PrimeGrid LLR Races Thread 2016

[VirtualLarry]

Just checked the scores before bed and thought it appropriate to say:

Nice Tony! We're actually ahead of HOCP, for however long that lasts, anyway.

 

[TennesseeTony]

41____10esseeTony__________TeAm AnandTech_____223625.68
87____Ken_g6_______________TeAm AnandTech____115536.40
99____VirtualLarry___________TeAm AnandTech______96261.24
132___geecee______________TeAm AnandTech______69203.35
161___waffleironhead________TeAm AnandTech______46184.99
162___Ryan________________TeAm AnandTech______46179.09
165___Orange Kid___________TeAm AnandTech______46024.09
202___zzuupp______________TeAm AnandTech______12560.89

Well, hello Ryan! Welcome!

 

[TennesseeTony]

And we just squeezed into 10th place as a TeAm.

 

[Ken g6]

Thanks for the stats, Tony! I think you might be out of range for me to catch you this time.

I think that Ryan is SomethingSketchy.

 

[VirtualLarry]

I'm throwing my Haswell Celeron G1820 rig with 16GB DDR3-1600 into the ring, with both cores.

Strangely, the estimated time for the Cullen and Woodall WUs are 52:22 and 55:44. On my G4400 @ 4.455, they were like 82hours estimated. I don't get it.

 

[Ken g6]

On my G4400 @ 4.455, they were like 82hours estimated. I don't get it.

Is it overheating and throttling? Is it unstable, so LLR is trying slower compute methods?

 

[VirtualLarry]

On the G4400 @ 4.455, 2x4GB DDR4-2400 @ 2520 or so, Windows 7 64-bit.

While it's crunching, the "Remaining" column actually counts UP, by 1 or 2 seconds every second. Until the "Progress" column ticks, then the "Remaining" goes down by a bit, but then starts counting up again.

On the Haswell G1820 rig, the "Remaining" column counts down, approx 1 per second.

Edit: I don't think the G4400 is throttling. I just installed HWMonitor, and CPUTIN shows 72-73C, fluctuating. Should be OK, I think Intel CPUs throttle at 100C core temps.

 

[TennesseeTony]

I'm working on the next 'surge' and will have precious few points the next 4 days, maybe 40-50k tomorrow and some sporadic points after that until just before the deadline.

Crunch on!

 

[Ken g6]

Day 5 stats:

Rank___Credits____Username
37_____385982___10esseeTony
89_____161977_____Ken_g6
118____120464_____VirtualLarry
126____100524_____zzuupp
146____92372______waffleironhead
148____92039______Orange Kid
164____69349______Ryan
166____69203______geecee

Rank__Credits____Team
7_____1823933____BOINC@Poland
8_____1397845____The Knights Who Say Ni!
9_____1267748____BOINC@MIXI
10____1091913____TeAm AnandTech
11____1015116____Team 2ch
12____849967_____[H]ard|OCP
13____634306_____AMD Users

Wow, 10th place! But I think we'll have to work to stay there.

 

[Ken g6]

A little late with the stats again - Day 6 stats:

Rank___Credits____Username
29_____549794___10esseeTony
105____185214_____Ken_g6
109____184162_____Orange Kid
140____133073_____VirtualLarry
146____115753_____Ryan
157____103920_____waffleironhead
160____100524_____zzuupp
197____69203______geecee

Rank__Credits____Team
7_____2250103____BOINC@Poland
8_____1773467____The Knights Who Say Ni!
9_____1584121____BOINC@MIXI
10____1441647____TeAm AnandTech
11____1383741____Team 2ch
12____1170380____[H]ard|OCP
13____1097452____Ukraine

Still battling for 10th place - we were 11th for awhile. Wow, Orange Kid, where'd you get all that power?

 

[QuietDad]

Looks like I'm not going to finish on time. BOINC ETAs have it running a half day too long

 

[Orange Kid]

I will get four more done. Another 10 will be real close, most likely will miss the deadline

 

[VirtualLarry]

Looks like I'm not going to finish on time. BOINC ETAs have it running a half day too long

If you're on a Core2-era platform (Xeon, right?), then memory bandwidth might be an issue if you are running four tasks. You might be able to set your compute preferences to only run 50% of available CPUs (two cores), and those tasks might complete faster, if you had four running at once.

I'm still trying to figure out how my 2.7Ghz Haswell G1820 (in Linux Mint 17.3 with an updated kernel), can crunch these WUs in 56hrs, and my G4400 Skylake @ 4.455 takes 80hrs.

Edit: My two G4400 rigs had one new WU queued up, to start in 2hrs, with 77hrs estimated. I figure that we have less than 72hrs left in the Challenge, so I aborted those WUs and set No New Tasks, because I don't think I can complete any more on these machines within the deadline.

I'm going to run my i3-6100 up until the very end, though, because those WUs take like 25hrs.

 

[Ken g6]

I will get four more done. Another 10 will be real close, most likely will miss the deadline

If you're on a Core2-era platform (Xeon, right?), then memory bandwidth might be an issue if you are running four tasks. You might be able to set your compute preferences to only run 50% of available CPUs (two cores), and those tasks might complete faster, if you had four running at once.

