5. BIONC Pentathlon 2014

[Assimilator1]

You Americans with the wrong date format! lol, just 3 days on Collatz then, can't even remember what that 1 does now.....

Re Pogs, I just can't catch Joe (in RAC)! lol, I fired up my 2nd PC & he's pulling away faster! Anyone know who he is?

 

[lane42]

He's been a long time on Seti, Oct. 99 but I can't place the name.

 

[Ken g6]

can't even remember what that 1 does now.....

Hailstone numbers. I actually tried a variation of the algorithm in "Optimizations" for PrimeGrid once. It was fast, but what I currently do is faster.

I can't place Joe either, unless it was Joe O.

 

[Assimilator1]

Interesting, we've probably spoken to him at some point then!

Hmm, just read about Collatz, unless I missed the point, their's a pointless project!
Is it the only GPU project in the race?

No one keeping track of stats here?
Well we're rank 20 & not far off the others.

Rank.... Team ........ Overall points ... Pogs

18 BOINCstats ......... 24 ...................24
19 BOINC@MIXI ....... 21 .................. 21
20 TeAm AnandTech ..18 .................. 18
21 ICRAR ................ 15 .................. 15

 

[Ken g6]

Hmm, just read about Collatz, unless I missed the point, their's a pointless project!

It's a math (dis-)proof project. Probably no more pointless than PrimeGrid, except it doesn't have any expected endpoint.

Is it the only GPU project in the race?

Yup. Well, probably - the grab bag ("showbag"? ) project could be a GPU project too.

No one keeping track of stats here?

I guess you are now. :twisted: Thanks for the stats. :thumbsup:

 

[GLeeM]

Re Pogs, I just can't catch Joe (in RAC)! lol, I fired up my 2nd PC & he's pulling away faster! Anyone know who he is?

LOL
Joe is #3 on the TeAm in BOINC :thumbsup:
http://stats.free-dc.org/stats.php?page=userteamcomb&team=TeAm+AnandTech

Joe has 12 projects over 1M and 33 projects over 100K points with 68 projects participated in!!

I don't know if he has posted to the forum, but I am sure he watches because of his participation in things like this.

A big HELLO and well done to Joe

 

[biodoc]

~650 million boinc points!

WTG Joe! :thumbsup:

 

[zzuupp]

FYI, if you haven't looked at the blog page lately, the GPU project is going to be Collatz. But we don't know when it will count yet, except that it won't be for at least three days from now.

Cool! Thanks for the heads up.

I'll try to find the threads on Collatz tomorrow. If you haven't done this one recently or forever, there's a bit of set up.
There are four flavors: micro, mini, solo, large. The default settings are very low. You'll need to set up a xxx.config file. here are his examples:

http://boinc.thesonntags.com/collatz/download/opt_config/

These are mine on the big GPU. (verbose can be 0, but for tweaking 1 is better)

verbose=1
threads=8
items_per_kernel=22
kernels_per_reduction=9
sleep=1

On an old laptop with an old NVIDIA Quadro NVS 140M (256MB) driver: 332.76 OpenCL: 1.00.

verbose=1
items_per_kernel=10
kernels_per_reduction=5
threads=5
sleep=1

 

[Ken g6]

Huh, interesting. I'm just using the CUDA app on my Nvidia card, since I don't want to dedicate a core to OpenCL. I see you're using it too. Have you tried the OpenCL app on that laptop? How does the speed compare overall?

By the way, Large with the CUDA app on Linux seems like a bad idea. That particular app has suspending problems. Which is why I built up a buffer on both Collatz and POGS and I'm holding it until (late) Friday.

Also, that Hawaii is amazing! 4.5 hours for a Large?! My GTX 460 would have taken 40 if I'd let it. It's taking 2.5 hours on Solos.

 

[petrusbroder]

127 Rosetta WUs dumped ... lets see if it matters ...

 

[Rattledagger]

Due to Rosetta@home running ancient server-code it's a little difficult to know how many reported, but by my rough estimate have returned roughly 500 of the 3 hour-types.

Of these had 2 errors, but thankfully this was bad wu's crashing after 20 seconds or something so no problem. Not so good was one validation-error, due to a bug in the ancient server-code.

Hmm, my results was reported at the same time as Petrusbroder so combined this accounts for the roughly 50k jump in the stats, if not anyone else also reported their bunker the same hour.

 

[GLeeM]

You'll need to set up a xxx.config file.

Thanks zzuupp for bringing this up. The file should already be in the project folder. It should be empty. All you have to do is open for edit and copy and paste!

The PPD should increase immensely with the right configuration. Examples below for large WUs:

On my GTX780 @1084:
OpenCL with empty config file = 16:34
OpenCL with filled in config file = ~11:00 (estimate after 1/3 done)
Cuda with filled in config file = 7:38 and 8:00

On my GTX560Ti Core 448 @797:
OpenCL with empty config file = 19:05
Cuda with filled in config file = 12:32

So Cuda does better for new and older Nvidia.

ATI does better:
7950@1000 = 7:15 to 7:35
6870@900 = ~22:00+

For my GTX780 and GTX560Ti Core 448 this is my config:
verbose=0
items_per_kernel=19
kernels_per_reduction=9
threads=8
sleep=1

I may be able to increase the items_per_kernel by 1 but they are at 98 & 99% usage now without any desktop lag so I think I am happy

More helpful info - Enigma only gave me a little over 200 WUs and said I reached the maximum. They are all crunched and awaiting the <min_rpc_time>.

