16/10-23/10 Who Will Win The Rosetta Rosette ? Rosetta@home

Post by **scole of TSBT** » Wed Oct 11, 2017 12:23 am

Usually the longer the WU runs the higher the credit per run sec. I compared your linux X5650 system to my win64 X5650 system. On that 20 WU sample mine did avg about twice the credit per run sec than yours but credit per run sec varies widely based on the type WUs. The roberta WUs pay the best and I've been getting a lot of those. They begin with rb_. I wish there was a way to download just those.

Here are some hosts to compare if you want...
https://boinc.bakerlab.org/rosetta/show ... id=3275469
https://boinc.bakerlab.org/rosetta/show ... id=3275460
https://boinc.bakerlab.org/rosetta/show ... id=3274758
https://boinc.bakerlab.org/rosetta/show ... id=3215829

Bryan and I tried to do a comparison but we couldn't find a clear pay advantage but Win64 may be paying better right now.

Post by **scole of TSBT** » Wed Oct 11, 2017 1:01 am

I'm going to roll the dice, let the active linux WUs finish, then boot to Win64 and reload.

Alez · Post by **Alez** » Wed Oct 11, 2017 1:34 am

Since my servers only run nix , guess I'll stick with it. Had to open a couple of systems with units about to expire. Think all I've got now should be good till the start. The GPU systems I'll close in a few days time. Not going to be huge, but will be my best bunker effort for a while. Rosetta credit is as crap as seti.

Post by **Bryan** » Wed Oct 11, 2017 1:57 am

Well we arrived in southern TX! Damn I'm glad that drive is over :)

I hope GC doesn't show up ... the last challenge without them was a pleasure. Maybe we should get their "whitelist" and only issue challenges on their black listed projects

I'm running the 12 hour WU also but I have the lid on so I haven't turned in any yet. All my systems are on Windoze. When we got here I tied into the home machines and everything is running and I think there is enough work to make it to the start. Of course if there isn't I'll just bring up more BOINC instances and get more.

Post by **Dirk Broer** » Wed Oct 11, 2017 3:34 pm

Bryan wrote: ↑Wed Oct 11, 2017 1:57 amI hope GC doesn't show up ... the last challenge without them was a pleasure. Maybe we should get their "whitelist" and only issue challenges on their black listed projects

http://www.gridcoin.us/Guides/whitelist.htm

Post by **Bryan** » Wed Oct 11, 2017 4:05 pm

I have 8000 valid WU from where I was running with resource share of 0 before I started my bunker. I looked at them this morning and found the same thing Scole did about the type of WU. The rb_ and 4hm pay about double so If you are in a position do so then you need to run them in this order:

rb_
4hm

Then everything else. I think I have more WU than I can handle before the start so I've suspended everything except those 2 types. I'll have to run some of the lower paying ones but at least all the higher paying WU will be ready to turn in

Alez · Post by **Alez** » Thu Oct 12, 2017 1:09 am

Had to open some systems again. Failed to get enough work for the full bunker, my own fault, guess I didn't leave enough time to download and messed up the cpu time. Oh well, at least we get the credit to get us away from the drop zone.

Post by **scole of TSBT** » Thu Oct 12, 2017 1:24 am

NO! Setup an additional client.

Alez · Post by **Alez** » Thu Oct 12, 2017 2:11 am

Too late, plus mucked up major with disk size, plus can't be arsed going out to tech shed and messing . Won't have time to check results as busy for next 2 days so ... . Apart from that , all good

Systems all buttoned up until the start.
Still no GC. Maybe they are following our boycott of all their pointless challenges.

Alez · Post by **Alez** » Thu Oct 12, 2017 11:02 am

McShane of AP wrote: ↑Thu Oct 12, 2017 4:53 amTalking about running extra clients. . Would appreciate a link please.

link

Post by **scole of TSBT** » Thu Oct 12, 2017 1:12 pm

When bunkering for rosetta you don't have to download 1000 WUs per system. The project allows you to choose the estimated runtime per WU, anywhere from 1 hr to 24 hrs. If you're going to bunker, set the estimated runtime to 24 hrs and load one per thread per days you need.

Post by **Bryan** » Thu Oct 12, 2017 1:46 pm

@Pete ... if you take the lid off your network it will upload your finished WU. IF YOU SUSPEND the project on those machines that have finished and DO NOT hit update then it should just upload the finished work but not report it. If you hit update you are screwed.

The easiest way to prevent that is use BoincTasks and set ALL computers to NETWORK not allowed. Then allow the new client to have network access and open up your firewall. After you load up the new machines then put the lid back on your firewall.

