10th annual Folding@Home Holiday Season Race

[QuietDad]

Mark, you sure about that? I thought for any DC that 1 CPU core was enough for 1 GPU crunching?
I've just been looking at my HD 7950's GPU load for a few minutes & it was nearly always at 100%.
Also, I stopped CPU crunching on my 2nd rig & didn't notice any change in GPU ppd, in the short term at least.......



Yea that's odd, my Q9550 @3.6 GHz, was only showing 1.2k ppd, although I forgot to mention that was only on 3 cores.
Your Q6600's output seems surprisingly high to me!

So what was the ppd of just your 2 Xeons?

Haven't really run it just the CPUs. Been doing Seti forever and recently with it's problems, I'm just getting away from it for the season. It was doing 9K Folding on the GPU and 7 cores. We'll see what 6 brings

 

[petrusbroder]

The stats are as of December 20, 2015, 21:00 UTC:

Individual stats:

The day-by-day-production:

The TeAm's production graph:

The PrimeGrid race is quite noticable ... :\

 

[petrusbroder]

We have, in 2015, produced as many points from December 1, 2015 to December 20, 2015 as in the whole race in 2014:

Not too shabby!

 

[Drsignguy]

Thanks for the stats Peter! Appreciate it.

 

[biodoc]

I can't remember where I read that, but it said always use a multiple of 2.

Yeah, I remember that too. I checked the FAH FAQ.

"We have seen reports of occasional problems during processing if FAH is using an odd number of processors greater than five. One and three are usually fine, but problems sometime occur on seven, nine, eleven, etc. numbers of cores. The problem relates to how Gromacs splits up the work and allocates it to each processor. We are working to resolve this unusual issue, which affects very few users. Changing the number of utilized cores is typically an advanced/expert option."

 

[TennesseeTony]

3. Short intermission for the The Winter Solstice PrimeGrid Challenge December 19 - December 22.

Thanks for the stats, but isn't the FAH race on hold? I'll never beat Pokey if it is not.

 

[Assimilator1]

Thanks for the stats Petrus
Err, regarding the monthly table you showed though, the points for 12/14 are just for that month aren't they? So 12/15 output is up ~536k from 12/14 isn't it?, still a big leap but not quite in the same ballpark .
Unless I've totally mis-read something, lol.

Haven't really run it just the CPUs. Been doing Seti forever and recently with it's problems, I'm just getting away from it for the season. It was doing 9K Folding on the GPU and 7 cores. We'll see what 6 brings

You don't have to just run it on the CPU to find CPU ppd, highlight the CPU slot or WU & you'll see it's ppd estimate .

biodoc
Ah thanks for that , so yea 1 core per GPU is fine, but crunching on 7 cores or higher odd numbers can cause problems.
I'm crunching on 7 atm! Although do they mean just real cores or virtual & real cores?
Hmmm, the FAQ doesn't say....
[update] I've created a thread asking about it.

But I did see this regarding GPU crunching :-

What about using both SMP and GPUs?

GPUs can also be used to process a separate FAH assignment while running SMP (see the GPU FAQs) but some CPU processing is required to support that GPU. If you also process with a GPU from AMD, that GPU support function consumes one CPU core and less is available for SMP processing. You can expect to higher SMP performance by dedicating fewer CPU-cores to SMP to avoid resource conflicts. NVIDIA GPUs do not currently require a full CPU’s resources.

 

[TennesseeTony]

I've been running at 27,25,23, 21, 13 and11 cores and had no issues that I know of.

 

[Assimilator1]

Wow!

 

[petrusbroder]

Thanks for the stats Petrus
Err, regarding the monthly table you showed though, the points for 12/14 are just for that month aren't they? So 12/15 output is up ~536k from 12/14 isn't it?, still a big leap but not quite in the same ballpark .
Unless I've totally mis-read something, lol.[/I]

The points 12/14 is for the whole month, the points for 12/15 are just for the first 20 days.
Thus we have produced in 20 days of December 2015 536 K more points than we did in 31 days of December 2014. That is the point ...
I predict that the production for the whole month of December 2015 will be in the order of 270 M points, almost 100 M points more than in 2014 ...

