philip165
Apparently I broke some rules the last time I tried this...so lets try not to make the mods angry again
Let me be clear, JUST BECAUSE THIS POST REQUIRES MENTIONING FUNDRAISING IN ORDER TO BE COMPLETE, DOES NOT MEAN THAT I AM ASKING FOR DONATIONS. I am merely trying to share a success story of human altruism.
So, let me begin with the backstory:
Folding@Home is a protein-folding project, hosted through Stanford, that aims to simulate the "folding" of proteins related to research into treatments/cures for diseases like cancer, alzheimers, zika, and ebola. Proteins have tertiary structure that is very difficult to directly measure. So, molecular dynamics simulations start with the "flat" chemical structure of the protein, and allow it to "fold" into its normal state, which involves all of the electro-static interactions between individual atoms and their electrons.
As you can imagine, this type of simulation on large proteins can be very computing-intensive. Because dedicated supercomputers are extremely expensive, the folding@home project utilizes a technique called distributed computing, in which everyday people donat...ugh...download software onto their own personal laptops/computers that allows the laptop to be used to run small parts of these folding simulations when it is not being used for other things. In this way, the folding@home distributed supercomputer has been running successfully for well over 10 years.
I am particularly passionate about this because I have lost several important people to the diseases on which folding@home focuses its efforts. I have been donat....err....running the folding@home client on my laptops and pcs for more than a decade.
The problem with this approach is that laptops and pcs are not very computing efficient. That is, they burn through way more power than they should for the amount of processing power they output. This is because they are not built to be single-application, and so there is a trade-off between helping out the medical community, and minimizing the carbon footprint of the project. As an example, 50,000 personal laptops running at 40W uses up a massive 2 million watts, and only outputting 150 TeraFLOPs (Floating Point Operations per Second).
Our team's project was simple, direct resources towards the folding@home project in the most efficient way possible. I was able to start the effort by...err...giving...some of my old gaming PC equipment to the cause (an Nvidia GTX 970 GPU card, a hard drive, a large computer case, and a 1000W power supply), and started a GoFundMe campaign. Our team of...donors....provided the $310 necessary to buy a refurbished motherboard (capable of future upgrades), a basic pentium G4400 processor (the processor only directs simulation packets to the GPU) and a minimal amount of RAM (4GB).
This machine currently uses (at maximum) 350W of power, and is able to consistently provide 3.5 TeraFLOPs of computing power. This means that the specific processing-to-power ratio is well over 100x that of the laptop equivalent. Furthermore, the machine is expandable to upwards of 8x its current processing power, for roughly 2.5x the power consumption. It would be able to provide over 1000 laptops-worth of computing power for less than 1/10th the carbon footprint, all for less than the cost of a typical mid-level university student laptop.
The rig went live 1 week ago, and since then our team has shot up to within the top 3% of folding teams, using (at worst) about 8% of the power consumption of an equivalent laptop group.
I would post a link to our folding@home stats page, which would provide you all with actual performance numbers, but I'm not sure if the mods would like that....so instead I'll post some pictures of the rig instead.
Anyway, I'm really happy with how this worked out. It gives me hope for humanity, and I hope our success will inspire you all. If this thread survives, I'll try to keep you guys updated with any performance milestones or future upgrades.
Thanks for reading!
View media item 240000External case
View media item 240001Internals of an early configuration, Note that the GTX260 (red card on the bottom) was not used in the final configuration, because there are no Nvidia drivers that support both a GTX 970 and a GTX 260, and the 970 performs better.
View media item 239998First life, the machine was later installed with Linux Mint as the OS, which seems to be resource friendly, and free to boot!
Let me be clear, JUST BECAUSE THIS POST REQUIRES MENTIONING FUNDRAISING IN ORDER TO BE COMPLETE, DOES NOT MEAN THAT I AM ASKING FOR DONATIONS. I am merely trying to share a success story of human altruism.
So, let me begin with the backstory:
Folding@Home is a protein-folding project, hosted through Stanford, that aims to simulate the "folding" of proteins related to research into treatments/cures for diseases like cancer, alzheimers, zika, and ebola. Proteins have tertiary structure that is very difficult to directly measure. So, molecular dynamics simulations start with the "flat" chemical structure of the protein, and allow it to "fold" into its normal state, which involves all of the electro-static interactions between individual atoms and their electrons.
As you can imagine, this type of simulation on large proteins can be very computing-intensive. Because dedicated supercomputers are extremely expensive, the folding@home project utilizes a technique called distributed computing, in which everyday people donat...ugh...download software onto their own personal laptops/computers that allows the laptop to be used to run small parts of these folding simulations when it is not being used for other things. In this way, the folding@home distributed supercomputer has been running successfully for well over 10 years.
I am particularly passionate about this because I have lost several important people to the diseases on which folding@home focuses its efforts. I have been donat....err....running the folding@home client on my laptops and pcs for more than a decade.
The problem with this approach is that laptops and pcs are not very computing efficient. That is, they burn through way more power than they should for the amount of processing power they output. This is because they are not built to be single-application, and so there is a trade-off between helping out the medical community, and minimizing the carbon footprint of the project. As an example, 50,000 personal laptops running at 40W uses up a massive 2 million watts, and only outputting 150 TeraFLOPs (Floating Point Operations per Second).
Our team's project was simple, direct resources towards the folding@home project in the most efficient way possible. I was able to start the effort by...err...giving...some of my old gaming PC equipment to the cause (an Nvidia GTX 970 GPU card, a hard drive, a large computer case, and a 1000W power supply), and started a GoFundMe campaign. Our team of...donors....provided the $310 necessary to buy a refurbished motherboard (capable of future upgrades), a basic pentium G4400 processor (the processor only directs simulation packets to the GPU) and a minimal amount of RAM (4GB).
This machine currently uses (at maximum) 350W of power, and is able to consistently provide 3.5 TeraFLOPs of computing power. This means that the specific processing-to-power ratio is well over 100x that of the laptop equivalent. Furthermore, the machine is expandable to upwards of 8x its current processing power, for roughly 2.5x the power consumption. It would be able to provide over 1000 laptops-worth of computing power for less than 1/10th the carbon footprint, all for less than the cost of a typical mid-level university student laptop.
The rig went live 1 week ago, and since then our team has shot up to within the top 3% of folding teams, using (at worst) about 8% of the power consumption of an equivalent laptop group.
I would post a link to our folding@home stats page, which would provide you all with actual performance numbers, but I'm not sure if the mods would like that....so instead I'll post some pictures of the rig instead.
Anyway, I'm really happy with how this worked out. It gives me hope for humanity, and I hope our success will inspire you all. If this thread survives, I'll try to keep you guys updated with any performance milestones or future upgrades.
Thanks for reading!
View media item 240000External case
View media item 240001Internals of an early configuration, Note that the GTX260 (red card on the bottom) was not used in the final configuration, because there are no Nvidia drivers that support both a GTX 970 and a GTX 260, and the 970 performs better.
View media item 239998First life, the machine was later installed with Linux Mint as the OS, which seems to be resource friendly, and free to boot!
