GPU don´t work

tontec1
tontec1
Joined: 1 Apr 17
Posts: 2
Credit: 1000206
RAC: 0
Topic 206592

Hi, I've just started working with E@H a few days ago. But I can not work with my GPU (Nvida GTX 1080) all drivers are up to date.

Log:

02.04.2017 20:34:27 |  | Starting BOINC client version 7.6.33 for windows_x86_64
02.04.2017 20:34:27 |  | log flags: file_xfer, sched_ops, task, cpu_sched_status
02.04.2017 20:34:27 |  | Libraries: libcurl/7.47.1 OpenSSL/1.0.2g zlib/1.2.8
02.04.2017 20:34:27 |  | Running as a daemon (GPU computing disabled)
02.04.2017 20:34:27 |  | Data directory: C:\ProgramData\BOINC
02.04.2017 20:34:27 |  | Running under account boinc_master
02.04.2017 20:34:27 |  | No usable GPUs found
02.04.2017 20:34:27 |  | Host name: PC
02.04.2017 20:34:27 |  | Processor: 8 GenuineIntel Intel(R) Core(TM) i7-7700K CPU @ 4.20GHz [Family 6 Model 158 Stepping 9]
02.04.2017 20:34:27 |  | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 fma cx16 sse4_1 sse4_2 movebe popcnt aes f16c rdrandsyscall nx lm avx avx2 vmx tm2 pbe fsgsbase bmi1 hle smep bmi2
02.04.2017 20:34:27 |  | OS: Microsoft Windows 10: Core x64 Edition, (10.00.14393.00)
02.04.2017 20:34:27 |  | Memory: 15.90 GB physical, 18.27 GB virtual
02.04.2017 20:34:27 |  | Disk: 223.02 GB total, 62.17 GB free
02.04.2017 20:34:27 |  | Local time is UTC +2 hours
02.04.2017 20:34:27 | Einstein@Home | URL http://einstein.phys.uwm.edu/; Computer ID 12512705; resource share 100
02.04.2017 20:34:27 |  | No general preferences found - using defaults
02.04.2017 20:34:27 |  | Reading preferences override file
02.04.2017 20:34:27 |  | Preferences:
02.04.2017 20:34:27 |  | max memory usage when active: 8138.25MB
02.04.2017 20:34:27 |  | max memory usage when idle: 14648.84MB
02.04.2017 20:34:27 |  | max disk usage: 65.11GB
02.04.2017 20:34:27 |  | max CPUs used: 6


 


CPU: Intel i7 7700k


GPU: NVIDIA GTX 1080


Windows 10


 


Tank´s for Help

Zalster
Zalster
Joined: 26 Nov 13
Posts: 3117
Credit: 4050672230
RAC: 0

Doesn't look you have any

Doesn't look you have any drivers for OpenCl.  Did Window's install your graphic drivers?  They are notorious for downloading their drivers which disables OpenCL.  Go to Nvidia's website and download the driver directly from them and do a clean install. Don't allow Windows to install your drivers in the future or next Tuesday you will find yourself right back where you started with no OpenCl.

tontec1
tontec1
Joined: 1 Apr 17
Posts: 2
Credit: 1000206
RAC: 0

I have just installed the

I have just installed the latest NVIDIA driver directly from NVIDIA. But the GTX 1080 is still not identify.
Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2139
Credit: 2752669592
RAC: 1465799

Johannes_35

Johannes_35 wrote:
Log:
02.04.2017 20:34:27 |  | Running as a daemon (GPU computing disabled) 

You will need to uninstall BOINC, and reinstall in user mode - go the the Advanced screen during installation and uncheck the service mode option.
[AF>France>Aquitaine>Cote-Adour-et-Gaves]Bernard 64250
[AF>France>Aqui...
Joined: 12 Mar 12
Posts: 3
Credit: 2747675
RAC: 0

You aren't the only one to

You aren't the only one to have problems.

On my Windows 7 laptop with a GeForce GTX 460M, all E@H GPU tasks fail after a few hours : the screen starts flahing on and off, there is a message telling the driver crashed and had to be recuperated, and the task fails.

Last one :

27/08/2017 01:01:05 | Einstein@Home | Computation for task LATeah0138L_1060.0_0_0.0_16336335_0 finished
27/08/2017 01:01:05 | Einstein@Home | Output file LATeah0138L_1060.0_0_0.0_16336335_0_0 for task LATeah0138L_1060.0_0_0.0_16336335_0 absent
27/08/2017 01:01:05 | Einstein@Home | Output file LATeah0138L_1060.0_0_0.0_16336335_0_1 for task LATeah0138L_1060.0_0_0.0_16336335_0 absent

 

Is there a recommended driver version? Do I need a specific SW package? Is there a special way to load Boinc? I can't find any information about it.

archae86
archae86
Joined: 6 Dec 05
Posts: 3145
Credit: 7023324931
RAC: 1821328

Johannes_35 wrote:I have just

Johannes_35 wrote:
I have just installed the latest NVIDIA driver directly from NVIDIA. But the GTX 1080 is still not identify.

Once when I had a persistent failure to identify it only got healed when I did a full driver uninstall, followed by an install of the current direct from Nvidia with the "clean install" option selected. I had already failed using the same driver, including use of DDU. I think in that specific case the clean install option removed some leftover setting or such which somehow contributed to the problem.

This is not to say that preliminary uninstall, use of DDU, or selection of the clean install option are always necessary. I have often had good success just running the downloaded installer with no special measures.

'tis a mystery, it is.