And the TeAm is now at position 17 with 51 points :thumbsup:

 

[Assimilator1]

I've decided to switch 2 cores over to R@H, loathe to lose pogs power but I've been meaning to crunch at least a little R@H for 8-9yrs! (assuming BOINC scheduler allows it!).

127 Rosetta WUs dumped ... lets see if it matters ...

Nice!, & RattleDagger too

Keng6
Not familiar with primegrid, re Collatz, it didn't list any end game practical uses, hence my comment .
Enigma's intriguing!

Guess I could rustle up some quick stats

Rank.... Team ........... Overall points ... Pogs ... R@H

16 Meisterkuehler.de Team ... 69 .......... 60 ....... 9
17 BOINCstats .................... 48 .......... 27 ....... 21
18 TeAm AnandTech ............ 48 .......... 15 ....... 33
19 Aggie The Pew ................ 42 .......... 24 ....... 18

We climbed 2 positions
Btw seti.germanys race stats rankings are messed up atm!

 

[Ken g6]

@Gleem, does the config file matter for CUDA? It looked like the config files only applied to OpenCL.

Thanks for the stats, A1!

 

[GLeeM]

@Gleem, does the config file matter for CUDA? It looked like the config files only applied to OpenCL.

Not positive, but I thought so. What percent GPU usage is it running at?
Is there an empty file named - large_collatz_windows_x86_64_cuda.config in the project folder? Or solo_collatz_???_cuda.config?
I would try a lower number like 16 or 17 or 18 instead of 19 for "items_per_kernel" for your 460.

 

[Ken g6]

I just set up the file with items_per_kernel=18. Speed almost doubled! And I didn't have to restart BOINC or anything - just waited for the next WU to start.

Thank you, GLeeM! :thumbsup:

 

[Ken g6]

The 5th and final of the BOINC Pentathlon in the discipline Showbag is:
POEM@HOME (http://boinc.fzk.de/poem/)

All credits granted between
05/14/2014, 0.00 (UTC) and
05/19/2014, 0.00 (UTC) will be taken into the BOINC Pentathlon validation.

I'm not sure if that project has GPU work or not. The front page says maybe, but the applications list says no.

Edit: Server status solves that: POEM++ OpenCL version: 0. But that might change, I guess.

 

[zzuupp]

... Have you tried the OpenCL app on that laptop? How does the speed compare overall?

....

No, I haven't. It's more dumb luck that those stats still exist in their system.

 

[biodoc]

I've been crunching Rosetta and Collatz and will dump WUs from both projects when Collatz goes live on the 10th, 00:00 UTZ.

 

[petrusbroder]

I've had connection problems for approx 22 hours. Now solved ...
Thus:
370 pogs-WUs "Validation pending" and 59 pogs-WUs valid ...
84 Rosetta-WUs valid
Lets see what thet does ...

 

[Assimilator1]

LOL
Joe is #3 on the TeAm in BOINC :thumbsup:
http://stats.free-dc.org/stats.php?page=userteamcomb&team=TeAm+AnandTech

Joe has 12 projects over 1M and 33 projects over 100K points with 68 projects participated in!!

I don't know if he has posted to the forum, but I am sure he watches because of his participation in things like this.

A big HELLO and well done to Joe

Forgot to say, wow a big gun then! No wonder I'm struggling ,lol
Good job Joe!

Rank.... Team ........... Overall points ... Pogs ... R@H

15 BOINC@AUSTRALIA ..... 69 ............. 42 ...... 27
16 Ukraine ...................... 69 ............. 30 ...... 39
17 TeAm AnandTech ........ 48 ............. 15 ....... 33
18 BOINCstats ................ 45 ............. 27 ....... 18

Err, not sure what's going on with the stats, some of the numbers look the same as yesterday

Must admit I haven't switched to Collatz, looks like hassle to set up & I haven't had time yet.
Oh & BOINC isn't letting R@H run , seem to be too many high priority tasks for Pogs, oh well at least it's still adding to the race .

Poem seems another good project!

Keng6

9) Ken_g6 .................... 132,988 ... 3,932.32 ... United States
10) [TA]Assimilator1 ....... 49,038 ... 3,928.45 ... United Kingdom

Didn't see that coming!, maybe I should stick with just Pogs .

 

[Ken g6]

Collatz isn't a hassle to set up. It's just a hassle to set up optimally. The settings aren't that hard to change late either.

 

[Fardringle]

I just set up a 6 core (12 virtual) Xeon E5-2430L server yesterday. It won't go into actual use for a couple of weeks (got delivered early) so for now it's just "stress testing" running POGS for me. It will be interesting to see how the Xeon CPU scores on the project.


edit: 10,399 points after just the first 24 hours running Pogs. Not too shabby at all!

 

[Ken g6]

maybe I should stick with just Pogs .

On second thought, I think you're right, at least for CPUs. We're at the bottom of a stack of several teams very close together, and a concerted effort should push us up several places.

 

[petrusbroder]

I have more than 40K points pending ... and I'll stick to pogs with the computers that are crunching pogs, and swithch the Rosetta@home computers to the other projects, I think ...