXBench 5

Graphics benchmark using modern GPU APIs and game-like scenes across multiple workloads.

Aztec Ruins High (4K):
192.00 FPS
16% of max score
Aztec Ruins Normal:
1039.00 FPS
53% of max score
Car Chase:
900.00 FPS
52% of max score
Manhattan 3.1 (1440p):
761.20 FPS
19% of max score
T-Rex:
3848.10 FPS
58% of max score

Blender

Rendering performance benchmark using Blender Cycles throughput as a practical content-creation metric.

Render Score:
5425 (16th)
36% of max score

GeekBench 6 ML

AI benchmark measuring common ML workloads like classification, detection, and super resolution.

ML Single Precision:
27166 (21st)
50% of max score
ML Half Precision:
40963
51% of max score
ML Quantized:
20332
52% of max score
Image Classification:
10246
54% of max score
Object Detection:
13134
47% of max score
Image Segmentation:
26153
47% of max score
Face Detection:
51337
45% of max score
Image Super Resolution:
32054
58% of max score
Style Transfer:
263477
38% of max score
Text Classification:
3703
71% of max score
Good
Machine Translation:
5484
63% of max score
Good
Depth Estimation:
50548
59% of max score
Pose Estimation:
124617
36% of max score

FurMark

Stress test designed to push sustained load for thermal stability and cooling behavior.

1080p Vulkan:
12540
37% of max score
1440p Vulkan:
8652
34% of max score
2160p Vulkan:
3650
25% of max score
1080p OpenGL:
11999
41% of max score
1440p OpenGL:
9077
39% of max score
2160p OpenGL:
4084
27% of max score

Performance Charts Of GeForce RTX 4070

These charts give a quick view of FPS scaling across resolutions and how this GPU compares against market top benchmark values.

Gaming FPS Analysis

Ultra settings performance of GeForce RTX 4070
1080p Ultra
1440p Ultra
4K Ultra

Performance Benchmarks

Grey is the market top value. Colored bar is GeForce RTX 4070.
Market Top
GeForce RTX 4070

Try to Compare GeForce RTX 4070 With Another GPU

You can search and find another gpu to compare.

More GPUs You Might Like to See

If you are still deciding, these are great options to look at next.

Image of GeForce RTX 4050 Laptop

GeForce RTX 4050 Laptop

NVIDIA

Image of GeForce RTX 4060 Laptop

GeForce RTX 4060 Laptop

NVIDIA

Image of GeForce RTX 4070 Laptop

GeForce RTX 4070 Laptop

NVIDIA

Image of GeForce RTX 4080 Laptop

GeForce RTX 4080 Laptop

NVIDIA

Image of MTT S70

MTT S70

Moore Threads

Image of Quadro RTX 4000

Quadro RTX 4000

NVIDIA

Image of Radeon RX 7600S

Radeon RX 7600S

AMD

Image of Radeon RX 7700S

Radeon RX 7700S

AMD

Image of GeForce RTX 4060

GeForce RTX 4060

NVIDIA

Image of Intel Arc A580

Intel Arc A580

Intel

Discover More GPUs

Frequently Asked Questions about GeForce RTX 4070

The GeForce RTX 4070 was officially released in April 12, 2023. It was launched as a graphics solution by NVIDIA.

The GeForce RTX 4070 scores 17,811 on 3DMark Time Spy, indicating its solid capabilities in modern gaming scenarios. This benchmark is the industry standard for measuring DirectX 12 gaming performance.

The GeForce RTX 4070 is equipped with 12 GB of VRAM, which is currently the sweet spot for smooth 1440p gaming and modern texture quality.

The GeForce RTX 4070 has a Total Graphics Power (TGP) of 200 W. This value dictates your power supply (PSU) requirements, so ensure your system can handle this draw effectively.

The GeForce RTX 4070 operates at a base clock speed of 1920 MHz. This is the minimum clock frequency the graphics card runs at under standard workloads before boosting.

The GeForce RTX 4070 features a boost clock speed of 2475 MHz. This clock speed represents the maximum frequency the graphics card can dynamically reach during intense gaming or rendering workloads.

The GeForce RTX 4070 is built on the Ada Lovelace architecture. This foundational design determines the GPU's power efficiency and support for modern features like hardware Ray Tracing and AI upscaling (DLSS/FSR).

With a performance score of 71%, the GeForce RTX 4070 is ranked as a high-performance graphics card in our extensive GPU database.

The GeForce RTX 4070 uses the PCIe 4.0 x16 interface, ensuring high-speed data transfer and minimal bottlenecking between the graphics card and the motherboard.

The GeForce RTX 4070 features 5,888 shading units (cores), which are directly responsible for rendering complex visual effects and lighting in modern games.

The GeForce RTX 4070 contains approximately 35.8 billion transistors, packed into its silicon to deliver massive parallel processing capabilities.

The GeForce RTX 4070 is manufactured on a 5 nm node. A smaller process node generally translates to better energy efficiency and improved thermal management.