BOINC Pentathlon 2021

[Gorgon]

16 hours ago, leadeater said:

Hmm I thought I tried that and it lost my progress 

 

Must have done something else, more than just suspend. Oh well I don't remember. Will give suspend a shot if I need to, will try it on something that doesn't have 6+ hours on it though.

Perhaps "suspend" is the wrong word. I have been adjusting concurrent tasks through app_config.xml and if I set the number of concurrent for SiDock less than the current number running then they just sit at "Waiting to Run"

 

Update - just tried a suspend then resume on a system with 23 SiDock tasks and it picked up where it left off

[leadeater]

9 minutes ago, Gorgon said:

Perhaps "suspend" is the wrong word. I have been adjusting concurrent tasks through app_config.xml and if I set the number of concurrent for SiDock less than the current number running then they just sit at "Waiting to Start"

Thanks, that's still very helpful as I'll be able to use that method. I should have enough RAM allocated to my VMs unless there is some really RAM hog projects selected. I provisioned most of them around 1GB per thread. If not a quick shutdown and increase can be done when the opportunity arises.

[Gorgon]

1 minute ago, leadeater said:

Thanks, that's still very helpful as I'll be able to use that method. I should have enough RAM allocated to my VMs unless there is some really RAM hog projects selected. I provisioned most of them around 1GB per thread. If not a quick shutdown and increase can be done when the opportunity arises.

Just tried suspend on a system with 23 running SiDock tasks and at resume it picked up where it left off.

 

Just don't restart the service and you should be OK

[Gorgon]

I've finished off of OpenPandemics and now have 120 threads on SiDock for the Marathon

 

Here's some run-times:

Intel i5-6400   3.1GHz 5.3 h/task Linux
Intel e3-1231v3 3.7GHz 6.0 h/task Win10
Intel i9-9900k  4.4GHz 5.6 h/task Linux
AMD   R5 3600x  4.1GHz 4.4 h/task Win10
AMD   R7 2700   3.3GHz 6.6 h/task Linux
AMD   R7 2700   3.3GHz 6.7 h/task Linux
AMD   R7 2700x  3.9GHz 6.1 h/task Linux
AMD   R9 3900x  4.1GHz 5.3 h/task Linux
AMD   R9 3950x  3.9GHz 5.1 h/task Linux

[leadeater]

Add another 36c/72t, just some older stuff of mine. 2x Dual X5650 & 1x Dual 2630v2

 

I also have another 5x Dual 2690v3 (120c/240t total) servers racked up at work that I need the network team to configure ports for me, hopefully will get those online tomorrow/today

[Windows7ge]

Dusted out the O.C.ed 3930K rig. Still toasty but temps are a little better.

 

[Gorgon]

1 hour ago, Windows7ge said:

Dusted out the O.C.ed 3930K rig. Still toasty but temps are a little better.

Still not seeing any results for you on SiDock. Still have your systems on WCG?
 

@Metallus97 You might want to change the Team Location to International

[Gorgon]

Averages over the day today on 10 or more valid tasks.

It appears that the autodock software used by SiDock is better optimized for Windows:

		 	 				Tasks 	 
	Points	Run-Time		Clock		/Day	Tasks	Points
Node    /Task   (h:m:s) CPU		(GHz)	Threads	/Thread	/Day	/Day
Win10	209.82	6:11:52	e3-1231v3	3.70	 6	3.87	 23.23	  4,875
Win10	230.22	4:25:25	r7-3600x	4.05	10	5.43	 54.26	  12,491
Linux3	207.42	5:23:33	i5-6400		3.10	 4	4.45	 17.80	  3,693
Linux4	223.33	7:24:35	r7-2700		3.34	11	3.24	 35.63	  7,957
Linux5	228.14	6:40:11	r7-2700x	3.91	13	3.60	 46.78	 10,672
Linux6	223.43	5:28:00	r9-3950x	3.97	29	4.39	127.32	 28,447
Linux7	215.37	5:32:12	r9-3900x	4.07	21	4.33	 91.03	 19,605
Linux8	220.45	7:10:58	r7-2700		4.06	13	3.34	 43.44	  9,576
Linux9	216.60	6:04:19	i9-9900k	4.40	13	3.95	 51.38	 11,130
				Totals		120		490.87	108,446

[leadeater]

1 minute ago, Gorgon said:

It appears that the autodock software used by SiDock is better optimized for Windows:

 

</pre><table border="0" cellpadding="0" cellspacing="0" style="border-collapse:collapse;width:387pt;" width="513">																														Node										Points /Task										Run-Time (h:m:s)										CPU										Clock (GHz)										Threads										Tasks /Day /Thread										Tasks /Day										Points /Day														Bashful										209.82										6:11:52										e3-1231v3										3.70										6										3.87										23.23										4,875														RhiDesk										230.22										4:25:25										r7-3600x										4.40										10										5.43										54.26										12,491														Fold3										207.42										5:23:33										i5-6400										3.10										4										4.45										17.80										3,693														Fold4										223.33										7:24:35										r7-2700										3.34										11										3.24										35.63										7,957														Fold5										228.14										6:40:11										r7-2700x										3.91										13										3.60										46.78										10,672														Fold6										223.43										5:28:00										r9-3950x										3.97										29										4.39										127.32										28,447														Fold7										215.37										5:32:12										r9-3900x										4.07										21										4.33										91.03										19,605														Fold8										220.45										7:10:58										r7-2700										4.06										13										3.34										43.44										9,576														Fold9										216.60										6:04:19										i9-9900k										4.40										13										3.95										51.38										11,130														 										 										 										 										 										120										 										490.87										108,446</

 

RIP readability.

