PPD down on SMP on FahMon

G

gnewbury

[H]ard|DCer of the Month - September 2007
Joined
May 4, 2001
Messages
2,544
Starting about Friday I noticed that my PPD were down on 4 instances of SMP across the board as reported by FahMon.
Instances on quad 1 were getting about 3500ppd, now about 3200
quad 2 (the tax machine) went from about 2100 to 1470
x2 went from about 800 to 700.

There's been no change on usage on quad 1 slight on quad 2 and I backed up a couple of directories on the x2.

Conversely my one machine that runs a single non SMP instance has skyrocketed, from 80ppd to 100ppd (big woop).

Anybody else using FahMon see something similar? It could just be that FahMon was reporting differently, but an overall drop of about 1K seems weird.

 
Check if you got a few 2619 WU. Those fucking bastard buggers fold for crappy points and like to fail at will... I'm pissed at them because yesterday, just because I wanted to restart my Q6600 to work on the guts (Watercooling), both instances trashed the WU for no reason. I restarted the laptop since I put a update and it got trashed as well...

Those WU caused a drop of close to 1500 ppd in my case...

 
There have been some "chewy" WU as they where called in another thread. I hope they are going to finish beta soon cuz this is getting sticky. :(

 
Kendrak said:
There have been son "chewy" WU as they where called in another thread. I hope they are going to finish beta soon cuz this is getting sticky. :(

Click to expand...

Aren't these the 306x WU's? I've been getting alot of those come down lately as well.


 
Xilikon said:
Check if you got a few 2619 WU. Those fucking bastard buggers fold for crappy points and like to fail at will... I'm pissed at them because yesterday, just because I wanted to restart my Q6600 to work on the guts (Watercooling), both instances trashed the WU for no reason. I restarted the laptop since I put a update and it got trashed as well...

Those WU caused a drop of close to 1500 ppd in my case...

Click to expand...

I've been getting killed by these 2619s. They take so much longer to finish if they finish at all. I switched to bigpackets=small and picked up another 2605.

 
alan2308 said:
I switched to bigpackets=small and picked up another 2605.
Click to expand...

The big packet switch is hard coded to big with the smp client due to the upload size.
If you could change it to a small setting, you would never be able to upload work.

I've found that with p2619 you need to backup your work folder and queue.dat file if your shutting down, so you can restore them after restarting.

Luck ............... :D
 
Yes, DEFINITELY BACKUP those WU when you want to stop the client (do it while it's running, not after stopping since it will be too late by then)...

 
I have one of the 2619s on my [email protected] and it has caused about a 100PPD drop on that machine compared to the 2605 and 2653s so it's not a big drop. That's the only machine I have those on right now.

The 3062 and 3064s on my main [email protected] is a different story. The PPD on those instances have dropped about 300PPD each totally 600PPD less on that machine.

They other two quads running VMs are still pulling in 2605 and 2653s and doing their normal PPD.

My problem at the moment seems to be lost work units. My production has plummeted the last few days. I thought it was because I took down a couple of my machines to swap them into new cases. However, the downtime doesn't even come close to equaling my production drop.

I looked at the log for all seven of the clients I have here at home and every one of them successfully turned in their work units. At least 4 of the clients turned in the work units between 10pm and 1am my time. It's now almost 2pm my time and I still don't have credit for a single one of these work units. I'm not exactly a happy camper and I'm suspecting that my drop in production over the last few days has something to do with lost work units rather than anything else. There's no other way to explain why my production has dropped to half of what it was before.

 
Tigerbiten said:
The big packet switch is hard coded to big with the smp client due to the upload size.
If you could change it to a small setting, you would never be able to upload work.

I've found that with p2619 you need to backup your work folder and queue.dat file if your shutting down, so you can restore them after restarting.

Luck ............... :D
Click to expand...

That may be, but over at FCF, there has been a few reports that it will keep you from grabbing a 2619. It could be just a coincidence though.

 
OK, update -
Now FahMon is reporting a little bit better ppd for my 2 instances on 1 quad, but what's amazing is that on my AMD 2500+ with 1 regular fold on a 2606 it's saying 184ppd!! That's a 230% increase over the 80ppd 2 days ago.
points are getting flaky!

 
I now have 3X 3065's and my Farm has gone from a peek of 11k ppd down to 9.5k

And Sunin is right behind me.... not fair!!!

 
Anyone here get any of those "Recovered" units? My production has been normal... Although one day I dipped without having any issues with Computers at all... 6.8k one day and my avg is 12k? Not sure what happened that day!

 
Sunin said:
Anyone here get any of those "Recovered" units? My production has been normal... Although one day I dipped without having any issues with Computers at all... 6.8k one day and my avg is 12k? Not sure what happened that day!
Click to expand...
A few members have including one or two myself. If you have unaccounted WUs they would have already shown up by now I believe. The spike in the production graphs was around Sunday. Unless more than one server was affected that I don't know about. How much do you think you're missing approximately?
 
I cannot track any missing WU so if it pop up, I wouldn't know :p

Show how much I'm producting...

 
Sunin said:
Anyone here get any of those "Recovered" units? My production has been normal... Although one day I dipped without having any issues with Computers at all... 6.8k one day and my avg is 12k? Not sure what happened that day!
Click to expand...

