Not 100% all the time

JAZJAZ UK
edited March 2004 in Folding@Home
As you can see from the cpu usage pic (sorry for the super widescreen effect) instead of using 100% cpu all the time Im getting these reverse spikes, now Im sure it doesnt make too much difference as they only last a second or 2 but over a long time its going to make quite a difference to my folding power.
I've put a task manager processes pic as well to show what else is running but theres nothing else I can see that accounts for the drop, Im sure this has only just started happening (my weekly production has dropped a litlle from 5-600 to 450-550).
I will be re-installing windows as soon as my Radeon 9800 pro arrives in 2 weeks or so (damb stock shortages) so maybe it will go back to normal after this but I just thought I'd post and see if anyone else has seen this before.
Im running FAH4Console with -service -advmethods -forcesse and Version 1.56.

XP with all updates
XP2500+ @ XP3200+ (200 FSB)
2 WD1200's
NF-7 S
512MB Corsair Platinum (1 stick of)

Comments

  • csimoncsimon Acadiana Icrontian
    edited March 2004
    No it won't go back no matter what you do. The 98% you see if obvious because taskmanager was taking up 2% ...soon as you close taskmanager FahCore_78.exe goes right back up to 100%.
    As for the spikes ...that is the way the particular wu you are folding works ...I noticed this about a month ago when this wu was beta and reported it to Stanford and they addressed it and said that this is normal operation. Also there are other wu's that operate the same way and some do the same but not so often. Why this is happeneing I do not recall but I assure you it is normal operation that you cannot help while folding this particular wu.
    So which wu are you folding?
    Just make an observation for the next few wu's to see if it is still happening if you get a different wu and report back what you find.

    edit:\ rest assured that you are not losing any points or time with these wu's since the spikes have been factored into the point system. Also ...I would show you a link where I reported the spikes but it's in the beta forum so unless you are a beta tester for wu's you wouldn't see it. If you want to become a beta wu tester contact one of the admins (like pythagoras) and I'm sure he'd grant you beta access.
  • JAZJAZ UK
    edited March 2004
    Right, thanks for that info Csimon.
    Im running a p258_NpN now (Project: 258 Run 19, Clone 64, Gen 11) and I'll keep an eye on the next few I get like you said, thanks for putting my mind at ease :)
  • csimoncsimon Acadiana Icrontian
    edited March 2004
    You're welcome ...also p259 & p257 (I think) does the exact same ...this is the one I reported and the points were increased very high (to around 70 pts from around 30)when they noticed how much longer the spikes made the wu take.
  • Straight_ManStraight_Man Geeky, in my own way Naples, FL Icrontian
    edited March 2004
    Since I do not have time frame for your spikes, what I have noted is that you might look and see how often the Core and Client save a percentage, and how often the client does a Check Point file. If the Client and Core are running well, the Check Point value can be cranked up to max, about 30. Then check points are saved less often. Check points take about 2-3 seconds to save, and my boxes take on average 2-3 seconds to save after each percentage completed to write all local files.

    IF your time from one percentage or group of ten to another percentage or group varys, then you are getting a calc variation as things are aggregated possibly, but you might want to note that here I do not get the client-related spikes of WU percentage to percentage time in any WU I have seen, rather a slow increase in times for the more complex ones and a decrease in times for the less complex ones. When I say client-related spikes, I mean any that cannot be related to apps running or O\S load that I can account for. This is on both the Barton and the P4.

    Also, with XP, my boxes fold almost constantly. I run the percentage of CPU avialable to WU at 97-98%, not 100%. I use true color video, the video card is grabbing a bunch of RAM at 1280x1024x32 bit color. These are both production boxes. By dropping the percentage to 97, I get effectiveness (rounded to three places) between .975-.984 variable for WU calc, meaning I am doing them in about 1.5-2.5% of credit limit time (deadline time, based on the WU bench doen at Stanford).

    John D.
  • mmonninmmonnin Centreville, VA
    edited March 2004
    Run a genome serviced in the background and it will stay at 100% all the time. Plus with the caching of genome WUs you can stay at 100% for several days like I did when the internet went down from Thru night until Tuesday morning. That kept a 2.33 athlon folding the whole time.
  • JAZJAZ UK
    edited March 2004
    The spikes are one every 55-65 seconds John, I don’t think I'll mess with those "Check Point files" as I don’t know how to and Csimon's reply said the reverse spikes should only happen with a couple of wu's anyway, this makes sense as I’m sure this has only been an intermittent problem (i.e. prob only when I got a 258/259)
    I realise that its never going to use true 100% and perhaps I should have said 100% of un-used cycles.
    Thanks.
Sign In or Register to comment.