Home
GPUs
GeForce RTX 5070

GPU • NVIDIA • BLACKWELL 2.0

GeForce RTX 5070

Q: When was the GeForce RTX 5070 released?

The GeForce RTX 5070 was officially released in March 4, 2025. It was launched as a graphics solution by NVIDIA.

Q: What is the 3DMark Time Spy score of GeForce RTX 5070?

The GeForce RTX 5070 scores 22,672 on 3DMark Time Spy, indicating its solid capabilities in modern gaming scenarios. This benchmark is the industry standard for measuring DirectX 12 gaming performance.

Q: How much VRAM does the GeForce RTX 5070 have?

The GeForce RTX 5070 is equipped with 12 GB of VRAM, which is currently the sweet spot for smooth 1440p gaming and modern texture quality.

Q: What is the TGP (power consumption) of GeForce RTX 5070?

The GeForce RTX 5070 has a Total Graphics Power (TGP) of 250 W. This value dictates your power supply (PSU) requirements, so ensure your system can handle this draw effectively.

Q: What is the base clock speed of the GeForce RTX 5070?

The GeForce RTX 5070 operates at a base clock speed of 2325 MHz. This is the minimum clock frequency the graphics card runs at under standard workloads before boosting.

Q: What is the boost clock speed of the GeForce RTX 5070?

The GeForce RTX 5070 features a boost clock speed of 2512 MHz. This clock speed represents the maximum frequency the graphics card can dynamically reach during intense gaming or rendering workloads.

Q: What architecture is the GeForce RTX 5070 based on?

The GeForce RTX 5070 is built on the Blackwell 2.0 architecture. This foundational design determines the GPU's power efficiency and support for modern features like hardware Ray Tracing and AI upscaling (DLSS/FSR).

Q: How does the GeForce RTX 5070 rank in gaming performance?

With a performance score of 80%, the GeForce RTX 5070 is ranked as a high-performance graphics card in our extensive GPU database.

Q: What bus interface does the GeForce RTX 5070 use?

The GeForce RTX 5070 uses the PCIe 5.0 x16 interface, ensuring high-speed data transfer and minimal bottlenecking between the graphics card and the motherboard.

Q: How many shading units does the GeForce RTX 5070 have?

The GeForce RTX 5070 features 6,144 shading units (cores), which are directly responsible for rendering complex visual effects and lighting in modern games.

GB205-300 Released: March 4, 2025

Compare

GeForce RTX 5070

GB205-300 Released: March 4, 2025

Memory

GB GDDR7

Boost Clock

2512

MHz

TGP

250

Technical Highlights

Architecture : Blackwell 2.0

Generation : GeForce 50

Shading Units : 6144

RT Cores : 48

About this GPU Card

GeForce RTX 5070 is a discrete graphics card from NVIDIA released on March 4, 2025 based on the Blackwell 2.0 architecture with 12 GB GDDR7 memory. It runs at 2325 base clock, reaches up to 2512 boost clock, and has a 250 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 5070

This summarizes how GeForce RTX 5070 scores across key areas like gaming, rendering, AI, and Energy Efficiency.

Overall

Gaming Performance

AI & Machine Learning

Hardware Architecture

Workstation

Energy Efficiency

General Info of GeForce RTX 5070

Vendor

NVIDIA

Release Date

March 4, 2025

Foundry

TSMSC

Architecture

Blackwell 2.0

Process Size

Excellent

5 nm

80% of the current best Process Size score

Transistors

31.1 billion

Launch Price

549

Graphics Processor of GeForce RTX 5070

Base Clock

Excellent

2325 MHz

87% of the current best Base Clock score

Boost Clock

Excellent

2512 MHz

80% of the current best Boost Clock score

Shading Units

6144

28% of the current best Shading Units score

Compute Units

Good

47% of the current best Compute Units score

Ray Tracing Cores (RTCs)

28% of the current best Ray Tracing Cores (RTCs) score

Tensor Cores

192

28% of the current best Tensor Cores score

Textured Mapping Units (TMUs)

192

28% of the current best Textured Mapping Units (TMUs) score

Render Output Units (ROPs)

