Hello !!

First of, I have to say that I made many many search about my problem, and I found nothing useful (or nothing was working).

So, here is my problem :
I recently bought a brand new video card for my gaming computer, an ATI R9 280X 3Go. This card worked only one time, on Milkyway@home project. It was 2 days ago.

But since I restarted my computer, my GPU is suddenly "incompatible" with BOINC.

When I freshly installed BOINC on my computer, the GPU was already not working. But I installed a software called AMD APP SDK (Accelerated Parallel Processing) and when I restarted BOINC, it "stopped working" due to a BEX64 event, but was still able to work (and compute) with my GPU.

I didn't know what was happening with my computer and BOINC, however it worked so I though everything was alright.

But when I restarted my computer, the BEX64 event was still there and the GPU computing was unavailable and an error message shows "ATU GPU R600 (R38xx) does not support OpenCL".

Here is the complete log from a fresh restart of BOINC to the moment it starts calculating (skipped resumed and suspended tasks event) :

23/01/2014 18:45:58 |  | Starting BOINC client version 7.2.33 for windows_x86_64
23/01/2014 18:45:58 |  | log flags: file_xfer, sched_ops, task
23/01/2014 18:45:58 |  | Libraries: libcurl/7.25.0 OpenSSL/1.0.1 zlib/1.2.6
23/01/2014 18:45:58 |  | Data directory: D:\ProgramData\BOINC
23/01/2014 18:45:58 |  | Running under account Camille
23/01/2014 18:45:58 |  | GPU detection failed. error code -1073741819
23/01/2014 18:45:58 |  | OpenCL: AMD/ATI GPU 0: Tahiti (driver version 1348.5 (VM), device version OpenCL 1.2 AMD-APP (1348.5), 2048MB, 2048MB available, 2688 GFLOPS peak)
23/01/2014 18:45:58 |  | OpenCL CPU: AMD FX(tm)-4100 Quad-Core Processor (OpenCL driver vendor: Advanced Micro Devices, Inc., driver version 1348.5 (sse2,avx,fma4), device version OpenCL 1.2 AMD-APP (1348.5))
23/01/2014 18:45:58 |  | Host name: Camille-PC
23/01/2014 18:45:58 |  | Processor: 4 AuthenticAMD AMD FX(tm)-4100 Quad-Core Processor             [Family 21 Model 1 Stepping 2]
23/01/2014 18:45:58 |  | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 htt pni ssse3 cx16 sse4_1 sse4_2 popcnt aes syscall nx lm svm sse4a osvw ibs xop skinit wdt lwp fma4 topx page1gb rdtscp
23/01/2014 18:45:58 |  | OS: Microsoft Windows 7: Home Premium x64 Edition, Service Pack 1, (06.01.7601.00)
23/01/2014 18:45:58 |  | Memory: 7.95 GB physical, 15.89 GB virtual
23/01/2014 18:45:58 |  | Disk: 931.51 GB total, 616.60 GB free
23/01/2014 18:45:58 |  | Local time is UTC +1 hours
23/01/2014 18:45:58 |  | VirtualBox version: 4.2.16
23/01/2014 18:45:58 | Milkyway@Home | URL http://milkyway.cs.rpi.edu/milkyway/; Computer ID 557409; resource share 100
23/01/2014 18:45:58 |  | General prefs: from http://www.worldcommunitygrid.org/ (last modified 01-Jan-1970 01:00:01)
23/01/2014 18:45:58 |  | Host location: none
23/01/2014 18:45:58 |  | General prefs: using your defaults
23/01/2014 18:45:58 |  | Reading preferences override file
23/01/2014 18:45:58 |  | Preferences:
23/01/2014 18:45:58 |  | max memory usage when active: 4068.67MB
23/01/2014 18:45:58 |  | max memory usage when idle: 7323.61MB
23/01/2014 18:45:58 |  | max disk usage: 616.54GB
23/01/2014 18:45:58 |  | max CPUs used: 3
23/01/2014 18:45:58 |  | (to change preferences, visit a project web site or select Preferences in the Manager)
23/01/2014 18:45:58 |  | Not using a proxy
23/01/2014 18:46:02 |  | Suspending computation - initial delay
23/01/2014 18:46:06 | Milkyway@Home | project resumed by user
23/01/2014 18:47:59 | Milkyway@Home | Restarting task ps_modfit_16_3s_bpl_128wrap_1_1382698503_12903973_0 using milkyway_separation__modified_fit version 128 in slot 7
23/01/2014 18:47:59 | Milkyway@Home | Restarting task de_modfit_86_3s_bplmodfit_128_wrap_2_1382698503_12893807_2 using milkyway_separation__modified_fit version 128 in slot 8
23/01/2014 18:47:59 | Milkyway@Home | Restarting task ps_modfit_16_3s_bpl_128wrap_1_1382698503_12903974_0 using milkyway_separation__modified_fit version 128 in slot 2
23/01/2014 18:48:08 | Milkyway@Home | Sending scheduler request: To fetch work.
23/01/2014 18:48:08 | Milkyway@Home | Requesting new tasks for CPU and ATI
23/01/2014 18:48:10 | Milkyway@Home | Scheduler request completed: got 0 new tasks
23/01/2014 18:48:10 | Milkyway@Home | No tasks sent
23/01/2014 18:48:10 | Milkyway@Home | Message du serveur: ATI GPU R600 (R38xx) does not support OpenCL
23/01/2014 18:48:10 | Milkyway@Home | This computer has reached a limit on tasks in progress
23/01/2014 18:48:30 | Milkyway@Home | project suspended by user

