Author |
Message |
|
Hello everyone,
After Proth Prime Search GPU application update (1.38), i am receiving tons of invalid wus for my ATI rig here:
http://www.primegrid.com/results.php?hostid=179118&offset=0&show_names=0&state=4&appid=
First i thought it was the driver but than i reformatted the whole system and installed Catalyst 10.10 APP. I believe there might be a clash between Cuda ones and ATI ones during validation. Please guide me on this issue.
Thank you.
NOTE: No OC, no voltage change. |
|
|
|
Sieve usually does not need to validated against a 2nd result. So your machine is sending back units that are somehow corrupt. Other than that, I am unsure what else to tell you. |
|
|
|
I take it you were doing just fine at version 1.37 (or whichever was being used previously on your machine)?
If that's the case then I'd bet it's an issue with the program. If other people notice the same problem, then we know what's the cause (but not necessarily why...)
Hm, and if you reformatted your system then I guess there's no need to reset PrimeGrid and simply try getting the data files again.
Perhaps it's something wrong on the server side. It might think you're sieving the wrong range, sieving it incompletely, or sieving things already sieved...but that's just a guess, since I can't see anywhere you may have gone wrong.
Confusing!
____________
|
|
|
|
I take it you were doing just fine at version 1.37 (or whichever was being used previously on your machine)?
If that's the case then I'd bet it's an issue with the program. If other people notice the same problem, then we know what's the cause (but not necessarily why...)
Hm, and if you reformatted your system then I guess there's no need to reset PrimeGrid and simply try getting the data files again.
Perhaps it's something wrong on the server side. It might think you're sieving the wrong range, sieving it incompletely, or sieving things already sieved...but that's just a guess, since I can't see anywhere you may have gone wrong.
Confusing!
I was doing fine with 1.35, then it was updated and my previous results were marked invalid which is completely fine. But as i have all v1.38, the wus are invalidated as nvidia ones gets the validation. That is why i am thinking about clash. |
|
|
Ken_g6 Volunteer developer
 Send message
Joined: 4 Jul 06 Posts: 915 ID: 3110 Credit: 183,164,814 RAC: 0
                        
|
It seems unlikely to me that TPSieve would send invalid results. It seems more likely that something might be wrong on the server.
However, I have been wrong before. :) And ATI on Win64 is about as far away from my testing scenarios as you can get. So could you please do the standard tests and let me know if anything goes wrong?
____________
|
|
|
|
It seems unlikely to me that TPSieve would send invalid results. It seems more likely that something might be wrong on the server.
However, I have been wrong before. :) And ATI on Win64 is about as far away from my testing scenarios as you can get. So could you please do the standard tests and let me know if anything goes wrong?
Ok, will do and will send you the results. |
|
|
|
It seems unlikely to me that TPSieve would send invalid results. It seems more likely that something might be wrong on the server.
However, I have been wrong before. :) And ATI on Win64 is about as far away from my testing scenarios as you can get. So could you please do the standard tests and let me know if anything goes wrong?
Ok, will do and will send you the results.
As per BOINC release notes:
Do not install BOINC in protected application execution mode on Windows Vista and Windows 7 if you plan to use a GPU for doing calculations.
Also ensure that .NET is correctly installed and that no screensaver is running.
Ensure that you can run OpenGL apps and Collatz (it should be up now). Then you should run the app from the command line.
|
|
|
|
It seems unlikely to me that TPSieve would send invalid results. It seems more likely that something might be wrong on the server.
However, I have been wrong before. :) And ATI on Win64 is about as far away from my testing scenarios as you can get. So could you please do the standard tests and let me know if anything goes wrong?
Ok, will do and will send you the results.
As per BOINC release notes:
Do not install BOINC in protected application execution mode on Windows Vista and Windows 7 if you plan to use a GPU for doing calculations.
Also ensure that .NET is correctly installed and that no screensaver is running.
Ensure that you can run OpenGL apps and Collatz (it should be up now). Then you should run the app from the command line.
Thank you for the advice. I'm pretty sure that NET is properly working.
But i don't know how to use Boinc command tool. I couldn't download it as well. or don't even know whether it comes with the usual software. If you treat me as a newbie on this issue, i will be able to provide the info:).
I checked the link given on the previous post but didnt know where to run it.
|
|
|
|
It seems unlikely to me that TPSieve would send invalid results. It seems more likely that something might be wrong on the server.
However, I have been wrong before. :) And ATI on Win64 is about as far away from my testing scenarios as you can get. So could you please do the standard tests and let me know if anything goes wrong?
Ok, will do and will send you the results.
As per BOINC release notes:
Do not install BOINC in protected application execution mode on Windows Vista and Windows 7 if you plan to use a GPU for doing calculations.
Also ensure that .NET is correctly installed and that no screensaver is running.
Ensure that you can run OpenGL apps and Collatz (it should be up now). Then you should run the app from the command line.
Thank you for the advice. I'm pretty sure that NET is properly working.
But i don't know how to use Boinc command tool. I couldn't download it as well. or don't even know whether it comes with the usual software. If you treat me as a newbie on this issue, i will be able to provide the info:).
I checked the link given on the previous post but didnt know where to run it.
I got it. No need for info |
|
|
|
Check out one my hosts please:
http://www.primegrid.com/results.php?hostid=182155&offset=0&show_names=0&state=4&appid=
What could be the reason of these invalidations? As posted previously, clean windows 7 with catalyst 10.9 stream 2.2 and i am pretty sure that there is no hardware problems. Invalidations became significant after the update of 1.38v
NOTE: The standart test did not give any error or unusual result. |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2649 ID: 1 Credit: 26,363,112 RAC: 0
                    
