BOINC Pentathlon 2024

[TheLANguy]

Just finished swapping everything to NFS!

[CWP]

Let me see if I can find some more senior CPUs to blindly swing their canes at NFS...

[Kilrah]

Said it we should have started on NFS earlier yesterday... since today would be the first bonus day the earlier we're done with NFS the better, but now it'll take all day, and if we don't make it it's alll for nothing and we have to repeat

[leadeater]

27 minutes ago, Kilrah said:

Said it we should have started on NFS earlier yesterday... since today would be the first bonus day the earlier we're done with NFS the better, but now it'll take all day, and if we don't make it it's alll for nothing and we have to repeat

I think we'll be fine, our MW hourly hasn't dropped by much and I've been on NFS for a while. I switched over my Linux systems few hours before the start, that 700k bump is when I released what I had.

 

Looking over my systems now though, I was expecting higher hourly than what we have been getting.

[justpoet]

My minor bunkers have been released.  I will say that it very much appears as if MW has had other folks drop bunkers today.  So bunkering there is still worthwhile after bonus fall off, even if not as effective as it normally is.  Unfortunately I still can't get my computer to grab more than a couple hours of WU at a time for it anymore.  

[porina]

4 hours ago, Baha said:

Is it possible to configure boinc to report results more often? I have no limits on network usage configured, but all of my machines are holding onto a number of hours worth of completed work at a time before automatically submitting, which makes my throughput much more erratic.

If they haven't changed it, there is a config file way to "report results immediately". Projects can also influence it. Most projects probably don't do that to reduce server connection load. 

[justpoet]

Eureka I've found it!  In the project specific preferences for MW they have a drop down for maximum number of tasks.  For some reason it was set to the same as max thread count.  This was overriding the amount of time to buffer setting.

[will0hlep]

Reporting back on this from a few days ago: On a 12900H, it is about 40% better to run 1 task at a time using only the P-cores, compared to running 2 tasks at a time using all cores.

[will0hlep]

We are now 37.21% of the way to the stopping point on NFS. Problem is that SG are basically at their stopping point already so should be switching back to MW any minute now.

[Justaphf]

36 minutes ago, justpoet said:

Eureka I've found it!  In the project specific preferences for MW they have a drop down for maximum number of tasks.  For some reason it was set to the same as max thread count.  This was overriding the amount of time to buffer setting.

Thanks for that, was wondering why I couldn't get more than 10-16 tasks per machine.  Grabbed over 100 now, though staying on Jav till we hit our goal.  I guess not enough bunkering happened for this throw.

[justpoet]

1 minute ago, Justaphf said:

Thanks for that, was wondering why I couldn't get more than 10-16 tasks per machine.  Grabbed over 100 now, though staying on Jav till we hit our goal.  I guess not enough bunkering happened for this throw.

Yeah, Jav looks like it needs some serious help or we wasted time doing this throw and will have to do the next.  I have my one system that can MW doing it and the others doing Jav....but I'm definitely in the geriatric CPU category there.  LOL

[HoldSquat]

we pretty much have no choice but to do this jav, or we might miss the next bonus day for steeplechase

[will0hlep]

The validation delay on some of these projects is insane. Apparently, I managed to earn 2k points on Ramanujan yesterday enough though the city run ended 3 days earlier.

18 minutes ago, HoldSquat said:

we pretty much have no choice but to do this jav, or we might miss the next bonus day for steeplechase

Problem is that (assuming the obstacle days are random) there is only a 1 in 6 chance that tomorrow (day 5 of jav) has a higher scoring bonus than today. There is like a 2 in 3 chance that today is the best remaining day to do MW.

 

Admittly, if the bonus days arn't random, then my maths dosn't hold and were into the realm of psychology.

But we've started now, and swapping to day 5 would mean all jav work today was wasted...

[Kilrah]

7 minutes ago, will0hlep said:

Problem is that (assuming the obstacle days are random) there is only a 1 in 8 chance that tomorrow (day 5 of jav) has a higher scoring bonus than today.

There are only 2 more steeplechase days after today, so 1 chance in 2 the 2nd bonus day is tomorrow...

I'd speculate they leave it for the last, but we won't know until tomorrow...