Alez · Post by **Alez** » Fri Oct 13, 2017 12:59 am

McShane of AP wrote: ↑Thu Oct 12, 2017 10:50 pmversion 7.8.2 should be skipped due to some serious problems like the slots not clearing.

Also has issues with removing projects with Bam. Can stick you into a neverending contact loop. Best avoided.

Think I read somewhere that Stop also has issues with WUProp.

3 days to go, 15 teams signed up. Tick tock ..... Anyone reading this on the team and not running Rosetta for TSBT at the moment, please consider joining us, even if only for a few days. Every little helps as a TV ad pointed out

Post by **Dirk Broer** » Fri Oct 13, 2017 10:49 am

Alez wrote: ↑Fri Oct 13, 2017 12:59 amThink I read somewhere that Stop also has issues with WUProp.

WUProp is getting annoying in it's steering behaviour. I have already stopped using WUProp on my QCN system and I bet he -Sebastien- is pissed about the resurrection of Goofyx too...

Post by **Bryan** » Fri Oct 13, 2017 11:43 am

Dirk Broer wrote: ↑Fri Oct 13, 2017 10:49 am
WUProp is getting annoying in it's steering behaviour. I have already stopped using WUProp on my QCN system and I bet he -Sebastien- is pissed about the resurrection of Goofyx too...

Yes, it is getting annoying. I had reached my 1st star on the 5th ring and had a couple of small stars over that level. Now I'm back to needing 1 more. I have no idea what he took away this time. A real PITA!

Alez · Post by **Alez** » Fri Oct 13, 2017 2:00 pm

Bryan wrote: ↑Fri Oct 13, 2017 11:43 am
Dirk Broer wrote: ↑Fri Oct 13, 2017 10:49 am
WUProp is getting annoying in it's steering behaviour. I have already stopped using WUProp on my QCN system and I bet he -Sebastien- is pissed about the resurrection of Goofyx too...
Yes, it is getting annoying. I had reached my 1st star on the 5th ring and had a couple of small stars over that level. Now I'm back to needing 1 more. I have no idea what he took away this time. A real PITA!

I've just stopped looking. No idea why he has suddenly got all boinc police

Post by **Megacruncher** » Sat Oct 14, 2017 1:49 am

2 days to go until the Rosetta kick-off. I've not managed a full on Bunker but I still have 3867 WU in progress most of which should be cooked in 49 hours. No doubt & hopefully others will eclipse this massively. Which is good!

I'm sure we will do well in the challenge but the main thing will be consolidating our top 100 status. Currently we are 97th (better than I'd have expected at this stage) with the low 90's a real possibility, something which would leave us secure on Rosetta & free to further improve our already much improved WCG standing.

Post by **Bryan** » Sat Oct 14, 2017 3:07 am

Megacruncher wrote: ↑Sat Oct 14, 2017 1:49 am 2 days to go until the Rosetta kick-off. I've not managed a full on Bunker but I still have 3867 WU in progress most of which should be cooked in 49 hours.

Great! Did you get the Threadripper yet?

No doubt & hopefully others will eclipse this massively. Which is good!

It is a safe bet that the bunker master, Scole, will do a little better - he's the guru

With luck I will add a little bit to our bunker drop also.

I'm sure we will do well in the challenge but the main thing will be consolidating our top 100 status. Currently we are 97th (better than I'd have expected at this stage) with the low 90's a real possibility, something which would leave us secure on Rosetta & free to further improve our already much improved WCG standing.

I think you might be a little on the low side of that estimate

How about the low 80s (or mid 70s) - especially if McShane gets things going. When focused McShane can lay down some serious credits.

Speaking of which, how it is going Pete ... having any luck?

Post by **Bryan** » Sat Oct 14, 2017 11:13 am

@Pete - you don't use the BOINC Manager to control it - use BoincTasks. You can actually use the manager but you have to give it the IP address followed by ":port number". ie if you set it up for port 2222 it would look something like 192.168.0.10:2222 That assumes my memory is correct since I haven't done that in years.

Have these in the options section of your cc_config:

<allow_multiple_clients>1</allow_multiple_clients>
<allow_remote_gui_rpc>1</allow_remote_gui_rpc>

In the gui_rpc_auth.cfg file put in the password you want to use for remote (ie what BoincTasks will use to get in).

Before BOINC will attach to a project and pull work you have to run the "CPU Benchmarks". In BoincTasks you find that under the EXTRA tab. If you have the command in your cc_config.xml:

<skip_cpu_benchmarks>0</skip_cpu_benchmarks>

make sure it is set to 0 otherwise BOINC won't run the benchmarks. It will not attach and pull work until it has run benchmarks.

