ERROR: cudaGetLastError() returned 30: unknown error

garo · 2013-04-14, 18:14

I just got this error while factoring M35798227 from 2^69 to 2^70. mfaktc-0.20-win32.exe, Windows7. Restarting from the checkpoint did not reproduce the error. So can I assume a hardware error is the cause?

TheJudger · 2013-04-14, 21:40

Hi garo,

could be anything I guess...
This is the code which prints the CUDA errors:

Code:

void print_last_CUDA_error()
/* just run cudaGetLastError() and print the error message if its return value is not cudaSuccess */
{
  cudaError_t cudaError;

  cudaError = cudaGetLastError();
  if(cudaError != cudaSuccess)
  {
    printf("  cudaGetLastError() returned %d: %s\n", cudaError, cudaGetErrorString(cudaError));
  }
}

The translation from "error 30" to "unknown error" is done with the function cudaGetErrorString() which comes from Nvidia. So if they don't know error 30 I don't know, too. Sorry!

Edit:

Quote:

Check /path/to/cudatoolkit/include/driver_types.h for error codes:
/**
* This indicates that an unknown internal error has occurred.
*/
cudaErrorUnknown = 30,

So they really don't want to tell us what is going wrong here...

Oliver

garo · 2013-04-16, 18:52

Thanks for your response. I will let you know if it happens again. Not that there is anything we can do about it.

Jatheski · 2013-04-16, 21:05

Quote:

Originally Posted by garo

I just got this error while factoring M35798227 from 2^69 to 2^70. mfaktc-0.20-win32.exe, Windows7. Restarting from the checkpoint did not reproduce the error. So can I assume a hardware error is the cause?

Sometimes I got the same error with mmff, but only with one of my two GPUs (GTX 560). I think it's a hardware problem.

patrik · 2013-04-23, 15:16

I got the same error using CudaLucas v2.03 after installing a few windows updates on Windows 7 yesterday.

Code:

Iteration 9520000 M( 33304223 )C, 0x70115f3ed0df6480, n = 1835008, CUDALucas v2.03 err = 0.2188 (0:36 real, 3.5385 ms/iter, ETA 23:22:25)
Iteration 9530000 M( 33304223 )C, 0x5e1f4ae22972beb7, n = 1835008, CUDALucas v2.03 err = 0.2188 (0:35 real, 3.5387 ms/iter, ETA 23:21:54)
Iteration 9540000 M( 33304223 )C, 0x212bc36fffd05cf5, n = 1835008, CUDALucas v2.03 err = 0.2188 (0:36 real, 3.5389 ms/iter, ETA 23:21:24)
CUDALucas.cu(695) : cudaSafeCall() Runtime API error 30: unknown error.

Today, I instead got

Code:

Iteration 9800000 M( 33304223 )C, 0xf927b0bc2c7cfc78, n = 1835008, CUDALucas v2.03 err = 0.2109 (0:34 real, 3.4199 ms/iter, ETA 22:19:26)
Iteration 9810000 M( 33304223 )C, 0x205331e8fa73c1ce, n = 1835008, CUDALucas v2.03 err = 0.2109 (0:35 real, 3.4122 ms/iter, ETA 22:15:52)
iteration = 9814801 >= 1000 && err = 0.5 >= 0.35, fft length = 1835008, not writing checkpoint file (because -t is disabled) and exiting.

It seems to have installed something called 3D Vision. I can also find three Nvidia items installed yesterday (NVIDIA 3D Vision drivrutin 311.06, NVIDIA Grafikdrivrutin 311.06 and NVIDIA-uppdatering 1.11.3).

Is it safe for me to try to uninstall these items? (Can I re-install them from somewhere if it fails? Or do they disappear completely from my system if I do so?) And which one is most likely to cause the problems?

kladner · 2013-04-23, 20:30

I have seen 3D Vision and Updater referred to as "bloat-ware." They are not required for CUDA-related processing. 3D Vision is only needed if you are gaming with nVidia 3D glasses. These items tend to complicate the driver install process, and they load things at Windows startup which are just consuming memory. I don't believe that they would cause the errors. That is more likely from an error in installing the Graphics Driver.

I suggest uninstalling all of the nVidia items, doing the Graphics Driver last. After the required reboot, get the latest driver from the nVidia site. It may help to run the installer as an administrator. It may also help to temporarily disable anti-virus software. Do a Custom Install and check only Graphics Driver and PhysX. Also check Clean Install.

You could try just uninstalling 3D Vision and Updater. That should not break anything. However, it may not fix your problems, either. nVidia drivers can be very finicky about installation. I actually follow a more severe uninstall procedure before I upgrade the driver, but I'm not sure that it is really necessary.