[justpoet]

6 minutes ago, will0hlep said:

The validation delay on some of these projects is insane. Apparently, I managed to earn 2k points on Ramanujan yesterday enough though the city run ended 3 days earlier.

Problem is that (assuming the obstacle days are random) there is only a 1 in 8 chance that tomorrow (day 5 of jav) has a higher scoring bonus than today. There is like a 3 in 4 chance that today is the best remaining day to do MW.

 

Admittly, if the bonus days arn't random, then my maths dosn't hold and were into the realm of psychology.

It is more that if we don't make it, and have to throw again, we've wasted a bunch of bunkering time already, making sure people would have to miss even more MW to bunker for the next throw.  In almost every scenario other than no more bonuses for MW, getting this throw to the threshold is the strategically better choice.

[Kilrah]

21 minutes ago, justpoet said:

 I have my one system that can MW doing it and the others doing Jav....

We need everything we can on jav, we've got 3.3M to go, currently 200k/hour, that's 16.5h, which is essentially the hours left in the day

 

...or give up now and prepare for tomorrow, doing what we should have done yesterday

 

  

On 5/14/2024 at 3:36 PM, leadeater said:

We do however need to bunker NFS 12 to 16 hours out from the start of the next throw.

but then he changed it to say not to

[will0hlep]

15 minutes ago, Kilrah said:

There are only 2 more steeplechase days after today, so 1 chance in 2 the 2nd bonus day is tomorrow...

I'd speculate they leave it for the last, but we won't know until tomorrow...

yes, but after they decide which day, they also have to decide the size of the bonus (20%, 30%, or 50%). Today, we got a 30%. The only way tomorrow can be more important to MW than today is if they choose it to be the bonus day (1 in 2) and they choose the bonus to be 50% (1 in 3), which means that if these bonus days are random, there is only a 1 in 6 chance that today was the right day to throw.

 

14 minutes ago, justpoet said:

It is more that if we don't make it, and have to throw again, we've wasted a bunch of bunkering time already, making sure people would have to miss even more MW to bunker for the next throw.  In almost every scenario other than no more bonuses for MW, getting this throw to the threshold is the strategically better choice.

But @justpoet is right, now we've started the throw, we can't stop.

[will0hlep]

41.78% of throw complete

 

Our current throw garentees us 7th

We need a throw of 2,315,301 for 6th

We need a throw of 4,179,395 for 5th

We need a throw of 5,261,595 for 4th

[mariushm]

I've left the client running at home and got up to nearly 80k points on that Ramanujam machine

 

What's the deal with the next projects, where should I move .... i guess steeplechase because javelin should be done by now 

 

 

[leadeater]

1 hour ago, Kilrah said:

but then he changed it to say not to

I said I  wouldn't because I knew I could generate a decent amount on the day, not everyone else not to 

[leadeater]

1 hour ago, will0hlep said:

which means that if these bonus days are random, there is only a 1 in 6 chance that today was the right day to throw.

Problem is we didn't know leading up that today was going to be a bonus day either. Main issue was that bunkering MW was more effective than it sounded like it would be and in either case we would have had to choose to bunker NFS or MW and at the time NFS was the correct choice, problem being not enough actually did so 

 

The absolute correct thing to have done was to bunker 5.8m NFS the day before, release and then spend the next 23 hours on MW (takes about an hour for me to get my NFS queue uploaded).

[will0hlep]

21 minutes ago, mariushm said:

I've left the client running at home and got up to nearly 80k points on that Ramanujam machine

 

What's the deal with the next projects, where should I move .... i guess steeplechase because javelin should be done by now 

We still need javelin throw atm. Switch everything to that.

[leadeater]

To make up for the above I'm going to provision 11 more dual Xeon 5118 servers to do NFS, they can only do today/tonight (10 hours ish) however

 

Edit: 

Done, all added

[mariushm]

Well, I got it working..  I'll stop ramujan when it's done in around 20 minutes when it's done with most units. Hopefully nfs at home will auto increase the threads.

 

 

 

[Kilrah]

Looks like HardOCP is aiming to validate their first 4.77M score instead of the 2nd 5.26M and not do tomorrow since they've basically stopped, thus saving us from having to push >5M