BOINC Pentathlon 2020

[FraktalU]

if we can keep performing like that at rosetta and P3D isn´t crunshing more than now, we will snatch silver in ~24h.

[rkv_2401]

Looks like RKN is going to catch up and pass us on Cross Country(AmicableNumbers) pretty soon. https://www.seti-germany.de/boinc_pentathlon/statistiken/challenge.php?challengeid=4

[GOTSpectrum]

1 minute ago, rkv_2401 said:

Looks like RKN is going to catch up and pass us on Cross Country(AmicableNumbers) pretty soon. https://www.seti-germany.de/boinc_pentathlon/statistiken/challenge.php?challengeid=4

 

@Den-Fi is going to throw a system on cross country for us! Gotta hope it will be enough. 

 

 

[miker07]

Doesn’t help that I have 1000 tasks pending validation. Also everybody might want to check their number fields bunkers for the javelin, we have over 90k points for throw4. It may be people crunching outside of the pent, but wanted to mention.

[Stormtrooper_007]

Can you help me with Amicable Numbers, my GPU is only doing one job at a time, but isn't using its whole power. Is there any way to do 3 or more jobs per gpu? I have 32gb of Ran, so Ram schouldn't be the issue

[cbigfoot]

18 minutes ago, Stormtrooper_007 said:

Can you help me with Amicable Numbers, my GPU is only doing one job at a time, but isn't using its whole power. Is there any way to do 3 or more jobs per gpu? I have 32gb of Ran, so Ram schouldn't be the issue

Create an app_config XML in the sech.me folder in C:\ProgramData\BOINC\projects\sech.me_boinc_Amicable with the following in it this is for amicable

for 2 tasks per GPU

<app_config>
<app>
<name>amicable_10_21</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
</app_config>

for 3 tasks per GPU 

<app_config>
<app>
<name>amicable_10_21</name>
<gpu_versions>
<gpu_usage>0.33</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
</app_config>

 

[Gorgon]

1 hour ago, Stormtrooper_007 said:

Can you help me with Amicable Numbers, my GPU is only doing one job at a time, but isn't using its whole power. Is there any way to do 3 or more jobs per gpu? I have 32gb of Ran, so Ram schouldn't be the issue

2nd method:

 

Create a file app_config.xml containing:

<app_config>
    <app>
        <name>amicable_10_21</name>
        <gpu_versions>
            <gpu_usage>0.5</gpu_usage>
            <cpu_usage>1</cpu_usage>
        </gpu_versions>
    </app>
</app_config>

and place it in C:\ProgramData\BOINC\projects\sech.me then select "Read Config Files" in the Options Menu in BOINC Manager

Edited May 15, 2020 by Gorgon
Original for NumberFields

[Stormtrooper_007]

Just now, cbigfoot said:

Create an app_config XML in the search.me folder with the following in it ths is for amicable

for 2 tasks per GPU

<app_config>
<app>
<name>amicable_10_21</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
</app_config>

for 3 tasks per GPU 

<app_config>
<app>
<name>amicable_10_21</name>
<gpu_versions>
<gpu_usage>0.33</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
</app_config>

 

 

Just now, Gorgon said:

Create a file app_config.xml containing:

<app_config>
    <app>
        <name>GetDecics</name>
        <gpu_versions>
            <gpu_usage>0.5</gpu_usage>
            <cpu_usage>1</cpu_usage>
        </gpu_versions>
    </app>
</app_config>

and place it in the amicable folder in C:\ProgramData\BOINC\projects the select "Read Config Files" in the Options Menu in BOINC Manager

Thank you for the quick answer  works perfektly fine

[Gorgon]

1 hour ago, miker07 said:

Doesn’t help that I have 1000 tasks pending validation. Also everybody might want to check their number fields bunkers for the javelin, we have over 90k points for throw4. It may be people crunching outside of the pent, but wanted to mention.

PSA:

 

If you are still running NumberFields then make sure you are bunkering (not submitting tasks upon completion but saving them) as we are NOT participating in today's NumberFields Javelin throw.

 

If you have GPUs and don't want to bunker you could run AmicableNumbers on them as long as you have 8GB RAM per GPU.

 

The last NumberFields Javelin Throw number 5 is the 18th-19th (the last day) and we are participating in that Throw and aiming to match the 1.5M points from Throws 2 & 3.

[SwedishNinja]

On Ibercivis, I have some jobs taking 2h10m. 13052020_R60_L000004 series

Then I have the ones that take 15h54m. 2604202_R04_L000007 series

Wish i knew which was the best one to run for credit!