 

[petrusbroder]

The stats are as of December 21, 2015, 19:00 UTC:

The Individual stats:

The day-by-day production, part 1, before the adjustment of the teams:

The day-by-day production, part 2, after the adjustment of the teams:

The TeAm's production graph:

I think we will produce more than 275M points during December and maybe even more during the race - as it continues for 12 more hours into 2016!

 

[Assimilator1]

The points 12/14 is for the whole month, the points for 12/15 are just for the first 20 days.
Thus we have produced in 20 days of December 2015 536 K more points than we did in 31 days of December 2014. That is the point ...
I predict that the production for the whole month of December 2015 will be in the order of 270 M points, almost 100 M points more than in 2014 ...

Ah ok, you said 'as in the whole race in 2014:', you're just missing 'December' in that post .
And yes an awesome jump in output

Thanks for the extra stats!

 

[TennesseeTony]

3. Short intermission for the The Winter Solarstice PrimeGrid Challenge December 19 - December 22.

Thanks for the stats, but isn't the FAH race on hold? ...

Are the points counting during the PG Challenge? No intermission?

Darn new guys....not understanding how the annual race works....

_

 

[Assimilator1]

AFAIK they carry on counting, but traditionally many take a little break for PGC.

I don't, I take the chance to do just a little catching up .

 

[TennesseeTony]

I see. Pokey needed this to help maintain his lead anyway.

"Intermission" may not have been the best choice of words in this case, as it indicates that the action is ceasing, that it's time to take a break.

But, that just makes catching up that much more fun/challenging. (I think there's enough time left....)

So...okay Pokey....6 more hours, then:

 

[Ken g6]

AFAIK they carry on counting, but traditionally many take a little break for PGC.

I don't, I take the chance to do just a little catching up .

In the past the counting has taken a break, at least sometimes.

 

[Assimilator1]

Ah fair enough, my memory isn't that great
I suppose any 1 of us could trawl through the previous threads to see, I can't be bothered though .

 

[biodoc]

I don't remember taking an official Intermission for the PrimeGrid race. It's a bit of a nightmare for the person keeping stats. In the past, it would not of been fair for those crunching bigadv WUs, since they took a couple of days to crunch and depending on the start/finish timing, would lose a lot of race points.

 

[GLeeM]

Looks like the first part of the month used up all of my allotted "anytime" bandwidth

I still have lots of 2:00 AM to 8:00 AM bandwidth but it would be hard to use with F@H I think. With BOINC I can set network use only during that time and not lose points.

Is the F@H download or upload size bigger? I might be able to do a WU a day? Sorry!

 

[TennesseeTony]

I'm certain the FAH client is set to under utilize the GPU, so users don't get mad about their computer being sluggish, and their youtube videos of cats chasing laser pointers being choppy.

Is there any advanced settings or flags that will allow FAH to use more GPU? Depending on the WU, I am seeing as little as 66% utilization.

 

[petrusbroder]

Well, the stats will allow to compare the numbers - with and without the intermission.
Since part of the race is to better the TeAm's standing, some of us continue folding ...

 

[theAnimal]

I'm certain the FAH client is set to under utilize the GPU, so users don't get mad about their computer being sluggish, and their youtube videos of cats chasing laser pointers being choppy.

Is there any advanced settings or flags that will allow FAH to use more GPU? Depending on the WU, I am seeing as little as 66% utilization.

No settings for that unfortunately.

 

[petrusbroder]

The stats are as of December 22, 2015, 21:00 UTC:

The individual stats:

The day-by-day production:

The TeAm's production graph:

The last leg of the race is on ...

The difference between the teams is now very small: only 1.40%!

 

[Drsignguy]

thank you for that stats peter! well done!

 

[theAnimal]

Looks like the first part of the month used up all of my allotted "anytime" bandwidth

I still have lots of 2:00 AM to 8:00 AM bandwidth but it would be hard to use with F@H I think. With BOINC I can set network use only during that time and not lose points.

Is the F@H download or upload size bigger? I might be able to do a WU a day? Sorry!

Upload is bigger, but GPU WU are only a few MB anyway. (My last 2 uploads were 16 & 6.)