Estimation on WU end absurd...

[vfrex]

It is claiming that I will finish my current work unit in Feb 07. I don't understand what the issue is. This WU has displayed a reasonable prediction of when the WU would be finished before. What is it doing now? I did play around with overclocking a bit and tried running it. Could that have screwed things up?

[GIBSON]

It is claiming that I will finish my current work unit in Feb 07. I don't understand what the issue is. This WU has displayed a reasonable prediction of when the WU would be finished before. What is it doing now? I did play around with overclocking a bit and tried running it. Could that have screwed things up?

meh, I wouldn't care about it, look at the log file and see how long it takes to get a percentage. Get your own ETA that way.

[Das Capitolin]

It is claiming that I will finish my current work unit in Feb 07. I don't understand what the issue is. This WU has displayed a reasonable prediction of when the WU would be finished before. What is it doing now? I did play around with overclocking a bit and tried running it. Could that have screwed things up?

I don't use, or recommend, the GUI client version. The console client does not make estimations, but it does give due dates.

[Bio-Hazard]

It is claiming that I will finish my current work unit in Feb 07. I don't understand what the issue is. This WU has displayed a reasonable prediction of when the WU would be finished before. What is it doing now? I did play around with overclocking a bit and tried running it. Could that have screwed things up?
As a bit of time goes by it will refigure the comp date back to where you were before, it only takes a bit of time. It'll do that everytime you you mess with your CPU and reboot as the GUI goes through its benchmarks.
I use the GUI version exclusively and perfer it to the console, I have the GUI completion dates and time to be right on the mark. It keeps me informed as to when I can plan for more benchmarking other items without interupting the current WU.
And it's just as fast....;)

[Das Capitolin]

And it's just as fast....;)

Except it won't run two instances, and a lot of people have HyperThreaded or dual-core CPU's. This is why I suggest console versions; along with less overhead.

[Bio-Hazard]

The overhead isn't really that much to even worry about........;) It doesn't effect the performance of the system what's so ever as you are already loading up the system with folding.
As for 2 at one time, I like having my second core free to do other thngs, like play games. While other folks shut down folding to play, my WU just keeps on folding..........:)

[werty316]

Don't believe what the ETA on the GUI version says.

[Bio-Hazard]

Works for me with in say a hour or 2.........;)

I plan things by it and I've been folding for differant teams for something like 4 years now.

Just to prove my point, I have one that's supposed to be finished by 1530 local time here, so if you check the standings after the 1600 Hr update, you'll see my WU and points added...............;)

[Kougar]

Each folding core has built in optimizations to take advantage of SSE SSE2 3dNow! and many of the other CPU instructions to greatly speed up processing time.

If your folding@home client is running when your machine crashes, upon being restarted it will detect this and assume it was the cause of the crash. It will then disable ALL of these instruction optimizations to try and prevent future crashes, which can turn a several day WU into a 2 week WU. Not sure if that is the case for you or not, but it's worth mentioning. If you did lock the machine trying to OC while it was running, then it's safe to say this is at least part of your reason why. ;)

I've seen some pretty absurd esitmations from the GUI before though, usually they'd fix themselves after it had time to run for awhile. I'd suggest making sure it was getting most of your CPU time though and that no other processes are hogging the CPU in the background.

[vfrex]

I didn't lock up on this particular instance (or any that I can think of). I'll give it more time when folding on the OC'd frequency and see what happens. The core was consistently taking 99% of my CPU time, so no problems there. Oh well.

[werty316]

I did some testing to see if have a CPU OC'd helped with Folding and I didn't notice an improvement.

[GIBSON]

To prevent happening that it shuts the optimisations flag off you should use the -forceasm flag. Read up on it in the F.A.Q. I've got it set up that way as I like things to go as fast as possible 8)

[Kougar]

I'd suggest only using that flag if you are 100% sure the system is stable and yer compeltely done trying to squeeze anymore performance out of it. ;)

[GIBSON]

I'd suggest only using that flag if you are 100% sure the system is stable and yer compeltely done trying to squeeze anymore performance out of it. ;)
You shouldn't be folding anyhow if you aren't sure. If you don't you might screw up the whole WU!

[Bio-Hazard]

I'm folding to idenical 20,000 WU on different A64 systems right now, the one on my main system is running at 11 sec per frame at 2.75 Gig OC and the other is running at 17 sec at 2.25 Gig. At least that's what it's showing. The slower system is just dedicated to folding and does nothing else and still is a bit slower, both folding programs are configured the same and both rigs have 2 gigs of ram. So at least on my rig, the OC does show a improvement.

