Equinox
09-06-2004, 06:13 PM
I gave this project a shot but I find it inefficient and easily exploitable.
Sieving, trial factoring, and double checking should be a part of the main project which they currently are not and this can easily lead to wasted proth testing which can take as much as two weeks. Personally I find this unacceptable.
The way statistics are computed is terribly flawed. The K factor has little effect on proth testing yet it is given a large weight in computing the stats which leads to easy exploitation of the entire stats records. I am not the brightest bulb in the world and I can very well imagine others have discovered this and restart their clients till they get the best K factor and hence better stats per work unit. I discovered this little foul up when I was trying to restore a unit which stopped upon a reboot. I accidentally entered the ID number (368007) of the unit in stead of the K value (10223) and ended up getting over 4 M cEMs'/sec.
When the client is first installed and set to work its first unit -will- be dropped and a new one started upon the next reboot. This is a known bug and only happens after the first reboot after install. Can you imagine how many wasted cycles this bug has undoubtably caused? The work-around for this involves editing of the windows registr. What the heck is it doing there?
The good news is that the developers seem to be aware of these issues and are addressing them in version 3. I will probably come back to this project during beta of the new client and try and help fix any issues with it, but until then I find this a complete waste of cycles.
Sorry if this comes off as a rant. I actually care a great deal about distributed computing and wish the best for every project. If I have made any inaccuracies in my assessment of this project please correct me.
I don't currently have access to the forum where the vote for September is going on so I have decided that I will move over to Distributed Folding and give them a boost during thier final month of operation.
Sieving, trial factoring, and double checking should be a part of the main project which they currently are not and this can easily lead to wasted proth testing which can take as much as two weeks. Personally I find this unacceptable.
The way statistics are computed is terribly flawed. The K factor has little effect on proth testing yet it is given a large weight in computing the stats which leads to easy exploitation of the entire stats records. I am not the brightest bulb in the world and I can very well imagine others have discovered this and restart their clients till they get the best K factor and hence better stats per work unit. I discovered this little foul up when I was trying to restore a unit which stopped upon a reboot. I accidentally entered the ID number (368007) of the unit in stead of the K value (10223) and ended up getting over 4 M cEMs'/sec.
When the client is first installed and set to work its first unit -will- be dropped and a new one started upon the next reboot. This is a known bug and only happens after the first reboot after install. Can you imagine how many wasted cycles this bug has undoubtably caused? The work-around for this involves editing of the windows registr. What the heck is it doing there?
The good news is that the developers seem to be aware of these issues and are addressing them in version 3. I will probably come back to this project during beta of the new client and try and help fix any issues with it, but until then I find this a complete waste of cycles.
Sorry if this comes off as a rant. I actually care a great deal about distributed computing and wish the best for every project. If I have made any inaccuracies in my assessment of this project please correct me.
I don't currently have access to the forum where the vote for September is going on so I have decided that I will move over to Distributed Folding and give them a boost during thier final month of operation.