NVIDIA
GPUDesktopReleased Aug 2, 2016

NVIDIA TITAN X (Pascal)

12 GB GDDR5X · 384-bit · 3,584 cores

The TITAN X (Pascal) was NVIDIA's flagship graphics card of 2016, sold directly by NVIDIA rather than through board partners and positioned between the GeForce line and the Quadro workstation range. It was the first consumer card to use the large GP102 chip, and its fast INT8 instructions made it a popular choice for deep-learning inference before dedicated Tensor Cores arrived. It dropped the GTX prefix of the similarly named 2015 Maxwell-based GTX TITAN X, and was followed by the GTX 1080 Ti and TITAN Xp.

General

Manufacturer
NVIDIA
Model
TITAN X (Pascal)
Segment
Desktop
GPU chip
GP102-400
Architecture
Pascal
Release date
Aug 2, 2016
Price

Memory

Memory size
12 GB
Memory type
GDDR5X
Bus width
384-bit
Bandwidth
480 GB/s
Memory clock
1,250 MHz

Clocks & silicon

GPU clock
1,417 MHz
Boost clock
1,531 MHz
Shader units
3,584 CUDA Cores
Tensor / AI cores
—
RT cores
—
Transistors
12 billion
Technology (nm)
16 nm
TSMC 16 nm FinFET
TDP / TBP
250 W

Interface & APIs

Bus interface
PCIe 3.0 x16
SLI / multi-GPU
SLI (2-way, HB bridge)
DirectX
12 (12_1)
OpenGL
4.6
Vulkan
1.3
Other APIs
CUDA Compute Capability 6.1, OpenCL 3.0

AI support

Precisions
FP8FP16BF16INT8INT4TF32FP64
AI accelerator
None (no Tensor Cores / RT Cores)
Peak AI TOPS
44 TOPS
Notes
No Tensor Cores and no RT Cores. NVIDIA quoted 44 INT8 TOPS at launch, via the DP4A/DP2A instructions of CUDA compute capability 6.1 and without sparsity, which Pascal does not support. FP16 is supported but runs at 1/64 rate on consumer Pascal (GP102/GP104/GP106/GP107/GP108); FP64 at 1/32. No BF16, TF32 or FP8.
All NVIDIA GPUs →

Desktop gpus from NVIDIA announced around the same time as the NVIDIA TITAN X (Pascal). Add any of them to your compare list to see the specs side by side.