Folding@Home, from a Bencher’s Perspective

Like most Team OCF Benchmarkers, I have amassed some fairly high end hardware in my quest for HWBot points and glory.  A quick look into the Folding@Home Team section of the Overclockers.com Forums points out that the exact same hardware is being put to work for glory and science, simulating how various proteins fold into the shapes they do.  This is highly important, as the shape of a given protein dictates it’s function, and even a slight variation from normal can cause large problems.  Mad Cow Disease, for instance, is caused by a mis-folded protein!

Some FAH Background:

Simulating how a protein folds requires a tremendous amount of computational power, there are hundreds of thousands to millions of atoms in a protein, and every single one of them and their interactions with all their neighbors and environment have to be simulated.

Up until ten years ago this was done almost exclusively on large scale supercomputers, and even then it took quite a while to get results back.  Plus, supercomputer time is not cheap!  In 2000 Folding@Home started working on a new method of simulating how proteins worked.  Rather then run on a singular computer, they split the protein simulation into a lot of little chunks and sent them off to be folded on home computers.  Individually the CPU of a home computer doesn’t hold a candle to a supercomputer, but if you get thousands and thousands of home computers folding proteins it is a very different story!  Now the Folding@Home users are part of one of the highest power supercomputers in the world, each home computer functioning as a node in the large whole.  Since that time Folding@Home has branched out to run on GPUs as well as CPUs, this change allowed access to the raw computational horsepower of modern GPUs, which is truly staggering.

One of the smartest things that the FAH project did was set up a team structure, and award points for completing “Work Units” and submitting them back to FAH HQ.  The overclockers.com team (team 32) is currently the fourth highest ranked team in the world, and is sneaking up on third place as we speak!

Onward to today!

I ran Folding@Home for quite some time, up until their new “A3” core came out and made my dual core CPUs fairly useless, FAH (Folding@Home) puts a lot of emphasis on getting the data back quickly, and a dual core cpu simply cannot do that, it lacks the power.  About a month ago however I stepped up to the Big Leagues in the Benching world and bought a Rampage III Extreme motherboard and an Intel Core i7 980x CPU to go with it.  Six cores plus hyperthreading!  It did not take long for me to realize that this CPU could fold proteins at a rather high rate, and make my previously dead slow quest for one million FAH credits a reality.

It did not take long for me to realize that the worlds of Benchmarking and FAH were quite different, despite both focusing on convincing your CPU or GPU to work as fast as possible.  This brings us to…

Stability!

The largest difference comes from what is considered “stable”.  I use quotation marks because “stable” means something different to everybody.  To a bencher, a setup is stable if it can complete the benchmark and gives you time to take and save a screenshot of the score.  If it crashes after that it is vaguely annoying, but not really an issue.  With this 980x cranked up as high as it will go and cooled by dry ice, that means it needs to sustain one core running a program at high speed for between six seconds and about four minutes, not very long at all, and only loading one core means it is much easier to complete the benchmark then it would be if it used more cores.  Not very many benchmarks use all the cores, and the most brutal of all is WPrime 1024m. It calculates the square root of the first billion numbers (ok, 1.024 billion, but that sounds sort of funny).  This takes a 980x between one and a half minutes and two and a half minutes in flat out subzero benching mode.  Still not very long, though definitely a large load!

To a normal user, stable means one of two things, either it means that the computer does not crash during normal use, or it means that it can run a stability test like Prime95 or Intel Burn Test for a certain number of hours.  Already things are different from the benching world, even the easiest stability test will quickly be failed by a cpu that has been cranked up to top benching levels!

To a FAH user stable has a much different meaning.  To be stable, the cpu has to be able to run a very heavy load for days, weeks, months, or years without crashing.  Even more importantly, it can’t have any small errors either, as a small error in computation of protein folding may not cause a crash, but it can result in the computer reporting that a given protein is shaped like a pretzel when it is really shaped like a dog bone.  Not only is this wrong, but if the computers that check the results don’t catch it, that protein will be passed off to the next computer in line still shaped like a pretzel.  The entire chain of work has to be thrown out and started over!  Now the FAH folks know this, and have quite a few safeguards in place, but I’m sure things still slip through.

To me, the concept of FAH Stable took some getting used to!  I got my CPU running at a speed I would be happy with for daily use, run some stability testing without any issues, started up FAH, and then moved on to browsing and posting on OCF.  Half an hour later, in the middle of a long post, my long involved description of how to solve some problem or another evaporated in a blue screen full of text and a memory dump!  I might have cursed.  This happened three times before I finally found functional settings!  Now however, I am one and a half hours away from hitting the one million credit mark!

It can take some doing to find a high speed overclock that is still capable of running Folding@Home (or other such programs, OCF also has SETI@home and Rosetta@home teams, those searching for extraterrestrial life and another way of looking at the problem of how proteins fold respectively), but once you do, you can sit back and bask in the feeling of doing something useful for humanity.  It’ll keep your house nice and warm in the winter, too!

—Bobnova, of the OCF Benching Team and FAH/Rosetta Teams.