You can find my complete config on my profile.

Thank you for your future help !!
- Europa

PS : I'm French so if you find mistakes, forgive me :)

It's a combination of AMD recycling their GPU codenames with a message set by the server at Milkyway@Home. They check for what GPU version is sent back by BOINC, to give you that message.

So at first you'll have to report this to Milkyway, as it's their message (Message from server).
You may also want to try to use BOINC 7.2.38 --the present alpha version. This one definitely has code that will put the R9 280X in the correct category:

case 20:
            gpu_name="AMD Radeon HD 7870/7950/7970/R9 280X series (Tahiti)";

For links to BOINC 7.2.38 and the change log leading up towards it, see post 52011 in the Change Log thread.

---

Hi again,

I tried to update to the new version of BOINC, but it doesn't affected any problems. My GPU still doesn't work with Milkyway, and all other projects that supports ATI GPUs doesn't support mine too.

I tried to work with World Community Grid, this didn't solved my problems.

And I wrote a message on Milkyway forums but nobody replied me for now.

Do I have to wait until projects update (or BOINC update) ? And why does my GPU worked one time and do not work anymore ?

Thanks,
- Europa

Hello !!

First of, I have to say that I made many many search about my problem, and I found nothing useful (or nothing was working).

So, here is my problem :
I recently bought a brand new video card for my gaming computer, an ATI R9 280X 3Go. This card worked only one time, on Milkyway@home project. It was 2 days ago.

But since I restarted my computer, my GPU is suddenly "incompatible" with BOINC.

When I freshly installed BOINC on my computer, the GPU was already not working. But I installed a software called AMD APP SDK (Accelerated Parallel Processing) and when I restarted BOINC, it "stopped working" due to a BEX64 event, but was still able to work (and compute) with my GPU.

I didn't know what was happening with my computer and BOINC, however it worked so I though everything was alright.

But when I restarted my computer, the BEX64 event was still there and the GPU computing was unavailable and an error message shows "ATU GPU R600 (R38xx) does not support OpenCL".

