If you use the machine at all I wouldnt do it with less than 512. If you have 256 then half you RAM is being used for folding, another 100 or so for the OS....there is not much else for your apps.
Just had a look at Sally's farm. Mostly they run at 11,500 - 15,000. One actually is way too high for some reason.....135,000. Got me beat as others the same Protien run 11,500!!!!
HyperThreading will give you 2 lots of FahCore_78.exe or FahCore_65.exe or one of each.
edit....the one using 135,000 is a P1310, which is one of those BIG ones. That is more than likely the reason why it is using so much.
Still scary this thread is giving you a long one and it is well-endowed (or in some cases, that's even doubled up)
Those big WUs are great with sufficient memory. Whether it's worth upgrading from 256 to 512 for those 'purely for FAH' rigs? Suppose it could be if you only got big units to fold, the amount of electricity used wouldn't be much higher but you can potentially get higher points over the same time period.
Only catch - if the WU goes wrong, what happens then
Brain hurtz :bawling:
0
Straight_ManGeeky, in my own wayNaples, FLIcrontian
edited November 2004
Well, if you get an Early_WU_End, you get points for what DID compute. Let's say you had a p1310 run though to 76% of complete. 100% is worth 316 points. So, .76 X 316 would be 240 points awarded when it got sent in. If it only got 30% done, you would still get 94 points. You would get points for your computer's donated calc time. AND, folding project admin would get to try and figure out why it EWE'd. The next gen might indeed get the problem fixed, if it was truely a WU problem.
The thing to watch out for, with big WUs, is to get them on a computer that can handle SSE2.
AND, if you use it for other things that are intense RAM use kinds of things, like gaming while you calc WUs, have 512 MB or more RAM in the computer. OTHERWISE, since the client is a Normal of lower process, the client could get pended, and so could the core. Pending, in this instance, means that teh client could get told to pause work. I'll give you an example here--- let's say that I run a religious research software pack, Corel Graphics Suite, Word Perfect Office, and my viral scanner while folding. This is a very high resource set box, as to CPU and RAM-- it is a 3 GHz P4 with a GIG of very fast RAM running at the equiv of DDR420 to DDR440 depending on temps in room where the computer is (I have to alter BIOS settings myself to adjust for heat trends, and simply do not have the funds to allow for buying a WC system for this box). I had an Amber beta running. Time per percent went to 5X normal, and STAYED there until I restarted the computer. RAM was 80-90% loaded, and swap was 15% of 1.5 GB loaded. This was a client 5.02 before 5.02 was officially released, with a core for Ambers that I cannot fairly talk much about yet.
so, one way to tell if you have not enough RAM or insufficient CPU, is to look at time trends whiel using box as normal for what you intend to do with that box. If just foldiong, and you have a 900 MHz of faster processor and 512 MB RAM or more in the box, then I recommend TRYING the big WUs and sometimes in a big WU fold trying to use the box intensely, and see if a WU breaks or radically slows down the folding effectiveness.
If you have a 700 MHz or lesser box, and 256 MB RAM or less (and I would not fold with less than 128 MB of RAM above and beyond frame buffer settings in BIOS in any Windows version), I would not fold Ambers or Double Gros, and one way to ensure this is to not use the -advmethods switch, not use the -forceasm switch, not disable Assembly optimizations in the configuration, and set the choice to accept large RAM workunits to NO(if this seems odd, run your client in -configonly switch mode, when it asks if you want to set advanced options reply with a Y or Yes or YES or yes (not a case sensitive query), and look at but do not play with defaults except as noted as they tend to be very overall-stable defaults.
"Slow, low resource box" notes:
NOTE, large RAM acceptance for workunits needs to be off for such a box unless you want real long folding times per WU and a good chance of EWEs with soem WUs you probably will get from time to time if you do not set things this way.
NOTE, choosing the GAH option for WU type will get you timeless tinkers more often than not with client 5.02 and 5.03, the Assignment servers have been set to connect your client to data servers that feed those when GAH is chosen when the timeless tinker servers are able to connect to your new WU request session and feed WUs. If the servers (last I heard, two of them could do this) that feed timeless tinkers are down or overlaoded, you are likely to find your "low resource" box getting something else than a timeless tinker until when it wants a new WU those servers are available to give you one.
