NVIDIA CUDA Toolkit 7.5 on Amazon Linux
NVIDIA | 1.0Linux/Unix, Amazon Linux 2016.03 - 64-bit Amazon Machine Image (AMI)
where is a simple "Quick Start" guide?
This looks like a really great resource. But, getting the samples to run has been a hit and miss opportunity.
Try running the "nbody" sample, which looks great. Instead, there are a host of libraries it can't find:
>>> WARNING - libGL.so not found, refer to CUDA Getting Started Guide for how to find and install them. <<<
>>> WARNING - libGLU.so not found, refer to CUDA Getting Started Guide for how to find and install them. <<<
>>> WARNING - libX11.so not found, refer to CUDA Getting Started Guide for how to find and install them. <<<
- Leave a Comment |
- Mark review as helpful
Disappointing
Pretty lackluster. For some reason, the whole thing is slow and nothing is installed by default, so I had to wait for scipy to compile while it charged me for time and of course none of the dependencies were installed. Even installing Keras takes forever, and then when I tried to go run my model, it's slower than just doing it on my home computer. None of the CUDA variables were even setup and I just lost my patience so I terminated it. I'm going to be on the look-out for another image.
Great AMI for learning CUDA programming
I only have a MacBook without a NVIDIA GPU, so I can't run CUDA on my laptop. This AMI turns out to be great for testing simple CUDA programs. No need to worry about software and hardware issues.
Kind of ok
It's really nice to have an AMI with late-ish drivers
What I really need is docker support. Nvidia-docker doesn't install on this AMI. Searching the yum repos for the two missing dependencies (docker-engine & systemd) yields nothing.
nvidia-smi doesn't work!
After fussing around with a few of the Deep Learning AMI's that were tedious to get the GPU backend running I thought I'd go to the source and load the NVIDIA AMI directly supported by the company that builds the hardware and would presumably know how to load the drivers for it....
WRONG!
After firing up a new p2.xlarge instance the most basic GPU command results in:
$: nvidia-smi
modprobe: FATAL: Module nvidia not found.
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.
REDICULOUS!
Come on AWS and NVIDIA...can you at least make your namebranded AMI work with your own code?