Large Work Units
mmonnin
Centreville, VA
Stanford has recently started to release "Large WU." They are upwards of 5mb to download and take up LARGE amounts of RAM.
The core takes up much more memory than the 20 or mb it usually takes up. I have one right now that is using 150mb of memory! Some take upwards of 200mb.
WHY would you or I want to do such a thing?
These new WUs have a large amount of atoms and require more RAM to do the calculations. To compensate for the need of more RAM, Stanford is giving out an extra 15% point bonus for taking up large amounts of RAM.
How do I get them you ask?
1.) Version 5.00 has an option for getting large WUs.
2.) With version 4.0 you can get them with -advmethods some of the time if you have more than 256mb of RAM.
500mb+ is preferred and not more than 1 client if you only have 256 or 512.
I...I...forgot what I was going say damnit.
The core takes up much more memory than the 20 or mb it usually takes up. I have one right now that is using 150mb of memory! Some take upwards of 200mb.
WHY would you or I want to do such a thing?
These new WUs have a large amount of atoms and require more RAM to do the calculations. To compensate for the need of more RAM, Stanford is giving out an extra 15% point bonus for taking up large amounts of RAM.
How do I get them you ask?
1.) Version 5.00 has an option for getting large WUs.
2.) With version 4.0 you can get them with -advmethods some of the time if you have more than 256mb of RAM.
500mb+ is preferred and not more than 1 client if you only have 256 or 512.
I...I...forgot what I was going say damnit.
0
Comments
KF
Sure hope it will improve things points wise as we lost some this week due to them self destructing. No reason why. Posted a message at stanford. No reason given as yet.
So if you are running the v4 client(either gui or console), you won't get one of these large wu's unless you have greater than 500 MB ram. With the v5 client, you can get them with just 256 MB ram though, but you can also specify if you want your machine to process the large wu's too with the new clients. I've been running 2 instances of large wu's on my main rig which has 512 MB ram but haven't had a much of a chance to test it while doing some kind of gaming since I had to go to work early. If you are concerned about performance while gaming, then only specify 1 client to run large wu's on HT P4's and don't use the large wu option on dedicated AMD machines if you are worried about gaming performance unless you have greater than 512 MB ram.
If you have a dedicated AMD folding machine and 256 MB ram, then install the v5 client and run large wu's as the performance with the present large wu's with 1 instance of large wu's is just fine for a dedicated machine and you can even do some light work or browse the net with no problem. On my dedicated folding P4 rig, I run 1 client specified with large wu and the other processing normal work with 256 MB ram with no problems.
p130_1RYP_AAAA_UM
151,432K Mem Usage
Project: 1301
/me still prefers tinkers.
Normal PPW production is just over 1000 for the 240 point one. For normal, smaller WU's it's lucky if it hits 800. Now I want to see what happens when I shut off the other client!
:smokin:
EDIT: Still like tinkers, EO?
I do like what v5 with the large wus have done for my production.
Actually thats phyiscally impossible since 8-9kb/s is over 56k. Actually 64-72k. That and dial-up is limited to 53k.