31% of the current best Render Output Units (ROPs) score

Die Size

263 mm²

Density

118.25 MTr/mm²

L1 Cache

128KB per cluster

50% of the current best L1 Cache score

L2 Cache

Good

48MB shared

50% of the current best L2 Cache score

Raw Performance of GeForce RTX 5070

Pixel Fill Rate

201.0 GPixel/s

42% of the current best Pixel Fill Rate score

Texture Fill Rate

482.3 GTexel/s

29% of the current best Texture Fill Rate score

FP32 Performance

30.87 TFLOPs

29% of the current best FP32 Performance score

Memory of GeForce RTX 5070

Memory Size

Excellent

12 GB

38% of the current best Memory Size score

Memory Type

Top tier

GDDR7

90% of the current best Memory Type score

Bus Width

192-bit

Memory Clock

Good

1750 MHz

67% of the current best Memory Clock score

Effective Speed

Top tier

28000 Mbps

93% of the current best Effective Speed score

Bandwidth

672 GB/s

38% of the current best Bandwidth score

Board Design of GeForce RTX 5070

Bus Interface

Top tier

PCIe 5.0 x16

100% of the current best Bus Interface score

PCIe Version

Top tier

5.0

100% of the current best PCIe Version score

PCIe Lanes

x16

TGP

250 W

Length

245 mm

Width

40 mm

Height

115 mm

APIs of GeForce RTX 5070

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

DisplayPort

Top tier

2.1b

100% of the current best DisplayPort score

CUDA

12.0

Ray Tracing

Yes

DLSS

DLSS 4

Benchmarks of GeForce RTX 5070

3D Mark

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

Steel Nomad Lite:

23186

44% of max score 44% of best score for Steel Nomad Lite

Time Spy:

22672 (13th)

48% of max score 48% of best score for Time Spy

Solar Bay:

106341

45% of max score 45% of best score for Solar Bay

Port Royal:

14348

38% of max score 38% of best score for Port Royal

Fire Strike:

58080

67% of max score 67% of best score for Fire Strike

Good

Wild Life Extreme:

43673

40% of max score 40% of best score for Wild Life Extreme

Night Raid:

175736

85% of max score 85% of best score for Night Raid

Excellent

GeekBench 6 OpenCL

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

Compute Score:

190967 (15th)

46% of max score 46% of best score for Compute Score

Background Blur:

254.4 img/sec

60% of max score 60% of best score for Background Blur

Good

Face Detection:

180.6 img/sec

61% of max score 61% of best score for Face Detection

Good

Horizon Detection:

7.91 Gpixels/sec

36% of max score 36% of best score for Horizon Detection

Edge Detection:

10.2 Gpixels/sec

30% of max score 30% of best score for Edge Detection

Gaussian Blur:

10.7 Gpixels/sec

31% of max score 31% of best score for Gaussian Blur

Feature Matching:

1.84 Gpixels/sec

73% of max score 73% of best score for Feature Matching

Good

Stereo Matching:

975.7 Gpixels/sec

34% of max score 34% of best score for Stereo Matching

Particle Physics:

27816.5 FPS

47% of max score 47% of best score for Particle Physics

Passmark Graphics

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

G3D Mark Score:

28787 (12th)

74% of max score 74% of best score for G3D Mark Score

Good

G2D Mark Score:

1304

92% of max score 92% of best score for G2D Mark Score

Top tier

DirectX 12:

106

60% of max score 60% of best score for DirectX 12

Good

GPU Compute:

14598 Ops/s

56% of max score 56% of best score for GPU Compute

GFXBench 5

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

Aztec Ruins High (4K):

1199.40 FPS

100% of max score 100% of best score for Aztec Ruins High (4K)

Top tier

Aztec Ruins Normal:

995.80 FPS

51% of max score 51% of best score for Aztec Ruins Normal

Car Chase:

859.00 FPS

50% of max score 50% of best score for Car Chase

Manhattan 3.1 (1440p):

4075.00 FPS

100% of max score 100% of best score for Manhattan 3.1 (1440p)

Top tier

T-Rex:

6657.70 FPS

100% of max score 100% of best score for T-Rex

Top tier

Blender

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

Render Score:

6160 (10th)

41% of max score 41% of best score for Render Score

GeekBench 6 ML

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

ML Single Precision:

