Please visit donation page to help the project cover running costs for this month
21) Message boards : Problems and Help : No WU for Sophie Germain Prime Search (LLR) via CUDA (Message 40483)
Posted 3729 days ago by uBronan
Indeed do not rush this, better get it working well before used by bigger group of users
Also must make sure people understand that it needs a 64 bit Operating System to run and not that when they have a 64 bit cpu means they can run it ;)
22) Message boards : Problems and Help : Ubuntu 10.10 64bit AMD Phenom II 6 core trouble with some projects (Message 40415)
Posted 3731 days ago by uBronan
My experience is that running unganged is in most cases faster
So in my view its not the setting of how the memory runs, your problem seems to be timing related
I have had similar problems with memory, thats why is is often wise to see what memory is supported with your mainbord
Funny enough this not allways gives a guarantee that timings problems won't appear.
This far all 3 of my home amd machines which i had running where more stable with only 2 modules or ran better with smaller memory modules

For instance i had 4x2 Gb =8 Gb DDR2 geil memory in my last X3-720 system, but whatever i did gave weird problems
Including booting slow, programs crash, weird results at boinc projects

So i took them out and took my intel systems memory 2 corsair 1066 modules 2x2 gb. This machine simply refused to boot in 1066 mode with them.
When i had those in my amd it suddenly ran smooth, and totally error free.

Now i put the 8 Gb in my intel system and poof solved the other issues as well.

If i wasn't running boinc i probably would have never had so many issues ;)
But just because the boinc projects are so demanding this issues come in the spot light
23) Message boards : Project Staging Area : llrCUDA testing (Message 40391)
Posted 3732 days ago by uBronan
Well i just noticed the started only llr.exe even though i have put the correct settings to the ini.


After i disabled the llr.exe in the ini it completely disregarded previous done work and started all over from scratch.
From what i can see it seems to go faster though then previous work done on the llr.exe application.
but time will tell the truth :D

Anyway when this unit is finished i will let it run on the other ports mentioned
Just am wondering why it did not want to start with the cuda version, and why previous work got deleted by the app

Yes it looks pretty promising this far, but maybe some more testing is needed before it gets widespread over boinc.
24) Message boards : Project Staging Area : llrCUDA testing (Message 40389)
Posted 3732 days ago by uBronan
I had installed the app and fixed the ini but somehow on 121 units constant started with the normal llr.exe
Does this mean that those units must be done by the llr.exe ?

Anyway i put // for the llr.exe in the ini and started again
Now the llrCUDA finally kicks in what you guys think after some inital slow start.
I think when everything runs smooth a port to boinc would be awesome
And i think this especially in the Seventeen Or Bust would make a huge impact
I saw people posting huge calculations getting done with awesome speeds
As i am watching it run i see the gpu load about 94%
But what is impressive is the load on the memory controller its almost constant at 55 to 58%
Before one ask you can see gpu mc load in a tool like aida 64 which has added this future in its recent product.
Sadly the test versions are limited so i am not sure if the test version will show this
25) Message boards : Project Staging Area : PSA credits to BOINC conversions (Message 40255)
Posted 3737 days ago by uBronan
I got some usefull information from your responces
At certain point while i was expecting to see the cuda client to be working with the cuda client noticed that it was switching between the different versions.
I did not see the genefercuda work but mostly saw the genefx64 and later the genefer80 application in process explorer.
Ofcourse i am not constant watching to what it does just noticed it when i had to stop a proces or was doing some test.
Should i change the stop option to delete all previous work, since it seems to finish well but get rejected by the server.

On the conversion to boinc i have an idea of what to expect now, that it takes a while i do not mind.
Today i restarted the client on my q6600 it started the unit it previously downloaded and did finish it completely

But now i followed the log to see something i did not notice before and could have been a reason why i seemed to have a low output.

What would be the reason for the job not being accepted as valid.
Since i am also alpha/beta tester in many different projects including os/software/game/boinc which sometimes causes my system to reboot or crash.
Most of the time when i going to test i stop the current running job by using the good old ctrl+c combination. In almost all cases i wait till it finished the job and shows downloading a new job.
But ofcourse sometimes when the system freezes a reboot is inevitable.
26) Message boards : Project Staging Area : PSA credits to BOINC conversions (Message 40233)
Posted 3737 days ago by uBronan
Well i just do not want to spend much power on this project.

Even with 9 cores on 3 machines and 2 cuda cards i saw such low rewards after running 28 hours, that the choice was not hard to switch back to only 1 core each.
To give an example i make easily 600,000 cobblestones on boinc in the same time, and probably even more if i was not trying to get some cpu projects to nr.1 like leiden.
After more then a week i received in boinc psa 13,000 cobblestones this includes the 28 hours running on all cores and gpu's and this without running boinc. And the rest of the time between 2 and 3 cores full on prpnet client.

Another reason is also because there are a lot of other projects where i am still not country number 1 or on signup date ;)

I do however set my goal to get some more badges, so when i reach them i probably need more power to stay on top on my other projects.

Ofcourse i have not much knowledge about this search for primes, i know many projects are trying to find the biggest including prime95 from mersenne which i ran some years ago (and ran some hours last month but i fear the $50,000 is not gonna be payed anytime soon :) )
Btw on prime95 i also never found a first prime in about a year or so.
At present its also not possible to run some prpnet projects cause they simply seem to be out of work because of some challenge.

