PrimeGrid Mathematics Awareness Month Challenge (Part 2 of the PrimeGrid Challenge Series)


[H]ard|DCer of the Year - 2014
Jan 29, 2006
Mathematics Awareness Month Challenge
From April 7th 12:00:00 UTC until April 22nd 12:00:00 UTC we will be having a 15 day challenge running the Prime Sierpinski Problem (LLR) Project.

This project is an LLR work unit and is CPU only. This means you will more than likely have better results with Hyper Threading turned off. Intels will outperform AMD. If you have your chips OC'd, make sure they are stable as the optimized mathematics apps will push your chips harder than any other work units.

Make sure to log into your account at their website and change the project preferences to only work on the PSP.
Anyone joining in for this second PG challenge of the year? This is a long task ~ 1 - 2 days depending on your CPU.

Oh if you happen to have Ryzen, maybe this is a good time to check if their AVX version will do well in this LLR PG project.
And before anyone gets any ideas, G34 chips will probably take forever on these work units. ~1-2 days is for Intel chips.
lelliott731, ChelseaOilman put your shiny new Ryzen to this challenge! Curious about their AVX performance... :)
Last edited:
I will be hitting WCG on CPU until I hit my goal on Smash Childhood Cancer and my GPU's are running Einstein for the next sprint for Formula-BOINC. However, I may switch my borgs (which aren't many these days) over to this to knock out a few.
lelliott731, ChelseaOilman put your shiny new Ryzen to this challenge! Curious above AVX performance... :)

Yeah me too, but from what I have read, the AMD AVX is still not up to par with Intel on PG LLR tasks.
But it is doing well on PG Sieve tasks.
Please do post up any findings.
Good luck on all the challenges going on/coming up.

If you have a 48/54 core AMD server and still need the heat, you should be able to get at least one batch in for the PG challenge.
Getting warm here only going with my fastest Intel CPUs and only running a couple rigs this go round.
Spring and car season is near and I want to do some mods on the Mustang.
Got to have priorities you know!
Good luck guys
However, I may switch my borgs (which aren't many these days) over to this to knock out a few
I don't have many borgs too but I do have many virtual machines (VMs) running WCG to optimize output...:D
Good luck to your team too Bill1024.

the AMD AVX is still not up to par with Intel on PG LLR tasks
. I read something similar in my earlier post here. I wonder why AMD did not pursue hardware AVX.
They have AVX pututu, at least it is listed they have it.
It is not implemented correctly, maybe it is a patent issue or something else, I don't know / understand what their problem is.
Anyone running PSP LLR project for this challenge? I've returned two WUs, so at least we are on the leaderboard now. Last check at 39th position.
For those wanting to reduce the task time per WU of PSP LLR, there is a good discussion here. In a nutshell, if you run one WU with one core, it takes X hours. You can create app_config.xml file to run as many cores per WU, so in this example if you decide to run 4 cores per WU, the time it takes to complete one WU is X/4 hours + overhead inefficiency as shown here. If you have ran Amicable Numbers before, it does about the same thing for CPU crunching, reducing task time per WU by running more cores/threads.

I tried on my E5-2670. Below is my app_config.xml file running 6/8 cores per WU and the result. It took about 8.45 hrs to complete on WU of PSP LLR.

<cmdline>-t 6</cmdline>


I'm currently running 7/8 cores but my CPU is reaching the power limit of 115W. Currently at 114.3W :oops:
...and the CPU freq (normally at turbo mode of 3.0 GHz sometimes throttles down to 2.9 GHz. Expected total time per WU is about 7.5 hours after more than 50% completion mark.

Great. Let's compare and publish results. I'm running some more studies and from reading Michael's post here and Mackerel's (right below Michael's post), you lose some efficiency but returns the WU faster, if that matters to some. However running PSP LLR, I'm running into max CPU thermal limit which throttles the CPU frequency down upon reaching that limit. Other than the overclockable CPU, I'm not sure if there is any way to force the Xeon CPU to run over its inherent power limit over sustain period of time. I can understand why Intel wants to do this.
After more than 5 days or a third into this challenge, we are hovering around twenty something position. Last check at 21st.
With less than 5 days to the end of this "marathon" challenge, we are at 23rd position. I'm aiming for Amethyst (1M) badge for PSP LLR....
We finished at 25th.

I'm at 81st with 1,157,118 credits, enough for the 1M badge.