Alez · Post by **Alez** » Sat Oct 14, 2017 12:32 pm

Megacruncher wrote: ↑Sat Oct 14, 2017 1:49 am2 days to go until the Rosetta kick-off. I've not managed a full on Bunker but I still have 3867 WU in progress most of which should be cooked in 49 hours. No doubt & hopefully others will eclipse this massively. Which is good!

Some of my systems have been bunkering since the 10th and I've got nothing like that number of units. I can only hope that running the 12 hour option is reflected in the points awarded. Probabily going to drop to the 4 hour option once we kick off unless there is a known balance point between points / time, but since this is credit screw I would have a good guess that the credit is totally random

Post by **Bryan** » Sat Oct 14, 2017 1:57 pm

I didn't realize they used credit screw ... my guess as to where we would finish in the standings is probably very optimistic then!

@Pete - when you get BoincTasks tied in the 1st time you obviously won't see anything under the Projects or Tasks tabs. However if you look at MESSAGES you should see the startup info from BOINC. When you hit the Run CPU Benchmarks you will see it start in the messages and it then will give you the results from the benchmarks.

NOTE: On one system I installed the "latest" version of BoincTasks and it seems to me that the Run CPU Benchmarks didn't actually work. I reverted to an older version and when I hit run cpu benchmarks things happened as they should. BOINC absolutely, positively will not attach and get work until the benchmarks have been run.

I'm also running 12 hours WU and plan on switching to shorter ones once I dump what is finished. I haven't decided what length I'll go to yet.

I'll sure be happy when we get past the bunker stage and start actually seeing some results. There are 366k WU in the wild so I'm guessing there is more than a little bunkering happening.

Alez · Post by **Alez** » Sat Oct 14, 2017 8:01 pm

Bryan wrote: ↑Sat Oct 14, 2017 1:57 pmI didn't realize they used credit screw ... my guess as to where we would finish in the standings is probably very optimistic then!

Rosetta grants credit based on the average claimed credit per decoy from all jobs submitted to date. So if you're the first to report, then you might get granted too much credit, but it will quickly get diluted by the reasonable credit claims from ever other machine that submitted results before yours did. Credit is low, seti low.
If, as I understand it, credit is an average, then I think that the bottom line is everyone gets the same credit per decoy, the difference being how fast you can complete a decoy.

Bryan wrote: ↑Sat Oct 14, 2017 1:57 pmI'm also running 12 hours WU and plan on switching to shorter ones once I dump what is finished. I haven't decided what length I'll go to yet.

Again the problem is this per decoy, makes it hard to gauge which is the best units to run. Some WUs run longer than they should, because sometimes a single decoy simply takes longer than your preferred time (you must return at least one decoy for a WU to be accepted). Rosetta gives your credits per decoy, not runtime.
it's luck of the draw with tasks to some extent, because some tasks produce very few decoys whereas others will produce many, and credit is assigned per decoy.

Unlike other projects, changing the run time preference in your account will also change the run time of the units already downloaded.

As far as I can tell, there's also the issue of the watchdog, which shuts a unit down if it hangs etc. Problem is that it does this after 4 hours past your run length, so a 12 hour hung unit will be shutdown after 16 hours and will gain 20 credits for running. Keep an eye on the cpu usage on boinc tasks.

Good chance we may grab 96 place before the comp starts.

Post by **scole of TSBT** » Sat Oct 14, 2017 9:41 pm

Any more thoughts between win vs linux or 32 bit vs 64 bit?

Post by **Bryan** » Sun Oct 15, 2017 12:01 am

@Alez, it will be interesting to see the bunker dumps. I ran for 24 hours at 0 resource share before I started to bunker and got 577k for that day. That isn't "Seti" like. Of course I don't do Seti but IIRC it pays something like 25 credits/hr

If it is a function of who finds a decoy 1st maybe the secret is to have no cache.

Post by **Megacruncher** » Sun Oct 15, 2017 12:48 am

WTF is a decoy? And more relevantly why TF?
Having started to pay attention again after a long period of crunching blind my impression is that Rosetta pays much better than it used to, and certainly much better than WCG or the other CPU projects I run.
And yes, the second Threadripper has been delivered and is running. Its a 1950X 16 core. It will run 24/7. Most of it's Rosetta work has yet to be returned but so far I have 351 completed WU on it netting 22.7K of Rosetta credit. That's 65/wu. Most of the WU take less than 2 hrs. So I might expect, decoy jiggery-pokery aside, to get about c.25K per day from this machine.
In 25 hrs we shall find out.

