• 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.

Help! Getting junky WU

B

budec

Gawd
Joined
Jul 10, 2002
Messages
748
In the last few days, I've been getting these really weird work units. They are a 100 frames, but are taking 2-4 days to complete (on both AMD and INTEL procs). Most of the other 100 workunits I could crunch out 1-3 per day per proc at apox 3-5 points per hour (according to EM Calc). This new work units are giving less the .5 points per hour!

Those really big ones (400 frames) where giving out 6-7 points per hour, how do I get some of them? These new 100 frames are taking longer then the big 400 frame ones and giving less points overall and giving less points per hour. I don't understand.

My AMDs are cranking on these right now:
Protein: 1000 p724_Abeta21-43-amberGS (time per WU: 20:56 @ 63.0 points)
Protein: 385 p558_BBA_pf (time per WU: 56:13!!! @ 47.0 points (wtf!))


How can I get it to force those big 400 frame WUs? Those are worth good points.... I have 6 procs (see sig + duel P4 3.0 Ghz I got today) "stuck" on these crapy work units. Is there a way I can cancel them and request another workunit that hopefully won't suck as bad?
 
i appear to have a couple of those myself
proj 559 - 160/3/5
it takes nearly 3 hours per frame and even though the machine is a slower one its rediculous, way below regular pts/hour
ive had 2 so far it takes forever
 
If your talking about the p559_BBA5_N...I'm seeing just a tad over 7 min per frame on a 2.0 gig AMD. Take a look and see if you have -forcesse turned on.

Curious

BillR
 
I dont think all of the 558 and 559 proteins are bad because ive got them on 3 of my systems now and they are taking between 9-30min per frame depending on the system.
 
I have -forceSSE on both AMD/INTEL and all 6 WUs are taking forever.
 
When did you last check for virus, spy ware,mal ware.

I do on average.
Points, Protien, Frame time, Total time, Speed
46, p559_BBA5_N, 00:06:44, 11:13:20, 2100 Mhz, Average from 6 runs.
47, p558_BBA5_pf, 00:06:42, 11:10:00, 2100 Mhz, Average from 20 runs.

Have you got the save.xyz box checked in EMIII.
That way you can bring each protien stats up and see how much variation in your folding speed you are getting.
The lowest rated protien I've had lately in terms of points per hour was a DGromac one which only gave me 3 pph at 2.1 Ghz.

So if your folding that slow I would guess that something is stealing cycles.
I would run a full system check.

Luck......... :D
u=Tigerbiten.gif
 
good advice, but its also the first thing done when i first saw it like 2 weeks ago. System running top notch and the foldingcore is getting all the cycles. Its currently on an Amd k6-2 500mhz and im wondering is its lack of instruction support whats making it take so long? otherwise, i cant come up with anything reasonable.
 
On my PIII 700 here at work, it takes me about 38 hours to chew through the p558_BBA5_pf protein if the system is not in use. That's at about 23 minutes per frame. I don't know about my machine. I haven't seen it since Wednesday night :( and don't know what it's chewin' on right now. Plus it was down for about 2 1/2 days before that. I will have to go visit it tonight so it knows I haven't abandoned it.
 
There are three cores that F@H uses when folding protiens.
Tinkers, Gromacs & DGromacs.
They tend to all fold at the same speed on an Intel system but a different speeds on an AMD.
I dont run an Intels so no data for them.
From what I've worked out on my AMD's.....
Tinkers fold around 3.5 points per hour per Ghz.
Gromacs fold around 2-3 points per hour per Ghz with SSE.
DGromacs fold around 1.5 points per hour per Ghz with SSE.
Tinkers dont use any optimizations so it makes no difference in speed if the SSE or 3dNow+ boost is turned on or not.
Gromacs with only 3dNow+ will fold around 1/3 slower.
Gromacs without any optimizations fold at 1/2 speed.
Only done 1 DGromac so I dont know if you will get then on a slower boxen and how well any optimization boost works.

So from that an 500 Mhz AMD without any optimization boost will fold a Gromac at around 0.5 points per hour.
I'm sorry to say you might be folding at full speed on that box.
If the CPU cannot use SSE then you may be better with the -forceasm flag as opposed to the -forceSSE one.
That will make sure that the 3dNow optimizations are turned on.

With Intels you want to add the -advmethods flag. That will make you more likely to download Gromacs/ DGromacs as opposed to Tinkers.

There is no way of downloading a particular work unit type/ protien number.
You just have to take what you are given.
But you can bias it slightly with the use/ no use of the -advmethods flag.

Hope this helps.

Luck........ :D
u=Tigerbiten.gif
 
budec said:
How can I get it to force those big 400 frame WUs?
Click to expand...
In theory you could block traffic to specific servers, and eventually Stanford's Assignment Servers would assign your box to a server you prefer (process of elimination). That could mean a lot of idle time waiting for a WU and if that server is handing out more than one project (and most do I think) its still a crap shoot. You'd have to stay on top of what projects are running on what servers, etc... a bit more administrating than I'd want to do. You wouldn't be doing the project any favours either.


budec said:
Is there a way I can cancel them and request another workunit that hopefully won't suck as bad?
Click to expand...
You get what you get, we're all in the same boat. Points ARE fun, but think of the bigger picture. :)
 
Weird. All those workunits finished in the middle of the night and the new ones I got are giving about 3-6 points per hour (depending on the CPU it's running under). Didn't change anything... they just came though. Very odd.


For furture reference though, for AMD procs I should run it with
-forceSSE

and for intel procs I should run it with:
-advmethods (and -forceSSE??)
 
budec said:
For furture reference though, for AMD procs I should run it with
-forceSSE

and for intel procs I should run it with:
-advmethods (and -forceSSE??)
Click to expand...
That's true, but I'd call it "for current reference" since it depends on how Stanford scores new WU's that come out. Just a couple of months ago we used -advmethods on everything, but the new Tinkers shot up in points so they became much more attractive to run on AMD's.
 
I used to notice on older machines that if a WU was shut down, or stopped improperly the remaining WU's would fold very slow. Can't explain it, but if I had that happen I would del the work folder and everything except the .exe program and the .cfg file and restart the program, all went back to normal. Haven't had that happen in a very long while, but it used to.

Just a thought, fold on bros

BillR
 
So what you're trying to tell me BillR is that should I feel the need to reinstall ... I could
just copy the .EXEs and the .CFGs and everything else (like F@Hlog.txt etc.) will be
regenerated automatically?
 
Hyperion, if you really want to reinstall, you should only carry over the .EXE file. You will also want to delete the registry entries pertaining to folding (search for pande group, I believe).
 
BillR said:
I used to notice on older machines that if a WU was shut down, or stopped improperly the remaining WU's would fold very slow. Can't explain it, but if I had that happen I would del the work folder and everything except the .exe program and the .cfg file and restart the program, all went back to normal. Haven't had that happen in a very long while, but it used to.

Just a thought, fold on bros

BillR
Click to expand...
That was the purpose of using the -forceSSE or -forceasm flags. If the client is improperly shut down it thinks that the CPU was at fault and did not handle the faster options.
So when restarted it would run without the options and usuall be a LOT slower.
 
You must log in or register to reply here.
Back
Top