9th Annual BOINC Pentathlon 2018 Strategy & Preparation

Here's how I've spread the systems for now after tracking down all of the boxes and getting things cleaned up:

8x systems on Asteroids
21x systems on NumberFields

Status for Asteroids:
State: All (15806) · In progress (8713) · Validation pending (1535) · Validation inconclusive (0) · Valid (374) · Invalid (0) · Error (5184)

Status for NumberFields:
State: All (11541) · In progress (5940) · Validation pending (0) · Validation inconclusive (0) · Valid (3236) · Invalid (0) · Error (2365)

Ignore the error count on both of the above; it's 99% from me aborting lots of WUs.

We'll see how things go on Asteroids over the next ~24 hours. With the whole wingman thing for this project, people would be idiots to be bunkering IMHO... so hopefully it'll be a good indicator of how teams are outputting.

NumberFields is a different story of course. No wingman and no reason not to be bunkering for the last few days to try and pull off an upset. Personally I'm not going to mess with that, on purpose at least, so my output should pretty much be WYSIWYG. ;)
 
Here is the comparison of the top 5 crunchers in NumberFields for TeamChina versus [H] as of today.

Their top 4 are outgunning our top 4. Hopefully this will change in the next day or so. They recruited wwei25 over the past 7 days and he/she is ramping up production in order to catch up with us. husq just power up his/her rigs since yesterday. These two are the big ones, so watch our back.

I just switch one more rig to NumberFields for a total of 3 full rigs.

We must continue to defend our medal and hopefully make history for the first ever BOINC pentathlon medal for the team!

Go [H]ordes!!!

upload_2018-5-14_21-16-39.png


upload_2018-5-14_21-15-10.png
 

Attachments

  • upload_2018-5-14_21-16-0.png
    upload_2018-5-14_21-16-0.png
    20.6 KB · Views: 0
Dropping bombs with dual booting machines over Teamviewer is a real pitfa. Totally worth it :)

The good news is that nearly all the Primegrid tasks validated.

The bad news is that only about half of the Asteroids did despite the tasks being picked up shortly after the event was announced. There are some serious bunkers coming our way from the teams behind us.
 
Currently have 3,318 tasks awaiting validation. This JMHO and am likely to post this in the open forum; but if teams are still bunkering this project, they're a bunch of idiots. There's absolutely NO guarantee that dropping a bunker now will result in a majority of tasks having a processed wingman, nor of being validated within a reasonable amount of time. Last I checked, the projects validator is running way behind and it's just going to get worse with bunker drops.
 
pututu, or two? Then again, maybe you know what project1 and project2 are all about. :)

That being said, their backlog is supposedly zero right now with no WU's waiting for validation or assimilation. Huh.
 
Well those stats are not always correct either.

But... lets say they are. The "waiting for validation" could be when they have both tasks and the server is validating them. Although, I feel this is most likely not the case and the stats page is just reporting incorrectly.

I released a bunker of CUDA units late last night that I downloaded last week. Their deadline was today so i went ahead and upped them. Had a 100% valid success rate on them.

Also had a few hundred (somewhere around 400) tasks that were CPU, most of those validated except the ones I downloaded when the project was first announced.

I think I had around 1200 validated tasks when it finished. Took around 30 minutes.
 
User has four computers with more tasks out than I have awaiting validation. I'd rather suspect this is a pretty serious bunker for their team.
 
User has four computers with more tasks out than I have awaiting validation. I'd rather suspect this is a pretty serious bunker for their team.

Still willing to bet those tasks never make it to completion. They'll time out and/or get aborted by the user.
 
motqalden, am guessing that's for PrimeGrid? Need some context. ;)

And, Skillz, maybe. S/He is sitting on over 4,400 WUs. At least three of those appear to be genuine systems, too. Guess we'll see. Would be nice if there was a way to punish people who do that. So meaning horde WUs and then cancel them at the end to deny other people valid credits. Some people think it's a legitimate strategy; I think it's pure bullshit.

Meanwhile the nine systems I've left on Asteroids are churning out the WUs and the pending pile keeps getting bigger. Thing about this project is that any dumped bunkers help other people, too.
 
BTW, a 4P Skylake Platinum was supposed to arrive today; but FedEx managed to fuck that up. Was really hoping to get that online tonight for either Ast/NF and I most likely won't be in tomorrow. Damn damn damn.

Box will have 192 threads w/ HT enabled.
 
Skillz Naw it isn't stable, even with 7.

It's the only rig i have that isn't built around a 'mining' board. Gonna swap it out for a H110 Pro BTC+ in a few here after i eat da suppaz.
 
Oh you have H110 Pro BTC+ handy for the swap? Skip dinner. It's not important man! We need the megahurtz meow!

You also got mentioned in the chat box on the seti-germany pent site. atp1916
 
Some crazy shit that people do in this pentathlon. http://asteroidsathome.net/boinc/show_host_detail.php?hostid=535490

S/he turns 12/24T machine to 64T and keep it running like this since the event started.

No way this rig can finish the 1800+ tasks by the end of this pentathlon. Look at the run time.

To be fair, I've 680 tasks in progress spread across 2 rigs that I won't be able to finish. So should I place Mr Nice guy and aborted the tasks that I can't finish or hold on till the pentathlon ends? What others are doing?

I bet that there are lots of people who are trying to bite off more than they can chew.

So this event is likely to have plenty of tasks with pending validation.
 
Is this normal or just a reporting bug from Afterburner?

My 7970 machine on an older driver and older afterburner is showing smoother, and seems to be getting more done than my fury

huh.png
 
rgMekanic make sure you are giving it enough CPU's to feed the gpu tasks. I ended up having to specifically set the number of cores for NF on one of my rigs because it was stealing from Prime too much and simply lowering the boinc CPU% didn't help. If you fully suspend all CPU tasks does it do better, and are you running any custom GPU config?
 
Ah well - this last rig won't go up. New mobo, same problems.

Could be anything - risers, cables, pcie connector, a short, blown wire...

Leave it be for now, on NumberFields.
 
Everyone needs to be on Numberfields. Ditch asteroids.
for numbers does it matter which application we are running? mine are picking up both
Get Decic Fields
Minimum Discriminant Septics

with the minimum discriminant septics having a way longer run time which I seem to have way more of
 
not really. The gain we have with PrimeGrid would be more than what the gain would be with Asteroids for the hardware in question.
 
For GPUs, are we at a point we should be dropping PrimeGrid and picking up Asteroids?

Primegrid ends 1 day before asteroids so don't put gpu on asteroids till they fully close out prime, then move to asteroids for gpu on last day
 
And again... we need CPU's moving to Numberfields as TC is nipping at our feet rather aggressively.
 
Our lead over TC is now only 1.5 mil but their production last hour was 3x ours. Getting kinda scary! Some of their other production on other projects has gone down , wondering if they are focusing on pushing numberfields now to secure a slot.
 
Our lead over TC is now only 1.5 mil but their production last hour was 3x ours. Getting kinda scary! Some of their other production on other projects has gone down , wondering if they are focusing on pushing numberfields now to secure a slot.
You bet they do ... at our cost :-(
 
When I looked at there team production it's almost all that one user with 3.5+ million points today....
 
I've been moving all my various systems back to NF and have opted to use the slightly shorter project on some of the higher core count boxes. I'll see what else I can do.
 
Back
Top