Here is the complete log from a fresh restart of BOINC to the moment it starts calculating (skipped resumed and suspended tasks event) :

23/01/2014 18:45:58 | | Starting BOINC client version 7.2.33 for windows_x86_64
23/01/2014 18:45:58 | | log flags: file_xfer, sched_ops, task
23/01/2014 18:45:58 | | Libraries: libcurl/7.25.0 OpenSSL/1.0.1 zlib/1.2.6
23/01/2014 18:45:58 | | Data directory: D:\ProgramData\BOINC
23/01/2014 18:45:58 | | Running under account Camille
23/01/2014 18:45:58 | | GPU detection failed. error code -1073741819
23/01/2014 18:45:58 | | OpenCL: AMD/ATI GPU 0: Tahiti (driver version 1348.5 (VM), device version OpenCL 1.2 AMD-APP (1348.5), 2048MB, 2048MB available, 2688 GFLOPS peak)
23/01/2014 18:45:58 | | OpenCL CPU: AMD FX(tm)-4100 Quad-Core Processor (OpenCL driver vendor: Advanced Micro Devices, Inc., driver version 1348.5 (sse2,avx,fma4), device version OpenCL 1.2 AMD-APP (1348.5))

It's not OpenCL support that is missing, But CAL support, i've seen two other hosts where CAL support was missing, but i think it was because their Bonaire GPUs didn't support CAL, while you get an error instead,
one of them was successfully crunching Seti OpenCL ATI work via anonymous platform (the project server wouldn't send work for the Stock app probably because of a requirement for a CAL driver version),
the other host was crunching Stock OpenCL ATI work at Einstein no problem.

I suggest you try uninstalling the APP SDK, and the catalyst driver, and reinstall the latest Catalyst driver (you don't need to install the APP SDK for crunching)

Claggy

HI,

I already installed AMD APP (v2.9) but it doesn't took effect on the project.

HI,

I already installed AMD APP (v2.9) but it doesn't took effect on the project.

You don't need the AMD APP SDK, the Catalyst driver comes with an APP runtime.

Claggy

It might help if you could download and run the application contained in

http://boinc.berkeley.edu/dl/clinfo.zip

and post the output here. That would help to show whether the problem is with your driver/runtime, or with BOINC's detection of the GPU's capabilities.

And please tell us what Catalyst driver you're got installed, if you're got one installed that is, and that you haven't got the driver from Windows update.

Claggy

@Richard Haselgrove
Your software has no output, I had no time to copy/paste anything. But I took a screenshot which may interest you : http://imageshack.com/a/img15/9853/az26.png

@Claggy
I have Catalyst 13.12, freshly reinstalled (downloaded from official website).

@Richard Haselgrove
Your software has no output, I had no time to copy/paste anything. But I took a screenshot which may interest you : http://imageshack.com/a/img15/9853/az26.png

Run clinfo.exe from a command line window.

Start->All Programs->Accessories->Right click Command Line->Run as administrator->Acknowledge.
In the command line window, go to the directory you saved and unzipped clinfo.zip in, then start clinfo.exe by typing clinfo and hitting Enter.

If you want the program to output to a text file, do clinfo >> clinfo.txt, which will output all the info that else scrolls in the command line window into a text file in the same directory as the clinfo.exe program starts from.

So, e.g. you have unzipped clinfo.zip to C:\temp\
You then do in the command line window:

cd\temp {Enter}
clinfo >> clinfo.txt {Enter}

You can now open clinfo.txt with Notepad in the normal way through Windows Explorer.

---

Number of platforms: 1
Platform Profile: FULL_PROFILE
Platform Version: OpenCL 1.2 AMD-APP (1348.5)
Platform Name: AMD Accelerated Parallel Processing
Platform Vendor: Advanced Micro Devices, Inc.
Platform Extensions: cl_khr_icd cl_amd_event_callback cl_amd_offline_devices cl_khr_d3d10_sharing cl_khr_d3d11_sharing cl_khr_dx9_media_sharing


