Jump to content

Recommended Posts

94Y7uYx.png

 

Project name: [email protected]

Project category: Physical Science

Sponsor: Rensselaer Polytechnic Institute

Goals: Creation of a 3d map of the milkyway galaxy using data gathered by the Sloan Digital Sky Survey. This project enables research in both astroinformatics and computer science.

Current whitelist status: Whitelisted!

SSL quality: A https://www.ssllabs.com/ssltest/analyze.html?d=milkyway.cs.rpi.edu

Links:

Official website: https://milkyway.cs.rpi.edu/milkyway/

Team info: http://milkyway.cs.rpi.edu/milkyway/team_display.php?teamid=6566

Team user rankings: http://milkyway.cs.rpi.edu/milkyway/team_members.php?teamid=6566

Boincstats link: http://boincstats.com/en/stats/61/team/detail/6566


^ Smash that upvote button! ;D

Follow me on:

Github   Twitter   Steemit   SoundCloud (Hangouts)

 

Share this post


Link to post
Share on other sites

[email protected] is an attractive and well-established BOINC project. However, if you plan to run it on your GPU, it's very important to remember that it requires double-precision compute capability (FP64) and consumer graphic cards today normally have very low FP64 perfomance, since FP32 (single-precision) is more than enough for gaming. FP64 is mostly relegated to the professional market and such graphic cards are very expensive.

 

You can check FP32/FP64 GFLOPS for your graphic card here:

http://www.geeks3d.com/20140305/amd-radeon-and-nvidia-geforce-fp32-fp64-gflops-table-computing

(clear and instructive, but newer cards are not listed)

 

https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units

(a bit harder to navigate, but ALL Nvidia GPUs are listed)

 

https://en.wikipedia.org/wiki/List_of_AMD_graphics_processing_units

(same remark, but for AMD/ATI)

Edited by Vortac

Share this post


Link to post
Share on other sites

HD7970 needs 9 seconds per task :)

 

Of course, most of the time is then spent contacting server, loading a new workunit, then contacting server again and uploading finished one. If you have such a GPU (which is powerful in FP64), the trick is to run multiple Milkway workunits per GPU. Four workunits per GPU seems to work best, at least in my case. When one workunit is getting downloaded or uploaded, three others are being processed. That way, you have a continuous flow of workunits and idle GPU time is close to zero.

 

To run multiple workunits per GPU, you need to create app_config.xml in Milkyway's folder. BOINC Manager will automatically detect it and load it upon start. It should look like this:

 

<app_config>

<app>

<name>milkyway</name>

<gpu_versions>

<gpu_usage>0.25</gpu_usage>

<cpu_usage>0.25</cpu_usage>

</gpu_versions>

</app>

</app_config>

Edited by Vortac

Share this post


Link to post
Share on other sites

It's probably due to recent network difficulties, both with Gridcoin and MilkyWay. My MilkyWay magnitude hasn't returned until today, after 2 or 3 days of waiting. I guess it will soon return to the whitelist leaderboard too.

Share this post


Link to post
Share on other sites

It's probably due to recent network difficulties, both with Gridcoin and MilkyWay. My MilkyWay magnitude hasn't returned until today, after 2 or 3 days of waiting. I guess it will soon return to the whitelist leaderboard too.

Is it still in the whitelist, yeah?


^ Smash that upvote button! ;D

Follow me on:

Github   Twitter   Steemit   SoundCloud (Hangouts)

 

Share this post


Link to post
Share on other sites
Is it still in the whitelist, yeah?

Yes, I am getting magnitude from MilkyWay now and it's shown in NN. I guess that wouldn't be possible if MilkyWay somehow dropped out of the whitelist. Gridresearchcorp is most likely still showing old data, from the incomplete superblock.

Edited by Vortac

Share this post


Link to post
Share on other sites

Hello!

 

i started [email protected] 3 days ago with pool and try to calculate grc earnings, and i run only gpu tasks with old HD5850 (400 GFlops DP). And want to calculate, is it worth to upgrade to GTX Titan GK110 (1500 GFlops DP) .

i found that tasks average time to complete on my system is 137 seconds, that equal or better than averege time per task for some top hosts with HD7970:

me  http://milkyway.cs.rpi.edu/milkyway/results.php?hostid=742847&offset=0&show_names=0&state=4&appid= , 137 sec per task

top host #1 https://milkyway.cs.rpi.edu/milkyway/results.php?hostid=591552&offset=0&show_names=0&state=4&appid=3   , 130 sec per task

top host #2 https://milkyway.cs.rpi.edu/milkyway/results.php?hostid=736335&offset=0&show_names=0&state=4&appid=   , 200 sec per task

 

so, first, is it possible to calculate ROI for device based on it`s fp64 performance running [email protected] 24/7

and

explane, how can it be, that old weak card shows equal or better performance than more modern cards?

 

Edited by boristsybin

Share this post


Link to post
Share on other sites

1. Upgrading to Titan is not worth it. You can upgrade to HD7970 at a much smaller cost for equal performance or nearly so.

2. Those hosts you are pointing out, they are probably running three or four tasks in parallel. So their average runtimes are longer,  but overall output is higher of course.

3. Modern consumer cards all have very weak FP64 performance, only professional cards get such niceties today. That's why 7970s or even 5870s and 5850s are better choice for Milkyway than a new shiny GTX 1080.

Edited by Vortac

Share this post


Link to post
Share on other sites

Those top hosts definetly run 24/7, i look deep into their task history. 

 

titan has 1500 GFlops dual precision peak performance, and costs about 260$

hd7970/r9280x has 1000 GFlops and costs about 170$

but the whole systems price (case+cpu+mb+ram+PSU+ssd+GPUs) differs just for 5%, and accounting for power consumption reduces that difference to zero. 

 

so, what about earnings? should i wait another week to get the right earning numbers or good simple online calculator exists somwhere?

Edited by boristsybin

Share this post


Link to post
Share on other sites

1. Top hosts run 24/7, no doubt. In addition to that, they probably run three or four tasks in parallel on every GPU. So, every 130 seconds, they complete at least THREE tasks. 7970 can complete a single task in about 30 secs.

2. True, Titan has better FP64 performance, at least on paper. However, Titans are not visible on the top hosts list. My guess is that Nvidia's implementation of OpenCL is less optimized and that 7970s are yielding better performance in the end, despite lower FP64 specs.

3. There is no simple online calculator, because BOINC computations are much more complex than the usual crypto-mining. Milkyway is a particularly good example, with older cards yielding better performance, something unheard of in the world of Proof-of-Work hashing. From my experience, a single 7970 can achieve RAC of 600k. That's good enough for 230 magnitude or about 40 GRC per day.

 

 

Edited by Vortac

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.