Its not quite a 4 way opteron, but both machines i'm putting up are dual HT xeon boxes, so i'll be dumping 8 einstein WUs every 11 or so hours, they still going to be legit for this comp?Berserker said:I'm also not going to worry about multiprocessors this time, although if someone comes up with a 4-way Opteron or something, then I may have to rethink that.
Team OcUK BOINC Crunchathon weekend *** DISCUSSION ***
- Thread starter Berserker
- Start date
OK, it looks like people are trying to find me out on the multiprocessors, so I'll have to separate them. I'll amend the rules later today when I've worked out the best way of handling things. 
Berserker said:OK, it looks like people are trying to find me out on the multiprocessors, so I'll have to separate them. I'll amend the rules later today when I've worked out the best way of handling things.
Maybe some sort of averaging out based on the boinc reported ops/sec, or a maximum that counts as one machine ?
Dual CPU machines tend to get lower ops/sec than singles (with the exception of Opterons). I also notice that 4.2x-4.3x clients appear to benchmark all CPUs simultaneously.
Mine, or theirs? 
Berserker said:Mine, or theirs?
The issue would appear to be at the CPDN end.
The CPDN site itself seems to have stopped showing trickles from sometime on Monday morning, as well as not showing an increased total credit.
Both BOINCstats and yourself show that there's been no new credit since then (indeed, yours says there's been no update to the stats since the 18th).
Soldato
Anyone?rich99million said:
I've started my two machines on Einstein already so it's all ready anyway - race on!!!!
rich99million said:Damn i hadn't thought of that - totally forgot about BOINC's silly validation process, hmmm that makes it a whole different prospect
Validation dosen't seem to be a problem just now. I only re-started BOINC at about 01.00 this morning and I have already gained 206 cobblestones on my previous total (9 out of 14 results validated so far).
Assuming the validators don't break, everything should be hunky dory by the time the comp starts
Stan
not 100% sure its possible to tell, but the 4.2x clients benchmark much lower than the earlier ones - by approx 20% or so - 4.25 should be minimum - though I guess self compiled clients are ok if thats possible to tell as well.
OK - How's this for dealing with multiprocessors...
Hosts with two or more processors are permitted, but will have a handicap applied to their outputs. This will be applied by multiplying the reported credit by a handicap factor once for each reported CPU. The handicap will be: 0.89 for Xeon systems showing more than 2 CPUs with speed of 2.0GHz or more, 1.00 for Pentium 4 systems showing up to 2 CPUs, 0.71 for Opteron-based systems, and 0.85 in all other cases.
Just to show how this works, let's take a Dual Opteron, Dual Athlon MP, and Dual Xeon HT, all with 1,000 credit:
Opteron: 1000 * 0.71 * 0.71 = 504.1
Xeon: 1000 * 0.89 * 0.89 * 0.89 * 0.89 = 627.4
Athlon MP: 1000 * 0.85 * 0.85 = 722.5
Single CPU: 1000 (no handicap)
The weightings may still need a bit more tweaking, so I'm open to suggestions on those. The Xeon weightings are intended for the Nocoma core. I can't tell the difference between, for example, Nocoma and Prestonia, so older Xeons will be at a disadvantage.
The other alternative is to keep the multiprocessors completely separate.
Hosts with two or more processors are permitted, but will have a handicap applied to their outputs. This will be applied by multiplying the reported credit by a handicap factor once for each reported CPU. The handicap will be: 0.89 for Xeon systems showing more than 2 CPUs with speed of 2.0GHz or more, 1.00 for Pentium 4 systems showing up to 2 CPUs, 0.71 for Opteron-based systems, and 0.85 in all other cases.
Just to show how this works, let's take a Dual Opteron, Dual Athlon MP, and Dual Xeon HT, all with 1,000 credit:
Opteron: 1000 * 0.71 * 0.71 = 504.1
Xeon: 1000 * 0.89 * 0.89 * 0.89 * 0.89 = 627.4
Athlon MP: 1000 * 0.85 * 0.85 = 722.5
Single CPU: 1000 (no handicap)
The weightings may still need a bit more tweaking, so I'm open to suggestions on those. The Xeon weightings are intended for the Nocoma core. I can't tell the difference between, for example, Nocoma and Prestonia, so older Xeons will be at a disadvantage.
The other alternative is to keep the multiprocessors completely separate.
Associate
Just to confirm, my single HT P4 stil only counts as one processor?
If not, I'll have to take the T-bird off the list.
Alan Woodford
If not, I'll have to take the T-bird off the list.
Alan Woodford
The restriction is two hosts (i.e. computers) per project, not two CPUs.
HT only counts as one CPU anyway.
HT only counts as one CPU anyway.
Associate
Guys - unfortunately I'm going to be out of work as of next friday - contract ends and found out today I'm not getting an extension this time round, been there two years so it had to end sooner or later .. wont be long before I'm back on track but it does mean I may have to De-Boinc what PC's I do have running there - even if thats the case I will be leaving at least two of them up for the crunchathon though i'm not sure if they will last long before they get spotted etc.
Soldato
unlucky mister rogue sir - i do hope those two will be running einstein so i can pound you into submission
muwahahahahahahha - errr sorry it's late and i'm sleepy
oh sorry to hear about the job - hope you get something else sorted soon
muwahahahahahahha - errr sorry it's late and i'm sleepy
oh sorry to hear about the job - hope you get something else sorted soon
RogueTLDW said:
That's very bad news
Hope you get another job soon. On the other hand, maybe I finally get the chance to catch you up As for this compo, unfortunately I'm not going to be around enough to massage my PCs, as I'll be buried in exams. Maybe next time.
OK - update on the stats. I'm going to start a reload of SETI@home stats sometime tomorrow. Hopefully that'll be done inside 24 hours, but time will tell. Other projects take a few hours each.
Then I just need to do the stats page. Hopefully Tuesday/Wednesday for that.
So, here's hoping I can get it all done in time.
Then I just need to do the stats page. Hopefully Tuesday/Wednesday for that.
So, here's hoping I can get it all done in time.
Berserker said:OK - update on the stats. I'm going to start a reload of SETI@home stats sometime tomorrow. Hopefully that'll be done inside 24 hours, but time will tell. Other projects take a few hours each.
Then I just need to do the stats page. Hopefully Tuesday/Wednesday for that.
So, here's hoping I can get it all done in time.
Loads of Pro-Plus washed down with Red Bull and you'll be fine
I'm sure we all appreciate your sterling efforts in organizing and running this competition (as well as the TeamOcUK stats site) and if we ever happen to wind up at the same OcUK meet, I will buy you a pint (or two).
Stan
At the moment, I'm trying to prevent the Team OcUK website from dieing for upwards of an hour during SETI@home updates (yes, it does - page load times of 20 minutes plus have been experienced personally).
As for meets, just don't think I'd cope on the BBQ/camping weekend, otherwise I'd be there in a flash. MB-organised events for the w1n.
As for meets, just don't think I'd cope on the BBQ/camping weekend, otherwise I'd be there in a flash. MB-organised events for the w1n.