FAH-Addict

  Welcome to FAH-Addict, the primary source of information about the Folding@Home project. Our goal is to provide:

- latest project news
- tutorials for installing or optimizing clients and OS's
- hardware reviews

Please enjoy your visit, and feel free to send us your suggestions and/or report errors.

After a long wait, it's finally here; version 6.30 of the SMP client, which sees the end of MPI support being required in order to run, among other improvements. Installation of the client is easier than before; there is no need to install it via a .BAT file or via the traditional program installation method, and Windows Vista/7 no longer launches an MPI installation in administrator mode, so the UAC should not whinge anymore. Installation of the client to run in service mode should now once again be possible.

Core A1 units are no longer officially distributed, but it's still possible for one to work its way get into your work queue, which might crash the client due to the absence of MPI. It is for this reason that the client is considered to be a beta. Note that you can still keep your MPI installation in order to reduce the risk of crashes.


To install the new client, you can perform either of two approaches:

• Simply replace the currently installed client executable with the new version. You should be able to then manually uninstall MPI (Deino or MPICH).
• Uninstall the old client and all of its components, perform a clean install of the 6.30 client.
  

You can find download links on the high performance clients page, and the official installation guide can be found here.

Please feel free to leave feedback and report your experiences in the comments section of this article.

Edited 9:35am:
The announcement has been made on the Official Blog.

Peter Kasson has released a new project for bigadv clients: p2685.

This project continues the work of the previous bigadv units, and has similar characteristics:


This project uses the a3 core, which due to bigadv-related bugs in the Linux version is currently being distributed to Windows and OS X clients only.

The Pande Group have been tracking the progress of the GPU3 client beta test, and have decided that it is time for the client to become more mainstream. As a result the 6.32 client required to access GPU3 client units has now been placed on the download page, to replace the older 6.23 client.

For now GPU3 is nVidia only, in the form of Core 15, but in the coming weeks it is hoped that it will be joined in testing by a second GPU3 core, imaginatively named Core 16.

The Pande Group are working hard on a port of GPU3 for ATI, which will use OpenCL. Sadly the code currently has performance issues and as such is not ready for testing. For the curious, more information on the code that will bring OpenCL support can be found at the OpenMM website. The source code which will enable ATI OpenCL support is available on the website under the LGPL licence, and anyone that wishes to take a look and/or contribute is welcome and encouraged to do so.

On a hardware note, the new client is required to fold with cards based on the Fermi chipset. It also features a much more up to date internal list of hardware IDs, so those of you who have been folding with ATI HD4xxx and HD5xxx cards will no longer need the forcegpu flag to fold with this client.

Source: Vijay Pande's Blog

Following an internal discussion, the Pande Group has decided to lengthen the deadlines for non-bigadv SMP units. The new deadlines are:

• Preferred Deadline: 6x the time to completion on the i5 benchmark machine
• Final Deadline: 10x the time to completion on the i5 benchmark machine
  

Projects 6701 and 6702 will have their preferred deadlines extended to fit the new model for all units issued under the new system. The final deadline will remain the same so as not to change the K factor. No other currently issued projects will be retroactively modified to the new system.

Bigadv work units will continue to use the previous, tighter deadlines. They are specifically designed for fast completion and so the Pande Group would prefer that any machines which are marginal for bigadv, fold standard SMP units instead.

Source: Folding Forum

The GTX480 was a cold shower for us all; outstanding performance but with power consumption never before seen on a publicly available card. In short, it was expensive, hot-running and power-hungry!

The 40nm process used by nVidia to manufacture chips seems to be very very poorly controlled on these semiconductors (more than 3 billion transistors). In order to sell its production waste, nVidia has launched a series of increasingly neutered cards based on the same chip (the GF100) but with more and more blocks of SPs disabled, the worst part of this being that the disabled SPs do not significantly reduce power consumption.

With the GTX460, we will abandon the monstrous GF100 chip in favour of a simplified and more streamlined version called the GF104.

We lose a full third's worth of transistors, but the losses are not uniform. It will be serious gamers who suffer the largest decline in performance. Where a GTX 480 has 480 SP, the new GTX still has 336.