 

Surprised Windows is the faster one though.

[Gorgon]

11 minutes ago, leadeater said:

RIP readability.

 

Surprised Windows is the faster one though.

Was editing it while you posted - RIP Code Block formatting

 

All of the systems have more than enough RAM and 1 thread reserved for the OS. If anything Linux should be doing better in this situation due to it's smaller footprint and better scheduling.

 

The other difference is the Linux systems are running a slightly older version of the client (7.9.3) versus 7.16 on Windows.

[leadeater]

16 minutes ago, Gorgon said:

The other difference is the Linux systems are running a slightly older version of the client (7.9.3) versus 7.16 on Windows.

I have some older client versions too, hope that's not affecting performance too much

[Gorgon]

1 minute ago, leadeater said:

I have some older client versions too, hope that's not affecting performance too much

The newest "Recommended" Linux Client is 7.4.22 with the Dev branch 7.16.6 and Ubuntu still reporting 7.9.3 as current

 

[Windows7ge]

2 hours ago, Gorgon said:

Still not seeing any results for you on SiDock. Still have your systems on WCG?

Correct, I will start switching them over prior to the beginning of the event. Though I aught to do it sooner in case I run into any issues.

 

What do the project file sizes look like for SiDock? 1GB/10GB/100GB? How about average RAM utilization?

 

Hoping my systems have enough of everything to handle it.

[Gorgon]

45 minutes ago, Windows7ge said:

Correct, I will start switching them over prior to the beginning of the event. Though I aught to do it sooner in case I run into any issues.

 

What do the project file sizes look like for SiDock? 1GB/10GB/100GB? How about average RAM utilization?

 

Hoping my systems have enough of everything to handle it.

The tasks are long-ish in duration (4.5-7hrs). They're dual-quorum with no check-pointing so if you stop or restart the client you lose any work in progress. The tasks have a two-day deadline and the project a hard-limit on two tasks per slot meaning you can only have 1 task running and one in the queue per thread.

 

The tasks themselves are fairly light-weight and the packages are 1.75-2.0MB in size and each seems to only use ~250MB RAM when running.

[Ithanul]

  

On 4/29/2021 at 4:32 PM, GOTSpectrum said:

Testing the 5960x at 4.7Ghz, will see how it goes, hopefully it's stable 

Hmmmmm, you know if it is a J Batch or not?

 

The deal though is keeping those cooled off.  My previous one (before it die in glory) could hit 4.8GHz across all cores.  But, that bugger was under a beefy loop and was still hitting into 70C.

 

Also, oh YESSSSSSSS.  Now, to work on getting that 5800X up.

 

 

I like me some big ol air coolers.

[Windows7ge]

32 minutes ago, Ithanul said:

I like me some big ol air coolers.

I like me the NH-D9L. Because it fits in 3U and up server chassis. Put some Delta fans on it and it cools like a hot damn.

[leadeater]

First Javelin has been announced

 

 

Question for those that know, CPU or GPU?

Answered: CPU

[Lightwreather]

Hey so, I joined using the link  you sent with the invitaiton code, so is this normal?
It's got no team name mentioned on it

[Lightwreather]

Also, is there a CLI variant of the manager but for macOS?

[Shlouski]

58 minutes ago, leadeater said:

First Javelin has been announced

 

 

Question for those that know, CPU or GPU?

 

Damn, I dunno why but I thought this event was starting later, guess I got my dates mixed up.

 

I have 4 or 5 rigs out of action, guess I know what I'm doing for the next couple of days.

[Gorgon]

48 minutes ago, J-from-Nucleon said:

Hey so, I joined using the link  you sent with the invitaiton code, so is this normal?
It's got no team name mentioned on it

Welcome aboard!

 

You need to Join your Account to the LinusTecTips_Team at the SiDock Site.

 

33 minutes ago, J-from-Nucleon said:

Also, is there a CLI variant of the manager but for macOS?

You can use the boinccmd tool but it isn't as nice as the GUI.

 

There should be a GUI for MacOS. If your running SiDock on another system and want to manage it from a Mac on the same LAN then you should be able to install the boinc app on the Mac but not run boinc on the Mac but just use the BOINCmanager tool

[Shlouski]

1 hour ago, leadeater said:

Question for those that know, CPU or GPU?

 

Looks like cpu only: Choosing BOINC projects (berkeley.edu)

[rkv_2401]

I'm not getting any new tasks for SiDock@home, and there don't seem to be any available either, going by the server statistics. Guess I'll just bunker WCG in the meantime. Thanks @Gorgon for the guide to bunkering.

[Red :)]

I'm in.

[RollinLower]

At work we're about to stop using 7 of these boxes. 12 blades per box, every blade has 4 cores. Nothing too fancy, but still 48 cores per box, with potentially 336 cores on the marathon!

 

(yeah i know, its a mess right now. but these boxes need to get cleaned up and taken out the rack soon.)