 

Anyone else has this issue? 

[Gorgon]

8 minutes ago, SwedishNinja said:

On Ibercivis, I have some jobs taking 2h10m. 13052020_R60_L000004 series

Then I have the ones that take 15h54m. 2604202_R04_L000007 series

Wish i knew which was the best one to run for credit!

 

Anyone else has this issue? 

When the project was announced the tasks at that time were long and they since released shorter tasks. I believe @leadeater found they both award more points for the longer tasks

[GOTSpectrum]

Managed to convince @Elektra57 to jump on the marathon, no idea what hardware she's throwing at it though. 

[Elektra57]

17 minutes ago, GOTSpectrum said:

Managed to convince @Elektra57 to jump on the marathon, no idea what hardware she's throwing at it though. 

Ryzen 5 3600x  Yes you little FAH and Boinc dealer.

[GOTSpectrum]

4 minutes ago, Elektra57 said:

Ryzen 5 3600x  Yes you little FAH and Boinc dealer.

Glad to have your beautiful ryzen chip getting in on the crunching action! 

[Gorgon]

5 minutes ago, Elektra57 said:

Ryzen 5 3600x  Yes you little FAH and Boinc dealer.

Cmom. You can try it. All the Cool kids are doing it. You can stop anytime. ... and so it begins

[Elektra57]

4 minutes ago, GOTSpectrum said:

Glad to have your beautiful ryzen chip getting in on the crunching action! 

I am still folding for LTT with my 2070 super

[Elektra57]

3 minutes ago, Gorgon said:

Cmom. You can try it. All the Cool kids are doing it. You can stop anytime. ... and so it begins

He came into the Pilot Chat on discord pedaling his wares. Put on the peer pressure.

[GOTSpectrum]

3 minutes ago, Elektra57 said:

I am still folding for LTT with my 2070 super

Beautiful!

 

3 minutes ago, Elektra57 said:

He came into the Pilot Chat on discord pedaling his wares. Put on the peer pressure.

I only do it on the more private parts of Discord... Can't do it in the open!

[Stormtrooper_007]

1 hour ago, Gorgon said:

PSA:

 

If you are still running NumberFields then make sure you are bunkering (not submitting tasks upon completion but saving them) as we are NOT participating in today's NumberFields Javelin throw.

 

If you have GPUs and don't want to bunker you could run AmicableNumbers on them as long as you have 8GB RAM per GPU.

 

The last NumberFields Javelin Throw number 5 is the 18th-19th (the last day) and we are participating in that Throw and aiming to match the 1.5M points from Throws 2 & 3.

Started bunkering NumberFields today  Any thoughts of how much ppd a 9900k@5ghz can produce?

[Gorgon]

2 minutes ago, Stormtrooper_007 said:

Started bunkering NumberFields today  Any thoughts of how much ppd a 9900k@5ghz can produce?

Sorry, I only do GPU on number fields. I did try running CPU tasks the other day but was looking at 8-10 hours per task. You'd have to submit 4 or 5 completed tasks to get an idea of what kind of points you'd get.

 

Can your 9900K actually sustain 5GHz all-core on compute? When I was testing mine with the 280mm EVGA CLC the best I could do with AVX enable small FFTs in mPrime (Prime85) was 4.4GHz as I had to use a -3 offset to keep that blast-furnace of a CPU under 95C.

 

It's currently crunching Rosetta and is sitting at 4425MHz and 65-69C and the VRM's at 70C which is pretty good for the rather weak VRM on the GigaByte z370 Gaming5

[Stormtrooper_007]

1 minute ago, Gorgon said:

Sorry, I only do GPU on number fields. I did try running CPU tasks the other day but was looking at 8-10 hours per task. You'd have to submit 4 or 5 completed tasks to get an idea of what kind of points you'd get.

 

Can your 9900K actually sustain 5GHz all-core on compute? When I was testing mine with the 280mm EVGA CLC the best I could do with AVX enable small FFTs in mPrime (Prime85) was 4.4GHz as I had to use a -3 offset to keep that blast-furnace of a CPU under 95C.

