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
1) Message boards : Number crunching : Wallis is Born Challenge (Message 100987)
Posted 1099 days ago by Ross*
Hi
Can anyone tell why AMD cards are so hopeless at this AP27?
Ross*
2) Message boards : Sicituradastra. message board (Message 85456)
Posted 1662 days ago by Ross*
Hi
great to see the team talk flowing.
the recent PG challenge was interesting to say the least.
I ran 23 GPUs . had discarded some 7970s and 1 780ti because of errors before the challenge. seems I have only lost/errored /invalided 3 WR tasks so far.
the GPUs were 8 280x of which 7 did 3 MR each, 6 780ti all doing 3 WR , 2 780s doing 2 WR and lots of smalls plus 2 Titan Zs doing 2 each and finally 2 980s doing 2 WR just. [they are useless] plus 3 7970s.

All ran without stop for the whole 8 days

Moon , you seem to be running to stand still. Take a breath and see the big picture and everything will come right. Good Luck

I will be into Poem challenge , but someone else can lead the charge. Will be very busy with other things in my life in the next couple of weeks.

a big thank you to all members for the great win on the latest PG challenge

Cheers
Ross*

3) Message boards : Project Staging Area : Wall-Sun-Sun (Message 84550)
Posted 1693 days ago by Ross*
Hi
thanks for the info
have 2 kingpins running in one box 20 secs per WU running @ 13m p/s
Cheers
Ross*
4) Message boards : Project Staging Area : 2015 PRPNet April Challenge (Message 84501)
Posted 1695 days ago by Ross*
Hi
Thanks for the info Roger.
What settings are used to get 2 GPUs to work?
what is good performance. 280x seem to be running at 7m p/s and ati 290x at 12 m p/s. on one task.
Is there any bunkering on this project?
Cheers
Ross*
5) Message boards : Project Staging Area : 2015 PRPNet April Challenge (Message 84478)
Posted 1696 days ago by Ross*
Hi Rodger
A couple of weeks ago I try out the Wall of Sun Project.
While I found the project ran well there seem to be a lack of use of the GPU.
After a little fiddling I copied the wwww file into the prpclient file and what a improvement
This is like the tweak for the WIEFERICH challenge.
Comments or other settings we need to know please.
Cheers
Ross*
6) Message boards : Project Staging Area : 2015 PRPNet March Challenge (Message 83775)
Posted 1721 days ago by Ross*
Bit of a spanner in my Bunkering effort. The server has a limit of 5 work units. That means I would complete them before the Challenge starts. Even if I hit Ctrl-C my completed work units get reported to the server before the challenge begins. After 1 day of crunching each core has done 3.5 work units.

To continue Bunkering I just close the Console windows (don't hit Ctrl-C). The completed work units are all check pointed. Once the challenge starts I just change to startoption=3 and restart the prpclient.exe and it should report the completed work units to the server. Next I still have more time so I copy prpclient-1 directory to a new number, say prpclient-7. Delete the temporary files and set instanceid=7 in the prpclient.ini and start prpclient.exe and repeat for each core. Now I've got the next set of clients continuing the Bunkering effort. I should have 15 prpclient's with completed work units by the time the Challenge starts.


Hi
thanks for that info, but I will be only running half of my boxes with a simple 5 bunker setup. The rest will start 1hr before the start.

Just to make sure that all those saved ? tasks sitting on the server get into the challenge, is the following right.

edit the master_prpclient and change the start and end options and update-prpclient-ini
This should upload all tasks? all the saved tasks in the pending will added to the challenge.

Cheers
Ross*
7) Message boards : Project Staging Area : 2015 PRPNet March Challenge (Message 83622)
Posted 1726 days ago by Ross*

Update
I have been running 6 cores of a 8 core 5960 for past 24 hrs plus 1 GPU 780ti
after a few hiccups they settled down. a bit of a conflict between 2 prpclient-1 s.
there is a big difference between an old 980x and a 5960 in processing power.

the 5960 users FMA3 and the average time per core is 102 mins, I had to decrease the OC to 3700 to keep the heat below 70c.

the GPU takes less averaging 80 mins , so not 4 times quicker

total tasks was 14 *6 cores = 84 for the cpu and 18 for the GPU = 102 per 24 hrs

so using the GPU adds about 20% to the output.

