I'm adding power whenever possible. currently have a P4 3.0 w/HT running, a P4 2.4 no HT running, and my 3200+ running (3200+ is temp while i wait on RMA from newegg)
perhaps some more P4s / P4-HTs coming too
all running under the dogs, of course.
I'm adding power whenever possible. currently have a P4 3.0 w/HT running, a P4 2.4 no HT running, and my 3200+ running (3200+ is temp while i wait on RMA from newegg)
perhaps some more P4s / P4-HTs coming too
all running under the dogs, of course.
Ok, need some help here!
I'm still getting a total of app 4 million on the 4 clients.
First system specs( yes, I know, you guys heard this 50 times BUT mabe someone who hasn't will see the problem.
Dual Xeon 3600/2mb(each)/800
Supermicro X6DA8-G2 MB
2X1 gig Viking/Corsair DDR2-400 @3.3.3.9
2-36 gig raptors in raid0
other stuff shouldn't matter..
Here's what I have in the config:
all defaults except changed from idle to normal in the attempt to see if that would help..does nothing..
So speak to me guys, I may not be here a lot longer and want to give you all I can.
Crunch with us, the XS WCG team
The XS WCG team needs your support.
A good project with good goals.
Come join us,get that warm fuzzy feeling that you've done something good for mankind.
Originally Posted by Frisch
Give me a taskmanager screenie! With processes! My rate on one instance shot back up to >5XXXXXX when i had downloaded a new block with k = 19249.
EDIT: Just to clarify.... approx +/- 1mil per Client?
Last edited by [XC] riptide; 08-26-2006 at 05:03 PM.
here you go:yes, to the 1 mil per client:
Crunch with us, the XS WCG team
The XS WCG team needs your support.
A good project with good goals.
Come join us,get that warm fuzzy feeling that you've done something good for mankind.
This is a shot in the dark. I'm miffed about this. But..... what affinity do you have set for each of the 4? Again this is a brainstorm! I don't know how SOB is handled by windows internally. Again you have 2 x 3600 with HT? Try setting each instance to use just one logical processor each! Do this through taskmanger now, before you start messing around with the service installer/command line. (PS Any change if any will take a while to reflect in the rate)I'll have my own DX tomorrow and will be able to get hands on eperience in a short time myself and will advise you more.
This is 2-3600's with HT,yes, the nice Irwindales with the 2mb cache, the $920.00 each Irwindales last august when I built it( I paid $1245.00 for the pair) I didn't set any infinity but I can..This is a shot in the dark. I'm miffed about this. But..... what affinity do you have set for each of the 4? Again this is a brainstorm! I don't know how SOB is handled by windows internally. Again you have 2 x 3600 with HT? Try setting each instance to use just one logical processor each! Do this through taskmanger now, before you start messing around with the service installer/command line. (PS Any change if any will take a while to reflect in the rate
HMMM//Won't let me set affinity..at least through task manager and thats how I usually do it..
access is denied
Last edited by Movieman; 08-26-2006 at 05:24 PM.
Crunch with us, the XS WCG team
The XS WCG team needs your support.
A good project with good goals.
Come join us,get that warm fuzzy feeling that you've done something good for mankind.
Hmmm thats cos they were started using the service sobsvc. Damn... then lets try 1 instance on its own. Kill them all, then restart one manually. Be insteresting to see what % of increase you get with just one going. Also when you restart you get a better idea of rate very quickly.
just over 1 mil..like it's hitting a stop..thats just one client running..other 3 stopped..in task manager shows one cpu at 100% and the other "3" at zero..
now at 1013800
whats this optimised I read about for SOB?
Last edited by Movieman; 08-26-2006 at 05:40 PM.
Crunch with us, the XS WCG team
The XS WCG team needs your support.
A good project with good goals.
Come join us,get that warm fuzzy feeling that you've done something good for mankind.