Platform Name: AMD Accelerated Parallel Processing
Number of devices: 2
Device Type: CL_DEVICE_TYPE_GPU
Device ID: 4098
Max compute units: 32
Max work items dimensions: 3
Max work items[0]: 256
Max work items[1]: 256
Max work items[2]: 256
Max work group size: 256
Preferred vector width char: 4
Preferred vector width short: 2
Preferred vector width int: 1
Preferred vector width long: 1
Preferred vector width float: 1
Preferred vector width double: 1
Max clock frequency: 1050Mhz
Address bits: 32
Max memory allocation: 536870912
Image support: Yes
Max number of images read arguments: 128
Max number of images write arguments: 8
Max image 2D width: 16384
Max image 2D height: 16384
Max image 3D width: 2048
Max image 3D height: 2048
Max image 3D depth: 2048
Max samplers within kernel: 16
Max size of kernel argument: 1024
Alignment (bits) of base address: 2048
Minimum alignment (bytes) for any datatype: 128
Single precision floating point capability
Denorms: No
Quiet NaNs: Yes
Round to nearest even: Yes
Round to zero: Yes
Round to +ve and infinity: Yes
IEEE754-2008 fused multiply-add: Yes
Cache type: Read/Write
Cache line size: 64
Cache size: 16384
Global memory size: 2147483648
Constant buffer size: 65536
Max number of constant args: 8
Local memory type: Scratchpad
Local memory size: 32768
Error correction support: 0
Profiling timer resolution: 1
Device endianess: Little
Available: Yes
Compiler available: Yes
Execution capabilities:
Execute OpenCL kernels: Yes
Execute native function: No
Queue properties:
Out-of-Order: No
Profiling : Yes
Platform ID: 5BEC2798
Name: Tahiti
Vendor: Advanced Micro Devices, Inc.
Driver version: 1348.5 (VM)
Profile: FULL_PROFILE
Version: OpenCL 1.2 AMD-APP (1348.5)
Extensions: cl_khr_fp64 cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_ext_atomic_counters_32 cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_media_ops2 cl_amd_popcnt cl_khr_d3d10_sharing cl_khr_dx9_media_sharing cl_khr_image2d_from_buffer


Device Type: CL_DEVICE_TYPE_CPU
Device ID: 4098
Max compute units: 4
Max work items dimensions: 3
Max work items[0]: 1024
Max work items[1]: 1024
Max work items[2]: 1024
Max work group size: 1024
Preferred vector width char: 16
Preferred vector width short: 8
Preferred vector width int: 4
Preferred vector width long: 2
Preferred vector width float: 8
Preferred vector width double: 4
Max clock frequency: 3901Mhz
Address bits: 32
Max memory allocation: 1073741824
Image support: Yes
Max number of images read arguments: 128
Max number of images write arguments: 8
Max image 2D width: 8192
Max image 2D height: 8192
Max image 3D width: 2048
Max image 3D height: 2048
Max image 3D depth: 2048
Max samplers within kernel: 16
Max size of kernel argument: 4096
Alignment (bits) of base address: 1024
Minimum alignment (bytes) for any datatype: 128
Single precision floating point capability
Denorms: Yes
Quiet NaNs: Yes
Round to nearest even: Yes
Round to zero: Yes
Round to +ve and infinity: Yes
IEEE754-2008 fused multiply-add: Yes
Cache type: Read/Write
Cache line size: 64
Cache size: 16384
Global memory size: 2147483648
Constant buffer size: 65536
Max number of constant args: 8
Local memory type: Global
Local memory size: 32768
Error correction support: 0
Profiling timer resolution: 262
Device endianess: Little
Available: Yes
Compiler available: Yes
Execution capabilities:
Execute OpenCL kernels: Yes
Execute native function: Yes
Queue properties:
Out-of-Order: No
Profiling : Yes
Platform ID: 5BEC2798
Name: AMD FX(tm)-4100 Quad-Core Processor
Vendor: AuthenticAMD
Driver version: 1348.5 (sse2,avx,fma4)
Profile: FULL_PROFILE
Version: OpenCL 1.2 AMD-APP (1348.5)
Extensions: cl_khr_fp64 cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_ext_device_fission cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_media_ops2 cl_amd_popcnt cl_khr_d3d10_sharing

That's great - you certainly have OpenCL, and a reasonably recent one (version 1.2) configured for your GPU - and your CPU as well, though that's less useful.

You mentioned installing a new driver, too. What does BOINC say at start-up now?

"BOINC stopped working." BEX64 event.

I choose to close the program but BOINC still starts up.
I though it was because the installer installs BOINC in Program Files (x86) by default, so I forced the installation in Program Files folder (64 bit one). The problem still there.