|
You GPU WUs are running entirely on CPU. I believe that something is wrong with the drivers and it falls back to CPU? Ken, can you confirm?
____________
|
|
|
|
You GPU WUs are running entirely on CPU. I believe that something is wrong with the drivers and it falls back to CPU? Ken, can you confirm?
I tried from 10.8 to 10.12 and 10.9 was the most stable for me. If you look at the valid ones, you will see that the cpu&gpu timings match also which indicates they rely on cpus as well. Isn't this normal since usage is 0.82? So I don't think it is a driver issue. Also i was doing fine with these drivers until 1.38 update. |
|
|
|
I updated the drivers to 10.10 APP after a format and still it gives errors. At least it is stable now. It only gives invalid wus on this project. Tried DNETC and milkyway. A guide would be appreciated though.
Please? |
|
|
|
I updated the drivers to 10.10 APP after a format and still it gives errors. At least it is stable now. It only gives invalid wus on this project. Tried DNETC and milkyway. A guide would be appreciated though.
Please?
Can you show the test output given in this thread (more the better)?
|
|
|
|
I updated the drivers to 10.10 APP after a format and still it gives errors. At least it is stable now. It only gives invalid wus on this project. Tried DNETC and milkyway. A guide would be appreciated though.
Please?
Can you show the test output given in this thread (more the better)?
I already sent it to Ken_g6 two or three weeks ago. It gave the same results as it was supposed to and showed no error.
If that important i can do it again but I just don't want to suspend the computing:) |
|
|
|
I opened this thread for seeking answers but it has been ignored and my rig has been producing producing %50 invalid wus out of total which is a disadvantage for both the project and myself.
I did the test and it produced correct answers.
Changing drivers didn't help. I even had format couple of times to handle the driver issues..
Observed no hardware failures. They run flawless on other computation processes. (Even changed with other gpus)..
Please inform me if you do something about it.
|
|
|
rroonnaalldd Volunteer developer Volunteer tester
 Send message
Joined: 3 Jul 09 Posts: 1213 ID: 42893 Credit: 34,634,263 RAC: 0
                 
|
Maybe the SP1 causes this?
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
Maybe the SP1 causes this?
The problem was there before SP1 installation. |
|
|