Jump to content
C.M

PrimeGrid

Recommended Posts

zUnJqhg.png

 

Project name: PrimeGrid

Project category: Mathematics

Sponsor: Private, supported by Rackspace

Goals: Search for prime numbers. Primes play a central role in the cryptographic systems which are used for computer security. Through the study of prime numbers it can be shown whether current security schemes are sufficiently secure.

Current whitelist status: Whitelisted!

SSL quality: B https://www.ssllabs.com/ssltest/analyze.html?d=www.primegrid.com&latest

Links:

Official website: https://www.primegrid.com/

Team info: http://www.primegrid.com/team_display.php?teamid=4469

Team user rankings: http://www.primegrid.com/team_members.php?teamid=4469

Boincstats link: http://boincstats.com/en/stats/11/team/detail/4469


^ Smash that upvote button! ;D

Follow me on:

Github   Twitter   Steemit   SoundCloud (Hangouts)

 

Share this post


Link to post
Share on other sites

Primegrid has plenty of GPU work at project: Proth Prime Search Sieve (PPS-Sieve)

 

This project pays very well so if a few GPU resources from other boinc projects are moved here Gridcoin can easily do more than 300M/day and therefore try to reach first spot at Primegrid without losing first spot on the other GPU boinc projects.

Share this post


Link to post
Share on other sites

If you are using a GTX-670 or better you may not be able to feed data to it efficiently using the project default of

 

<avg_ncpus>0.078836</avg_ncpus>
<max_ncpus>0.078836</max_ncpus>

 

I do not know what the minimum cpu requirement is, but changing the cpu% from ".078836" to "1.0" caused my time to complete to change from 2 weeks to hours or less on gennifer prime work units.

 

To fix this put the bottom text "app_config.xml" into your primegrid project directory.  If that does not work then you will have to edit sched_request_www.primegrid.com.xml and do a mass substitution for that .0788

This problem is discussed here

http://www.primegrid.com/forum_thread.php?id=7776

 

Possibly this is a windows 10 scheduling problem and if you are running Linux (or whatever) maybe your scheduler does a better job.

 

app_config.xml follows

 

<app_config>

<app>
<name>genefer</name>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>

<app>
<name>ap26</name>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>

<app>
<name>pps_sr2sieve</name>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>

<app>
<name>genefer15</name>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>

<app>
<name>genefer16</name>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>

<app>
<name>genefer17low</name>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>

<app>
<name>genefer17mega</name>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>

<app>
<name>genefer18</name>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>

<app>
<name>genefer20</name>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>

</app_config>

Share this post


Link to post
Share on other sites

I've pointed my new GRC/BOINC rig at primegrid with it's 2x 1080's.  So far so good...able to crunch 2x WU per GPU @ ~4min average per with 99% GPU usage.  Only issue I had is when I tried to point my spare cores from my ryzen 1800x at [email protected] the GPU WU increased to 16min per and I couldn't figure out why.  Thrashing CPU memory perhaps?  I am only using 80% of my cores with HT enabled so I should have one full physical core dedicated to my GPU's still running primegrid.  I ended up just switching all my other cores back and the GPU performance improved back to 4min average per WU.

Share this post


Link to post
Share on other sites

Are you crunching for grcpool?  I am doing the pool until I get enough coins to solo.   Following that discussion (above) with a PrimeGrid staff member, I found I was unable to download any sub-projects other than pps sieve (cuda).  In addition I can only download no more than 1 WU for every gpu in a system.  If I set <gpu_usage>0.5</gpu_usage>  (or ngpus) on a system with 2 gpus, I can get 2 PrimeGrid WUs but they run only on one gpu with the other gpu empty.  I didnt have these problems when using BAM!  The staff member made it clear he didn't like grcpool's "random" distribution of hosts.  However I read that PrimeGrid had given away 5000 in gridcoin from their own wallet so some of the staff must like them: http://tinyurl.com/yakxzcnk

 

You might do better with gpugrid as primary (set its resource=100) on your 1080 paired system and use einstein with resource=0 and 0.25 for both cpu and gpu as a fallback as gpugrid runs dry occasionally.

 

I cannot put more than a single 1070 in one of my systems as most are core 2 quad and dont have enough cores to feed a full up nvidia system.  Attached to this post is the cc_config from my gtx1070ti and gtx770 system.  The file allow 1 gpugrid to run on the 1070 but also allows 4 einstein to run if gpugrid is out of data.  It also allows 3 milkyway to run on the 770 as that board does not work well with gpugrid or einstein and milkyway uses very little % core which is a perfect match for a Q9650.  BTW, I attempted to buy another 1070 from newegg over new years but my order was canceled by them.  The 1070 was going for $435 and I already had a matching evga hybrid cooler for it.  When I checked a week later the price had more than doubled and anything under 800 was out of stock.  I still have the liquid cooler but no video board for it.

cc_config.7z

Edited by BeemerBiker

Share this post


Link to post
Share on other sites
3 hours ago, BeemerBiker said:

Are you crunching for grcpool?  I am doing the pool until I get enough coins to solo.   Following that discussion (above) with a PrimeGrid staff member, I found I was unable to download any sub-projects other than pps sieve (cuda).  In addition I can only download no more than 1 WU for every gpu in a system.  If I set <gpu_usage>0.5</gpu_usage>  (or ngpus) on a system with 2 gpus, I can get 2 PrimeGrid WUs but they run only on one gpu with the other gpu empty.  I didnt have these problems when using BAM!  The staff member made it clear he didn't like grcpool's "random" distribution of hosts.  However I read that PrimeGrid had given away 5000 in gridcoin from their own wallet so some of the staff must like them: http://tinyurl.com/yakxzcnk

 

You might do better with gpugrid as primary (set its resource=100) on your 1080 paired system and use einstein with resource=0 and 0.25 for both cpu and gpu as a fallback as gpugrid runs dry occasionally.

