Advanced search

Message boards : Number crunching : Quarantine@Home - GPU project

Author Message
Aurum
Avatar
Send message
Joined: 12 Jul 17
Posts: 259
Credit: 9,908,781,269
RAC: 3,516,839
Level
Tyr
Scientific publications
watwatwat
Message 54727 - Posted: 17 May 2020 | 18:42:02 UTC
Last modified: 17 May 2020 | 18:46:25 UTC

Keep your GPUs warm with Quarantine@Home: https://quarantine.infino.me/
For first time nvidia-docker users these 2 commands do nothing, skip to install. Make sure all nvidia-dockers are removed first:

sudo docker volume ls -q -f driver=nvidia-docker | xargs -r -I{} -n1 docker ps -q -a -f volume={} | xargs -r docker rm -f
sudo apt-get purge nvidia-docker

Install nvidia-docker 2.0. First install the repository for your distribution:
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/ubuntu18.04/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
sudo apt-get update
sudo apt-get install nvidia-docker2 -y
sudo pkill -SIGHUP dockerd
sudo docker run --runtime=nvidia --rm nvidia/cuda nvidia-smi
git clone https://github.com/cjmielke/quarantineAtHome && cd quarantineAtHome

Run Quarantine@Home:
sudo docker build -t quarantinegpu -f Dockerfile.gpu . && sudo nvidia-docker run -it -e ME=Aurum quarantinegpu

biodoc
Send message
Joined: 26 Aug 08
Posts: 167
Credit: 1,669,740,969
RAC: 957,617
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 54738 - Posted: 19 May 2020 | 14:51:00 UTC

Thanks for the guide Aurum. It worked perfectly for me on Linux mint 19.3.

10esseeTony
Send message
Joined: 4 Feb 15
Posts: 6
Credit: 1,042,752,088
RAC: 45
Level
Met
Scientific publications
watwatwatwatwatwat
Message 54743 - Posted: 20 May 2020 | 1:41:27 UTC

Thanks! Gold man!

Gunde
Send message
Joined: 6 Jan 15
Posts: 38
Credit: 5,670,234,369
RAC: 3,494,119
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 54788 - Posted: 20 May 2020 | 23:02:52 UTC
Last modified: 20 May 2020 | 23:03:11 UTC

Thanks for guide and interesting gpu project to run autodock in a docker. Sadly i could add docker2 on 20.04LTS.


ME=Aurum
default is userid
ME=anonymous

Greg W.
Send message
Joined: 2 Apr 20
Posts: 3
Credit: 19,294,680
RAC: 27,849
Level
Pro
Scientific publications
wat
Message 54789 - Posted: 21 May 2020 | 3:16:51 UTC

Is there a blacklist/whitelist of GPUs usable for quarantine@home, as there is for F@H? Are there tasks that can only run on a CPU, or would it be more efficient to use only the GPU for quarantine@home while reserving the CPU for something else?

I have quarantine@home running on a virtual instance of Debian Buster right now, CPU only, just to try it out. When I get back home, I'll put together a machine with a GPU; we'll see how it works out.

Are you aware of WCG's Open Pandemic project, and the Ibercivis project? Both seem to be doing similar work. The three seem to compliment one another, as well as Rosetta, F@H, and GPUGrid.

Jim1348
Send message
Joined: 28 Jul 12
Posts: 745
Credit: 1,498,076,391
RAC: 535,067
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 54796 - Posted: 21 May 2020 | 12:16:14 UTC - in response to Message 54789.
Last modified: 21 May 2020 | 12:30:37 UTC

Are there tasks that can only run on a CPU, or would it be more efficient to use only the GPU for quarantine@home while reserving the CPU for something else?

It seems that the same tasks run on either a CPU or GPU.

As for the time, let's put it this way: the average Windows or Linux CPU is in the range of 200 to 400 seconds. My GTX 1070 under Ubuntu 18.04.4 is averaging 8.5620 seconds. (It helps to reserve two CPU cores though.)

EDIT: But it randomly stops running and I have to restart it every few hours or even minutes. I leave my machines on 24/7, so I think they are dropping the connection.

Greg W.
Send message
Joined: 2 Apr 20
Posts: 3
Credit: 19,294,680
RAC: 27,849
Level
Pro
Scientific publications
wat
Message 54876 - Posted: 22 May 2020 | 12:53:21 UTC - in response to Message 54796.

Thanks. I've been getting about 160s/task. I have 2 instances of Q@H running, both in that same VM, to which I assigned 2 cores. (The host laptop has 4.)

So, from your experience on both GPUs and CPUs, it looks like the project is parallel-able. Or "Embarassingly parallel?" (Please excuse my ignorance of the correct terminology; still new to this.) So it'll run better on a GPU. It works on CPU, but that should be reserved for projects that are not parallel-able, or that simply don't have a GPU option, such as Rosetta, WCG, etc.

Sound about right? I'll try it out on some GPUs when I get home.

I'm also doing some due diligence. It's common sense to check out a new project to make sure it's not really a bitcoin mining scam, or that the code isn't malware. So far it looks good.

Speedy
Send message
Joined: 19 Aug 07
Posts: 34
Credit: 9,747,703
RAC: 12,771
Level
Ser
Scientific publications
watwatwatwatwatwatwat
Message 55055 - Posted: 16 Jun 2020 | 7:18:15 UTC

I will be contributing when they get a Windows client built. Hopefully there is one in the works but I am unsure wishful thinking on my part

Post to thread

Message boards : Number crunching : Quarantine@Home - GPU project