[vfrex]

Ok, it just got weirded. The date in windows was set to September 2005 :???:

[GIBSON]

Ok, it just got weirded. The date in windows was set to September 2005 :???:
Aha! So around a year should be correct then eh! :lol:

[vfrex]

Well, I guess it isn't folding related, but I don't understand why my clock was off like that. I couldn't seem to get the NTP option in windows to go. Could that be related to instability of the overclock? F@H also threw an error last night, raising doubt that my overclock is as stable as I previously thought.

[Kougar]

Well, a few times when I'd leave my system on overnight heavily loaded with torrents and other stuff running I'd come back to find the system's clock lagging a few hours behind... never did figure out exactly why. ;)

I can certainly say having the windows clock time off can throw stuff out of whack, Gmail Notifier won't work unless the system time matches the server authentication certificates for instance. ;)

[GIBSON]

Well, a few times when I'd leave my system on overnight heavily loaded with torrents and other stuff running I'd come back to find the system's clock lagging a few hours behind... never did figure out exactly why. ;)

I can certainly say having the windows clock time off can throw stuff out of whack, Gmail Notifier won't work unless the system time matches the server authentication certificates for instance. ;)

Yer, wrong systemclock can screw up a lot of things. Wasn't there even this bug that would let you run pirated windows copies by changing something in the clock or so?

[Kougar]

I don't think that bug went quite THAT far... or if it did then I've never heard of it until now which is possible... ;) I knew it worked well enough for time- based demo programs though, roll the clock back a few months, install, put it back to the current time and whatever program it was would usually work just fine.

[Kougar]

I did some testing to see if have a CPU OC'd helped with Folding and I didn't notice an improvement.

OCing apparently has only a small effect... I ran two F@H clients on a Core 2 Duo, then tightening the RAM timings after doing a 55% overclock. Both WUs were taking 18 to 19 minutes a cycle at stock, and OCed they both improved only to 15-16 minutes per cycle.

[GIBSON]

I don't think that bug went quite THAT far... or if it did then I've never heard of it until now which is possible... ;) I knew it worked well enough for time- based demo programs though, roll the clock back a few months, install, put it back to the current time and whatever program it was would usually work just fine.
Don't you mean forward as I don't quite see it working the way you described?

OCing apparently has only a small effect... I ran two F@H clients on a Core 2 Duo, then tightening the RAM timings after doing a 55% overclock. Both WUs were taking 18 to 19 minutes a cycle at stock, and OCed they both improved only to 15-16 minutes per cycle.
Now that's odd, isn't it supposed to scale with the clock?

[Kougar]

Don't you mean forward as I don't quite see it working the way you described?

I think yer right... It's been a long time since I was in high school doing that stuff anyway. :wink:

Now that's odd, isn't it supposed to scale with the clock?

That is the catch, almost nothing will scale linearly with an overclock. Alot of it is the rest of the system, and the exact same processor design being used. After looking up the first two projects I have this rig folding the "slow" progress got me grinning again as both of these WUs are worth 600 points apiece. :twisted:

[GIBSON]

So, from what I understand the OC actually is a lot faster?

[werty316]

So, from what I understand the OC actually is a lot faster?
Just read Kougar's posts here:
http://www.bjorn3d.com/forum/showpost.php?p=45550&postcount=24

If a WU takes 3 days and an OC'd CPU saves 1-2hrs its still 3 days...

[GIBSON]

I was referring to the fact he was grinning again. I remember him posting the non-overclocked system did about 17 mines for a cycle but the OC'ed system did 15. Now I was thinking that 17min might be for a low point WU and therefore the mentioned the two WU's he's folding now are 600 points each. That would explain why he started grinning again as this would mean it is indeed a lot faster. It's just an interpretation I guess, Kougar, clear this one up will ya :-D

[Kougar]

lol, sorry Gibson!

I can't honestly say either way definitively, because it's just subjective... For some people that kind of an increase is good, for others it would mean almost nothing.

My system finally went unstable and knocked out one of the F@H programs. So much for 4 hours of dual Prime95ing meaning anything... Long story short, it's back up and running with a very much improved airflow design. But the new F@H WU it downloaded is also a 600 pointer, Project 2413. I can say it is definitely faster, but it's nothing ground shattering all the same. ;)

[GIBSON]

This kind of got me thinking. Would the extra cache of a 6600 help with folding? Or is folding not that cache hungry?

[Kougar]

I am sure doubling the cache would definitely help. It's definitely a memory intensive program. But again any improvement would be by a small percentage (I want to guess 5%?), nothing that great of an improvement as ya may think either way.