FP32 8.19 IS DERIVED, NOT PUBLISHED. NVIDIA never printed a TFLOPS figure for this card, in the product page's Full Specs panel, in the 2017-10-26 announcement blog, or in the 2017-11-02 GeForce News article. 8.186 is its own published 2,432 CUDA cores times two times its own published 1,683 MHz boost clock, which is the method its GP104 whitepaper states and uses to arrive at the GTX 1080's printed 8873 GFLOPs. spec_confidence is vendor_claimed on that basis: the board specs are NVIDIA's, the throughput figure is arithmetic on them. FP32 IS ALSO THE ONLY PRECISION ROW, for the reason recorded on the GeForce GTX 1080. REFUSED: FP16, FP64 and INT8; sm_count is derived and l2_cache_mb is refused outright, because the whitepaper's 64 ROPs and 2048 KB L2 are full-GP104 figures and this is a harvested die that NVIDIA never characterised. THE SOURCE IS AN ARCHIVED NVIDIA PAGE because the live URL now redirects to the current GeForce landing page; captures from 2017 are server-rendered so the Full Specs panel survives intact. Anyone re-scraping it should know the 10-series template leaves a stray hidden row reading "336.5 Memory Bandwidth (GB/sec)" after the correct 256; 336.5 is the GTX TITAN X's figure, not this card's.

NVIDIA GeForce GTX 1070 Ti 8GB

Pascal PCIe 2017 16nm Spec confidence: Vendor claimed
FP32
8.2
TFLOPS
Memory
8 GB
GDDR5
Bandwidth
256 GB/s
memory
TDP
180 W
45.5 FP32 dense TFLOPS/kW of TDP
Download datasheet

Overview

The GeForce GTX 1070 Ti is the card NVIDIA slotted between the 1070 and the 1080 in late 2017, and on paper it is much closer to the 1080: 2,432 CUDA cores against 2,560, the same 1,683 MHz boost as the 1070, and the same 180 W board power as the 1080, for about 8.19 TFLOPS FP32. What it does not get is the 1080's GDDR5X, so its 8 GB runs at 8 Gbps on the same 256-bit bus for 256 GB/s rather than 320. For anything memory-bound that 20 per cent bandwidth gap is the difference between the two cards, not the small step in shader count. Like every Pascal GeForce part it has no Tensor Cores, no structured sparsity and no NVIDIA-published half-precision rate, so FP32 is the whole story.

Performance

Peak theoretical throughput by precision type

PrecisionPeak
FP64
No verified data available
FP32
32-bit floating point
8.2TFLOPS
TF32
No verified data available
BF16
No verified data available
FP16
No verified data available
FP8
No verified data available
FP6
No verified data available
FP4
No verified data available
INT8
No verified data available

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.

Specifications

Architecture

Pascal

Form Factor

PCIe

Launch Year

2017

Process Node

16nm

Memory

8 GB GDDR5

Bandwidth

256 GB/s

TDP

180 W

Max power

180 W

Transistors

7.2 billion

CUDA Cores

2,432

Spec Confidence

Vendor claimed

Full Specifications

Compute Engine
CUDA Cores 2,432
Streaming Multiprocessors 19
Base Clock 1.61 GHz
Boost Clock 1.68 GHz
Chip Design
Transistors 7.2 billion
Die Size 314 mm²
Process Node 16nm
Memory
Memory 8 GB
Memory Type GDDR5
Bandwidth 256 GB/s
Interface Width 256-bit
Memory Clock 8 GT/s
Interconnect & I/O
PCIe 3.0 x16
Power & Thermal
TDP 180 W
Max power 180 W
Power Connector 8-pin
Cooling Active
Enterprise Features
ECC Memory No
Sparsity No
Compute APIs CUDA, DirectX 12, Vulkan, OpenGL
Physical & Media
Card Length 266.7 mm
Width 2-slot
General
Form Factor PCIe
Architecture Pascal
Process Node 16nm
Launch Year 2017

Datasheet & Resources

Data Provenance

Every figure traced to a source

Primary Source

Publisher
NVIDIA
Published
No verified data available

Data Quality

Spec confidence
Vendor claimed
Clock basis
Boost
Core precisions with figures
1 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

NVIDIA Titan Xp 12GB

PCIe · 2017
FP32: 12 TFLOPS
Compare vs Titan Xp

NVIDIA Tesla P40 24GB

PCIe · 2016
FP32: 12 TFLOPS
Compare vs Tesla P40

NVIDIA P100 SXM2 16GB

SXM · 2016
FP32: 11 TFLOPS
Compare vs P100

NVIDIA P100 PCIe 16GB

PCIe · 2016
FP32: 9.3 TFLOPS
Compare vs P100

Frequently Asked Questions

How many TFLOPS does the NVIDIA GeForce GTX 1070 Ti have?

The NVIDIA GeForce GTX 1070 Ti delivers 8.2 TFLOPS FP32 at peak. Flopper does not currently have verified FP16 and FP8 throughput figures for it.

What is the power consumption of the NVIDIA GeForce GTX 1070 Ti?

The NVIDIA GeForce GTX 1070 Ti has a TDP (Thermal Design Power) rating of 180 watts.

How much memory does the NVIDIA GeForce GTX 1070 Ti have?

The NVIDIA GeForce GTX 1070 Ti is equipped with 8 GB of memory with 256 GB/s of memory bandwidth.

What architecture is the NVIDIA GeForce GTX 1070 Ti based on?

The NVIDIA GeForce GTX 1070 Ti is based on the Pascal architecture, launched in 2017.

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