But no i have several issues if i run the prpnet client along with normal boinc, especially when together with for instance einstein or some other gpu projects which also need a free core to run well
That makes it almost a day time job to toggle between projects and the ones who get into problems when something else runs along.
Some projects slow down almost to double/triple the normal time needed or in worst case scenario crash or stall.
I try to avoid running those projects together with those which do not like each other much.
Sadly that makes you have to choose what you should run instead of run them all together and let boinc choose.
27) Message boards : Project Staging Area : PSA credits to BOINC conversions (Message 40226)
Posted 3738 days ago by uBronan
well with my luck, i think i have more change on finding a 5 euro bill on the street then to find a prime.
I see people having found prime in a few units but for me it probably won't happen in the next months or years and if i do find one it probably will be just to verify another.
Please do explain how to get the best psa credits because i have no idea what i am doing :D
Ofcourse i am somewhat limited since i try to get other badges from boinc cpu projects as well
28) Message boards : Number crunching : Computation errors on ATI (Message 40205)
Posted 3740 days ago by uBronan
About ati 3xxx/4xxx hardware supported

What i think was meant is that while opencl is developing at some point you get new programs who does not support the older models anymore like cuda
I think both sides are at version 4.x now

Gpugrid for instance needs almost allways the latest and fastest cards, simply because the program makers did make their application for cuda 3.2 or 4.0
Since older cards mainly support cuda 1.x or 2.x they can not finish the jobs in time so they are simply no longer usable on this project.

Any hardware is made for version x.x of these so its up to the application builders if they still support the older versions or not.
As with cpu projects some enhanced instruction sets make a huge impact for calculations others hardly do anything, so in cpu projects you see some difference also.
Look at raistmer's seti enhanced special builds for almost every instruction set from SSE2 up to AVX and more coming each time new processor designs get introduced making sometimes huge speed increases possible.
This will be seen on the upcoming cal/cuda/opencl development as well.
I expect in time we will see the gpu getting more and more very good and advanced calculation devices, which is very good for d.c.

Did you say 3200 cores? That is unfeasible. They must be a lot smaller than CPU cores?!?

they are completely different then cpu cores, gpu cores are simple cores which can do limited tasks, but since they have so many and all can work on this 1 or 2 tasks to complete. Thats why they will do it extremely fast.
cpu is with complex instruction sets and can do many different tasks and is able todo anything you want.
Gpu can do only a few different task since they have simple instruction sets
Also though some projects allows more instances to be running at the same time untill now i think the limit is still maximum 2 real jobs(streams) at one time.
The other difference is that a cpu can stop tasks and start them at any time and does in real time switch between many tasks
Now the gpu is limited in this as well in most cases when it gets started on a job needs to finish it no matter what happens.
In most cases if it get stopped it will have to do some or all of the work done over
29) Message boards : Problems and Help : how to reduce GPU usage to 50% (Message 40163)
Posted 3741 days ago by uBronan
No it is actually kinda easy
Especially on newer models
Get yourself a copy of msi afterburner at
{ before one ask it works on every brand / model ati/nvidia }

When you installed it set it to start with windows
Then you see some bars which to set memory and core speed
To get our cards making less heat no extremes have to be done
Mine runs normally at 850 and memory at 1200
To get it to run withing proper cool settings i just set the core to 700 and i lowered the memory to 750 which gives a huge drop of temps
Remind yourself the output drops as well
But to honest many boinc projects do not need high memory speeds, this is also where i see the biggest drop in temperature (collatz is a project which needs high memory speeds)
Now do not set it below a certain minimum or your system will crash
The lowest at which my card still runs ok is 450 on core and 400 at memory ( effective this is 1600 mhz since its ddr5) some set it at 300 but for me the desktop becomes sluggish which i do not like
Each time when you change something you need to apply
Now if you found the speeds you like best temperature wise you can save it as a default with the numbers down under in my case i have 3 settings as preset
1. performance where i set it to actually run faster 890/1260
2. high core low memory 890/400
3. low temp running 450/400

If you need any help just send me a message
As a sidenote do not mess with voltage and such if you do not know what your doing with this tool
30) Message boards : Project Staging Area : PSA credits to BOINC conversions (Message 40133)
Posted 3742 days ago by uBronan
Credits in a day ..... lol yeah right.
I am running it for about a week or so and nothing showed up on the free-dc psa primegrid stats
Not even my name shows up.
I even ran it some 20 hours or so on my cuda cards but took them off quickly cause 10k for 2 cuda cards for more then a day is kinda .... nothing
I think one of the cards did not do any work even though i had started the 2 cards start file.

If i would calculate the 18/1 ratio i would have made in 5 or 6 days about 24k on credits... i make more in less then 30 minutes with the cuda devices on other projects
To compare something i get after about 22 hours run time psa/boinc 10,000/555 credits on a gts450 card with 2 core cpu active as well.

So i turned them quickly back to better giving projects, then i tried running 3 machines with in total 9 cores, that did speed up a little but after some hours running i saw even this does not really make it worth while.

So i turned back to now 2 cores running on seperate machines but these psa points seems slow as hell and no points showing up in a for me understandable stats page makes it not my favorite

In boinc most good projects make a calculation for credits based on processing power done in the projects.
Ofcourse there are many issues with credit awarding since we have lots if people who are for a fair system based on the M/G flops spend on the projects, and other seem favor more to put it into some weird low credits system.
These people tend to say well its nice you use a more powerfull tool but that should not give more points .....
At some points people even was saying gpu projects do the same work so they should get the same points.
They simply disregard that a 100 till 350 watt device is pumping calculations like mad through it, and pumps up electra bills
Hell some persons on boinc spend enough electricity on this to feed an entire village for a full day on their electra usage.
So yes its hard to say what is a fair credit system but for me collatz is the one which seems to have the most fair rewarding system.

Next 10 posts
[Return to PrimeGrid main page]
Copyright © 2005 - 2021 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 2.30, 1.69, 1.25
Generated 6 Dec 2021 | 12:24:40 UTC