A near-full AD104: 56 of the die's 60 SMs, against 46 on the plain RTX 4070. Same 12 GB on a 192-bit bus, 20 W more power.

NVIDIA GeForce RTX 4070 SUPER 12GB

Ada Lovelace PCIe 2024 4nm Spec confidence: Official
FP8 (dense)
142
TFLOPS
FP32
35
TFLOPS
VRAM
12 GB
GDDR6X
TDP
220 W
161.3 TFLOPS/kW

Overview

The GeForce RTX 4070 SUPER takes the AD104 die most of the way to its limit, with 7,168 CUDA cores and 224 fourth-generation Tensor cores against the plain RTX 4070's 5,888 and 184, a 22 percent increase for 20 W more board power. Memory is unchanged at 12 GB of GDDR6X on a 192-bit bus, which is the card's real constraint for AI work. It reaches 35.5 TFLOPS of FP32 and 141.9 TFLOPS of dense FP8, putting it within reach of the older RTX 4070 Ti at a lower price. It is a sensible second-hand choice for quantised models up to about 13B, and a poor one for anything that needs more than 12 GB.

Performance

Peak theoretical throughput by precision type

PrecisionDense2:4 Sparse
FP64
64-bit floating point
0.6TFLOPS Structured sparsity is a tensor-core feature; this vector precision has no sparse form
FP32
32-bit floating point
35TFLOPS Structured sparsity is a tensor-core feature; this vector precision has no sparse form
TF32
TensorFloat-32
35TFLOPS The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision
BF16
Brain Float 16
71TFLOPS The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision
FP16
16-bit floating point
71TFLOPS The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision
FP8
8-bit floating point
142TFLOPS The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision
FP6
No verified data available The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision
FP4
No verified data available The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision
INT8
8-bit integer
284TOPS The part supports 2:4 sparsity, but the vendor publishes no sparse figure for this precision

The GeForce RTX 4070 SUPER 12GB in the GPU landscape

Peak FP16 TFLOPS (dense) against TDP, single-GPU parts tracked by Flopper

06001,2001,8002,4003,0000 W250 W500 W750 W1000 W1250 W1500 WInstinct MI355XGeForce RTX 4070 SUPER 12GB

Higher and further left is better: more half-precision throughput for less power.

Specifications

Architecture

Ada Lovelace

Form Factor

PCIe

Launch Year

2024

Process Node

4nm

Memory

12 GB GDDR6X

Bandwidth

No verified data available

TDP

220 W

Max Power

220 W

Transistors

35.8 billion

CUDA Cores

7,168

Spec Confidence

Official

Full Specifications

Compute Engine
CUDA Cores 7,168
Tensor Cores 224 (4th Gen)
Streaming Multiprocessors 56
Base Clock 1.98 GHz
Boost Clock 2.48 GHz
Chip Design
Transistors 35.8 billion
Die Size 295 mm²
Process Node 4nm
Memory
VRAM 12 GB
Memory Type GDDR6X
Bandwidth No verified data available
Interface Width 192-bit
Interconnect & I/O
PCIe 4.0
Power & Thermal
TDP 220 W
Max Board Power 220 W
Cooling Active
Enterprise Features
Sparsity Yes
Physical & Media
Card Length 244 mm
Width 2-slot
General
Form Factor PCIe
Architecture Ada Lovelace
Process Node 4nm
Launch Year 2024

Datasheet & Resources

Data Provenance

Every figure traced to a source

Primary Source

Publisher
NVIDIA
Published
2024-01-08

Data Quality

Spec confidence
Official
Clock basis
Boost
Core precisions with figures
7 of 9
Normalization
All values in TFLOPS

Rent this GPU

$0.07 / GPU-hr

Lowest of 1 live listing across 1 provider

See all rental prices →

Price history and every provider on the pricing page. Some links are affiliate links.

Similar GPUs

NVIDIA L40S 48GB

PCIe · 2023
FP32: 92 TFLOPS
Compare vs L40S

NVIDIA L40 48GB

PCIe · 2023
FP32: 91 TFLOPS
Compare vs L40

Frequently Asked Questions

How many TFLOPS does the NVIDIA GeForce RTX 4070 SUPER have?

The NVIDIA GeForce RTX 4070 SUPER delivers 35 TFLOPS FP32, 71 TFLOPS FP16 and 142 TFLOPS FP8 at peak.

What is the power consumption of the NVIDIA GeForce RTX 4070 SUPER?

The NVIDIA GeForce RTX 4070 SUPER has a TDP (Thermal Design Power) rating of 220 watts.

How much memory does the NVIDIA GeForce RTX 4070 SUPER have?

The NVIDIA GeForce RTX 4070 SUPER is equipped with 12 GB of VRAM. Flopper does not currently have a verified memory bandwidth figure for it.

What architecture is the NVIDIA GeForce RTX 4070 SUPER based on?

The NVIDIA GeForce RTX 4070 SUPER is based on the Ada Lovelace architecture, launched in 2024.

Stay Updated on GPU Releases

Get notified when new GPUs are added or specifications are updated.

Loading verification...

No spam, unsubscribe anytime.

Back to GPUs