Wow! I would expect it to be 4mil on its own! Even at your already low score, dropping to 1 instance per physical CPU should almost double the rate. Its curious that it can handle 2 per CPU at 1 mil and also 1 at 1 mil. There is something serious wrong! Did you restart SOB. I mean not just kill 3 and leave the 4th. Did you kill all 4 and restart 1 of them. Hardly the program is corrupted?
EDIT: What optimized?
EDIT2: Baaaah optimized for amount of instances vs proc numbers(incl. HT). I had proposed that a 1.6 Xeon (512kbcache) would work better with 3 instead of 4 instances. He tried and reckoned that 3 was indeed 5% better. Well itsa also in the readme about Xeon and instances, but I think it's especially true for 512 cache procs.
Last edited by [XC] riptide; 08-26-2006 at 06:00 PM.
I opened the client windows(all 4) stopped 3 and exited those. The 4th I stopped and then restarted. The other day Kaptain Blazzed mentioned something about optimisations..maybe he menat the dos commands with the :p4..etc.Wow! I would expect it to be 4mil on its own! Even at your already low score, dropping to 1 instance per physical CPU should almost double the rate. Its curious that it can handle 2 per CPU at 1 mil and also 1 at 1 mil. There is something serious wrong! Did you restart SOB. I mean not just kill 3 and leave the 4th. Did you kill all 4 and restart 1 of them. Hardly the program is corrupted?
EDIT: What optimized?
be right back, going to boot and shut of HT..
Crunch with us, the XS WCG team
The XS WCG team needs your support.
A good project with good goals.
Come join us,get that warm fuzzy feeling that you've done something good for mankind.
In terms of you're last post, I'm truely confused! Maybe th program is corrupted or something. Do you bench? Would you do super pi run? At this time. Might throw up something? After that I'm confused! Really connfused. More confused that the rate did not change with only 1 instance.
get this: with HT shut off and still 4 instances running they are all in the high 900K range..just a little less than with HT
2 instances now..climbing past 1.3 mil each and still going
ok, UNCHecked that box so that they will NOT transmit between blocks..that was stopping them from climbing beyond 1.5 mil per client..now over 1.6 mil each on 2 clients
Last edited by Movieman; 08-26-2006 at 06:18 PM.
Crunch with us, the XS WCG team
The XS WCG team needs your support.
A good project with good goals.
Come join us,get that warm fuzzy feeling that you've done something good for mankind.
Go go go!
EDIT: Its in my guide to uncheck that box if you notice it idling at all! The client stops crunching for a split second and some cases for more. Even if you have it reporting to SBQueue proxy server on the same MACHINE!!!!!
Last edited by [XC] riptide; 08-26-2006 at 06:20 PM.
it was doing one block, would get to a certain speed like 1.2 mi, then back off, opened task manager and cpu went to zero while that happenned..now it is flying almost 1.75 mil each..again, thats on 2 clients
Yup, still climbing..1.8 mil each now..what threw me is that box and the length of time it takes to max..I wonder what the max is here as it keeps climbing
Last edited by Movieman; 08-26-2006 at 06:27 PM.
Crunch with us, the XS WCG team
The XS WCG team needs your support.
A good project with good goals.
Come join us,get that warm fuzzy feeling that you've done something good for mankind.
UPDATE! With the help of that rabid dog from the Emerald Isle,Riptide I Think we have made some headway on making my DX3600 at least get some reasonable numbers: Just checked and with 2 clients I'm totaling just under 8 million cEMs/sec..Still not showing the xeon optimisation message in the status box so I have my doubts as to whether it's actually using the SSE3 instruction set. Beats the crap out of the 4 mil total I was seeing before he helped so thank you Riptide for all your help pal!
Crunch with us, the XS WCG team
The XS WCG team needs your support.
A good project with good goals.
Come join us,get that warm fuzzy feeling that you've done something good for mankind.
