4 petaflop, what does that mean?

[chengbin]

I saw on the stats page that F@H has hit 4 petaflops. What does that exactly mean? Does that mean F@H has completed that many flops? Or does that mean it has 4 petaflops of computation ability every (day, minute, second?)

[cbuchner1]

It means 4.000.000.000.000.000 floating point operations per second actively contribute to folding protein molecules every single second.

[chengbin]

At this rate of speed increase, about how long will it take to finish all the work?

[Tohya]

Its the number of floating-point operations per second.

A petaflop is 10 to the 15th power or 1,000,000,000,000,000


There is an unlimited amount of work to do.

[chengbin]

How come it is unlimited?

[Foxery]

How come it is unlimited?

You may wish to read up on some of the project's basic FAQs to better understand what it is that we are processing here. There are thousands of unique proteins in the human body, and each one takes several months to compute. The project isn't truly "unlimited," but the end is nowhere in sight.

[alancabler]

There are over 1,000 proteins just in human spit. Science has only just begun to understand proteins and protein folding.
We do not yet even know the extent of that which we do not know...

[sdack]

Congratulations!!

... the one & only meaning of it.

[smASHer88]

It means we're getting closer to 5 Petaflops!

Nah ya see we will probably never understand how 'far' we are into this project and how each extra Petaflop helps out, but we can just assume, the more power the better, and the more 'active' folders the better.

[ElectricVehicle]

One measure of the progress or Scientific output of Folding at Home are the published Research Papers. Each one of these papers represents a huge amount of work. It can take months or years of research for a single paper. The revolution, is that such simulation is now possible within a single human lifetime and even within the couple year time frame of a graduate student's study at a University (Say, for example, Stanford)!

You can't experimentally watch a physical protein fold. It's way, way, way too fast and way, way, way too small. You can try and simulate it on a computer. Then you validate those results with what little you can observe experimentally about protein folding and behavior.

Interestingly, much of the GPU1 folding was not usable scientifically but it was hugely successful in finding out what wouldn't work and developing the GPU2 client which is making large contributions to the science now and even more in the future. So, it's a long haul from the computation to the scientific research papers of months or years.

For short term, instant gratification, just follow the Points! The points themselves are also very useful short term to help optimize computing clients to produce more scientific results over the long term.

[chengbin]

How often does the client page updates?

We got from 3 petaflop to 4 petaflop in 1 month! I hope we can keep this up! I'm madly folding right now as I got 2 computers (one P4 2.8GHz, one with Q6600 and 8800GT folding). I hope more and more people with GPUs can contribute.

BTW, sorry to whoever makes the work units at stanford for giving you so much work load making us work units.

Also, I hope they can make SMP on CPU easier to install. I think the installation process throw people off and scares them away (me included). I just fold with 1 of my 4 cores. I can contribute and not drag my computer down too much.

[Beberg]

Sadly there is not a PFLOPS to papers conversion ratio. So it just means more horsepower to make scientific progress with hopefully less waiting around.

[anandhanju]

I may be jumping the gun here, but have we reached a stage where turning in results happens faster than assimilating, summarizing and drawing conclusions out of them (i.e., research papers)? I understand that except for cruching and completing work, there isn't much that clients can do in terms of results-to-paper conversions, but if we ever run into a scenario wherein our horsepower exceeds our manpower, does FAH plan to make the raw results available to like minded institutions (non-profit, open domain) for their usage so that they can analyse results while you folks continue to live on the bleeding edge? Just a thought.

[John Naylor]

I just fold with 1 of my 4 cores. I can contribute and not drag my computer down too much.

You could fold with multiple uniprocessor clients... it's not hard Just give each one its own folder and when configuring make sure each one has a different machine ID.