Folding@home QMD FAQ

Table of Contents

The Folding@home (FAH) developers (primarily Young Min Rhee) have been working to incorporate a new quantum molecular dynamics method into Folding@home.

Note: The principle investigator for QMD work units, Young Min Rhee, graduated in 2006, so this area of research is temporarily on hold. QMD work units are not available at this time.

As a participant of Folding@home, is there anything I need to do?

Yes, we are only allowing QMD work units on machines which have "opted in." QMD has demanding CPU and memory requirements and so we will not assign these WUs to just any CPU. See details below.

How can I get QMD Work units?

You must have two settings to get QMD's. You must have both the "BigWU" flag (allowing for large work units) and the advanced methods (-advmethods) flag on.

How can I prevent my machine from getting QMD Work units?

If you want to prevent getting QMD's, just make sure you don't have both the Big WU flag and the advanced methods flags on. If you still want big WUs, but not QMD's, just make sure not to have the advanced methods flags on.

Could you give some examples?

Sure, here are some possible configurations:

• 2 QMDs? If so, run both clients with bigWU+adv
• 1 QMD + 1 big WU? bigWU+adv and bigWU alone (no adv)
• 2 big WUs (no QMD)? both clients with bigWU alone (no adv)
• 1 QMD + 1 non big WU? bigWU+adv + normal FAH .

What is peculiar about QMD core?

For other molecular dynamics simulation packages, interactions between atoms are described by an empirical potential energy (“force field approach”), which is an analytical function of atomic coordinates. In QMD core, there is no force field – atomic interaction is calculated using quantum chemical method or by solving Schrödinger equation.

What program are you using for QMD core?

This core is based on the CPMD program. The program was modified for Folding@home supercluster environment.

Why include QMD? What do you gain?

For the scientists using Folding@home, the results will be interesting. When carefully conducted, quantum chemical calculation gives a very reliable potential energy surface. Also, it can naturally incorporate multi-body interactions. In force field approach, such interactions are usually accounted through polarizability.

If it can be very reliable, why didn’t you use the method from the first time?

The biggest problem of quantum chemical method is its speed and memory dependence. Simply, accurate methods usually take longer time and more memory. Such computational capability came in handy only very recently.

If it's slow, will the participants get less points?

No, the points in the stats will still be benchmarked by the CPU time, so your stats won't be affected either.

How much memory does it use?

It will have strong molecular system dependency, but for many systems it will be over 512 MB. We are planning to simulate some systems that will require up to 1GB of main memory (or even more).

What about SIMD (SSE/SSE2/3D-Now) support?

QMD core uses double precision only. Therefore, only SSE2 will be useful. Currently, SSE2 on Intel CPUs are supported.

What about AMD support?

We are currently using the Intel Fortran compiler and libraries. This software intentially runs slowly on AMD CPUs (SSE2 is not supported). While there have been hacks around this issue, such modifications violate Intel's EULA. At FAH, we follow software EULAs and are thus bound to this limitation imposed by Intel.

That's not fair to AMD chips!

Yes, it isn't fair and AMD is taking actions against this. We hope that things will change, but our hands are tied. We could give these WUs to AMD machines, but they would have a much worse points per day (less than half of what one gets on Intel hardware) and would not perform well in general.

What about new chips such as the PentiumM or Pentium 4M?

New CPU id's will be supported once the CPUID code is updated in the v6 client.

Why not just use a compiler that legally supports SSE2 on both Intel and AMD CPUs?

Currently, the best compiler (speed wise) for QMD is the Intel compiler (by a factor of 2x). The AMD compiler is pretty good, but is limited (license-wise) to AMD chips; however, it is not as fast or as optimized as the Intel compiler. We have not found a compiler which has no restrictions (runs on both Intel and AMD) which close in speed to that of the Intel or AMD compilers.

What about Mac OS X support?

We are considering adding OS X support. We have not found the compiler/library support sufficient so far. At this point, there is not much for us to do, but to wait for new compiler and library releases to see if this improves.

What about the current scientific codes? What will happen to them?

Other cores will continue to perform well for most calculations. In fact, the system of interest with QMD core will be very different from other cores (in terms of size).

I'd like to learn more about CPMD. Where should I look?

First, check out the CPMD web page. Also, a general theory of this program was firstly described in: R. Car and M. Parrinello, Phys. Rev. Lett. 55, 2471-2474 (1985).

For More Information, Please See:

• FAH FAQ
• CPMD web page.
• Folding Community Forum
• FAH WIKI: FAH CORES

Last Updated on January 24, 2008, at 06:24 AM