Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Generalized Fermat Prime Search :
High quorum
| Author |
Message |
|---|
 Dave 
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,430,963,941
RAC: 4,051,704
                           
|
|
Hi all.
Just topping up a few more CUDA while SAH is down & FYI just downloaded a _7:
http://www.primegrid.com/workunit.php?wuid=252950130
Last time I downloaded a high quorum unit like that my GTX580 worked where everyone else failed, so I'd like to think I can deliver on this occasion also.
All the best
Dave | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,430,963,941
RAC: 4,051,704
|
|
Come on Mr 202799... Done my bit. | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 476,993,571
RAC: 281,271
                          
|
|
Thanks for the heads up on this. I fear what you're seeing there is going to be a growing trend now that Nvidia seems to consider 295.xx as the current released driver. Their automatic driver install program is now asking to install 295. 295 doesn't work with CUDA if you have your computer configured to turn off the monitors when they're not in use -- which is the way most people configure them.
Read this post about the problem.
Now I know why I've been seeing so many of those errors lately. :(
____________
My lucky number is 75898524288+1 | |
|
|
|
|
Guess it's a good thing I never setup my monitors to be turned off. Never did like that idea of a pc controlling pieces of the system but I suppose to save on electricity it's what some people do. | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,430,963,941
RAC: 4,051,704
|
|
Thanks Micheal, glad to be off assistance on this matter.
I have full control over what my monitor does - via my hand & the power button ;). + I don't currently want to unduely compromise my record-50-day uptime for that machine.
I tried 295.51 due to someone saying they were having trouble & they wanted to know if it was still the same with an HDMI connection. As it happens I use an HDMI due to my telly/DVR using its composite option, so was in an ideal place to innovate.
SAH is downloading more units now so my participation in this may reduce over the coming days, but it remains the most interesting backup project.
Anything else you want me to try give me a shout. I will always be able to make a core or 3 available as I like to micromanage (hence kicking the _7 into work as soon as I saw it).
Regards
Dave. | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,430,963,941
RAC: 4,051,704
|
|
Hi all.
While SETI down it's too cold in the den, so got a bit of PG to keep things busy this morning. I assume n=22 given the size, which will keep things busy for a while. But is it normal for them to go round the houses 16 times with qora of_16?!??! never seen anything like it. | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 476,993,571
RAC: 281,271
|
Hi all.
While SETI down it's too cold in the den, so got a bit of PG to keep things busy this morning. I assume n=22 given the size, which will keep things busy for a while. But is it normal for them to go round the houses 16 times with qora of_16?!??! never seen anything like it.
Unfortunately, the answer is yes right now. There's two BIG problems at the moment, both causing a lot of "compute errors". The first is the 295 driver under Windows, (which is responsible for most of the errors you see), and the second is overclocking.
The driver problem will go away as people either roll back the driver, disable the power saving feature that turns the monitor off, or Nvidia releases a fix.
Because this problem affects ALL CUDA programs, you will see this right now on any BOINC project with CUDA applications. It's especially noticeable on projects that run ONLY CUDA applications such as the GFN World Record project. Off the top of my head I think GPUGRID is the only other CUDA project that doesn't have a CPU or ATI application. I suspect they're seeing similar failure rates because of the driver problem.
The overclocking problem is a user education task more than anything else.
Also, with ANY long WU you tend to see lots of failures. SoB also gets a lot of WUs with many, many errors. The difference here is that the driver problem is causing the WUs to fail almost instantly, so the errors are accumulating in a matter of hours or days instead of weeks or months. It's a very efficient bug. :)
____________
My lucky number is 75898524288+1 | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,430,963,941
RAC: 4,051,704
|
|
I am aware of the graphics bugs as have seen it discussed on other threads, however I have no GPU overclocking & the screen is not set to power down. I have 295.51 with BOINC also beta @ 7.0.18. but the machine is 99.9% left to its own devices, showing extensive screen lag upon redraw. I am conscious that even maximising/minimising the client could have a calculational effect! 18% after 14 hours, so hopefully only 3 days per GPU (GTX580s :)). | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 476,993,571
RAC: 281,271
|
I am aware of the graphics bugs as have seen it discussed on other threads, however I have no GPU overclocking & the screen is not set to power down. I have 295.51 with BOINC also beta @ 7.0.18. but the machine is 99.9% left to its own devices, showing extensive screen lag upon redraw. I am conscious that even maximising/minimising the client could have a calculational effect! 18% after 14 hours, so hopefully only 3 days per GPU (GTX580s :)).
Hmmmmm.
We originally thought screen lag was a significant problem. On further investigation, everyone who came forward with a lag problem was able to solve it.
Here's what I know about screen lag. Hopefully this will be useful. This applies only to Windows.
1) GeneferCUDA is capable of running, full speed, without any noticeable screen lag.
2) GeneferCUDA is also capable of creating incredibly annoying screen lag. The difference between the two is what else is running on the computer at the same time.
3) Everyone I know of has been able to configure their computer so that the lag problem goes away.
There are ways that some programs -- including several popular Microsoft programs -- interact with the video system that do not play nicely with CUDA. If none of those programs are running, you'll get no screen lag. If at least one is running, the lag will probably be terrible.
Programs that I'm aware of that cause this problem are:
* Windows UAC pop up window. Yes, that window that dims the screen and asks for the admin password is one of the programs that causes lag. It's a real PITA.
* Microsoft Live Mail.
* If memory serves, Internet Explorer also has this problem.
A bunch of other Microsoft tools also cause the problem, but I don't remember them all. They must have a toolkit or something that's interacting with Aero in a way that CUDA just doesn't get along with.
