Jump to content
Search In
  • More options...
Find results that contain...
Find results in...


  • Content Count

  • Joined

  • Last visited


About Macaw2000

  • Title

Contact Methods

  • Discord

Recent Profile Visitors

383 profile views
  1. 335,000 points in a week was my best week by far. Scaled the cluster way down for now though. Fun times.
  2. AWS.. spot market, containers based on their highest value GPU instances. Deploys anywhere in the world depending on where it is the cheapest.
  3. Some observations from folding in the cloud... As many of you know, I fold entirely in the cloud. I've evolved my software to scour the planet for the cheapest GPU instances and grab them on the spot market. When the price is cheap, my software will spin up to 10 GPU containers on g4dn.xlarge and commence folding. What I've discovered this week is how dynamic the market can be. It's been a nice exercise in Infrasctructure As Code, software development, DevOps, and git. Mumbai is a complete tease. Crazy cheap GPU instances that last for a couple hours then get interrupted. There must be some big customer workloads that run a few times a day and interrupt the spot market. Canada has been steady eddy. I suspect there is more capacity than demand in Canada. It's not the very cheapest but was never interrupted. West coast was a mixed bag. Ran about 85% of the time uninterrupted and at a good rate. After the contest, I'm dedicating most of my GPU budget back to my crazy moonshot project. Real-time deep fakes at high quality. Would anyone be interested in a livestream of how to cloud fold or do deep fakes or other interesting cloud stuff? Gauging interest. My daughter is a professional twitch streamer so I'm sure she can get me started.
  4. Hah thanks! just noticed top 10. Must have just happened.
  5. Hah! I do own a little AMZN but my folding is an exercise in extreme cheapness. I've programmed my cluster to go look for the cheapest GPU instances and grab them on the spot market. When the contest started Mumbai India (of all places) was having a fire sale on GPU spot instances so my cluster spun up bigtime there. But, then another customer came along and needed GPU (during their daytime hours) and my cluster was nuked. Hours later, the compute costs got cheap again and it spun up. Seen in the chart below: Folding also took place in Oregon, California, Ohio, Virginia, Toronto, Paris, and Stockholm but Mumbai was the real star for me today.
  6. Start your engines! @Den-Fi @GOTSpectrum @rodarkone
  7. @rodarkone Here's my battle station. Very little heat generated
  8. Thanks for organizing @GOTSpectrum. Looking forward to learning and trying new things in this event.
  9. Nice! All my containers run on ECS on spot instances. I could link to some S3 bucket with the config but haven't messed with that.
  10. Seems like you could just edit the config.xml and have only a slot for the GPU in there... <slot id='0' type='GPU'/> And have it added to your Dockerfile ADD config.xml /etc/fahclient/config.xml
  11. Very nice! When is the June event. Would be fun to race you
  12. If Docker is interesting... it's a far cleaner solution than an AMI. Nice example of the Docker config here: https://github.com/raykrueger/FoldingAtHomeContainer Edit: Docker cleaner, AMI much easier barrier to entry and works well!
  13. Oh and the results can be quite good. Here's the contest we had last month and I ran entirely on AWS