 

I cannot put more than a single 1070 in one of my systems as most are core 2 quad and dont have enough cores to feed a full up nvidia system.  Attached to this post is the cc_config from my gtx1070ti and gtx770 system.  The file allow 1 gpugrid to run on the 1070 but also allows 4 einstein to run if gpugrid is out of data.  It also allows 3 milkyway to run on the 770 as that board does not work well with gpugrid or einstein and milkyway uses very little % core which is a perfect match for a Q9650.  BTW, I attempted to buy another 1070 from newegg over new years but my order was canceled by them.  The 1070 was going for $435 and I already had a matching evga hybrid cooler for it.  When I checked a week later the price had more than doubled and anything under 800 was out of stock.  I still have the liquid cooler but no video board for it.

cc_config.7z

Yes I'm participating in grcpool (pool 2 to be exact).  You should be able to download more than 1 WU per GPU.  Have you tried adding a cc_config.xml  in the main BOINC folder with...

<cc_config>

<options>

<use_all_gpus>1</use_all_gpus>

</options>

</cc_config>

 

It really shouldn't need that option set in the cc_config.xml any more as the newer boinc client will detect multiple GPU's and assign it least 1 WU to each until you modify the gpu_usage like you have above, but give it a try anyways if you haven't yet.  I know the pains of buying cards...I got INSANELY lucky two weeks ago when I decided to just randomly build a rig from scratch for BOINCing/grc mining.  I managed to snag an EVGA 1080 SC FTW and an EVGA 1080 SC Hybrid for $550 and $600 off newegg.  Very next day they were listings for like $1000 and $1600.  Just stupid.  I had no idea the GPU market was that bad right now.  Those 1070's are crazy expensive because of their low TDP too.  Electricity ain't cheap near me so I need to be semi-cost conscious though I can afford to burn $100-200 a month to run a few rigs 24/7.

 

I'll definitely take a look at GPUgrid.  I really liked the science around that project.  I just happened to pick Primegrid to get my rig up and going and it's built up about 700K in RAC in just a few days so far.  Thanks for the config settings I'll definitely take a look and see if I can integrate it going forward.  I'm pretty new so I haven't dealt with running out of WU yet, but it's always good to be prepared.

Share this post


Link to post
Share on other sites

Except for PrimeGrid, all my other projects easily download multiple WU for my GPUs and I maintain a queue of about 8 hours of work for all other projects.  Unaccountably, after I had a discussion with a PrimeGrid staff member, I could no longer download any tasks other than cudaPPSsieve and only a single one at a time, one for each GPU in a system.  I am not sure if this is something they implemented or if I have screwed up something on my end.  I had complained to them that when I attempted to exclude some of the problem "genefer" sub-projects  that the sub-project (like genefer) was being downloaded and then deleted.  I said that was a waste of bandwidth to allow BOINC to download a work unit and then immediately delete it.  

 

I did some checking against my "wingman" on a PrimeGrid genefer 21 task that I had to abort as it had been running for several days and had several weeks left.  What I noticed was that 5 other systems "wingmen" had also either aborted the same task or had generated an error.  One of the systems had the newer titan and I suspect that if a titan could not finish the job in 2 weeks that my really slow gtx670 could run for several months on the same low credit "genefer" tasks.  Typically genefer tasks take days to weeks to complete while those pps_sr2seive take only minutes for the same credit amount on my gtx670 systems.

 

I no longer download any genefer type PrimeGrid tasks but yet I do not have any exclusions to prevent their download.  I find this suspicious.

Share this post


Link to post
Share on other sites
On 1/20/2018 at 3:19 PM, Zedy44 said:

 Only issue I had is when I tried to point my spare cores from my ryzen 1800x at [email protected] the GPU WU increased to 16min per and I couldn't figure out why.  Thrashing CPU memory perhaps?  

 

Yes:  Ryzen has 512kb of second level cache for each core.  Compare that to older core 2 quad  

 

For comparison, 2 of my systems with older cpus:

MSI Platinium & X5470 has only 4 cores but 6mb of level 2 cache for a pair of cores and  I get about %13 cpu utilization for each of 6 tasks running on a total of two GPUs

Dell Z400 & X5680 has 6 cores 12 threads but only 256k level 2 cache for a single core.  I get %96 cpu utilization for each of only 4 tasks running on a total of two GPUS

 

I suspect my MSI Platinium is sharing 6 mb of second level cache on each GPU among the 3 Einstein while the Dell is thrashing.

Edited by BeemerBiker
second level was 2*6mb

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Recently Browsing   0 members

    No registered users viewing this page.

Cryptocurrenytalk Logo

 

News, information, and discussions about cryptocurrencies, blockchains, technology, and events. Blockchaintalk is your source for advice on what to mine, technical details, new launch announcements, and advice from trusted members of the community. Cryptocurrencytalk is your source for everything crypto. We love discussing the world of cryptocurrencies.

 

   
×

Important Information

By using CRYPTOCURRENCYTALK.COM, you agree to our Terms of Use.