FAH-News

[Patrick Keller]

Wieso etwas spät dran ?
Das Topic in deren Forum wurde am 16.Juli erstellt ... Gab also genügend Informationen

[X1900AIW]

Glaube das bezog sich auf unser Forum bzw. mich, sorry für mein spätes (Doppel)Posting, EdmundBlackadder, hatte den ganzen Thread FAH und das Pandelab Team !!! nicht wahrgenommen. Mea culpa !

[EdmundBlackadder]

Glaube das bezog sich auf unser Forum bzw. mich, sorry für mein spätes (Doppel)Posting, EdmundBlackadder, hatte den ganzen Thread FAH und das Pandelab Team !!! nicht wahrgenommen. Mea culpa !

"FAH und das Pandelab Team !!!" ist auch ein denkbar ungeeigneter Titel. Hätte der gute Bommer vielleicht besser formulieren können.

[X1900AIW]

Appell vom 20.08.09:
Importance of software and data integrity

• "(...) It is therefore extremely important not to modify the scientific cores or data for Folding@Home calculations. Returning any work results with modified cores or data can taint both your results and the results of any WU's calculated on the basis of those. Thus even well-meaning optimization can potentially hurt the scientific value of the project substantially.
  (...)
  
  Please use only the controls provided by the client and the operating system; do not modify the cores or Folding@Home data in any way. Doing so is the equivalent of providing a tainted donation. When we detect such tainted donations, we may remove points awarded, decline further donations by blocking work assignment, or take further action. (...)"

[Michael H.W. Weber]

...das ist ganz, ganz übel. Schrauben da also welche ungefragt am Client herum...

Michael.

[X1900AIW]

Mit Veröffentlichung des neuen Nvidia-Treibers 190.62 (nur für Windows7, aber für sämtliche GPUs, wie ich es verstehe) mit dem integrierten Feature "DirectCompute" wird die alte Diskussion um Nutzung für DC evtl. aufleben, für FAH gab es bereits vor Monaten einen Thread zum Thema mit einer (inoffiziellen) Einschätzung ...
Directx 11 with GPGPU. Future client with these libraries?

• "The GPU1 code was based on DirectX and it was discontinued because DirectX was not reliable enough for the needs of scientific computing. Whether that will be fixed by DX11 or not remains to be seen, but I don't think there's any reason to opt for DX11 if oCL actually works as advertised.
  
  On the other hand, oCL would provide a common code-base which might replace the current CAL/CUDA implementations . . . provided it's mature enough and reliable enough platform to support FAH's needs. It's not possible to predict when that might be or whether Stanford will seriously consider the time rewriting the code once again to be worth the benefits of replacing two clients with one. (...)" (bruce am 20.04.09)

(Siehe dazu die Aussage von Prof. Dr. Kun Zhou, woanders gepostet.)

[X1900AIW]

Ausserhalb der üblichen offiziellen Verlautbarungen ein Hinweis auf FAH-Addict (französisch & englisch), "the first dedicated web site to Folding@Home project" (toTOW), ab dem 01.09.2009 sollen dort eine Reihe Informationen, News, Artikel etc. veröffentlicht werden. Z.Zt. noch wenige Inhalte.

• "Our goal is to provide :
  - latest project news
  - tutorials for installing or optimizing clients and OSes
  - hardware reviews"

Beispiele:

• Pande Group Researchers - Vorstellung mit Text, Bild, Video u.ä. (Quelle ist die offizielle News-Serie "Meet FAH team member ...").
• How many atoms are there in a BigAdv unit? - Gegenüberstellung zum normalen SMP-Client
• GPU environment variables - anschaulich erklärt, wie Umgebungsvariablen definiert werden (habe ich so noch nirgends gefunden ! )

[X1900AIW]

