Formula Boinc 2014 News

[zzuupp]

Can somebody tell me of changes I can make to the Moo computing preferences to further increase PPD?

There is no easy way. If all of your tasks are pretty close in time, then most likely nothing. Their auto-picker is working.

 

[StitchExperimen]

I've switched to Milkyway with 1 GTX 460 and 14 cores at 3.4 GHz.

Leiden is fine. Nobody pursuing.

Constellation is fine. Nobody pursuing.
SETIKAH@KOREA is back -338,158 that's 48,000 a day so I'd say count them out.

 

[VirtualLarry]

Are we in danger of an overtake in MW@Home? My two HD4850 cards have been idle for probably at least two days, to allow my Thuban to crunch WCG for that race.

Should I fire MW@Home back up?

Hmm, I fired MW@Home back up, and the cruncher was still running, but CoreTemp said that the min temp was 0C, and the max temp was 192C.

Do you think that was a legit temp, and how is the system still running?

 

[GLeeM]

the max temp was 192C

192C=378F! Not possible - you will cook your turkey at a lower temp!

Should I fire MW@Home back up?

They are catching up faster than I thought. But, you could probably start after the WCG race.

EDIT: Yikes! After looking closer, I think I will switch over too.

 

[StitchExperimen]

My math has Milkyway taken over in 3 days from final results in 6 days, 7 hours, 3 minutes.
Por fa vor more power, please!

 

[salvorhardin]

I'm finishing my last tasks in moo and switching both my gpus over to milkyway.

 

[GLeeM]

My math has Milkyway taken over in 3 days from final results in 6 days, 7 hours, 3 minutes.
Por fa vor more power, please!

I should have been watching closer - they added ~2.5 Million per day!

I suspended the MOO WUs and started Milkyway. I added 150 to the OC on the 7950 for 50,000 more ppd. The GTX 780 was finishing WU in 205 seconds and only at 55% usage so I tried two at once ... they finish in 225 :thumbsup:
My total ppd calculates to 563,312!

BUT we still need more - who else can help?

 

[StitchExperimen]

Latest guesstament is from preliminary numbers are we will last 4 1/3 days of 6.

 

[GLeeM]

Latest guesstament is from preliminary numbers are we will last 4 1/3 days of 6.

How much more do we need to get through to the end of the month?

 

[VirtualLarry]

Ok, I've told PrimeGrid no new tasks, and allowed MW@Home tasks, on my two Q9300 @ 3.0 w/GT630 384SP Kepler cards.

One of them is still running NumberFields@Home on CPU, should I tell that no new tasks? Are we in danger of an overtake on that project?

Edit: While I've been waiting for the PrimeGrid CUDA WUs to finish up, I've been web browsing, and I ran a CDM benchmark on my SSD. Not long after, the machine froze hard. No keyboard / mouse, even hooking up a wired keyboard and mouse didn't work. Mouse wouldn't light up, keyboard wouldn't toggle LEDs. Hit the reset button on the chassis, and it eventually restarted. Temps didn't seem like they were the problem. Unsure. This PC's mobo is from 2007, maybe it's finally starting to get old. Or maybe I need to notch the vcore up one setting. (Currently 1.22 or 1.23v-something).

CPU temps on the Q9300 @ 3.0 got as high as 74C, video card is getting to 69C, that might be somewhat high for a passive kepler 384SP card. Whoops, just hit 70C on the video card, doing "PPS (Sieve) 1.39 (cudePPSsieve)".

Edit: Froze up again, while I was Aborting the rest of the GPU tasks, so I could get started on MW@Home sooner.
Took it apart, uses some compressed gas on it, wasn't altogether too dusty, although the tips of the fan blades in the PSU had some dust bunnies. Would have to take PSU apart to get those out. Took the side off, before placing PC back into desk cubby.

Edit: Specs for that GPU are here:
http://www.geforce.com/hardware/desktop-gpus/geforce-gt-630/specifications
Max temp 90C, so it's not hitting max temp. (You might think it would throttle if it hit close to that temp.)

 

[GLeeM]

Thanks VirtualLarry :thumbsup:

Are we in danger of an overtake on that project?

We are still getting a larger lead in NumberFields.

What you could do to improve ppd is to set your Milkyway Preferences to only get regular WUs. The Modified Fit WUs do not give as good PPD on (EDIT: some) GPUs, yours included, (64,000 vs 50,000).
Also make sure to leave a free CPU core/thread for the GPU to use.