 

solling2
solling2
Joined: 20 Nov 14
Posts: 219
Credit: 1563256630
RAC: 48358

_AF_France_Aquitaine_Cote-Ado

_AF_France_Aquitaine_Cote-Adour-et-Gaves_Bernard 64250 wrote:

You aren't the only one to have problems.

On my Windows 7 laptop with a GeForce GTX 460M, all E@H GPU tasks fail after a few hours : the screen starts flahing on and off, there is a message telling the driver crashed and had to be recuperated, and the task fails.

Last one :

27/08/2017 01:01:05 | Einstein@Home | Computation for task LATeah0138L_1060.0_0_0.0_16336335_0 finished
27/08/2017 01:01:05 | Einstein@Home | Output file LATeah0138L_1060.0_0_0.0_16336335_0_0 for task LATeah0138L_1060.0_0_0.0_16336335_0 absent
27/08/2017 01:01:05 | Einstein@Home | Output file LATeah0138L_1060.0_0_0.0_16336335_0_1 for task LATeah0138L_1060.0_0_0.0_16336335_0 absent

 

Is there a recommended driver version? Do I need a specific SW package? Is there a special way to load Boinc? I can't find any information about it.

Laptops generally tend to be susceptible to overheating, especially low cost producers may skimp on fans or power supply units. Did you monitor temperature, memory load or overclocking recently? No special Boinc installation required to my knowledge, other than a recent release (which generally seems to apply to all software).

[AF>France>Aquitaine>Cote-Adour-et-Gaves]Bernard 64250
[AF>France>Aqui...
Joined: 12 Mar 12
Posts: 3
Credit: 2747675
RAC: 0

In fact my GPU is a NVIDIA

In fact my GPU is a NVIDIA GeForce GT 540M (2048MB) driver: 385.41.

There is no overheating : my laptop has a very big fan, not silent at all. GPU tasks are running on it from years, including 7 x 24 use.

I'm currently crushing SETI GPU tasks (CUDA and OpenCl) 7 x 24, and they are running without any problem.

The strange thing is that Einstein GPU tasks fall into error just before reaching 100%. Overheating should stop them at any time.

Result: My computer works hours and hours for nothing.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109386730141
RAC: 35925423

_AF_France_Aquitaine_Cote-Ado

_AF_France_Aquitaine_Cote-Adour-et-Gaves_Bernard 64250 wrote:
There is no overheating : my laptop has a very big fan, not silent at all. GPU tasks are running on it from years, including 7 x 24 use.

How do you know there is no overheating?  A fan making a lot of noise is a sign of a fan needing to work hard.  If you have been running "for years" how often have you inspected/cleaned/serviced the heat sink and fan assembly?  I used to run tasks on just the CPU cores of a couple of laptops a few years ago.  I found I needed to remove fluff/dust from fins/filters etc, about every 6 months.  Even then, both died prematurely from the stress of crunching.  Laptops are not designed to cope with crunching loads 24/7.

Quote:
I'm currently crushing SETI GPU tasks (CUDA and OpenCl) 7 x 24, and they are running without any problem.

It's a fallacy to think that because a different project works OK, the problem must be with this project.  Different apps put different stresses on devices.  I run lots of desktop style GPUs and my experience is that when tasks fail there are two main culprits - heat and quality/adequacy of power.  Occasionally, problems may be caused by a particular driver/software update but when that happens it's usually rather obvious because lots of people have the same issue.  At the moment, that doesn't seem to be happening.  After heat and power, the third most likely cause of problems is hardware issues, particularly as components (eg capacitors) age.

Quote:
The strange thing is that Einstein GPU tasks fall into error just before reaching 100%. Overheating should stop them at any time.

You currently have 5 failed tasks showing.  Two were around the 11k and 16k mark.  Three were over 40k.  What %complete was showing for the longest run time?  The estimate stops increasing at 89.997% on all of mine and then jumps to 100%.  Initial crunching has actually finished at this point.  There is a followup stage which assesses the 10 highest ranking candidate signals that have been found in the initial pass.  Double precision is used in this followup stage.  If the GPU isn't capable of double precision, the followup stage is done on the CPU.  I imagine this is the point of highest stress so perhaps it's not surprising that tasks may fail at this point if the hardware is not 'robust' enough.

I know you probably don't want to hear all this but I agree with Solling2's assessment.

 

Cheers,
Gary.

Richie
Richie
Joined: 7 Mar 14
Posts: 656
Credit: 1702989778
RAC: 0

Your host is running Windows

Your host is running Windows so you could test if downclocking the GPU would change the situation some way. MSI Afterburner for example can be used to set maximum core clock somewhat slower than default. It depends on the GPU model how much the default speed could be tweaked.

https://www.msi.com/page/afterburner##downloads

TThrottle is another software you could use to set a maximum temp limit for the GPU.

https://efmer.com/tthrottle/

Both softwares can be used to monitor the load temperatures of both the CPU and GPU. I would try setting GPU core clock 200MHz lower than default (if possible) or setting the max temp to 70 C and then see if tasks would still error out.

tullio
tullio
Joined: 22 Jan 05
Posts: 2118
Credit: 61407735
RAC: 0

Latest attempt on my Windows

Latest attempt on my Windows 10 PC, GTX 1050 Ti, nVidia driver 385.41

ERROR: /home/bema/fermilat/src/bridge_fft_clfft.c:948: clFinish failed. status=-36
ERROR: opencl_ts_2_phase_diff_sorted() returned with error 0
10:12:06 (1624): [CRITICAL]: ERROR: MAIN() returned with error '-36'
FPU status flags:  PRECISION
10:12:18 (1624): [normal]: done. calling boinc_finish(28).
10:12:18 (1624): called boinc_finish

Tullio

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.