30221 (18th)

56% of max score 56% of best score for ML Single Precision

ML Half Precision:

30221

37% of max score 37% of best score for ML Half Precision

ML Quantized:

23474

60% of max score 60% of best score for ML Quantized

Good

Image Classification:

12270

65% of max score 65% of best score for Image Classification

Good

Object Detection:

16709

60% of max score 60% of best score for Object Detection

Good

Image Segmentation:

30237

55% of max score 55% of best score for Image Segmentation

Face Detection:

62662

55% of max score 55% of best score for Face Detection

Image Super Resolution:

35585

65% of max score 65% of best score for Image Super Resolution

Good

Style Transfer:

246598

36% of max score 36% of best score for Style Transfer

Text Classification:

4198

80% of max score 80% of best score for Text Classification

Excellent

Machine Translation:

6233

72% of max score 72% of best score for Machine Translation

Good

Depth Estimation:

58300

68% of max score 68% of best score for Depth Estimation

Good

Pose Estimation:

145190

42% of max score 42% of best score for Pose Estimation

FurMark

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

1080p Vulkan:

17361

51% of max score 51% of best score for 1080p Vulkan

1440p Vulkan:

12595

49% of max score 49% of best score for 1440p Vulkan

2160p Vulkan:

7899

54% of max score 54% of best score for 2160p Vulkan

1080p OpenGL:

16477

56% of max score 56% of best score for 1080p OpenGL

1440p OpenGL:

12295

53% of max score 53% of best score for 1440p OpenGL

2160p OpenGL:

6339

42% of max score 42% of best score for 2160p OpenGL

Performance Charts Of GeForce RTX 5070

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 5070

1080p Ultra

1440p Ultra

4K Ultra

Performance Benchmarks

Grey is the market top value. Colored bar is GeForce RTX 5070.

Market Top

GeForce RTX 5070

Try to Compare GeForce RTX 5070 With Another GPU

You can search and find another gpu to compare.

GPU 1

Select GPU 2

More GPUs You Might Like to See

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

GeForce RTX 5050 Laptop

NVIDIA

GeForce RTX 5060 Laptop

NVIDIA

GeForce RTX 5070 Ti Laptop

NVIDIA

GeForce RTX 5070 Laptop

NVIDIA

GeForce RTX 5080 Laptop

NVIDIA

GeForce RTX 5090 Laptop

NVIDIA

Intel Arc B570

Intel

Radeon RX 9060 XT (8GB)

AMD

Radeon RX 9060 XT (16GB)

AMD

GeForce RTX 5060 Ti (16GB)

NVIDIA

Discover More GPUs

Frequently Asked Questions about GeForce RTX 5070

The GeForce RTX 5070 was officially released in March 4, 2025. It was launched as a graphics solution by NVIDIA.

The GeForce RTX 5070 scores 22,672 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 5070 is equipped with 12 GB of VRAM, which is currently the sweet spot for smooth 1440p gaming and modern texture quality.

The GeForce RTX 5070 has a Total Graphics Power (TGP) of 250 W. This value dictates your power supply (PSU) requirements, so ensure your system can handle this draw effectively.

The GeForce RTX 5070 operates at a base clock speed of 2325 MHz. This is the minimum clock frequency the graphics card runs at under standard workloads before boosting.

The GeForce RTX 5070 features a boost clock speed of 2512 MHz. This clock speed represents the maximum frequency the graphics card can dynamically reach during intense gaming or rendering workloads.

The GeForce RTX 5070 is built on the Blackwell 2.0 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 80%, the GeForce RTX 5070 is ranked as a high-performance graphics card in our extensive GPU database.

The GeForce RTX 5070 uses the PCIe 5.0 x16 interface, ensuring high-speed data transfer and minimal bottlenecking between the graphics card and the motherboard.

The GeForce RTX 5070 features 6,144 shading units (cores), which are directly responsible for rendering complex visual effects and lighting in modern games.

The GeForce RTX 5070 contains approximately 31.1 billion transistors, packed into its silicon to deliver massive parallel processing capabilities.

The GeForce RTX 5070 is manufactured on a 5 nm node. A smaller process node generally translates to better energy efficiency and improved thermal management.