I think I ended up down around 20x p2619 work-units over a period of 5 days.
I got them all back in one update.

Luck ........... :D

 
Xilikon said:
Check if you got a few 2619 WU. Those fucking bastard buggers fold for crappy points and like to fail at will... I'm pissed at them because yesterday, just because I wanted to restart my Q6600 to work on the guts (Watercooling), both instances trashed the WU for no reason. I restarted the laptop since I put a update and it got trashed as well...

Those WU caused a drop of close to 1500 ppd in my case...

Click to expand...

What descriptive terminology Xililon :D
I agree with you whole heartily, that 2619 WU is a PITA (and that's an understatement :rolleyes: )

What really bums me out, I just had two boxen folders 24/7 go down in flames (my ppd went to sh*tzz), finally got one back up and what the f*uck do you think my first WU was, a freakin' 2619 :mad:

Welp, it's all for the cause anyway :)

 
BTW, jws2346, Dave from OC.com says hi. Thought I'd pass it along since I saw your post. :)
 
Tigerbiten said:
I think I ended up down around 20x p2619 work-units over a period of 5 days.
I got them all back in one update.

Luck ........... :D

Click to expand...

Somehow I didn't get a single 2619 so I guess I was not affected by this.

 
Sunin, the 2619 WU is Linux and OSX only so if you don't have Linux SMP boxes, you will never get them. Speaking of this, I might reconfigure my main PC with 1 VM and 1 WinSMP to limit the PPD loss with the 2619 that I'm getting all the time on linux boxes.

 
APOLLO said:
BTW, jws2346, Dave from OC.com says hi. Thought I'd pass it along since I saw your post. :)
Click to expand...

Thanks APOLLO

Tell Dave a big hello back for me, please.

While I think OC.com has a great team, IMO team 33 is much better :D

 
Xilikon said:
Sunin, the 2619 WU is Linux and OSX only so if you don't have Linux SMP boxes, you will never get them. Speaking of this, I might reconfigure my main PC with 1 VM and 1 WinSMP to limit the PPD loss with the 2619 that I'm getting all the time on linux boxes.
Click to expand...
Ah, that explains why I haven't received any 2619s either. I was going to configure multiple VMs on my dual quads since they're only running XP, but these systems have only 3GB of RAM or less each. I don't know what to do now...

jws2346 said:
Thanks APOLLO

Tell Dave a big hello back for me, please.

While I think OC.com has a great team, IMO team 33 is much better :D
Click to expand...
Yes, this team has been #1 so long for very good reason. OOC, what made decide to fold for [H]?
 
APOLLO said:
Ah, that explains why I haven't received any 2619s either. I was going to configure multiple VMs on my dual quads since they're only running XP, but these systems have only 3GB of RAM or less each. I don't know what to do now...

Yes, this team has been #1 so long for very good reason. OOC, what made decide to fold
for [H]?
Click to expand...

Ahh.., first off I realize this is very off topic for this thread. I want to appologize to my fellow members. :)

Now as far as what caused my decision to leave team 32 and start folding for team 33 is a very complicated question, one I think should be left in the past and buried. It's all forgotten now and should be. (we're talkin' more than a million pernts ago) Suffice to say, I had personality differences with some of the people on team 32. (this is in retrospect and I may just be imagining this :rolleyes: )

Now tryin' to get back on topic.

Does anyone think these 2619s might be the WU of the future for LinSMP? Would adding more or faster memory be worth the cost? :confused: (I think DDR2 is pretty cheap ATM)

Somewhere or somehow this WU is going to be outfoxed and when it is, it'll probably be done by one of the members of [H] folding team 33. No wonder team 33 is #1, it's members have more than the usual amount of computer intelligence, thanks you all for letting me fold for team 33. :p

Don't any other team member that's not lucky or smart enough to fold for team 33 get their "shorts in a twist", I said team 33s members had more than the usual amount of computer intelligence. I know all teams have their computer geniuses, team 33 just has more of them :D

* Disclaimer *
Of course team 33 is numero uno ATM and the above is just my opinion, but you can't deny a "stone cold truth".

 
jws2346 said:
Somewhere or somehow this WU is going to be outfoxed and when it is, it'll probably be done by one of the members of [H] folding team 33. No wonder team 33 is #1, it's members have more than the usual amount of computer intelligence, thanks you all for letting me fold for team 33. :p

Don't any other team member that's not lucky or smart enough to fold for team 33 get their "shorts in a twist", I said team 33s members had more than the usual amount of computer intelligence. I know all teams have their computer geniuses, team 33 just has more of them :D
Click to expand...
Yeah, like this member and his setup in the networking forum. Nuts doesn't even come close... :eek:

I understand about what you said concerning other teams. MPC's forum is OK but rather slow. OCAU can't be read without registering. DL.TV, an up and coming team doesn't even have a F@H forum. OC.com is a good place, though. I still want to see woods' mass borg succeed.

 
Sunin said:
Anyone here get any of those "Recovered" units? My production has been normal... Although one day I dipped without having any issues with Computers at all... 6.8k one day and my avg is 12k? Not sure what happened that day!

Click to expand...

I got a few "recovered" units considering I had a 26k+ point update at one time which from my rough calculations would be about right.

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