New SMP WUs

mmonninmmonnin Centreville, VA
edited May 2008 in Folding@Home
Lately all I have seen are the 1760 pointers but I remember some smaller ones back when I first started SMP.

I looked at my points for today and I had 2 WUs within the same update for 4,288 points and 2 1760 point WUs cant equal that ammount.

So I look on the WU lising page and I see one listed for 2539 points (p3060_BBA5_1). I dont seem to have any more but does anyone know if its new or how well it folds?
«1

Comments

  • mmonninmmonnin Centreville, VA
    edited March 2008
    Looking back through some of the logs I see I've had some other WUs:

    p3062_lambda5_99sb - 1732 points - 1719 PPD Q66@3.24GHz
    p3064_lambda5_2003 - 1753 points - 1713 PPD E66@3.15GHz
    p3065_lambda5_99sb_big - 2144 points - 1510 PPD E66@3.15GHz
    p3065_lambda5_99sb_big - 2144 points - 1599 PPD Q66@3.24GHz

    The times were calculated with EM3 manually entering the time per % and point value. I usually get 2000-2100 PPD on each Q66 client and ~1900 on the E66 via the FahMon client so thats a nice point drop. :(
  • DanGDanG I AM CANADIAN Icrontian
    edited March 2008
    there's a couple of the new WU's that are killing me on points. Particularly a couple of the revisions of the 3064's, same with a couple of the 3062's. Down around 1200ppd when running them on SMP with the F@H affinity changer on a Q6600 @ 3.04.

    Not liking these new WU's...
  • mmonninmmonnin Centreville, VA
    edited March 2008
    I think I had been getting these for awhile now on one the clients on the quad. On Fahmon one client was around 2k and the other dropped down to around 1500-1600. Then they both went to that level and I restarted the clients. One was on the 2144 that I mentioned earlier and I lost it at 57%. I'd rather have the 1760 pointers as well.
  • QeldromaQeldroma Arid ZoneAh Member
    edited March 2008
    I've gotten so spoiled by the 1760 pointers that I've almost forgotten there are other SMP WUs out there ....

    These take a little longer and don't have quite the ppd. However, they may have other WUs that may be better point producers too- just haven't seen any yet.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited March 2008
    Thanks for bringing this up, Mmonnin. I'm on again/off again concerning keeping up with Stanford developments, so I had noticed the 'new' SMP units. Sure enough, production (well, at least points) is way off with 3062 and 3064 as compared to good ol' 2653. Such is the way it goes.
  • scottscott Medina, Ohio Icrontian
    edited March 2008
    I also have some of these "slow producers" did a little digging and found that they are a Dan Ensign project. As some of you may remember his gromac projects that took our socket A machines from 200ppd to 80ppd. His projects are harder, usually have less points and shorter deadlines. Here is a short discussion at the Folding Forum.

    http://foldingforum.org/viewtopic.php?f=8&t=1604

    Scott
  • QeldromaQeldroma Arid ZoneAh Member
    edited March 2008
    I had a previous summary that noted that the points on a 3065 was 1837 points and I was only getting about 960ppd on FAHMon. It looks like it was recently revised to earn 2144 points and that has helped my ppd production go up to almost 1100ppd.
  • mmonninmmonnin Centreville, VA
    edited March 2008
    Hmm I dont know what their test machine is to set WU points is anymore but it must not be any type of Core 2 machine as the PPD difference doesnt make sense.
  • QeldromaQeldroma Arid ZoneAh Member
    edited March 2008
    As the FCF thread discusses: I don't think I find the ppd potential as disconcerting as the fact that they are issuing WUs with 32 hour deadlines to machines that can't do them in less than 40, etc. I'm thinking of shutting off the advmethods flag to possibly prevent that.
  • mmonninmmonnin Centreville, VA
    edited March 2008
    I dont think forceasm or advmethods has an effect on the SMP client. That choice has made by using the client. I thought I read that in FAQ somewhere but it was probably on the old Fah Community site.
  • QeldromaQeldroma Arid ZoneAh Member
    edited March 2008
    You're probably right, but it was suggested without rebuff in the FCF forum and basically- there is a way to find out ... done. May know within the week :D .
  • SPIKE09SPIKE09 Scatland
    edited March 2008
    mmonnin wrote:
    I dont think forceasm or advmethods has an effect on the SMP client. That choice has made by using the client. I thought I read that in FAQ somewhere but it was probably on the old Fah Community site.
    some discussion a few weeks back about forceasm allowing wu's to survive powercuts etc and even though the entire SMP prog is beta wu's have the same evolution, internal testing/ beta / advmethods and then vanilla SMP wu's.
  • mmonninmmonnin Centreville, VA
    edited March 2008
    Its been beta for sometime and they are still quite unstable. /shrug
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited March 2008
    I can say this though - the new beta SMP client is much better for stability than the previous one. I always backup the Folding client folders before turning off machines, but I now rarely have to use the backed up files. It used to be that one out of every three or four shut downs the work units would be corrupted, and yes, that was with by-the-book shutdowns.
  • mmonninmmonnin Centreville, VA
    edited March 2008
    "by-the-book shutdowns" was never needed with their normal clients for years. Hitting the X, C+A+D or just turning off the PC did not corrupt the WU. Its unstable as hell compared to the normal client.
  • QeldromaQeldroma Arid ZoneAh Member
    edited March 2008
    mmonnin wrote:
    "by-the-book shutdowns" was never needed with their normal clients for years. Hitting the X, C+A+D or just turning off the PC did not corrupt the WU. Its unstable as hell compared to the normal client.

