This image shows the issue. Look at the claims at the top vs the claims at the bottom..WAY off of what they should be,,

[IMG]http://img96.imageshack.us/img96/7333/exq0.jpg[/IMG][/URL]

I checked this and no, this is single. IE: No wingman.
I also think you missed what I was saying. On these WU my machine is claiming app 10 WCG points per core per hour vs the 24-25 it was claiming before. I "beleive" this to be a function of the WU itself not my machine. Then if any more proof of this is needed is it credible to beleive that this machine which for all reality has the top cpu's money can but, 24 cores at 3226 MHz, IB arch, is only getting what a decent 6 core would get? Just no logic behind this that I can see and thats why I asked for help here..

I am not Lawrence, but I can offer another observation. I too have noticed this with my machines running Windows, and not only the ones running Linux. I have a q6600 running Vista and it gets very poor points, about 7 to 9 points per hour. At one time it was doing about 15 to 18 points per hour. I have done nothing different with the machine so I don't know why the difference. On the other hand I have an AMD 9150e with Win 7 which has consistently done 11 to 15 points per hour. It seems unaffected by whatever is affecting the Linux machines and the Vista machine.
I agree with you that something is amiss. My first thought is I do not think this has to do with your machine but with the way the server(?) is awarding points. But, after looking at your times, I am discarding this notion.Your times for these units should be much faster with the kind of horsepower you have in those processors, even given the Windows penalty you are paying. I know the units are inconsistent with times, but there seems to be a bottleneck someplace.
The only thing I can think of right off hand would be to cut the hyperthreading and see if there is a difference. Then try cutting the number processors in half and see if there is a difference. If the WU's show a rebound in how they perform, then you need to start looking elsewhere for the bottleneck. If there is no rebound, then maybe it is a software/OS problem. At the moment I feel like I am shooting in dark.
Good luck.
Cheers

Well, I have no qualifications to justify a response, but I'll dive in with some observations of my own anyway :

• The points that your machine claims are determined by the server, not your client.
  
• If a client is new (and maybe if it hasn't been running a project for a while, too) it tends to claim high to start with.
  
• After a few results the "levelling" process brings it down to a more consistent level.
  
• The mix of science you run on a client can affect the way any particular client claims points. (I assume that this is because different machines are more or less suited to specific implementations of the algorithms used. If one science app claims high on a particular client I think it has a knock-on effect to what the other science apps claim too.) I guess this could drag numbers either up or down, depending on the mix of app and architecture.
  

But the main point is: new boxes/apps tend to claim high at first and then come down. This seems to be what you're seeing.

What do others think?

Movieman,

In your first post you said "After two days of running ..." so I assumed that that was all the time the box (installation) had been up. The levelling process takes a number of WUs to sort itself out, so it may not be exactly time dependent. And I certainly don't know the details, but I wouldn't be surprised if it's based on some statistics which are gathered every 12 or 24 hours. Either way, my own feeling is that two days is actually pretty quick to sort out the levelling.

If I misunderstood I'm sorry. And I certainly have no other explanation for what you're seeing.