Dual core setup for FAH console
nonstop301
51° 27' 24.87" N // 0° 11' 38.91" W Member
I just wanted to bring up a small query with respect to the settings of the FAH console on a dual core computer.
I haven't set up the FAH console yet on the dual core computer I use, but I'd like to know whether it's possible (or even necessary for that matter) to adjust the console settings so that both cores are being used to complete a WU at a faster rate instead of just having the one core carrying out the task with the other one remaining idle for most of the time.
If I configure the FAH console for the dual core computer in a similar manner to what I have done with my single core machine, does it mean that I will automatically use both cores to complete a task ?
Thank you once agan for your help and any advice you provide is always gratefully received.
I haven't set up the FAH console yet on the dual core computer I use, but I'd like to know whether it's possible (or even necessary for that matter) to adjust the console settings so that both cores are being used to complete a WU at a faster rate instead of just having the one core carrying out the task with the other one remaining idle for most of the time.
If I configure the FAH console for the dual core computer in a similar manner to what I have done with my single core machine, does it mean that I will automatically use both cores to complete a task ?
Thank you once agan for your help and any advice you provide is always gratefully received.
0
Comments
You'll then be working on two different WU's simultaneously - and collecting points accordingly.
Thanks for that Prof
Does that mean that if I leave the Machine ID=1 for both consoles, I'll run into instability or will it allow both cores to work on one WU ?
I'm just a little curious to see how much faster 2 cores complete a WU compared to the single core computer
I do intend to subsequently get each core to run a different WU independently.
There is a way to let the cores work together in SMP. It requires running Linix. Though it looks like people have had good sucess running it through winows in a virtual machine.
The CPU doesn't actually do more calculation this way, but it can handle bigger single chuncks of work and that helps out F@H. Since it help the project you get more points.
Thanks for your replies edcentric and profdlp
From what I understand then, if I set up two FAH consoles on the dual core computer, I will have to also use a separate folder for the second console and also make sure that for the second console the Machine ID=2
If I just the run the one console on a the dual core computer and use the same configuration settings as the single core machine, then I should notice that each task will be completed a lot faster (almost twice as fast) than the single core computer. Both cores would be employed for the task without having to adjust the console settings to make that happen.
NB. The dual core processor is the Athlon64 X2 3800+ and on the single core machine I have an overclocked Sempron64 2600+ at 2.3 GHz
...but this part is not. If you only run one instance it will only use one core and Fold at a rate commensurate with a single-core machine.
Think of your dual-core system the same as you would think of two separate computers. Installing FAH on one and not the other will limit your production to half of what it could be.
Thanks for the clarification Prof
In terms of FAH performance, the most sensible thing to do with the dual core computer is to set up two consoles then. Otherwise the tasks will be completed at the same rate as a single core machine.
Well, I thought 2 hands do a single job faster than one, but in this case 2 hands want to do the single job with one hand only
It makes you wonder a little bit about the dual core concept but at the end of the day it permits you to do 2 jobs at the same time.
I'm not an expert on the subject, but I believe that there is a Linux OS which does something along the lines of what you were asking about, as in letting both cores crunch the same WU simultaneously. (And with great success, from what I've heard.) In any case, I think you'll be pleasantly surprised at how much Folding you'll get done under the two-WU system.
There's no doubt it will be nice to see two tasks being completed within approximately the same time frame using the dual core machine
I'm still very new to Folding at Home and I'm exploring the possibilities because I will have more machines at my disposal and I just want to have a better idea of what is the best way to make them contribute both for Team 93 and for the overall Stanford project
I have noticed that on the Sempron machine, the recent tasks always originate from the Gromacs core (FahCore78.exe) and involve the same protein p2124_lambda_5way_melt_4_10011.
The first task that was assigned came from the Amber core (FahCore82.exe) but once that was completed the next 4 WUs only came from Gromacs.
I'm not sure what difference each FAH core makes and I just want to know if I'm better off using the Amber or the Gromacs core on the Sempron machine.
Thanks again for all your help and suggestions.
The hitch is that Stanford releases the different projects in batches and needs to receive the data within a certain time frame for it to fit into their research needs. The bottom line is that the best way to support the science is to take what you get. Sometimes that means knocking out a WU in record time and collecting a fat pile of points quickly. Other times that means slogging through a "slow" WU for the sake of the research. It all evens out over the long run.
I'm not so concerned about the tasks that are being sent and I was just thinking whether I'm stuck in a vicious circle of Gromacs and p2124_lambda_5way_melt_4_10011.
I didn't know that each protein project can require more than one work unit in order for the calculations to provide an accurate set of results. The p1814 task from the Amber core only needed one work unit and then I was given p2124 so I assumed the same thing would apply.
Like you said though, still early days and it's just 5 WUs I've completed
The linux console will fold SMP (symetrical multi-processor). The one WU will be worked on by both cores. Right now these get bonus points, not becuase they are faster, but because this lets you fold bigger more complicated work.
I haven't set up a machine to do this yet. Once I get my C2D box folding I may try it.
I'm going to start FAH on the dual core computer tomorrow and set up a console for each core in the way you and Prof have described.
The Linux option is interesting and a couple of the computers that I'll be using to fold in the coming weeks have Linux on them and carry dual core processors. I will consider using SMP with them and then compare it with the two console configuration I will perform with the dual core tomorrow.
As always though, with folding it's all about sticking to it rather than what method you use
The work unit of the single core was interrupted and then it received a new task once each console on the dual core machine had started to fold.
Should the machine IDs in this case have been different with the FAH console on the single core computer as ID 1 and then the consoles on the dual core computer as ID 2 and ID 3 ?
Thank you once again for your replies and any further suggestions you wish to add.