There are various drivers and tools and other stuff running on everyone's computers in the background, and something may be interacting with GeneferCUDA in such a way as to cause the lag.
One person reinstalled Windows and the lag was completely gone.
Besides the those programs, I've found that running *small* video windows is ok, but running full screen video causes lag (and very choppy video). But that's not completely unexpected or unreasonable. Why an email program should have that kind of interaction with the video subsystem is beyond me, but it does.
Hopefully this is helpful and you can discover what's interacting with GeneferCUDA on your computer to cause the lag.
____________
My lucky number is 75898524288+1 | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,430,963,941
RAC: 4,051,704
|
|
Hi
The lag is merely an observation - it is not a problem as it is purely a crunching machine, not my daily driver. Everything is closed or minimised. | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,430,963,941
RAC: 4,051,704
|
|
Both _16s completed & validated:
http://www.primegrid.com/results.php?hostid=242806&offset=0&show_names=1&state=0&appid= | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 476,993,571
RAC: 281,271
|
|
Grats!
____________
My lucky number is 75898524288+1 | |
|
|
|
Both _16s completed & validated:
http://www.primegrid.com/results.php?hostid=242806&offset=0&show_names=1&state=0&appid=
Very interesting indeed..
I'm still a bit confused with the point's - scoring system:
Laufzeit(sek) // CPU Zeit(sek) // Punkte // Anwendung
206,618.46 // 18,857.70 // 204,790.61 // Genefer (World Record) v1.06 (cuda32_13)
299,534.75 // 266,151.73 // 4,000.00 // Woodall (LLR) v6.10
416,694.52 // 361,245.76 // 6,000.00 // Prime Sierpinski Problem (LLR) v6.10
234,255.52 // 7,261.02 // 230,599.06 // Genefer (World Record) v1.06 (cuda32_13)
But anyhow .. congrats to your efforts !
Best regards .. parabol
| |
|
|
|
Both _16s completed & validated:
http://www.primegrid.com/results.php?hostid=242806&offset=0&show_names=1&state=0&appid=
Very interesting indeed..
I'm still a bit confused with the point's - scoring system:
Laufzeit(sek) // CPU Zeit(sek) // Punkte // Anwendung
206,618.46 // 18,857.70 // 204,790.61 // Genefer (World Record) v1.06 (cuda32_13)
299,534.75 // 266,151.73 // 4,000.00 // Woodall (LLR) v6.10
416,694.52 // 361,245.76 // 6,000.00 // Prime Sierpinski Problem (LLR) v6.10
234,255.52 // 7,261.02 // 230,599.06 // Genefer (World Record) v1.06 (cuda32_13)
But anyhow .. congrats to your efforts !
Best regards .. parabol
no explenation for this ??
____________
I'm a prime millionaire !
9*2^3497442+1 | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 476,993,571
RAC: 281,271
|
Both _16s completed & validated:
http://www.primegrid.com/results.php?hostid=242806&offset=0&show_names=1&state=0&appid=
Very interesting indeed..
I'm still a bit confused with the point's - scoring system:
Laufzeit(sek) // CPU Zeit(sek) // Punkte // Anwendung
206,618.46 // 18,857.70 // 204,790.61 // Genefer (World Record) v1.06 (cuda32_13)
299,534.75 // 266,151.73 // 4,000.00 // Woodall (LLR) v6.10
416,694.52 // 361,245.76 // 6,000.00 // Prime Sierpinski Problem (LLR) v6.10
234,255.52 // 7,261.02 // 230,599.06 // Genefer (World Record) v1.06 (cuda32_13)
But anyhow .. congrats to your efforts !
Best regards .. parabol
no explenation for this ??
Explanation for... what exactly? It's not obvious to me what needs to be explained.
____________
My lucky number is 75898524288+1 | |
|
|
|
Both _16s completed & validated:
http://www.primegrid.com/results.php?hostid=242806&offset=0&show_names=1&state=0&appid=
Very interesting indeed..
I'm still a bit confused with the point's - scoring system:
Laufzeit(sek) // CPU Zeit(sek) // Punkte // Anwendung
206,618.46 // 18,857.70 // 204,790.61 // Genefer (World Record) v1.06 (cuda32_13)
299,534.75 // 266,151.73 // 4,000.00 // Woodall (LLR) v6.10
416,694.52 // 361,245.76 // 6,000.00 // Prime Sierpinski Problem (LLR) v6.10
234,255.52 // 7,261.02 // 230,599.06 // Genefer (World Record) v1.06 (cuda32_13)
But anyhow .. congrats to your efforts !
Best regards .. parabol
Comparing llr tasks with genefer is comparing apples to oranges. Gpu work has always got more credit than cpu one. Ten minutes of gpu pps sieve worth more credit than a three days woodall.
If you compare sieve credit with genefer credit you would notice that the latter is not the most rewarding one if you only consider credit.
| |
|
|
|
|
Sorry for that..
mixed up my fruit basked too much..
put my apples to the left .. and the oranges to the right now ..
.. bananas on seperate basket ..
Thank you !
parabol
____________
I'm a prime millionaire !
9*2^3497442+1 | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,430,963,941
RAC: 4,051,704
|
|
I'm back.
More like a fruit salad this end: got a _21!!! Put it into work now. All those people who have problematic 295.xx + set their monitor to power down... | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,430,963,941
RAC: 4,051,704
|
|
Ok job done:
http://www.primegrid.com/workunit.php?wuid=255438033 | |
|
Message boards :
Generalized Fermat Prime Search :
High quorum |