News vom 25.09.09:
Update on new FAH cores and clients

• SMP2: Gromacs and Desmond.
  (...) The main goal here was to make it MUCH easier to use. In order to do that, it meant getting rid of our use of MPI. We have had two approaches to this. Both ditch MPI by using threads instead. One was to switch to a new piece of software for the core. This has led to the "Desmond" core, based on software from DE Shaw Research. The second approach was to communicate the MPI issues with the Gromacs developer team and work with them to push for a threads-based Gromacs implementation. Both of these are coming along well and we are testing cores in house. You should hopefully see these cores "in the wild" (i.e. running on FAH) in a month or two, assuming that tests go well.
• Normal Mode Langevin (NML) Dynamics in the Protomol core.
  (...)
• GPU3: Next generation GPU core, based on OpenMM.
  We have been making major advances in GPU simulation, with the key advances going into OpenMM, our open library for molecular simulation. OpenMM started with our GPU2 code as a base, but has really flourished since then. Thus, we have rewritten our GPU core to use OpenMM and we have been testing that recently as well. It is designed to be completely backward compatible, but should make simulations much more stable on the GPU as well as add new science features. A key next step for OpenMM is OpenCL support, which should allow much more efficient use of new ATI GPUs and beyond.

Seit langem angekündigt, scheint der neue SMP2-Client bald Wirklichkeit zu werden. Auch die ATI-Besitzer werden (hoffentlich) endlich zufriedener gestellt sein.

[Txt.file]

News vom 25.09.09:
A key next step for OpenMM is OpenCL support, which should allow much more efficient use of new ATI GPUs and beyond.

das klingt doch echt mal sehr gut ... mal sehen was draus wird und vor allem wann

[X1900AIW]

Zum Thema "Will the OpenCL client take less development time?" zwar keine News, aber interessante Einsichten von ihaque:

• "... if FAH uses doubles. In the current version, it does not. Not even all the CPU cores use doubles; currently, none of the GPU cores do. It's possible that a future core will require DP support, but not all of them will. Most announced consumer-level hardware either doesn't support doubles at all (NVIDIA G80/G92 chips, Juniper and smaller versions of the Cypress chip from AMD), or takes a performance hit for using doubles. Policy will probably continue to be as it currently stands - doubles only where necessary.
  
  Also, one distinction that's important to remember (both in this thread and for folding in general) is that the client and individual simulation cores are distinct. The client is responsible for talking to the FAH servers and starting cores; the cores are what do the actual simulation. As a consequence, it doesn't really make sense to ask whether the client will require DP; it's a question about the cores. Like I said above, the "typical" core will probably be single precision, and if necessary double precision cores can be written.
  
  Core/client is a lesser-known distinction on the GPU folding side of things because there simply aren't as many alternative cores. If you exclude beta core types (which are only used for testing), ATI/AMD only has one - core 11 - and NVhas two - 11 and 14. CPU and SMP folders are already well aware of the diversity of simulation cores out there" (10.10.09)

[Peter Fanghänel]

Im foldingforum wurde ein interessantes Statement von ihaque, einem PG-Mitglied zum Thema "Is it worth folding on ATI cards?" gepostet. Die Diskussion kam natürlich ziemlich schnell auf den obligatorischen nvidia-ATI-Vergleich und der (theoretischen) Überlegenheit der ATI-Architektur, weshalb der Autor nur auf diese Thematik eingeht. Das Fazit ist:

So after a very long discussion, what's the upshot?

Don't compare hardware on the basis of peak teraflops. While the Nvidia hardware has a lower peak teraflop count, for the applications were interested in, that does not make them "clearly technically far inferior", because the architecture makes it easier to actually get close to that peak performance than does the AMD architecture.

Wen die technischen Zusammenhänge interessieren, möchte sich bitte den (englischsprachigen) Post selbst durchlesen. Für mich war es interessant zu erfahren, wohin die Reise mit ATI geht, da momentan ein neuer GPU-Client entwickelt wird und ich im nächsten Jahr die Anschaffung eines neuen PC plane. Mein bereits feststehender Entschluss dieses mal nvidia zu verbauen und nicht auf die Umsetzung der theoretischen Überlegenheit der ATI-Architektur zu warten wurde damit bestätigt.

[duftkerze]

und das heisst?

ati ist ne mogelpackung was gpu-computing angeht?