Test4Theory has point problems... I suppose many do. Please read this thread.

http://lhcathome2.cern.ch/test4theory/forum_thread.php?id=1453&postid=16419#16419

It would seem a slow dual core CPU get more points than a 4.5GHz I5-3570 system on 24 hour runs...

I would REALLY REALLY like to know what's going on...

8-)

All the current tasks are supposed to run 24 hours and most do. I'm running the same tasks that the slower systems are running and I have NO PROBLEM getting tasks.

In fact, you completely fail to address the issue specified; that being why do two computers running THE SAME TASKS get different points!!!

Remember, the tasks are continuous and run based on a FIXED TIME... 24 hours per WU for now. The slower systems (look at the statistics) often get 20% or MORE points doing provably LESS work. This isn't always the case... some get 50% less! The point system is a complete whack job it seems.

I have to say your answer seems like gibberish to me... no flame intended, just that you seem to have completely missed the point and not looked at things or you would have discovered the link I provided IS IN FACT on the T4T website!

Sheesh... waiting for intelligent response from developer if possible.

8-)


UPDATE: It would appear that the person(s) getting higher points per task are running a slightly different application... a Multi-Core application... So the seemingly GROSS error isn't as bad as one would think. In fact, if I can test run those same MT apps, I would probably discover I did much better than the example system. We will see...

:D

Can you provide a link showing the difference in credits you're talking about? If you are looking on a "Workunit details" page, that shows different replications of a workunit on different hosts you might be looking at the "Claimed credit" instead of "Granted credit". In the case of workunits with multiple replications, the final "Granted" credit is averaged out by using the "Claimed credit" values from each of the participating hosts.

---

My Detailed BOINC Stats

at that point the t4t server will check the number of cern jobs done and grant credits based on the number of cern jobs done by the vm.

That's exactly what is not happening...

... and why Test4Theory provides his own stats and Billionares Club

Can you provide a link showing the difference in credits you're talking about? If you are looking on a "Workunit details" page, that shows different replications of a workunit on different hosts you might be looking at the "Claimed credit" instead of "Granted credit". In the case of workunits with multiple replications, the final "Granted" credit is averaged out by using the "Claimed credit" values from each of the participating hosts.

UPDATE: It would appear that the person(s) getting higher points per task are running a slightly different application... a Multi-Core application... So the seemingly GROSS error isn't as bad as one would think. In fact, if I can test run those same MT apps, I would probably discover I did much better than the example system. We will see...

Update-2:

I setup my system as the higher point people do it and I did in fact get higher points, but not as high as expected. The reason was the older wrapper uses multi-threading and the points get multiplied by the number of cores used.
However, it's still not equivalent to the difference in CPU speed... On a 24 hour run, I was able to get around 850 points and with the MT task, got 1400. Seems to me the points should have been closer to double at around 1600... but, I did notice that the tasks were not using all the CPU power available in the VM. For some reason, the VM seemed to be throttling to 75% usage or so and that could be the final reason.

8-)