A picture worth a thousand words:

Click image to enlarge

Cast your eyes to the energy consumption values; this is where the new Fermi will change everything.

For more information, we recommend reading PCINpact's article on the GTX460, which should be of special interest to hardcore folders.

33% of FAH-Addict readers use Windows XP, which is why we're taking the time to remind you that from tomorrow, Microsoft is officially bringing to an end its support of Windows XP SP2.

What does this mean for users? Windows Update will no longer supply any machine running XP SP2 with security patches after 13 July 2010. This means that as of July 14, each bug and security hole found will not see the a fix released for this version of Windows. Seeing as how there is little shortage of critical vulnerabilities of all sorts, such as the ones that enabled the Blaster Worm to operate, it would be wise to switch to a more recent version of Windows in order to be afforded better protection.

Of course, one can always just install Service Pack 3 to postpone the inevitable, or switch over to an alternative OS if one is so inclined.

Peter Kasson has just released two new SMP2 projects: p2630 and p2631.

These projects implement a new set of features, setting them apart from the majority of current projects. Previously unseen variations in performance between machines may be observed as a result.

These projects are distributed by the server at 171.67.108.24, and have the following properties:

• 102,731 atoms
• preferred deadline of 3 days
• final deadline of 4 days
• worth 462 points
• bonus factor of 3.1
  

These projects have proven very stable on quad-core machines during preliminary tests, though a few problems have been observed on machines registering 8 cores or more, probably due to the special properties of the simulated systems in these projects, which can cause instabilities relating to the parallelization of calculations. Therefore, these projects are initially being distributed only to machines using quad-cores and running with the -advmethods flag enabled.

Updated 11/07 at 9:10pm:
In the wake of complaints regarding the size of the results files (approximately 114MB) these projects have been suspended. We do not yet know whether they will be re-released as BigAdv projects (whose results files are >100MB) or if they will return in a more streamlined and efficient version which generates smaller results files.

As a result of a bug affecting A3-core BigAdv units on Linux, Peter Kasson has suspended the assignment of these units to Linux machines.

The cause of this bug remains unknown, but pending its resolution, you should fall back on non-BigAdv units for the time being.

If you have one handy, you can still try your hand at folding BigAdv units under a Windows installation, as this capability has just been activated for the OS. As with the Linux client, running it with the -smp and -bigadv flags on a version 6.29 client or higher will allow you to obtain BigAdv units.

The current stock of BigAdv units will remain fairly limited until new projects are started with corrected cores. In the event of a shortage, your Windows machine should fall back on A3 projects.

To ensure the stability of the GPU3 client, yslin has added several new projects to test new simulation parameters on the Villin protein.

Stanford has also added some redundancy to the GPU3 project by spreading these units across two servers.


The GPU3 beta test is going well, but due to the limited supply of units, most are going to owners of Fermi cards (who cannot run the older GPU2 units). As a result, do not be surprised if you continue to be assigned GPU2 units if you own older cards.

As many of you will know, a new version of the client (dubbed v7) is under development which is a ground-up rewrite of the client code. This is intended to make the client easier to maintain, more reliable, and much more adaptable to add features that donors request. The first versions will not have everything that donors have asked for, but some significant changes will be noticed.

The most obvious of these will be that the client will be the first truly all-in-one client the project has had since uniprocessor was the only client type available. The client will be able to manage single-core CPU, SMP, and GPU cores from just one executable, with any number of uniprocessor and GPU cores controlled from one client executable (subject to the limitations of the hardware available, obviously). Also instantly noticable will be a new GUI, and for developers, much better tools to support monitoring the client. In time, it is hoped that it will be possible to use one copy of the client to manage other clients and cores on multiple machines, but it should be stressed that this is a longer-term aim.

The first console version of the v7 client has been through internal testing and is now undergoing limited closed alpha testing. Whilst there are definitely some rough edges, Vijay hopes that no showstoppers will appear, and that something ready for open beta will be ready within the next month or two, possibly sooner.

Source: Official Blog