The RTX 2070 SUPER is a cut TU104, not a full TU106, which is why it has NVLink where the plain RTX 2070 has none. NVIDIA published a single clock bin for the SUPER refresh, so there is no reference-versus-Founders-Edition split. Its precision ladder is transcribed from the Ampere GA102 whitepaper, which prints it beside the RTX 3070 for comparison.

NVIDIA GeForce RTX 2070 Super 8GB

Turing PCIe 2019 12nm Spec confidence: Official
FP32
9.1
TFLOPS
VRAM
8 GB
GDDR6
Bandwidth
448 GB/s
memory
TDP
215 W
42.3 TFLOPS/kW
Download datasheet

Overview

The GeForce RTX 2070 SUPER moved the 70-class card from the TU106 die to a cut-down TU104, which brought a bigger tensor array and, unusually for the tier, NVLink support. It delivers 36.3 TFLOPS of FP16 tensor throughput accumulating in FP32 and 145 TOPS of INT8, on 8 GB of GDDR6 at 448 GB/s and 215 W. Turing has no BF16, no TF32 and no structured sparsity, so those are dense figures that need no adjustment before you compare them with a newer card. As with every 8 GB part of this era, the frame buffer rather than the arithmetic is what decides whether a model fits.

Performance

Peak theoretical throughput by precision type

PrecisionPeak
FP64
64-bit floating point
0.3TFLOPS
FP32
32-bit floating point
9.1TFLOPS
TF32
No verified data available
BF16
No verified data available
FP16
16-bit floating point
36TFLOPS
FP8
No verified data available
FP6
No verified data available
FP4
No verified data available
INT8
8-bit integer
145TOPS
INT4
4-bit integer
290TOPS

Every figure here is dense. The vendor documents no structured (2:4) sparsity mode for this part, so there is no second number to quote.

The GeForce RTX 2070 Super 8GB 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 2070 Super 8GB

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

Specifications

Architecture

Turing

Form Factor

PCIe

Launch Year

2019

Process Node

12nm

Memory

8 GB GDDR6

Bandwidth

448 GB/s

TDP

215 W

Max Power

215 W

Transistors

13.6 billion

Interconnect

50 GB/s NVLink

CUDA Cores

2,560

Spec Confidence

Official

Full Specifications

Compute Engine
CUDA Cores 2,560
Tensor Cores 320 (Turing)
Streaming Multiprocessors 40
Base Clock 1.60 GHz
Boost Clock 1.77 GHz
Chip Design
Transistors 13.6 billion
Die Size 545 mm²
Process Node 12nm
Memory
VRAM 8 GB
Memory Type GDDR6
Bandwidth 448 GB/s
Interface Width 256-bit
Memory Clock 14 GT/s
Interconnect & I/O
GPU-to-GPU NVLink
Interconnect Bandwidth 50 GB/s
Interconnect Links 1
PCIe 3.0 x16
Cache
L2 Cache 4 MB
Power & Thermal
TDP 215 W
Max Board Power 215 W
Power Connector 6 pin + 8 pin
Cooling Active
Enterprise Features
ECC Memory No
Sparsity No
Compute APIs CUDA, DirectX 12, Vulkan, OpenGL
Physical & Media
Card Length 266.74 mm
Width 2-slot
General
Form Factor PCIe
Architecture Turing
Process Node 12nm
Launch Year 2019

Datasheet & Resources

Data Provenance

Every figure traced to a source

Primary Source

Publisher
NVIDIA
Published
2019-07-02

Data Quality

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

Compare cloud providers offering on-demand GPU instances for AI training, inference, and HPC workloads.

Browse GPU Cloud Providers

Similar GPUs

Frequently Asked Questions

How many TFLOPS does the NVIDIA GeForce RTX 2070 Super have?

The NVIDIA GeForce RTX 2070 Super delivers 9.1 TFLOPS FP32 and 36 TFLOPS FP16 at peak. Flopper does not currently have a verified FP8 throughput figure for it.

What is the power consumption of the NVIDIA GeForce RTX 2070 Super?

The NVIDIA GeForce RTX 2070 Super has a TDP (Thermal Design Power) rating of 215 watts.

How much memory does the NVIDIA GeForce RTX 2070 Super have?

The NVIDIA GeForce RTX 2070 Super is equipped with 8 GB of VRAM with 448 GB/s of memory bandwidth.

What architecture is the NVIDIA GeForce RTX 2070 Super based on?

The NVIDIA GeForce RTX 2070 Super is based on the Turing architecture, launched in 2019.

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