Ya check the stats. You were at 160 for today already! Dave, that message will not stickie there. It would soon be written over by another message about uploading blocks, server busy etc. I reckon that there is some sort of delay happening in the use of SSE2.UPDATE! With the help of that rabid dog from the Emerald Isle,Riptide I Think we have made some headway on making my DX3600 at least get some reasonable numbers: Just checked and with 2 clients I'm totaling just under 8 million cEMs/sec..Still not showing the xeon optimisation message in the status box so I have my doubts as to whether it's actually using the SSE3 instruction set. Beats the crap out of the 4 mil total I was seeing before he helped so thank you Riptide for all your help pal!
For those interested it seems for MM that the message that appears in the status box at start up about ' recognising Pentium 4 3.6Ghz blah blah "Enabling specific CPU optimiations"' was not apprearing. It did however appear after he used CrystalCPZ (sp?) like cpuz. But maybe that was conincidence. Dave I'll remind you again. You should see about 80% of your final rate within the 1st few seconds. When I went to bed, it was still quite low and was rising slowly. SSE2 ops must have initialized during the night and given a few hours, your full rate shows!
EDIT: I'm not sure it uses SSE3 Dave, SSE2 I know it does!
your probably right, my eyes aren't even functioning yet! Hey, it's better, we;ll get the team some decent points with this!
I think the key was when I used that little taskassign app and set the affinity for both cpu's. That made the beech start to cook!
Last edited by Movieman; 08-27-2006 at 02:37 AM.
Crunch with us, the XS WCG team
The XS WCG team needs your support.
A good project with good goals.
Come join us,get that warm fuzzy feeling that you've done something good for mankind.
MM,
just a possibility but you may have some DNS issues when it connects to transmit each block. I personally have my own DNS server locally so the connect time is minimal..
You could try putting this into your c:\windows\system32\drivers\etc\hosts file
69.9.160.188 sbp.pns.net
so that it doesn't go to your DNS server every time.
Note, This is presuming you are using this one and not the sb.pns.net.
Bok
Bok
Welcome. Dave wasn't getting the "Enabling CPU ops etc etc" message. He got it after it had been running sometime.... and it just appeared! So not sure what was going on there.... but... how does SOB pick up the CPU. From software or harware?
rippy
no idea where it picks it up from...I'd presume software.
I've got a similar system, dual xeon 3.6 nocona's here, though it has a Gentoo 64bit build on it (development stats server :p ) and it runs just fine with 3 clients. I'm sure it just picks it up by parsing a 'cat /proc/cpuinfo', but whether it does something similar on windows, we'd have to ask the SoB guys...
Bok
Thanks for the input Bok. Seems to have just peaked at just under 4 mil per client. Oh, just posted at freedc, you might want to read that..no idea where it picks it up from...I'd presume software.
I've got a similar system, dual xeon 3.6 nocona's here, though it has a Gentoo 64bit build on it (development stats server :p ) and it runs just fine with 3 clients. I'm sure it just picks it up by parsing a 'cat /proc/cpuinfo', but whether it does something similar on windows, we'd have to ask the SoB guys...
Bok
Crunch with us, the XS WCG team
The XS WCG team needs your support.
A good project with good goals.
Come join us,get that warm fuzzy feeling that you've done something good for mankind.
Ukraine dropped another >40T this morning! I wonder how long they'll keep this!
Hey! Are the SOB servers down! I can't flush my work and I notice our last update was..... poor to say the least!
EDIT: Servers are back!
Last edited by [XC] riptide; 08-28-2006 at 03:06 PM. Reason: who cares!
Can someone explain what this means in the client status box:
n.high=5064708
Thanks..
PS: Still not showing the xeon optimizations in that box
Crunch with us, the XS WCG team
The XS WCG team needs your support.
A good project with good goals.
Come join us,get that warm fuzzy feeling that you've done something good for mankind.
n.high I believe is the range of n numbers on that particular k that have been tested. Still worried about the lack of optimization signs!
Posting Permissions
Reply With Quote
Bookmarks