GPU3 reaches internal testing phase

By , the in Folding@Home Project - No comment
GPU3 reaches internal testing phase
In a post on the official forum, Vijay Pande leaked that the GPU3 client (aka Core 15) has entered the internal testing phase. He estimates that the client might be released soon, in a matter of weeks from now.

However, the schedule might be delayed if major issues arise in the testing phase … hence the release date can’t be guessed accurately, but our first estimates weren’t that far from reality.

Statistics system outage

By , the in Folding@Home Project - No comment
Update 8H UTC (11/24/09) :
The point recredit has started, the missing point from the 11/12th november will be fully given back soon.

Update 17H UTC (11/12/09) :
The stats server is back online. Missing credits will be entered manually, but it might take a couple of days, to make sure nothing went missing.

Original news 15H10 UTC (11/12/09) :
The statistics generation process has been frozen for the last 24 hours. This outage also affects the generation of the Server Status page.

The PandeGroup team have been notified, and are currently looking at the issue. We'll keep you informed as soon as we get more information.

In the meantime, don't worry if the stats don't show your uploaded WUs; everything is recorded, and your points should show up when the stats system catches the backlog.

The "psummary" file is now bonus-compatible!

By , the in Folding@Home Project - No comment
The "psummary" file is now bonus-compatible!
The K(ay) factor! The "psummary" file is used by FahMon and other third-party monitoring software to determine the points value of a protein, deadlines, etc. It has now had its structure changed slightly.

A column called "Kfactor" has made an appearance. This is a factor that - depending on the WU - adjusts the bonus on "BigAdv" units using the new method of points calculation. We await the relevant modifications to the various third party monitoring applications that will allow us to see this info, without manually opening the psummary file in a text editor.

The making of Folding@Home videos

By , the in Folding@Home Project - No comment
The PandeGroup offers a long string of Youtube videos explaining the intricacies of both protein folding itself and using the folding software.

Sometimes, however - true to their geek heritage - their attempts at coherent communication fail miserably. Our favourite nerds are sometimes made... quite uncomfortable. Our undying love for them must shine a little too bright.

Mock not lest ye be mocked first.

New Projects (p6313-6316) for uniprocessor client

By , the in Folding@Home Project - No comment
Vincent Voelz has just put online a new set of projects for the Folding@Home uniprocessor client (single processor). These projects require the "Big WUs" setting to be set, and their value is doubled compared to the usual benchmark. Each unit reported 336 points and should preferably be returned within 36 days, and no later than 52 days.

The SH3 domain is a molelcule found throughout many diverse biological systems. It is usually part of a larger complex that helps to integrate a cellular signals. Its main function is modulate and regulate signaling by transiently binding a certain class of polyproline motifs. The specificty of different forms of SH3 for certain peptide motifs can vary widely, and great insights into celluar signaling networks could be gained by computational prediction of sequence-dependent affiities.

These are simulations designed to test a new sampling method that we have been developing, called the multi-scale fragment potential (MSFP) method. In MSFP, the problem of computing how a peptide chain binds to a receptor surface is broken down into parts that can be efficiently parallelized. Instead of simulating the interaction between a large polypeptide chain and a receptor, we break the peptide chain into small fragments, and exhaustively scan the receptor surface to construct a potential of mean force for each fragment. Using this detailed chemical information, we then use a coarse-grained chain model to sample peptide conformation and sequence space.

The simulations in these projects are of single-amino fragments anchored along the surface of test protein, Abl SH3 domain. These tests will let us test some of the fundamental assumptions of the MSFP method.

Points and deadlines:

  • project 6313: 336 points, timeout 36 days, deadline 52 days
  • project 6314: 336 points, timeout 36 days, deadline 52 days
  • project 6315: 336 points, timeout 36 days, deadline 52 days
  • project 6316: 336 points, timeout 36 days, deadline 52 days

(NOTE: Because these are large WUs (>5 MB), the statcredit point value is doubled from the value obtained by our standard benchmarking scheme).