 

[salvorhardin]

So the regular WUs give better ppd? I've been running modified fit wus since they constantly crash on my 5850 and didn't want risking the same on my 7950.

 

[StitchExperimen]

One of their crunchers is down for now.
...
Still down 4 hours later.
...
Still down 4 hours later. (Last update : 11/25/2014 16:02 (CET).)

 

[GLeeM]

So the regular WUs give better ppd? I've been running modified fit wus since they constantly crash on my 5850 and didn't want risking the same on my 7950.

Hmm ... OK, some GPUs give better ppd with regular WUs. Check which does better on your card.

 

[biodoc]

Ok, I suspended GPUGrid on my nvidia cards and moved them to MilkyWay.

 

[StitchExperimen]

Preliminary 4 hour numbers with them down one cruncher has us stable.
We've only lost 18,000 points of 3,955,961 in ~4 hours.
~5 days left.

 

[GLeeM]

Ok, I suspended GPUGrid on my nvidia cards and moved them to MilkyWay.

Thanks biodoc :thumbsup:

Boing ... Boing ... Boinc! Kinda like ping pong

My GTX 560 Ti 448 core does a regular WU in 148 seconds.
The GTX 970 in 119 seconds.
GTX 780 does two in 225 seconds. (It had less than 60% usage with one WU)
The 7950 finishes in 26 seconds

I'm not sure if 980 or 780Ti needs to run two WUs.
Here is my app_config.xml file:

<app_config>
<app>
<name>milkyway</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.02</cpu_usage>
</gpu_versions>
</app>
</app_config>

 

[StitchExperimen]

US-------------------Them----------3,955,961 difference
286,282,290-------282,262,787
-----571,670------------508,128----4,019,503 difference
They are still down one cruncher.
~4 days 12 hours

Ours <15K Theirs per 4 hours difference.

 

[GLeeM]

Hey Stitch - have you checked Collatz recently?

We had 31M lead, but over the last three days we lost 16M to [H]! They seem to have slowed down for now. Maybe we'll be OK - must watch closely

Is it possible they are bunkering finished or almost finished WUs? groan

EDIT: I had my numbers wrong for MW WU times:
GTX560Ti 448 core is 119 seconds
GTX970 is 148 seconds

The 560 does better than 970?

 

[waffleironhead]

Hey Stitch - have you checked Collatz recently?

We had 31M lead, but over the last three days we lost 16M to [H]! They seem to have slowed down for now. Maybe we'll be OK - must watch closely

Is it possible they are bunkering finished or almost finished WUs? groan

EDIT: I had my numbers wrong for MW WU times:
GTX560Ti 448 core is 119 seconds
GTX970 is 148 seconds

The 560 does better than 970?

Does mw depend on double precision?
The 560ti 448 has 1/8 dp so 1371/8=171.375 dp rate
The 970 has 1/32 dp so 3494/32= 109.1875 dp rate

 

[biodoc]

Yes, Milkyway does require Double precision.

My nvidia numbers for the "short" WUs:

GTX570: ~126 seconds (1 WU at a time)
GTX660Ti: ~407 seconds (2 WU at a time)
GTX780Ti: ~181 seconds (2 WU at a time)
GTX980: ~127 seconds (1 WU at a time)

Good thing I normally focus on GPUGrid.

 

[StitchExperimen]

Hey Stitch - have you checked Collatz recently?

No
I have just kept on Riojas, Leiden, Constellation & Milkyway@home and for a time on Moo,
Riojas I look at their server doling out wu.

 

[GLeeM]

Teams are splitted in 3 divisions, after calculation of their average RAC of November the year before. RAC Division thresholds are fixed to 0, 1 million and 10 millions credits.

From the Formula Rules.

The TeAm did so much Collatz (high paying) this month that we are over 10 Million RAC for November!
We would be in League #1 next year

But it looks like we won't have to push in Collatz, they slowed down a little. I will bunker some WUs ready if we need them. If we don't, then I will submit them in December

How fast does RAC respond? Is there time to get it below 10M and still maintain our Formula position? It is at 10,289,937 now.

 

[salvorhardin]

It should be going down since some of us switched over to milkyway from moo. we're 2 million a day lower now and that rac is still at 1 million. I checked right now and our rac is at 10,109,011 now. It looks like we'll keep our position in milkyway, we're still around 4 million points ahead.

 

[biodoc]

I switched from GPUGrid (1.5 million RAC) to MW and am now doing 250K per day.