You pay attention to this too, OK. Even if you don't have a Core2. All the quad-cores seem memory-limited on this.

Fortunately, I expect to get 15 more done, so I don't expect you to catch me. :sneaky:

P.S. Did you notice Cullens seem slightly faster than Woodalls? With proportionately less credit, I assume.

I'm still trying to figure out how my 2.7Ghz Haswell G1820 (in Linux Mint 17.3 with an updated kernel), can crunch these WUs in 56hrs, and my G4400 Skylake @ 4.455 takes 80hrs.

I still say try dropping your OC to 4 or 4.2GHz. You might be surprised at the result. I suspect your system is unstable, and LLR is trying older computing methods, or restarting a lot.

 

[VirtualLarry]

I still say try dropping your OC to 4 or 4.2GHz. You might be surprised at the result. I suspect your system is unstable, and LLR is trying older computing methods, or restarting a lot.

I dropped the OC from 4.455 @ 1.300V, to 4.125 @ 1.25V.

Still doing the same thing.

RAM is at DDR4-2333 17-17-17-38 2T 1.25V, dual-channel mode

 

[Ken g6]

OK, I'm surprised. Double-check your RAM is in dual-channel mode?

 

[QuietDad]

You pay attention to this too, OK. Even if you don't have a Core2. All the quad-cores seem memory-limited on this.

Fortunately, I expect to get 15 more done, so I don't expect you to catch me. :sneaky:

P.S. Did you notice Cullens seem slightly faster than Woodalls? With proportionately less credit, I assume.


I still say try dropping your OC to 4 or 4.2GHz. You might be surprised at the result. I suspect your system is unstable, and LLR is trying older computing methods, or restarting a lot.

Dropped both PCs to 50% of the CPUs and we'll see what happens. One is a q6600 quad, but the Xeon is a twin quad (2x5355) or 8 cores. We'll see what happens.

 

[Ken g6]

Day 7 stats:

Rank___Credits____Username
37_____597018___10esseeTony
107____233456_____Ken_g6
126____196068_____VirtualLarry
133____184162_____Orange Kid
136____150936_____zzuupp
150____138786_____Ryan
151____138563_____waffleironhead
213____69203______geecee

Rank__Credits____Team
7_____2810605____BOINC@Poland
8_____2401442____The Knights Who Say Ni!
9_____1848023____BOINC@MIXI
10____1708195____TeAm AnandTech
11____1649062____Team 2ch
12____1395577____[H]ard|OCP
13____1369592____Ukraine

Still hangin' on to 10th. And my slow WUs are coming in.

 

[geecee]

Thanks for the stats. I tried juggling in some extra units, but it looks like they are not going to finish in time (27 hrs left vs 31+ hrs estimated to finish, plus I need 2 hours or so for the suspended WUs to finish). Sorry folks. I'm only running 1 box right now, as I've shut down one for the season and the other for repairs (and prob for the season as well).

 

[Ken g6]

Day 8 stats:

Rank___Credits____Username
28_____818311___10esseeTony
105____316321_____Ken_g6
129____232962_____VirtualLarry
137____207382_____Orange Kid
146____185371_____Ryan
156____162514_____zzuupp
162____150111_____waffleironhead
201____104070_____geecee

Rank__Credits____Team
8_____3110480____The Knights Who Say Ni!
9_____2352353____BOINC@MIXI
10____2189977____Team 2ch
11____2177047____TeAm AnandTech
12____1625767____Ukraine
13____1550316____[H]ard|OCP
14____1432090____BOINCstats

Like I said, fighting for 10th. Right now we're behind. I've got 4 more WUs to complete, but it's probably going to be close either way.

 

[Orange Kid]

I'm done, no more will make it and I had two that were invalid

 

[zzuupp]

I've got 3 cullens left. I will make it. One might. The last, prolly not.

Bad news is that I have one woodall showing inconclusive.

 

[Ken g6]

Yeah I got an inconclusive Cullen myself. And neither my machine nor the wingman has any invalid work. That machine of mine is seven years old. So I'm not sure what to think.

 

[Orange Kid]

Another bad one. I'm going backward now. These machines a very stable on everything else and nothing is overclocked. Not going to run this project again. WU's take too long and then invalid? I don't mind a few hours wasted but days of work don't cut it, give me a partial credit for trying.

 

[Ken g6]

Preliminary final stats:

Rank___Credits____Username
38_____818311___10esseeTony
109____365874_____Ken_g6
143____232962_____VirtualLarry
144____232315_____zzuupp
149____219389_____waffleironhead
152____208493_____Ryan
159____195879_____Orange Kid
196____138956_____geecee

Rank__Credits____Team
8_____3580093____The Knights Who Say Ni!
9_____2692455____BOINC@MIXI
10____2465024____Team 2ch
11____2412182____TeAm AnandTech
12____2405222____Ukraine
13____2260532____[H]ard|OCP
14____1817399____AMD Users

Well, we didn't catch back up to 2chan, but at least we're still ahead of Ukraine. Of course, there may be more bad WUs that will change one or both things.

I asked about all the bad WUs on the PrimeGrid forums, and another person agreed it was a little strange, but no official answer yet.