That is what I know about using slower boxes with Folding. In theory a 300 MHz box with 128 MB of RAM can fold a timeless tinker and since there is not credit deadline, that box can contribute some. NOTE, timeless tinker projects are limited, they are used for things that do not have to be figured out in a short time. SO, there are both limited numbers of Timeless Tinkers AND limited number of slower servers feeding them right now. The highest capacity data servers are feeding very complex and large WUs.
Comments
I'l leave it then
HyperThreading will give you 2 lots of FahCore_78.exe or FahCore_65.exe or one of each.
edit....the one using 135,000 is a P1310, which is one of those BIG ones. That is more than likely the reason why it is using so much.
Those big WUs are great with sufficient memory. Whether it's worth upgrading from 256 to 512 for those 'purely for FAH' rigs? Suppose it could be if you only got big units to fold, the amount of electricity used wouldn't be much higher but you can potentially get higher points over the same time period.
Only catch - if the WU goes wrong, what happens then
Brain hurtz :bawling:
The thing to watch out for, with big WUs, is to get them on a computer that can handle SSE2.
AND, if you use it for other things that are intense RAM use kinds of things, like gaming while you calc WUs, have 512 MB or more RAM in the computer. OTHERWISE, since the client is a Normal of lower process, the client could get pended, and so could the core. Pending, in this instance, means that teh client could get told to pause work. I'll give you an example here--- let's say that I run a religious research software pack, Corel Graphics Suite, Word Perfect Office, and my viral scanner while folding. This is a very high resource set box, as to CPU and RAM-- it is a 3 GHz P4 with a GIG of very fast RAM running at the equiv of DDR420 to DDR440 depending on temps in room where the computer is (I have to alter BIOS settings myself to adjust for heat trends, and simply do not have the funds to allow for buying a WC system for this box). I had an Amber beta running. Time per percent went to 5X normal, and STAYED there until I restarted the computer. RAM was 80-90% loaded, and swap was 15% of 1.5 GB loaded. This was a client 5.02 before 5.02 was officially released, with a core for Ambers that I cannot fairly talk much about yet.
so, one way to tell if you have not enough RAM or insufficient CPU, is to look at time trends whiel using box as normal for what you intend to do with that box. If just foldiong, and you have a 900 MHz of faster processor and 512 MB RAM or more in the box, then I recommend TRYING the big WUs and sometimes in a big WU fold trying to use the box intensely, and see if a WU breaks or radically slows down the folding effectiveness.
If you have a 700 MHz or lesser box, and 256 MB RAM or less (and I would not fold with less than 128 MB of RAM above and beyond frame buffer settings in BIOS in any Windows version), I would not fold Ambers or Double Gros, and one way to ensure this is to not use the -advmethods switch, not use the -forceasm switch, not disable Assembly optimizations in the configuration, and set the choice to accept large RAM workunits to NO(if this seems odd, run your client in -configonly switch mode, when it asks if you want to set advanced options reply with a Y or Yes or YES or yes (not a case sensitive query), and look at but do not play with defaults except as noted as they tend to be very overall-stable defaults.
"Slow, low resource box" notes:
NOTE, large RAM acceptance for workunits needs to be off for such a box unless you want real long folding times per WU and a good chance of EWEs with soem WUs you probably will get from time to time if you do not set things this way.
NOTE, choosing the GAH option for WU type will get you timeless tinkers more often than not with client 5.02 and 5.03, the Assignment servers have been set to connect your client to data servers that feed those when GAH is chosen when the timeless tinker servers are able to connect to your new WU request session and feed WUs. If the servers (last I heard, two of them could do this) that feed timeless tinkers are down or overlaoded, you are likely to find your "low resource" box getting something else than a timeless tinker until when it wants a new WU those servers are available to give you one.
That is what I know about using slower boxes with Folding. In theory a 300 MHz box with 128 MB of RAM can fold a timeless tinker and since there is not credit deadline, that box can contribute some. NOTE, timeless tinker projects are limited, they are used for things that do not have to be figured out in a short time. SO, there are both limited numbers of Timeless Tinkers AND limited number of slower servers feeding them right now. The highest capacity data servers are feeding very complex and large WUs.