[Ubuntu] Setting up big-bigadv (P6903) for dedicated 2600k - Page 2

TBH I am not sure.. I haven't used 10.04 myself, guide may work, may not.. I can have a look at it when I have some time. My concern would be what about 10.10 isn't working for you?

I'm not sure if ImgBurn will do it for USB's... I just burnt a DVD because I like having it ready at a moments notice.. maybe try that if you have a blank one around.

Yes, I know there is a shortage, and yes I understand the implications of users doing things like this in the interest of points.

Just like running regular bigadv on x6's and 2500k's, people are going to do it anyways. Especially when a forum such as this has competitions in which we are all fighting for an edge in our point production.

I completely understand your point on this.. Part of me agrees with you for sure. Like I said, unless Pande group comes up with a failsafe way of ensuring these only go to those with the desired hardware, people are going to do it anyways. Might as well teach them the most efficient way of doing it so the impact on the science is reduced as much as practically possible.

Good point there sstnt...

Ok, I'm working on pointers to add to the guide for x6's and 2500k's doing bigadv as well and make this more of a general guide rather than specific to 2600k's and big-bigadv's. I will also indicate what steps a Hexacore needs to do, and what parts to ignore. Give me a bit of time to work all that out...

I still need to do a bit of testing to confirm, but I believe the current guide for for 2600k's could use a slight tweak as well.

If anyone uses this, I would like them to try editing the terminal profile for fah6smp to give the client the flag -smp 8 instead of just -smp. Reasoning here, is that the client uses the total # of cores detected to determine what WU's you can pull, the number of threads we specify to use with the -smp flag does not affect what WU you get. Now, when we run just -smp, it will launch 12 threads. Now an 8 threaded chip running with 12 threads is going to be less efficient as one running just 8 threads overall.

Currently I only have a P6900 running, and after switching to -smp 8 the TPF improved. I didn't have a chance to try this out while running the P6903, but I suspect it should improve the performance there as well. I also need to make sure for certain (going off others feedback on this ATM) that the client will still grab a P6903 with -smp 8 set, as long as the client detects the minimum 12 cores.

If anyone is able to do some testing here for me that would be greatly appreciated. (note, carefull switching the flag in the middle of a WU, has worked fine for me a few times, but has also caused the WU to fail once)