Maybe, if the GPU "won't work" it's because Milkyway has no GPU tasks for now. But it's same for every other projects.

I though it was because the installer installs BOINC in Program Files (x86) by default, so I forced the installation in Program Files folder (64 bit one). The problem still there.

Actually, the installer checks the registry for if an earlier BOINC was installed, and if so where. If you had a 32bit BOINC installed before, it would've installed to the C:\Program Files (x86)\BOINC\ directory and stored that value into the registry.

The installer doesn't know what the contents of the database are that it carries. It cannot see if it is a 32bit or 64bit program. So when it finds a value already filled in at this point, it'll take that and will try to install to that directory.

That's where the user comes in to check everything before the actual install, and as such you can press Advanced in the third screen in the installer and change the install paths of the programs directory and the data directory, as well as set some flags for when you want to use the screen saver, install BOINC as a service and give permissions to others to control BOINC.

---

Hello,

I installed the service with BOINC v7.2.38 (beta version) which does not detect any usable GPU. There is no crash at startup, but GPU computing still not available :

27/01/2014 18:23:28 | | Data directory: D:\ProgramData\BOINC
27/01/2014 18:23:28 | | Running under account boinc_master
27/01/2014 18:23:28 | | No usable GPUs found
27/01/2014 18:23:28 | | Host name: Camille-PC
27/01/2014 18:23:28 | | Processor: 4 AuthenticAMD AMD FX(tm)-4100 Quad-Core Processor [Family 21 Model 1 Stepping 2]


I will try to uninstall/reinstall latest stable version (7.2.33) (with service)

EDIT : Still the same thing with 7.2.33

I didn't say that you had to install BOINC as a service. Have you read the description of the Service Install? If not, I refer to the image I gave above.
"Run project applications under a privileged account. This provides increased protected from faulty applications, and on Windows, it will prevent the use of any applications that use the graphics chip (GPU).

Before you ask when this will be fixed then, it isn't something for BOINC to fix. It's a security measure instated by Microsoft.

Or as a good (nameless) friend of mine wrote some time ago:

The benefit of installing as a service is that BOINC can run even if no one is logged onto the machine as all services are started at bootup before user logon.

Therein lies the issue with Windows Vista and newer; Microsoft has tried to reduce the number of crashes that bring the entire system down, so they moved the sound driver and video driver out of the kernel space (Ring zero) and into the user space (Ring three) so that, even if you have a driver crash, it can be recovered without rebooting the entire system.

Because of this change, only a basic video driver is loaded during boot to display the boot process. The enhanced functions of video cards aren't loaded until after the user logs into the machine. Thus, the crunching capabilities of a GPU aren't loaded until after the user logs on, and since BOINC is already running, it cannot use those extra functions.

In order for device manufacturers to have their drivers certified to run on Windows, and Microsoft has enforced driver signing so that poorly written drivers that do not pass Windows Hardware Quality Labs (WHQL) testing, each vendor must submit their driver for testing to Microsoft and meet Microsoft's requirements so as to improve the stability of the platform.

Since it is cheaper and easier to submit a single driver, most vendors have been getting their drivers passed and back-porting the certified driver to older OSes, such as Windows XP (which preferred driver signing, but didn't enforce it).

---

OK.. So this was not a solution to my problem ? Why does others can use their GPUs ? I had an old GPU (Radeon HD6700) that was working on BOINC, but since I swapped my graphics, BOINC can't compute with it.

When you swapped, did you uninstall the videocard drivers, power down, swap out the cards, restart the computer, install the new videocard drivers?

---

I uninstalled my first driver in Safe mode, and then I restarted my computer to install a brand new (freshly downloaded) video driver.

Maybe BOINC does not support R9 280X GPUs ??

I uninstalled my first driver in Safe mode, and then I restarted my computer to install a brand new (freshly downloaded) video driver.

Maybe BOINC does not support R9 280X GPUs ??

That's what we were trying to explore with those clinfo logs, and suchlike.

It's beginning to look a little like that, isn't it? Though perhaps we should say "BOINC does not support R9 280X GPUs yet"