Comparing the performance/suitability of GeForce GTX Titan Z vs GeForce GTX TITAN vs TESLA M60 vs TESLA M40

Accelerated Computing CUDA CUDA Setup and Installation
1

I’m down to comparing 4 cards that I intend to buy on the second hand market (eBay) for Tensor Flow programming.

We’re a group of tinkerers who are going to use this platform for understanding Tensor Flow and CUDA for high performance computing.

We’d like to understand the value/price proposition of the following 4 cards. It seems to us that the M40 > Titan > Titan Z > M60 – does that sound about right?

TESLA M40
Second hand price: $450
Architecture: Maxwell
RAM: 16GB
GFLOPS (FMA): 60 double precision
Threads: 3072

GeForce GTX TITAN
Second hand price: $400
Architecture: Kepler
RAM: 6GB
GFLOPS: 1500 double precision
Threads: Couldn’t figure this out

GeForce GTX Titan Z
Second hand price: $850
Architecture: Kepler
RAM: 12GB
GFLOPS: 2700 double precision
Threads: Couldn’t figure this out

TESLA M60
Second hand price: $3,200
Architecture: Maxwell
RAM: 16GB
GFLOPS (FMA): 300 double precision
Threads: 4096

2

Factual correction: Titan and Titan Z are sm_35 devices, i.e. Kepler architecture. There are two kinds of Titan X, one based on Maxwell, the other based on Pascal (the latter is referred to as Titan XP by some users to avoid the confusion created by NVIDIA’s naming).

3

@njuffa – I mean seriously – could Nvidia have created a more confusing branding scheme – lol