PrimeGrid Double Top Secret Mystery Challenge March 1-16

[Ken g6]

The Seventeen or Bust race will not be seen this month, but may be rescheduled for next year. Instead, please stay tuned for the following special race announcement.

Since Seventeen or Bust is jumping to a new WU size, PrimeGrid took a straw poll on what to do this month. And it was decided that this month's race will be...the Prime Sierpinski Problem LLR. The race is still 15 days, like the last one, from March 1 to March 16, starting and ending at 18:00 UTC. (Fortunately, this is the last 15-day race this year.) But this time GPUs won't help you. :ninja:

What you need:

• One or more fast x86 processors. See below for the best kinds.
• All of those processors running stably. One good way to check that is to run PPSE or SGS LLR for a few days. (Although Prime95 may be a more reliable test.) If you get any errors, your processor (or RAM) probably isn't stable at your current clock/voltage.
• Windows, Linux, or MacOS 10.4+.
• BOINC, attached to PrimeGrid (http://www.primegrid.com/).
• Your PrimeGrid Preferences with only "Prime Sierpinski Problem (LLR)" selected in the Projects section.
• Patience! I honestly don't know how long these WUs will take, but they could be long. Though unlike with Seventeen or Bust, almost all computers should be able to complete at least one WU in the 15 days allotted.

What might help:

• AVX on Intel i3/5/7 processors. It nearly doubles your speed.
• A 64-bit OS. (I think it helps a little with the latest version.)
• Fast RAM. (DDR3-2133 may be 10-15% faster than DDR3-1600.)
• Having hyper-threading off on i3/5/7s is said to help. (Some laptop i5s do have hyper-threading.)
• Following [post=30943822]Ken's WU Juggling Guide[/post] toward the end might help too.

What won't help (but won't hurt either):

• A large amount of RAM.
• Any video cards.

Work not downloaded and uploaded within the challenge will not be counted. Should you not be able to be in front of one or more computers at that time, see the challenge strategy thread for ways to start and finish on time. (I think my Bash script for Linux is up-to-date.)

Good luck - and patience - to all!

 

[Fardringle]

Speaking as someone that isn't really familiar with Primegrid, does the project have have checkpoints so work units can resume when interrupted?

My computers are office machines so BOINC is frequently forced to suspend itself while the computers are in use, so I could only participate if the work units can be resumed without starting from the beginning of the work unit again since one of the notes in the announcement says that they could be pretty long tasks.

 

[GLeeM]

I'm in

Saturday Noon Central Time

Question: When will it end Central Time seeing as US has Daylight Savings time change on March 9? I suppose I could just watch the clock on PG homepage.

 

[Ken g6]

does the project have have checkpoints so work units can resume when interrupted?

Absolutely. Feel free to set them to suspend to RAM, or even to disk.

Edit: P.S. @GLeeM I'm no good with time zones. :$ But I'm guessing 1pm?

 

[Ken g6]

Bump! 15 minutes till race time. Start your engines, if you haven't already!

 

[salvorhardin]

count me in. as soon as it starts i'll be switching 7 cores on my i7 to it. too bad gpus don't work on the project.

 

[somethingsketchy]

I'm in, crunching some work now.

 

[biodoc]

I'm in too for 15 days of stress testing my CPUs.

 

[GLeeM]

I'm in too for 15 days of stress testing my CPUs.

Yeah, me too.

But I turned off HT so the cores are running ~5*C cooler than with HT and A@H even though I upped the OC from 3948 to 4053Mghz

 

[biodoc]

Yeah, me too.

But I turned off HT so the cores are running ~5*C cooler than with HT and A@H even though I upped the OC from 3948 to 4053Mghz

Thanks for reminding me about turning off HT. I also discovered one of the fans on my radiator was unplugged!

 

[Ken g6]

All of those processors running stably.

I upped the OC from 3948 to 4053Mghz

I think 4053 million gigahertz might be a little much. I hope 4053MHz isn't.

 

[VirtualLarry]

My too-slow-for-challenges 1035T may or may not be participating.

 

[Rudy Toody]

6 cores made it to .44% before everything froze. So, I'm back to POGS. Sorry.

 

[Fardringle]

I was just reading the challenge announcements again and noticed this:

AVX on Intel i3/5/7 processors. It nearly doubles your speed

I'm not finding a whole lot online about AVX except one page that says it can only be unlocked on Windows 7 SP1. Is there something specific that enables this, or is it automatically used if the computer is running Win7 SP1?

 

[Ken g6]

My too-slow-for-challenges 1035T may or may not be participating.

My Core 2 Q9400 is going to get two sets of WUs done in ~14 days. Your 1035T should be able to get something done.

6 cores made it to .44% before everything froze. So, I'm back to POGS. Sorry.

