« NVIDIA core upgrade to 1.06 | Main | Modification to Assignment Server code »

July 08, 2008

Comments

rpfaas

It would be great if one day we could all download just one simple package to install that can easily be configured to use one core or all available cores, or just the graphics processor, or maybe even just turn on the lot and make that computer squirm :D

i know it will propably be some time before we see that, but i do hope it is the final goal of this project to integrate all clients into one.

i for one am happily donating the processing power of my 3 computers, but i'm not "technical" enough to get into it too deep...

to be honest my computers have WAY more free time then i do, so in the end i'm running 3 "standard" clients but that does mean only one machine is running at it's full potential.. (two are multi core, and one even has a compatible ATI graphics card)

i can't fathom what it would mean in terms of work-to-be-done, so i'll just wait and hope for an "all inclusive" client that's as easy to configure as the "standard" one ;)

greetz and keep up te good work!

Richard

Ibrahim

I agree, it seems like it would be great to use all your cores in SMP and your GPU at the same time. By the way Richard, the GPU client is just as easy to configure as the standard one if I recall the standard one correctly, it might even be easier. You just download the installer and run it. You should definitely try it out.

Niels Madsen

Yeah make a controlpanel for it and add Fahmon to that, allowing you to add any type of client you want to, that handles updates, usernames etc. That would be the best way to go about it, so when i buy my x core processor and my new GPU in a couple of years years i can just set it to x cores + GPU, and it will figure out the core affinities etc. This will be a lot easier to sell to new users, and would make it all a lot easier.

Sneakers55

The optimum method would be "plug and play" where there is a single executable that finds out how many cores you have and whether or not there's a supported GPU. On the other hand, there are some very slow dual core machines and you'd have to figure out that they're better served running two copies of the CPU client.

William T. Stephens

Thanks! Anything you can do is appreciated. I'm looking forward to the day we can have "one stop" installing of a single, "unified" client that'll do what's best for the installed hardware and my system can start producing the maximum possible degree of quality output, a sort of "launch and leave" program. I'm honestly afraid to try a beta at the moment, it seems too likely I could gum up the works and have no way to clean out the debris and do a fresh install. I'll keep my fingers crossed; I'm pulling for all of y'all and the project. Thanks

Stephen Dewey

@ everybody above,

While I speak only for myself, I think it is pretty clear that Stanford is moving in the direction(s) you have all suggested (which are all in the same vein as each other). V6 includes the basic code necessary to enable an all-in-one client. As the V6 FAQ states:

'[V6] added code support for an "all-in-one" console client. For example, running the console client with no switch will run CPU work units. Using the -smp flag on eligible hardware will run SMP work units*. And in the future, using the -gpu switch will run GPU work units (-gpu not supported yet, use the GPU specific client until then).'

Presumably, once the flags are working, this can also be enabled through a more user-friendly settings page in a window-based client.

However, it looks like Stanford is making the choice to get the individual clients for the various computer architectures (single-core, SMP, GPU) working more smoothly before combining them, which makes sense. Fortunately, there are only so many architectures out there that would provide meaningful results for FAH, so it will happen sooner or later.

Stephen Dewey

I just realized that my excerpt from the FAQ might confuse you if you don't also note the following (i.e., your options for V6 are somewhat more constrained right now):

"*Note: We regret that a v6 Windows/SMP client is not yet available, so -smp is not available for the Windows clients [use v5.91 or v5.92 SMP clients for Windows]. The Linux client is the first available client with this option. And there is no single core CPU client for the Macintosh, so all v6 Mac clients will use -smp."

You can read the full V6 FAQ here:

http://folding.stanford.edu/English/FAQ-v6

Note also that some of the V6 clients are still considered "beta".

rpfaas

Stephen,

thanks very much for the feedback, it looks like we're all on the same track with this and that's a good thing :D

for now i've added a "check the F@H beta's" to my "things to do when i have the time" list ;)

just one more thing... a regret if you will...

it's such a shame these new multi core processors aren't set up in such a way that the multithreading etc happens on-chip, but have to be software driven...

it would have been much better (smarter, propably much faster, etc) to create a multicore processor that can be treated by the software as a single core one...

software should just send instructions on WHAT to calculate, the HOW should be a hardware thing...

it's a shame the folks at AMD and M$ didn't look at the multicore architecture of mainframes and supercomputers before they started slapping multiple cores on one chip...

we can only dream about the performance that would have brought us all!

John Willaford

I'd like a good article on what we are contributing to as far as what is being done in multi gpu processing that we are helping accomplish. I have my Q6600@3.2ghz on 1.42v stable with Ubuntu 8.04 64bit 8gb DDR2800 and running Compiz, finally. ATi card patiently waiting for Linux client. Also, are there any flags i need to throw to help it use more RAM? i see where you can set the available ram in MB, i have it set to the numeric equivalent of 4gb ram, 4096.

Lowell Ray Anderson

How can I contact Folding@Home with a question?

The comments to this entry are closed.