Karl M Johnson · 2013-04-24, 06:06

Huh?
I cant find Forceware v311.06 on NV's website.
I suggest you use the latest release Forceware, which is 314.22, from here.

Keldor · 2013-04-27, 11:05

Same error on my computer. After running CL for a while the screen turns black for about 2 seconds, I get a message that the video driver didn't react and had to be restored, and CL throws out error 30.

Strange: This seems to happen only at iteration numbers ending with a "2".

mfactc runs without problems.

owftheevil · 2013-04-27, 13:53

garo, do you happen to remember what line in mfaktc the error came from?

garo · 2013-04-27, 18:18

No it did not give a line number.

owftheevil · 2013-04-27, 19:01

After seeing your answer, I looked at Oliver's code snippet above and realized how pointless my question was. Oh well, it would have been nice to know. The reason I asked is that the CUDALucas errors I've seen along these lines all happened at calls that did little else except sync the host and device.

2013-04-14, 18:14	#1
garo Aug 2002 Termonfeckin, IE 2768₁₀ Posts	ERROR: cudaGetLastError() returned 30: unknown error I just got this error while factoring M35798227 from 2^69 to 2^70. mfaktc-0.20-win32.exe, Windows7. Restarting from the checkpoint did not reproduce the error. So can I assume a hardware error is the cause?

2013-04-16, 18:52	#3
garo Aug 2002 Termonfeckin, IE 2⁴·173 Posts	Thanks for your response. I will let you know if it happens again. Not that there is anything we can do about it. Last fiddled with by garo on 2013-04-16 at 18:52

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
ERROR: cudaGetLastError() returned 4: unspecified launch failure	MacFactor	GPU Computing	0	2017-12-22 16:04
Error returned on P95 submission	kladner	PrimeNet	4	2017-09-30 19:49
Software error or hardware error	GuloGulo	Software	3	2011-01-19 00:36
Error 5 causes userid change and error 17 updates for exponents	Old man PrimeNet	PrimeNet	0	2006-02-05 02:27
ERROR: Primenet error 2252. Q: which tcp/ip ports are being used for the transfer?	nevillednz	PrimeNet	15	2004-05-17 23:08

2013-04-23, 20:30	#6
kladner "Kieren" Jul 2011 In My Own Galaxy! 2×3×1,693 Posts	I have seen 3D Vision and Updater referred to as "bloat-ware." They are not required for CUDA-related processing. 3D Vision is only needed if you are gaming with nVidia 3D glasses. These items tend to complicate the driver install process, and they load things at Windows startup which are just consuming memory. I don't believe that they would cause the errors. That is more likely from an error in installing the Graphics Driver. I suggest uninstalling all of the nVidia items, doing the Graphics Driver last. After the required reboot, get the latest driver from the nVidia site. It may help to run the installer as an administrator. It may also help to temporarily disable anti-virus software. Do a Custom Install and check only Graphics Driver and PhysX. Also check Clean Install. You could try just uninstalling 3D Vision and Updater. That should not break anything. However, it may not fix your problems, either. nVidia drivers can be very finicky about installation. I actually follow a more severe uninstall procedure before I upgrade the driver, but I'm not sure that it is really necessary.

2013-04-24, 06:06	#7
Karl M Johnson Mar 2010 3×137 Posts	Huh? I cant find Forceware v311.06 on NV's website. I suggest you use the latest release Forceware, which is 314.22, from here.

2013-04-27, 11:05	#8
Keldor Feb 2013 2³ Posts	Same error on my computer. After running CL for a while the screen turns black for about 2 seconds, I get a message that the video driver didn't react and had to be restored, and CL throws out error 30. Strange: This seems to happen only at iteration numbers ending with a "2". mfactc runs without problems.

2013-04-27, 13:53	#9
owftheevil "Carl Darby" Oct 2012 Spring Mountains, Nevada 3²·5·7 Posts	garo, do you happen to remember what line in mfaktc the error came from?

2013-04-27, 18:18	#10
garo Aug 2002 Termonfeckin, IE 2⁴·173 Posts	No it did not give a line number.

2013-04-27, 19:01	#11
owftheevil "Carl Darby" Oct 2012 Spring Mountains, Nevada 3²×5×7 Posts	After seeing your answer, I looked at Oliver's code snippet above and realized how pointless my question was. Oh well, it would have been nice to know. The reason I asked is that the CUDALucas errors I've seen along these lines all happened at calls that did little else except sync the host and device.