• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

FAHSpy Points Not Adding Up

W

Waypo

Weaksauce
Joined
Oct 11, 2005
Messages
66
I've got two networks with computers at my disposal to run Folding. Naturally I'm taking advantage of this. Network 1 has 52 computers on it the majority of which are P4 2.4GHz 512 RAM, nothin special. Network 2 has 15 machines on it the majority of which have 1Ghz P3s.

The problem is FAHSpy is telling me that the machines on network 1 should be producing 4500ppd to 5000ppd. Some machines get turned off here and there not a big deal. Network 2 according to FAHSpy should be producing about 2000ppd.

Now if you look at my stats you'll see a drop in points about a week ago. That is because I lost internet at my apartment and my two fastest machines are sitting idle. But, shouldn't I still be getting about 6000ppd if FAHSpy is even close to correct.

http://www.hardfolding.com/index.php?go=38&tm=33&id=147162
 
I have been using FAHMon, but it seems to always overestimate PPD for me. I really just use it to see that everything stays up and running. I probably wouldn't worry about it.


 
One big problem with the folding monitors is that they don't take into account the amount of time it takes to upload/download a new work unit. I am not saying that is your entire problem, because with uniprocessor clients you don't spend that much time uploading/downloading, but that is a contributor. With 52 computers, depending on the internet connection speed you might be getting choked a bit because of multiple clients trying to use the network at the same time.

Also, if these computers are getting used much they will loose points per day pretty quick. Are they in a lab where there are classes or anything?

 
The usage of all of those machines is saddly low. I would like to think that I could get maybe 60-70% or what FAHSpy is reporting. That seems reasonable to me.

Additionally the lab with 52 is getting upgraded to Core2Duos in December. How many points would you guys estimate that to be?
 
Would you have administrator access to these Core 2 Duos? With administrator access you could put the SMP client on there which gets a higher point allocation per unit by far. However, that client can tend to be a bit buggy and with 52 clients, you would be spending a good bit of time restarting computers and trying to figure out errors quite probably.

You could easily double your points per day with core 2 duos at the least. But, if you can put up with the hassle and have administrator access you could do far better with the SMP units.

 
Perhaps SMP when it's no longer beta. The trouble I've had with it on my Xeon makes me say definitly no for now.
 
I don't know if this helps or not I read somewhere FAHmon was about 10% optimistic. What I do is take the FAHmon reading and subtract 10% to give me a more accurate reading. I've also read whare using the "last 3 frames" type of monitoring was more accurate. Like I said I don't know if this helps and to be honest I don't know if FAHmon or any F@H monitor is accurate. I still get different readings with 3 different boxen with 3 different FAHmon programs (go figure) :confused:

I don't claim to be anywhere near being an expert so would one of our generous and knowledgeable folders please chime in ?:confused:

Thanks :D

Folding and WCGing for a CURE

 
Another thing to consider, a while back there was an error on the psummary page and it caused fahmon to show double the PPD for that particular project. If there's any little glitch in the variables it's using to calculate PPD (even if its not as large of an error as that), that can really add up when multiplied by 52 machines.


216
 
You should be doing roughly 40 PpDpGhz.
So (52 x 2.4 + 15) x 40 = 5,500 PpD.
Plus any bonus work-units you happen to grab.

How did you install the clients ??
Did you install the clients by hand or did you clone the drives with FaH installed ??
How long have you been running the farm ??
Has every box turned in work and when ??

Remember that the monitoring programs work on what is being crunched while the stats work on whats been sent back in.
So if you get some of the bigger work-units then it could take a week or three for the slower boxen to turn work in and your stats will show a nice spike when they do.

Luck ....... :D
 
I've just looked up your stats over at Stanford and it only shows 22 active clients in the last 7 days.
Your PpD is about right for that number of machines running.
22 x 2.4 x 40 = ~2,000 PpD.

How did you install the clients ????????
Have you ghosted a drive with the client installed ??

Look in the top of all your FAHlog.txt file and record the User ID shown there.
Each one needs to be unique.
If not then you will be downloading the same work-unit to multiple boxen, they will all crunch and upload it, but you will only get credit for the first upload.

Luck .......... :D
 
Wow thanks guys!

I did ghost the machines so I'll check the user IDs.

If the IDs are wrong how would I go about changing them?

Thanks again.
 
You need to edit the registry to remove your User ID from it.

Start -> Run -> regedit
Scan the registry for folding@home.
Should be ............
HKLM -> Software -> Wow6432node -> PandeGroup -> folding@home. <- 64 bit.
HKLM -> Software -> PandeGroup -> folding@home. <- 32 bit.

On your first box just record the User ID and which Work-Unit your crunching, Project ( Run, Clone, Gen ) numbers.
Now on all your other boxen, stop the client and lookup you User ID & Protien.
If the User ID is a duplicate then just delete it.
If the Work-Unit is a duplicate then delete the work folder and Queue.dat file, otherwise record the Project ( Run, Clone, Gen ) numbers.
Once you've cleaned out the duplicate bits, just restart folding and it will download a new User ID.

You will probably have to delete all bar one of the User ID numbers but you should be able to save around 20 part crunched work-units doing it that way.

Next time you ghost a drive with folding on it.
Remove the User ID from the registry and delete the Work folder and Queue.dat files from your folding folder, before you ghost it.

Edit. That above is for Windoze.
Linux keeps a machinedependent.dat which has your User ID in, just delete it.

Luck .............. :D
 
Looks like the User ID was the problem. They were all the same and most were working on the same WU.
 
Glad you got to the bottom of it. Things should start looking up for you shortly.


2.png
 
You must log in or register to reply here.
Back
Top