Alez · Post by **Alez** » Sun Oct 15, 2017 1:02 am

Bryan wrote: ↑Sun Oct 15, 2017 12:01 am @Alez, it will be interesting to see the bunker dumps. I ran for 24 hours at 0 resource share before I started to bunker and got 577k for that day. That isn't "Seti" like. Of course I don't do Seti but IIRC it pays something like 25 credits/hr

If it is a function of who finds a decoy 1st maybe the secret is to have no cache.

We will see in about 24 hrs

, but roughly 25/hr for me before closing my systems. I very much doubt I will be generating 577K/day. Still if that's the case we will be in awesome shape.

Rosetta is quorum of 1 so it's not who finds a decoy first, it's how long it takes you to find a decoy on the unit.

[mention]Megacruncher[/mention] what setting do you have on your account for run time ? Mine was 12 hrs and now 8 hrs. Should I be setting less ?

Post by **Megacruncher** » Sun Oct 15, 2017 1:21 am

Alez wrote:
@Megacruncher what setting do you have on your account for run time ? Mine was 12 hrs and now 8 hrs. Should I be setting less ?

It's set at 2 hrs. I don't recall ever setting it at that or understanding why I might want to change it.

Alez · Post by **Alez** » Sun Oct 15, 2017 1:26 am

[mention]Megacruncher[/mention] I think the current default is 8 hrs on the project. Going to back mine off as the comp progresses down to 2 hrs so that the boinc manager doesn't loose it's mind by backing off too quick.

Post by **scole of TSBT** » Sun Oct 15, 2017 2:06 am

Megacruncher wrote: ↑Sun Oct 15, 2017 1:21 amIt's set at 2 hrs. I don't recall ever setting it at that or understanding why I might want to change it.

For one, it pays a little better credit per run sec on the longer WUs. The main reason though, if I recall, this the project has a limit to the number of WUs it will issue per day to a system. For most systems you won't reach that limit but the 56, 64, 72 thread systems some us run can reach that limit running 1-2 hr WUs. Your 1950X system could be affected too. Also, when loading a 5-6 day bunker it takes fewer WUs to load that bunker and it's easy to calculate a 1 WU per day per thread bunker.

Post by **Bryan** » Sun Oct 15, 2017 2:48 am

scole of TSBT wrote: ↑Sun Oct 15, 2017 2:06 am it's easy to calculate a 1 WU per day per thread bunker.

Well I'm mister smarty pants ... I took higher level math in college so I was able to figure out how many I needed at 2 WU/thread/day

Now I'm really anxious to see our bunker drop. Either we are going to be ecstatic and move up the standings quickly or we are going to be sorely disappointed! I'm hoping we get to do our "happy dance"

Alez · Post by **Alez** » Sun Oct 15, 2017 3:17 am

Bryan wrote: ↑Sun Oct 15, 2017 2:48 amWell I'm mister smarty pants ... I took higher level math in college so I was able to figure out how many I needed at 2 WU/thread/day

I did maths at university and the best I can do is drop the units then suck it and see.... what does that say about Boinc credit ?

Alez · Post by **Alez** » Sun Oct 15, 2017 4:05 pm

10 hours until the kick off. Looks like we are just shy of taking 96th before the start. I expect that after the bunker drops we should hit low 90's. Here's hoping. Going to be very intresting to see how much we have all generated.

Post by **Bryan** » Sun Oct 15, 2017 5:26 pm

Even with "Seti" level credits I think we will be in the 80s. Scole has a full bunker!

I should say that I'm going to be very very disappointed if we don't get to that level after we drop. But you can never count on BOINC credits giving you what you rightfully deserve

We are special you know!

No GC ... yet.

Alez · Post by **Alez** » Sun Oct 15, 2017 8:37 pm

Current state of play after last update, prior to kick off.
5 hours until the [mention]scole of TSBT[/mention] carpet bomb run commences.

Rosetta@Home position

Position	Team	Points
88	NDTC Rosetta	49,588,779
89	Team-SciFi	48,649,805
90	California State Universities (CSU)	48,539,167
91	www.Au-Ja.de	48,108,291
92	IBM	47,771,887
93	Israel	47,274,770
94	The Final Front Ear	47,148,547
95	BOINC@Finland	46,971,804
96	Picard	46,587,070
97	The Scottish Boinc Team	46,530,255
98	ALS-tutkimuksen tuki ry	45,483,181

Post by **Megacruncher** » Sun Oct 15, 2017 11:37 pm

Alez wrote: ↑Sun Oct 15, 2017 8:37 pm Current state of play after last update, prior to kick off.
5 hours until the @scole of TSBT carpet bomb run commences.