    I'm not sure, Leo, but I've shut down SMP on my laptop a dozen different ways- from CTL-C to power-button shutdown (sometimes I simply don't have the time) and I've rarely lost a WU- and when I did I could not finger a shutdown reason. I've tried to peg this issue and the closest I could guess is network stability. These WUs ARE crazy flakey on a wireless system ... or almost any network where it has to find itself negotiating a new IP now and then. However, I'm not certain. My other guess is your overclocking, but heck- we almost all do that.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited March 2008
    Let's try this again. Sorry if I didn't make it clear.

    The latest SMP client is much more stable than the previous client. I almost never get a corrupted unit anymore at shut down, whether the shutdown is Ctrl-C, Windows shut down without stopping the clients, or even a hard restart with the power button or abrupt power interruption.
    These WUs ARE crazy flakey on a wireless system
    Previously I did have problems with that. After I assigned static IPs to all the computers on the LAN, poof, that problem disappeared.
  • QeldromaQeldroma Arid ZoneAh Member
    edited March 2008
    I guess my point is then: I don't think the SMP client is any more stable ... still. We've simply changed our environments to make it run more stable. Me- I use ONLY wired connections. You use static IPs.

    IMHO: This new client works hardly better (or worse) on a wireless network than ones they released months ago. It also craps out just about everytime I unplug my wired network cord. I think it has to do more with negotiating the IP address and not with the IP address itself because it craps out whether it negotiates the same IP address via DHCP or not.

    I don't have the time or skill to debug all of FAH's bugs, otherwise I would have more exhaustively tested it and wrung out their source. So take it as speculation from what I've observed and read: I think this is an (perhaps THE major) unresolved problem with the SMP client- and it's plagued them since day one.

    I actually don't remember WU corruption- perhaps that's because I started SMP months after many of you did (April last year) and have been folding inconsistently. My major SMP hangs have basically been: 1. Time restricted Beta updates and 2. Network drop-outs.

    It seems like the FAH community is aware of it but, to date, there appears to be no fix.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited March 2008
    The only machine on which I get corrupted work units now is a machine that is controlled remotely through Ultra VNC. If I have to shut it down, it's 50/50 odds the work units will be trashed. All the work units on my other LAN connected machines, Ethernet and wireless, are nearly bulletproof. Sorry about your difficulties.
  • mmonninmmonnin Centreville, VA
    edited March 2008
    EUEs happen on stock SMP machines. That doesnt happen with the normal client. There is a reason SMP is still a beta client. SMP is better than what it used to be but its nowhere close to clients of the past.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited March 2008
    The corrupted work units I was referring to were all SMP units. I think I've had only one EUE with an SMP in the last three months with five computers, eight clients folding continuously.
  • mmonninmmonnin Centreville, VA
    edited March 2008
    OMG 171.64.65.63 is full of the ****ty WUs(3050 and 3052). All I could get were the horrible 1440 point WUs.
  • mmonninmmonnin Centreville, VA
    edited April 2008
    I'm about to turn off my computers since I keep getting all these ****ing ****ty WUs that I dont even get credit for. 19k this week when I should be doing 50K+ and I don't think I've lost any WUs either.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited April 2008
    Hey, Marc, don't sweat it. I think everyone is getting those zingers now, so the proverbial playing field is still level for us Win SMP folders.
  • QeldromaQeldroma Arid ZoneAh Member
    edited April 2008
    Leonardo wrote:
    Hey, Marc, don't sweat it. I think everyone is getting those zingers now, so the proverbial playing field is still level for us Win SMP folders.

    Yeah- had my first EUE in months just this morning. On a 3064.
  • SnarkasmSnarkasm Madison, WI Icrontian
    edited April 2008
    One of my clients has been looping around 34% on one because of a damn "long 1-4 interaction" report. It hits 34%, long interaction, it goes 3 hours without a checkpoint and restarts. Supposedly the only thing to do to fix it is to restart some short time before the report and continue it.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited April 2008
    Yeah- had my first EUE in months just this morning. On a 3064.
    I was about to roll my eyes, then thought I'd check the logs. Sure enough, I had a unit crash - but not EUE - on one of my dual core (Pentium D) machines yesterday. And that's the first lost unit not attributable to my actions in a couple months.
  • mmonninmmonnin Centreville, VA
    edited April 2008
    No EUEs, just no credit from crappy 306x WUs. I shut of my desktop as its the most power hungry and it just got a new WU.
  • QeldromaQeldroma Arid ZoneAh Member
    edited April 2008
    Sounds like my EUE is more coincidental than anything else- so I'll do the :rolleyes2 for you, Leo. It's just been so long since I had one I thought it might be related.

    I'm pretty diligent about cleaning out my systems regularly- clean them all about the same time, and right now I've wound them down for the summer but can't monitor them regularly with my schedule. However, I'm not seeing a lot of abberation in the WUs I've had working recently- many of them 306xs- and my production is about dead-on with what I'd expect.

    mmonnin- I'm half tempted to suggest your system/environment may need a more in-depth look-see.
  • DanGDanG I AM CANADIAN Icrontian
    edited April 2008
    I'm missing out on ~2000 ppd with these 306x WU's. Not impressed.
Sign In or Register to comment.