New beta bigadv coming soon needs 12 cores

Anyone can get the extra big WUs if they use the -bigbeta flag, are running a suitable SMP client, and have selected 'big' units during the configuration. The machine must satisfy the hardware criteria.

I'm using Linux but Windows users don't seem to be having any trouble [correction (see below) - currently only Linux machines]

The v.6.34 client is sufficient, most earlier SMP clients will not pick up these jobs, since there was a significant update in v.6.34. v.7 is not required.

You have to meet criteria for the Quick Return Bonus scheme to get the huge bonuses.

I'm getting about half-and-half P6903 and others, currently on P2865, which is not an extra big WU.

 

what hardware specs do you guys think would be a minimum requirement for adding the -bigbeta flag? I can't imagine I'd manage to complete one, but with such high points at stake I just had to ask!

 

May just be the turn of phrase used but to clarify you can only pick up the uber bigadv 6903 units if you're using Linux. Other bigbeta units such as 2689 will be picked up if using the flag in Windows.

 

Ah, that had escaped my notice - ignore what I said above re Windows

What would be really useful is a word to properly distinguish these WU from the 'smaller' bigadv... but uber bigadv seems good to me

 

I've got a couple of hex-core HT machines picking up the Linux bigbeta P6093 - long may they last

 

Will adjust my configs to pick up the bigbetas tomorrow, have one 24-core server plus a couple of other servers (hex core plus HT) that could do it. Any idea what the average time to complete is for a bigbeta?

 

My i7 980X's (@ stock, because of temps) do 1% about every 50mins.

 

Thanks, got a 24-core server (dual 12-core AMD Opteron 6176) running through 1% every 30 minutes, FAHmon estimates ETA at 2 days 5:30 and 182K PPD (408K for the WU) - should be fun

 

Wonder if you can fudge getting them and processing on an 8 core machine by using the virtual machine method

I wonder.....

 

If you can get 50mins per frame on an i7 980X at stock speed, would there be any chance of completing one of these on an i7 920 at say 3.7GHz or higher? The amount of points available is phenomenal!

 

no, you need 12 cores/threads for these 'uber' bigadv, a 920 only has 8 threads

 

There are ways around that restriction....

http://forums.bit-tech.net/showpost.php?p=2714410&postcount=49

 

n'awww

I best start saving up for some more cores then!

 

Not liked by Stanford. Using unusual set-ups that have not been QA'd can damage the science. Not liked by Forum mediators and some Donors.

From Vijay's Blog

Folding@home (FAH) is a major scientific endeavor, but is also a kind of contest for some donors to see who can donate the most points. In order to keep a sense of fair competition, we asked donors to help us establish a list of rules, summarized on this FAQ page: http://folding.stanford.edu/English/FAQ-BestPractices .

 

Hmm missed this arrival whilst I was on holiday, should fit nicely with the changes to the farm I'm working on at the moment.

 

You should really get this sorted out. There must be something very wrong that you can only run at stock. My i7 980x @4.2 folds in the low 60s, with a Noctua D14 cooling it. TFP 42.19 for 120K.

 

Yep, you're absolutely right - a Corsair H50 (with Arctic Silver paste) in an Akasa Eclipse 62 case seems incapable of keeping the CPU below 70DegC (with an ambient of 25DegC).

Fortunately this, and a few other 'Farm mods', will all change in the next month or so

 

I've got a similar problem with my 980. It clocks up to 4.3, but the amount of heat it kicks out is amazing. I've got 2x 2.120 plus a 1.140 rad all with fans in push config and only the CPU under water (well a GTX 295, but that's not folding due to heat issues with the CPU).

In the same board I've tried a 970 and that's much cooler - a good 10-15 degrees for the same clocks & voltages.

[QUOTE Not liked by Stanford. Using unusual set-ups that have not been QA'd can damage the science. Not liked by Forum mediators and some Donors. #Snip# [/QUOTE]

Stanford also discouraged the use of i7's to fold - their requirement was to have 8 CORES not 8 THREADS. Technically all of the i7 range fall out of this one, and it's only become accepted practice because it's been so widely used and PROVEN to work.

As for Damaging to the science, as long as big chunks of data aren't lost or corrupted, I don't subscribe to that argument. How can a machine with 8 threads achieve different numerical results to those with 12 cores, using the same logic... It's all 1's and 0's and for a given input you will receive a given output, unless there's an unknown or element of chance in there - that just doesn't stack up.

As it is, I'll be able to run the new BIGADV units on my home server as that's going to have the 970 fitted this weekend instead of the 980X - my rooms just too darned hot! and I'll see if I can get everything up and running in a VM on my Windows 7 server... The VM software I use on my 2600K rig is Virtualbox, and I don;t think you can assign any more cores that are physically available, so that point's moot.

What Linux O/S would people recommend to set up as a virtual machine to try and fold the BIGBETA units on?

 

I would also like to try VM Linux as these 6903s are only 64bit Linux ATM & I can't wait for the Windows equiv forever!

 

For the O/S, choose any flavour you like - Ubuntu, CentOS, Fedora, Mint, ad infinitum. I fold on Ubuntu 11.04 at work and CentOS 5.6 at home (both x64).