Next experiment will be trying to run 8 tasks on 8 cores, which would give 114 tasks per 24 hrs if the 102 mins per core holds up. ``

looking at the stats page there seems a huge variation in tasks done and points awarded. why is this? My current points per tasks is 22037 to 22100.
I can not control what task I get, so what is the difference?

also why do some tasks go into iteration ? some don't.

Cheers
Ross*






Update
8 cores on a 5960 will run 102 mins per task = 14 tasks * 8 =112 per day
I have achieved 15 with a higher OC.

On a 5930 running @ 4100mhz it comes down to 92 mins av on 6 cores
that's over 15 per day *6 =90 and I could add 1 GPU of 18 per day

yes its not that efficient to use GPUs but if your CPU takes longer that 3 hrs it may be a option.

Yes I see the comments about using GPUs [and only 1 will work per box] but good GPUs do DP like 780ti and Titans. all my GPU tasks have been accepted so far.
On 3930k using AVX they take 130 mins, 11 a day per core , 66 per box per day.
So everyone to there own way of doing the work.
To me the 59 box out perform any combination of GPUs, but the slow CPUs like my old 980/970 they need a good GPU to boost the output.
Cheers
and happy crunching
Ross*
8) Message boards : Project Staging Area : 2015 PRPNet March Challenge (Message 83547)
Posted 1729 days ago by Ross*

Update
I have been running 6 cores of a 8 core 5960 for past 24 hrs plus 1 GPU 780ti
after a few hiccups they settled down. a bit of a conflict between 2 prpclient-1 s.
there is a big difference between an old 980x and a 5960 in processing power.

the 5960 users FMA3 and the average time per core is 102 mins, I had to decrease the OC to 3700 to keep the heat below 70c.

the GPU takes less averaging 80 mins , so not 4 times quicker

total tasks was 14 *6 cores = 84 for the cpu and 18 for the GPU = 102 per 24 hrs

so using the GPU adds about 20% to the output.

Next experiment will be trying to run 8 tasks on 8 cores, which would give 114 tasks per 24 hrs if the 102 mins per core holds up. ``

looking at the stats page there seems a huge variation in tasks done and points awarded. why is this? My current points per tasks is 22037 to 22100.
I can not control what task I get, so what is the difference?

also why do some tasks go into iteration ? some don't.

Cheers
Ross*




9) Message boards : Project Staging Area : 2015 PRPNet March Challenge (Message 83535)
Posted 1730 days ago by Ross*
Hi
yes llrCUDA does work, and is much quicker , how much exactly I have not worked out yet. At this stage just experimenting.

I was hoping to get more definite info from my first post, but not yet.

I have tried 2 tasks on one GPU but as the GPU is running over 90% load it did not seem to be any better.
Also ran a Titan Z, and managed to do 5 tasks in 12 hrs. one GPU runs 90% load and the other at 60% . will have to play with the setting in the next week.

Next try will running 6 cores and leave 2 cores for the 1 gpu on a 5960 to see what happens.

after that perhaps with HT

The other query is bunkering, It appear, may be wrong on the January challenge that some jumped out to a lead right from the start.

If we can get all options or settings out for everyone, the more tasks get done.

Cheers
Ross*





10) Message boards : Project Staging Area : 2015 PRPNet March Challenge (Message 83512)
Posted 1731 days ago by Ross*
Hi
I am trying out the 27 prime search project before the challenge starts.
It would appear this is a CPU project using llr64 program. may be wrong, however seems to be working.
setup 2 tasks on a old 980x 6 core but the tasks take forever.

I then used a box with a 780ti and ran the llrCUDA app and that works as well but 4 times faster , after 12 hrs the gpu has done 8 tasks and the cpu 2

Are they both doing the same tasks but by different ways?

The 780ti is running quite and cool at 1200mhz. will it do 2,3 at a time ?

Are there any more settings that need to be amended to achieve better performance?
Cheers
Ross*


Next 10 posts
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2019 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 2.36, 2.71, 2.45
Generated 21 Nov 2019 | 15:28:47 UTC