Rosetta@Home position
Position Team Points

88 NDTC Rosetta 49,588,779

89 Team-SciFi 48,649,805

90 California State Universities (CSU) 48,539,167

91 www.Au-Ja.de 48,108,291

92 IBM 47,771,887

93 Israel 47,274,770

94 The Final Front Ear 47,148,547

95 BOINC@Finland 46,971,804

96 Picard 46,587,070

97 The Scottish Boinc Team 46,530,255

98 ALS-tutkimuksen tuki ry 45,483,181

Sheesh, who decided on a 2am kickoff?

Nice table again but it's now out of date!
As of now:
TSBT have 46,623,940
Picard have 46,600,423
For those without the benefit of a maths degree, our number is bigger than theirs which means that we have moved up to 96th!

Post by **Bryan** » Sun Oct 15, 2017 11:42 pm

Megacruncher wrote: ↑Sun Oct 15, 2017 11:37 pm
For those without the benefit of a maths degree, our number is bigger than theirs which means that we have moved up to 96th!

Alez · Post by **Alez** » Mon Oct 16, 2017 12:35 am

Megacruncher wrote: ↑Sun Oct 15, 2017 11:37 pm
Sheesh, who decided on a 2am kickoff?

Nice table again but it's now out of date!
As of now:
TSBT have 46,623,940
Picard have 46,600,423
For those without the benefit of a maths degree, our number is bigger than theirs which means that we have moved up to 96th!

Jeez, everyones a critic

Sadly I don't have a maths degree, only an engineering one, so fully qualified to hit things with a selection of hammers

I would bother my arse to update the table but I suspect it's going to change radically in about 90 mins

Still, very happy my prediction of 96th before kickoff is valid.

Alez · Post by **Alez** » Mon Oct 16, 2017 1:18 am

McShane of AP wrote: ↑Mon Oct 16, 2017 1:13 am
Alez wrote: ↑Mon Oct 16, 2017 12:35 am
Megacruncher wrote: ↑Sun Oct 15, 2017 11:37 pm
Sheesh, who decided on a 2am kickoff?

Au contrare, it is an excellent start time. It will only be 11AM here.
At least someone appreciates my sleeping for 4 to 5 hours at the most

Alez · Post by **Alez** » Mon Oct 16, 2017 2:02 am

I do believe we are off and running. Bombs away gentlemen.

Post by **Bryan** » Mon Oct 16, 2017 2:43 am

It's going to take quite a while to upload!

We ought to get a picture, the top 5 users are TSBT!!

Alez · Post by **Alez** » Mon Oct 16, 2017 2:52 am

Bryan wrote: ↑Mon Oct 16, 2017 2:43 am It's going to take quite a while to upload!

painfully slow

Bryan wrote: ↑Mon Oct 16, 2017 2:43 am We ought to get a picture, the top 5 users are TSBT!!

This ?

Post by **Megacruncher** » Mon Oct 16, 2017 3:01 am

And we have an enormous lead over everybody else: 1.3 Million vs 21K !
Very encouraging start but it'll be a while before a reliable picture emerges.

Post by **Bryan** » Mon Oct 16, 2017 3:20 am

From what I've seen so far it appears that Alez was correct ... Seti like credits. CRAP!

Alez · Post by **Alez** » Mon Oct 16, 2017 3:25 am

2 Million on last update. That should put us up to at least 89th now.

Suberb effort by everyone involved.
My credits so far for a 6 day bunker are ...poor....what I actually think is not fit for a family friendly forum. :text-imsorry:

Alez · Post by **Alez** » Mon Oct 16, 2017 3:43 am

Bryan wrote: ↑Mon Oct 16, 2017 3:20 am From what I've seen so far it appears that Alez was correct ... Seti like credits. CRAP!

Indeed, I did warn you, but makes our massive move up the world rankings in just over 1 week all the more to celebrate.

A great team effort.

Post by **Bryan** » Mon Oct 16, 2017 3:55 am

Don't forget you can follow the team HERE

Alez · Post by **Alez** » Mon Oct 16, 2017 4:03 am

To date our best ever position was 92 at 2011-03-20. We've smashed that.
2.7 million v's 100k after 2 hours. Next update should put us at 88th at least.

Post by **Bryan** » Mon Oct 16, 2017 5:15 am

I'm done - all in. We are in 88th but that will improve when Scole gets around to dumping the rest of his bunker

Alez · Post by **Alez** » Mon Oct 16, 2017 10:55 am

I'm all in too. Took nearly 3 hours to dump all the units. I had hoped for a bit more credit for that effort.

Post by **scole of TSBT** » Mon Oct 16, 2017 11:15 am

That's all I have. Based on daily output before the challenge I thought it would be more.