PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
11) Message boards : Number crunching : Conjunction of Venus & Jupiter Challenge (Message 124350)
Posted 410 days ago by Profile FishFryProject donor
In each chart below...


Thank you, Michael!

This is so helpful as most of us have varied CPUs crunching. Could Kaby Lake be that more efficient on a slower CPU? I wonder if the memory configuration influenced these samples:

| 4 | Intel(R) Core(TM) i5-7400 CPU @ 3.00GHz [Family 6 Model 158 Stepping 9] | 1 | 30.335277777777776 | 30.335277777777776 |
| 4 | Intel(R) Core(TM) i7-7700 CPU @ 3.60GHz [Family 6 Model 158 Stepping 9] | 11 | 30.432777777777776 | 101.5621571111111 |


Regardless, very useful benchmarks and your efforts are tremendously appreciated!

12) Message boards : Number crunching : Winter Solstice Challenge (Message 123931)
Posted 424 days ago by Profile FishFryProject donor
Guys - we're all nuts! ^^

Couldn't be happier to be part of the insane asylum :)

Really impressed by all that hardware pointed at GFN-21 and the advancement made. Top 10 results are just astounding. Congrats to all those who cracked 20MM+ !
13) Message boards : Number crunching : PrimeGrid is Moving!!! (Message 122848)
Posted 453 days ago by Profile FishFryProject donor
Well done to the team. 4.5 hours with all that data meant a lot of foresight. Kudos!
14) Message boards : Number crunching : 2019 Challenge Voting (Message 122774)
Posted 456 days ago by Profile FishFryProject donor
I'm guessing there's enough sieving done every day, so I vote for GFNs and LLRs:

llrTRP
llrESP
llrSGS
genefer 17 mega
llrPSP
llrGCW
genefer 18
15) Message boards : Number crunching : Leonids Shower Challenge (Message 122484)
Posted 462 days ago by Profile FishFryProject donor
Thanks Rick!
16) Message boards : Number crunching : Leonids Shower Challenge (Message 122481)
Posted 462 days ago by Profile FishFryProject donor
Tasks will take ~29 hours on average for CPUs and ~75 minutes on average for GPUs.


I never ran AP27 on CPUs. Anyone know if this application works well with multi-threading like LLR?

Good luck everyone!
17) Message boards : Number crunching : Halloween Challenge (Message 121500)
Posted 479 days ago by Profile FishFryProject donor


Check my computers for this:

http://www.primegrid.com/hosts_user.php?userid=63648

I use all cores on each CPU (usually 4) to run one WU. There is no other Boinc Application or program running in the background. So i am very confused why my run times are so inconsistent. I am using the app_config file as posted in the thread.


My Run and CPU times look just as inconsistent; however, I have been adjusting thread allocation in app_config.xml throughout the challenge. For all my hyper-threaded processors, running 50% is most efficient as Roger indicated in the challenge description - especially for llrWOO. I noticed some of your machines are running all threads, not 50%.

If the CPU/Run time calculation references -oThreadsPerTest in Stderr XML per task, I've made a mess of the time values. :)
18) Message boards : Generalized Fermat Prime Search : Points: 17 Low vs 17 Mega (Message 120822)
Posted 508 days ago by Profile FishFryProject donor
In both pages where averages are displayed (the task selection page and the fastest GPUs page) I've tried to make the data as reliable as possible, but there's only so much I can do to normalize all the real-world conditions that affect run times.


Totally understand. While ranges of hardware and their averages can be listed (including whether the user multithreads or not), the webpage would appear noisy per project. Too many hardware options exist for a clean breakdown. Less is more.

In the least, it's a benchmark to try beating. :)
19) Message boards : Generalized Fermat Prime Search : Points: 17 Low vs 17 Mega (Message 120815)
Posted 508 days ago by Profile FishFryProject donor
If your data includes older GFN17 tasks (i.e., b < 8,388,608), then for GPU tasks, GFN 17 Low used to use a more efficient transform (OCL3) than does GFN 17 Mega (OCL2). Same amount of work, but it took less time. Therefore, same amount of credit, but it takes less time.

Both use the same algorithm on the CPU, however. If you were running those on you CPU, you should see almost identical credit/hour rates for both 17 Low and 17 Mega.


Hi All,

Closest thread I could find on my question. I can't determine why the averages for GPU on the higher base number (b>=42,597,774) takes 1/6 of the average time than a lower base. Moreover, the CPU average time is ~1/5 less time, which makes sense.

Only thing I can think of is PrimeGrid crunchers use more powerful GPUs for Mega, which may skew the averages PrimeGrid compiles. If crunchers have a powerful CPU, they may opt to run Low over Mega.

Listed Averages:
Generalized Fermat Prime Search n=17 (GFN-17-Low or Genefer 131072 Low)
b131072+1 (or b217+1)

Deadline: 4 days (up to 30 days)

Recent average CPU time: 11:31:23
Recent average GPU time: 2:01:14
CPU

Generalized Fermat Prime Search n=17, b>=42,597,774 (GFN-17-Mega or Genefer 131072 Mega)
b131072+1 (or b217+1), b>=42,597,774

Deadline: 4 days (up to 30 days)

Recent average CPU time: 14:55:38
Recent average GPU time: 0:23:09


Cheers,
Brendon
20) Message boards : Number crunching : BOINC Manager - Remaining (estimated) Value (Message 120128)
Posted 544 days ago by Profile FishFryProject donor
Hello Crunchers,

I've read many posts surrounding this topic, but none addressed my observation. BOINC Manager shows properties for any downloaded task. One of the property values is "Estimated computation size", which is measured in GFLOPs.

Perhaps BOINC estimates the time it'll take to run GFLOPs rather than periodically recalibrating the progress percentage to the elapsed time. That's basically how we can estimate the duration of a particular task. The number of cores/threads directly impact the amount of GFLOPs dedicated to a task. Thus, we see the estimated time fluctuate erratically.

Since BOINC runs multiple projects, the developers may have justified measuring GFLOPs as consistent for estimating time across all projects, but sometimes inaccurate for PrimeGrid.

Cheers,
Brendon


Next 10 posts
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2020 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 3.02, 2.99, 3.00
Generated 20 Feb 2020 | 11:46:29 UTC