PrimeGrid LLR Races Thread 2016

[Ken g6]

Day 2 stats:

Rank___Credits____Username
130____46830______Ken_g6
160____35924______waffleironhead
164____32870______zzuupp
166____32333______GLeeM
167____28336______Ryan
197____18800______Orange Kid
210____18519______SlangNRox

Rank__Credits____Team
13____289103_____Rechenkraft.net
14____283614_____Ukraine
15____261224_____SETI.USA
16____213615_____TeAm AnandTech
17____212659_____US Navy
18____203275_____Russia
19____190171_____Canada

Still no Larry?

 

[VirtualLarry]

Still no Larry?

ETA: 2hrs

 

[TennesseeTony]

My ETA is 14 hours for the first wave, 26 hours for the second wave.

Maybe I should quit calling WUSS and BumbleBee twins... Not sure why they have such different completion times, everything else tends to run fairly equal, with WUSS having a slight lead. WUSS's lead this time is quite substantial.

Stranger yet: WUSS is bouncing from 2.4 to 2.5GHz, Bee is a steady 2.6GHz, but falling behind. WUSS however is pulling an extra 3 watts on the RAM, 34w vs 31w for Bee. Both have identical brand/model of memory, purchased at the same time.

Further details: Bee has the better case, so a chilly 40C, WUSS only has a third of the front fans, so 50C. I don't think XEONs throttle at such low temperatures.

Ah, what is this Row Refresh Cycle Time (tRFC) that CPUID says is vastly different? 278 for WUSS, 347 for Bee. Otherwise memory shows the same specs.

 

[TennesseeTony]

Well I found the BIOS setting for tRFC and changed it to match the timing of WUSS. And now Bee's RAM is also pulling the extra 3 watts, matching WUSS.

 

[Orange Kid]

Most interesting, Tony. Are the mother boards the same and bios the same version?
In other words are all things equal on both machines other than that one setting?
Could all come down to quality control in the manufacturing process

 

[TennesseeTony]

Yep, everything is equal, really equal now that I've changed that one memory setting. Will have to see if there's any gap between the two in some other test, as WUSS has dumped it's load of PG tasks already.

 

[Ken g6]

On the PrimeGrid thread for this race, Mackerel found that, "Throughput peaked with 12 cores of 14 running," on his Xeon.

 

[Ken g6]

Day 3 stats:

Rank___Credits____Username
84_____141405___10esseeTony
132____84745______Ken_g6
133____84603______Orange Kid
156____65998______zzuupp
158____65056______GLeeM
163____56745______Ryan
182____54564______waffleironhead
222____36554______SlangNRox
291____9410_______VirtualLarry

Rank__Credits____Team
10____769877_____BOINCstats
11____734691_____Team 2ch
12____656815_____Rechenkraft.net
13____599083_____TeAm AnandTech
14____593268_____BOINC@Poland
15____529941_____Ukraine
16____479626_____Christians

Looks like I'm the "hare", and Tony is the "tortoise". I expect Orange Kid to pass me soon too, but not yet.

 

[TennesseeTony]

Thanks for the stats and the tip in the previous post, Ken!

Round 2 (BumbleBee) is incoming soon. A few tasks sprinted ahead of the rest and made it into today's update, but the rest will count for tomorrow's stats.

 

[TennesseeTony]

I have briefly jumped from 84 to 44, (TeAm rose up one to 12) since BumbleBee just finished all tasks.

I was hoping to finish top 50, but I don't think that's realistic with so much time left. (I'm done with the race).

Maybe I'm not done with the race....Need to pit the Twins against each other, with 12 tasks each, now that I've tweaked Bee's BIOS, to see if there is still a difference between them. Purely as a test, of course. Has nothing to do with competitiveness.

 

[waffleironhead]

This having a LLR race in september is pretty hard on my systems. Hopefully everything validates.

 

[Ken g6]

Day 4 stats:

Rank___Credits____Username
61_____248452___10esseeTony
139____114078_____Ryan
143____113208_____Ken_g6
153____99459______zzuupp
157____94082______Orange Kid
160____93483______GLeeM
224____55703______SlangNRox
228____54564______waffleironhead
338____9410_______VirtualLarry

Rank__Credits____Team
11____1130533____BOINCstats
12____1068525____BOINC@Poland
13____922791_____Rechenkraft.net
14____882444_____TeAm AnandTech
15____737805_____Ukraine
16____724345_____Christians
17____658332_____US Navy

The race is almost over! Maybe set your computers to get no new work.

 

[TennesseeTony]

Thanks for the stats Ken!

Interesting, but confusing: I raced Bee and WUSS against each other with 12 tasks each, this time Bee was the winner by about 2.5 hours.

 

[Ken g6]

It can be hard to compare these work units to one another, since they vary in size so much. Speaking of which, I think I didn't time the end of the race well. I don't think I'm going to get these last WUs in on time.

 

[Ken g6]

Preliminary final stats:

Rank___Credits____Username
55_____372750___10esseeTony
152____142809_____Ryan
153____142001_____Ken_g6
155____141828_____Orange Kid
158____138588_____waffleironhead
162____133072_____zzuupp
194____98241______GLeeM
227____74826______SlangNRox
373____9410_______VirtualLarry

Rank__Credits____Team
12____1400952____Ukraine
13____1394931____BOINC@Poland
14____1360374____Rechenkraft.net
15____1253528____TeAm AnandTech
16____891676_____US Navy
17____818347_____Christians
18____692640_____Russia

Oh, well, I guess a couple of extra WUs wouldn't have mattered anyway. Except for my rank within the TeAm.

 

[TennesseeTony]

Thanks Ken, for the stats these five days. It was an interesting experience, learned a bit about memory timings.

 

[Orange Kid]

Thanks for the stats.

 

[QuietDad]

Second Prime Grid race I couldn't get anything finished in time

 

[Orange Kid]

I've had two invalid WU's and am dropping like a rock in the standings
Hope the five pending are going to be good

 

[Ken g6]

I've had two invalid WU's and am dropping like a rock in the standings
Hope the five pending are going to be good

At least we have a buffer as a TeAm overall.

Apparently, errors are more common with Skylake.

 

[TennesseeTony]

Not another Pentium bug I hope. Lol, that was SOME time ago, eh?

 

[Orange Kid]

It's an AMD rig. I'm thinking it is a RAM issue. It seems the Skylake problem has more to do with memory stability and over clocking than the architecture its self. Non-overclocked systems with "slower" memory are not giving errors. It's those with "fast" memory or overclocked CPUs that are showing more errors. None of my systems are overclocked these days. Maybe I'll look for some fresh stuff.

 

[Ken g6]

Bump for the last LLR race of the year!

 

[TennesseeTony]

Getting closer...3 and a half more days, Monday at noon EST (USA). No time to lose on this one, only a 24 hour race!

Get your preferences set for SGS-LLR, Sophie Germain Prime Search, CPU only.

Fast tasks this time, 22 minutes recent average.

 

[TennesseeTony]

I've done the math again for the start time and it doesn't start until 2pm, not noon as posted above, for the Eastern US.