 

It's currently crunching Rosetta and is sitting at 4425MHz and 65-69C and the VRM's at 70C which is pretty good for the rather weak VRM on the GigaByte z370 Gaming5

No problems  i've just asked. I had really luck with binning, so i could get my 9900k to arround 5,8ghz with my custom waterloop. For one task i need between 0.5h to 1.5h and im getting between 40 and 60 Points. So if my math is right (1h,  50points average and 8 threads =24*8*50=9600) i could bunker around 30k points  Thought it would be more  

[Gorgon]

I've been playing around with GPU settings on NumberFields to see just how utilized the lower-tier cards in my 6 GPU rig are when running just 1 task per GPU.

 

The 1060 6GB on one task runs about 93-94% and both the 1660Ti and 1070Ti sit at about 90-91% utilized. No real conclusions here. I'm back to running 2 tasks per GPU as it's just easier to manage one bunker per system than 2.

 

Did however confirm that in multi GPU rigs the cards in BOINC are enumerated fastest to slowest based, I'm guessing, on the benchmark run at startup. Haven't tried a mixed AMD/NVidia, no idea how that would work out.

 

To remove my slowest card from the bunker I used in the option of the cc_config.xml:

<ignore_nvidia_dev>5</ignore_nvidia_dev>

and on the test bunker had to use:

<ignore_nvidia_dev>0</ignore_nvidia_dev>
<ignore_nvidia_dev>1</ignore_nvidia_dev>
<ignore_nvidia_dev>2</ignore_nvidia_dev>
<ignore_nvidia_dev>3</ignore_nvidia_dev>
<ignore_nvidia_dev>4</ignore_nvidia_dev>

[Gorgon]

5 minutes ago, Stormtrooper_007 said:

No problems  i've just asked. I had really luck with binning, so i could get my 9900k to arround 5,8ghz with my custom waterloop. For one task i need between 0.5h to 1.5h and im getting between 40 and 60 Points. So if my math is right (1h,  50points average and 8 threads =24*8*50=9600) i could bunker around 30k points  Thought it would be more  

5.8GHz - nice. After this is all done try running Prime95 Small FFTs with AVX enabled and let me know what all-core you can maintain. This is usually pooh-poohed as an unrealistic test for gaming and productivity workloads but F@H does appear to use AVX as I've noticed running F@H the AVX offset is applied and mine drops from 4.7 all-core down to 4.4.

 

Instead of getting the 9900k I should have cut my losses and just gotten another 3900x. That thing is a beast at compute tasks and with what I ended up having to spend on the 280 CLC to jeep the 9900k cool I would have been most of the way there for another x570 motherboard that has a decent VRM.

 

That's probably about right now that you mention it I did notice that the points awarded seemed similar for a CPU task as what a GPU got for the same task which seemed to make the only advantage to running a GPU is more tasks per hour. My 2070 Supers take about 8 minutes running two tasks concurrently and the 6-GPU rig averages out to around 12 minutes per task (2070A, 2060S, 2060, 1660Ti, 1070Ti, 1060 6GB).

[Stormtrooper_007]

6 minutes ago, Gorgon said:

5.8GHz - nice. After this is all done try running Prime95 Small FFTs with AVX enabled and let me know what all-core you can maintain. This is usually pooh-poohed as an unrealistic test for gaming and productivity workloads but F@H does appear to use AVX as I've noticed running F@H the AVX offset is applied and mine drops from 4.7 all-core down to 4.4.

 

Instead of getting the 9900k I should have cut my losses and just gotten another 3900x. That thing is a beast at compute tasks and with what I ended up having to spend on the 280 CLC to jeep the 9900k cool I would have been most of the way there for another x570 motherboard that has a decent VRM.

 

That's probably about right now that you mention it I did notice that the points awarded seemed similar for a CPU task as what a GPU got for the same task which seemed to make the only advantage to running a GPU is more tasks per hour. My 2070 Supers take about 8 minutes running two tasks concurrently and the 6-GPU rig averages out to around 12 minutes per task (2070A, 2060S, 2060, 1660Ti, 1070Ti, 1060 6GB).

5.8Ghz is Prime95 stable, but (sadly) only at larger FFTs and if my GPU isnt outputting 280 Watts.

But this is probably because its in a Midtower with a single 280mm rad (45mm thick)

 

But with 5.8Ghz it can suck up to 480watts which would make my powerbill tooooo high lol

 

I'm currently running at 5Ghz all Core without AVX offset

Also my 1080ti is outputting 280Watts so a total of arround 500Watts,

so temperatures are ok i guess

[Stormtrooper_007]

Just messured my water temps. Seems like the high cpu temp comes form the bad intel thermal paste  . i'm thinking of delidding the cpu and apply some proper thermal paste