Must've overheated. That's why it would have helped to run a test with PPS or PPSE LLR before the race.

I'm not finding a whole lot online about AVX except one page that says it can only be unlocked on Windows 7 SP1. Is there something specific that enables this, or is it automatically used if the computer is running Win7 SP1?

Any machine with Win7SP1 or newer, or a sufficiently modern version of Linux, and a CPU that supports AVX, will use AVX. This only works on Core i-series Sandy Bridge or later processors; not on Pentiums or Celerons.

 

[Fardringle]

Any machine with Win7SP1 or newer, or a sufficiently modern version of Linux, and a CPU that supports AVX, will use AVX. This only works on Core i-series Sandy Bridge or later processors; not on Pentiums or Celerons.

I thought so, but wanted to be sure. All of my computers currently running PrimeGrid are i5's, so I figured I better get everything I can out of them. Thanks!

 

[geecee]

Sorry, got a late start on this one. I am in now with half my cores until my other BOINC work finishes up, then you'll have all my cores.

 

[Fardringle]

There is definitely a huge difference in productivity between CPU types on this project. My i5 machines are all showing run times of about 70-75 hours per work unit, while my Core2Duo 4600 is estimating about 820 hours remaining after running for 12 hours, so I think I'll kill that one and just let the i5's play with the challenge tasks.

 

[Icecold]

I am in but at a very limited capacity. Is this challenge particularly stressful on cpu cores or ram? I was running Proth Prime Search and Seventeen or Bust seemingly without any issues on multiple machines previously but this one is causing me issues on multiple non-overclocked machines. I am running mem test on my 8350 now and will try to return it to the challenge if I can figure out what's going on with it(it just reboots randomly around 2-10 hours. No BSOD and nothing in event viewer other than it losing power. Does not seem heat related as I put a box flan blowing in the side and it sitll happened despite temps seeming fine. I'm going to try a different power supply but think it may be something with the motherboard / VRMs. Power supply seems fine as I can run BOINC on 2 GTX 480's without issue, but I'l still try another. I took a 480 out as well to troubleshoot.)

Fortunately I have quite a few parts I can use to troubleshoot.. but I am probably not bringing much to this challenge. I should be turning in 4 WU's in around 9-10 hours.

 

[Ken g6]

my Core2Duo 4600 is estimating about 820 hours remaining after running for 12 hours, so I think I'll kill that one

Right, you haven't been running BOINC lately. Don't trust the estimated times. Use the running time so far and the percentage done to compute the total time a WU will take. It should get at least two or three WUs done with juggling. (See the first post for a link to juggling instructions. )

I am in but at a very limited capacity. Is this challenge particularly stressful on cpu cores or ram? I was running Proth Prime Search and Seventeen or Bust seemingly without any issues on multiple machines previously but this one is causing me issues on multiple non-overclocked machines.

I don't know why that would be. This shouldn't be any harder than SoB.

But your "limited capacity" has you as the first TeAm member on the board.

 

[Ken g6]

But your "limited capacity" has you as the first TeAm member on the board.

And now it's time to make that official with Day 3 stats:

Rank___Credits____Username
50_____55944______Icecold

Rank__Credits____Team
21____70531______University of California, Berkeley
22____70310______USA
23____56356______SwissTeam 1
24____55944______TeAm AnandTech
25____55388______Hardware Canucks
26____43387______Bulgaria
27____42803______Keep The Fire Alive!

 

[biodoc]

Congrats Icecold! First on the board for the TeAm! :thumbsup:

 

[Ken g6]

Just because three more people have appeared in the stats, I'm doing a double top secret extra stats update.

Rank___Credits____Username
43_____68391______biodoc
48_____57115______plimogz
58_____55944______Icecold
105____27251______Ryan

Rank__Credits____Team
12____279694_____BOINCstats
13____264935_____BOINC Confederation
14____220910_____Dutch Power Cows
15____209407_____TeAm AnandTech
16____170973_____BOINC@Poland
17____151464_____Ukraine
18____142900_____UK BOINC Team

 

[Plimogz]

Thanks for the double top secret extra stats update.

 

[Fardringle]

Right, you haven't been running BOINC lately. Don't trust the estimated times. Use the running time so far and the percentage done to compute the total time a WU will take. It should get at least two or three WUs done with juggling. (See the first post for a link to juggling instructions. )

The estimated time was exactly on using completed time (12 hours) and % completed (just under 1.4%). I let it run as long as I did hoping the time would drop, but it didn't. It doesn't really matter, though. I have a bunch of i5s running PrimeGrid while my Core2 is back to running "little" random projects for me. Now I just need those i5s to hurry up and return some results (they're a bit slower than they should be since BOINC gets suspended while they are in use during business hours).