Not seeing bonus points

[ctblackm1]

My GTX 780 on 13000 get around 75k each run, but also need 11h

This seems to confirm what I am seeing/saying. From my observations, it is rare to get over 50k, with most in the 30k range. And, mine take over 11h to complete as well. So, GTX 980 is pretty awesome for gaming. Not so much for folding.

80k Per WU is more than possible with CPU's - not necessarily 1 cpu but it is doable - I get about 275k per WU - now these are the BA units that will be going at the end of January so it won't last for ever. My stats to show what I mean.

http://folding.extremeoverclocking.com/ ... =&u=510459

Yeah. Not possible with my single CPU setup I guess. So, if you have multiple CPU's in your machine, they all work on the same work unit at the same time? Too bad on the BA units going away. (not sure if my rig would even qualify)

And, that site is awesome. I wish folding@evga did that. Impressive stats too. What is your CPU setup and clocks? I notice that core 17 and core 18 projects also make heavy use of a CPU thread as well. I wonder if I could get any improvements by pushing my CPU a little harder.

[Nathan_P]

Yes, you can set multiple CPU's in one machine to work on the BA units or you could set multiple SMP projects up to work on each CPU. It has always been recommended though to run one project on as many cores as you can. The overriding goal is to get each unit back as quickly as possible, quicker turnaround for the project and more points for the folder

BA units require a minimum of 24 threads, currently a pair of x5650 lga 1366 xeons is the bare minimum to make the deadlines but power consumption on those is heavy. My rigs are all 2 cpu machines, dual E5-2665's 8c/16 at 2.4ghz, Dual E5-2670's 8c/16t at 3.0ghz and dual E5 v2 12c/24t at 2.4ghz.

any improvement in clock speed will help get more points but the folding cores love more cpu cores - my v2 E5's at 2.4 are 23% quicker than the 2670's at 80% of the clock speed

[bruce]

Most people talk more about PPD than P/WU.

Suppose I get a WU that takes N hours followed by another WU that takes N more hours. Now suppose on another day, I get a WU that takes 2N hours. It seems I shoulc earn the same number of total points for the first two WUs as I did for the one WU the next day under the policy of EQUAL PAY FOR EQUAL WORK.

Points can never be perfectly aligned like my example, but I think it is still a good idea to be close to that.

[ctblackm1]

Yes, you can set multiple CPU's in one machine to work on the BA units or you could set multiple SMP projects up to work on each CPU. It has always been recommended though to run one project on as many cores as you can. The overriding goal is to get each unit back as quickly as possible, quicker turnaround for the project and more points for the folder

That makes sense. It would be kind of cool if they could find a way to get 2+ video cards working on the same WU. (Maybe someone is already working on this)

BA units require a minimum of 24 threads, currently a pair of x5650 lga 1366 xeons is the bare minimum to make the deadlines but power consumption on those is heavy. My rigs are all 2 cpu machines, dual E5-2665's 8c/16 at 2.4ghz, Dual E5-2670's 8c/16t at 3.0ghz and dual E5 v2 12c/24t at 2.4ghz.

Wow. Nice. I am guessing from the mobo's in your rigs that you built these yourself? Got any pictures?

Most people talk more about PPD than P/WU.

Suppose I get a WU that takes N hours followed by another WU that takes N more hours. Now suppose on another day, I get a WU that takes 2N hours. It seems I shoulc earn the same number of total points for the first two WUs as I did for the one WU the next day under the policy of EQUAL PAY FOR EQUAL WORK.

Points can never be perfectly aligned like my example, but I think it is still a good idea to be close to that.

Right. I don't see many people (or any people) touting their P/WU numbers. I was just comparing numbers with some of the top folders and couldn't understand how it was even possible. Curiosity more than anything. (and because I want more points PPD, so I need my video cards to churn out WU's faster )

[bruce]

It would be kind of cool if they could find a way to get 2+ video cards working on the same WU. (Maybe someone is already working on this)

True, it would be cool.

Some tests have been run by the guys at Gromacs and probably at Stanford, too, and so far, the results have been less than ideal. Because the RAM being used by multiple CPUs is shared, they can cooperate more fully than when moving data from video-memory to another bank of video-memory, a lot of time is wasted. I'm sure somebody will fix that problem someday, but not necessarily soon.

Then, too, CPUs in a single system are generally symmetric, whereas it is not uncommon for multiple GPUs to be different, adding the probabilty of bottlenecks caused by their speed differences.