Information gathered from the project description

nVidia Core 14 lands in version 1.26

By , the in Folding@Home Project - No comment
After the improvements to the nVidia-centric Core 11, which were announced in this news, it's the newly public beta'd Core 14's turn to undergo a similar fix in version 1.26, enabling it to function on machines that have different cards (number of shaders disputes or cards of different generations). However, unlike Core 11 1.31, the latter does not include changes in the benchmarking system.

The procedure for updating is quite simple. To take advantage of Core 14 1.26, stop your client, delete the core file (Fahcore_14.exe) and restart the client. The next time you are assigned a unit requiring Core 14, the new version will be downloaded automatically.

Source: Folding Forum

10 years since Folding@home's inception!

By , the in Folding@Home Project - No comment
10 years since Folding@home's inception!
On October 22, 1999, Vijay Pande and Adam Beberg met for the first time.

Vijay was the newest professor of BioChemistry and Adam was a Ph.D. student of Computer Science, who had already proved himself in developing Their meeting was the beginning of a distributed computing project working on the folding of human proteins.

And 10 years from the start of Folding@home? It is a mere year away, marking the public launch of Folding@home.

Source: Folding Forum

News on the Protomol and GPU3 cores

By , the in Folding@Home Project - No comment
Currently, the Pande Group is focussing on adapting the ProtoMol core to the standard of the other cores used in Folding@home. Indeed, as with any core developed on a new codebase, it must ensure it is able to produce files compatible with the existing clients and servers, and that it can correctly communicate with the client (error codes, progression, checkpointing). Vijay Pande announced today that progress is encouraging and that the core is continuing the process of internal validation. Note that the development of this core has been entrusted to a professional programmer, Joseph Coffland.

On another note, the GPU3 core also continues on its path. Vijay has said it will be called Core 15 (the logical next number in the series of GPU cores). Recall that the main change to this core is to the codebase: it now runs around OpenMM. At first, OpenMM was based on the codebases of the current and past GPU cores (GPU1 and GPU2) but it has since undergone many additions and modifications to make it much more reliable and stable. Unfortunately, this stability is at the expense of performance, however the benchmark will take this slight new decline in scientific effectiveness into account. You shouldn't even notice this in your PPD output. This approach is made possible by the fact that GPU3 projects will start with a clean sheet, not continuing any of the GPU2 work.

Unfortunately we have no release date for these new cores as per Pande Group policy, but careful study of the OpenMM Roadmap can speculate on the release date of Core 15:

Beta (October 2009) includes:

Release 1.0 (January 2010) adds:
  • Finalized API
  • Extensive unit tests and validation suite
  • Custom bonded interactions

If you believe this roadmap, considering that core 15 is developed in parallel and it will require some additional validation tests during the integration of version 1 of the OpenMM code, the latter stages of the first quarter of 2010 seem to be a realistic arrival date for the new core.

We will post more as and when we get it ;)

Largely based on: Updates on new cores: Protomol and GPU3 core

Shipment of servers bound for Stanford!

By , the in Folding@Home Project - No comment
Shipment of servers bound for Stanford!
Next week, Stanford University will be taking delivery of some new storage servers. Each server will have 24 hard drives, with 2TB capacity to each, which will be a great help in providing some breathing room for the Pande Group. Unfortunately, however, the official newscast did not specify the number of servers.

This is obviously good news, as in recent weeks, disk space was teetering at a critical level, which could have lead to reduced activity on some projects, as well as a lack of units being assigned. This near-crisis is a stark reminder of the benefits that Storage@Home will soon be providing to the project in the form of distributing quite a bit of stored data amongst S@H users for off-site safekeeping.

Source: The official blog

Atlas Folder leads the way

By , the in Folding@Home Project - 3 Comments
Atlas Folder is the most active contributor to the project. He is heavily involved in the fight against Huntington's Disease.

Atlas Folding's cluster was already very powerful. However he decided to further enhance its contribution! "Phase 2" has been assembled with a total of 31 9800GX2s and 23 GTX295s.

Two videos to capture the excesses of Atlas Folder!

All the teams of Folding@home tip their hat to Atlas Folder for his impressive contributions to the project.

The Atlas Folding site: click here