Author |
Message |
|
Hi everyone,
My 5770 is getting errors on all WUs ... here's an example of the message I get in the BOINC client :
05/12/2010 23:03:53 PrimeGrid Aborting task pps_sr2sieve_4674925_0: exceeded elapsed time limit 1814.408691
And here is the full report in the task list :
<core_client_version>6.10.58</core_client_version>
<![CDATA[
<message>
Maximum elapsed time exceeded
</message>
<stderr_txt>
Sieve started: 3019078000000000 <= p < 3019081000000000
Thread 0 starting
Detected 160 multiprocessors (800 SPUs) on device 0.
Sieve started: 3019078000000000 <= p < 3019081000000000
Resuming from checkpoint p=3019079905786881 in tpcheck3019078e9.txt
Thread 0 starting
Detected 160 multiprocessors (800 SPUs) on device 0.
Unhandled Exception Detected...
- Unhandled Exception Record -
Reason: Breakpoint Encountered (0x80000003) at address 0x7C91120E
Engaging BOINC Windows Runtime Debugger...
********************
BOINC Windows Runtime Debugger Version 6.10.17
Dump Timestamp : 12/05/10 23:03:54
Install Directory : F:\BOINC\
Data Directory : F:\BOINC\Data
Project Symstore :
Loaded Library : F:\BOINC\dbghelp.dll
Loaded Library : F:\BOINC\symsrv.dll
Loaded Library : F:\BOINC\srcsrv.dll
LoadLibraryA( F:\BOINC\version.dll ): GetLastError = 126
Loaded Library : version.dll
Debugger Engine : 4.0.5.0
Symbol Search Path: F:\BOINC\Data\slots\4;F:\BOINC\Data\projects\www.primegrid.com
ModLoad: 00400000 0003d000 F:\BOINC\Data\projects\www.primegrid.com\primegrid_tpsieve_1.35_windows_intelx86__ati13ati.exe (-nosymbols- Symbols Loaded)
Linked PDB Filename : C:\Documents and Settings\Ken\Desktop\ppsieve-cl-source\TPS-BOINC-Release\tpsieve-cl-boinc-x86-windows.pdb
ModLoad: 7c910000 000b9000 C:\WINDOWS\system32\ntdll.dll (5.1.2600.5755) (-exported- Symbols Loaded)
Linked PDB Filename : ntdll.pdb
File Version : 5.1.2600.5755 (xpsp_sp3_gdr.090206-1234)
Company Name : Microsoft Corporation
Product Name : Syst�me d'exploitation Microsoft� Windows�
Product Version : 5.1.2600.5755
ModLoad: 7c800000 00106000 C:\WINDOWS\system32\kernel32.dll (5.1.2600.5781) (-exported- Symbols Loaded)
Linked PDB Filename : kernel32.pdb
File Version : 5.1.2600.5781 (xpsp_sp3_gdr.090321-1317)
Company Name : Microsoft Corporation
Product Name : Syst�me d'exploitation Microsoft� Windows�
Product Version : 5.1.2600.5781
ModLoad: 10000000 00010000 C:\WINDOWS\system32\OpenCL.dll (1.1.0.0) (-exported- Symbols Loaded)
Linked PDB Filename :
File Version : 1.1.0
Company Name : Khronos Group
Product Name : Khronos OpenCL ICD
Product Version :
ModLoad: 77da0000 000ac000 C:\WINDOWS\system32\ADVAPI32.dll (5.1.2600.5755) (-exported- Symbols Loaded)
Linked PDB Filename : advapi32.pdb
File Version : 5.1.2600.5755 (xpsp_sp3_gdr.090206-1234)
Company Name : Microsoft Corporation
Product Name : Syst�me d'exploitation Microsoft� Windows�
Product Version : 5.1.2600.5755
ModLoad: 77e50000 00093000 C:\WINDOWS\system32\RPCRT4.dll (5.1.2600.6022) (-exported- Symbols Loaded)
Linked PDB Filename : rpcrt4.pdb
File Version : 5.1.2600.6022 (xpsp_sp3_gdr.100813-1643)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 5.1.2600.6022
ModLoad: 77fc0000 00011000 C:\WINDOWS\system32\Secur32.dll (5.1.2600.5834) (-exported- Symbols Loaded)
Linked PDB Filename : secur32.pdb
File Version : 5.1.2600.5834 (xpsp_sp3_gdr.090624-1305)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 5.1.2600.5834
ModLoad: 7e390000 00091000 C:\WINDOWS\system32\USER32.dll (5.1.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : user32.pdb
File Version : 5.1.2600.5512 (xpsp.080413-2105)
Company Name : Microsoft Corporation
Product Name : Syst�me d'exploitation Microsoft� Windows�
Product Version : 5.1.2600.5512
ModLoad: 77ef0000 00049000 C:\WINDOWS\system32\GDI32.dll (5.1.2600.5698) (-exported- Symbols Loaded)
Linked PDB Filename : gdi32.pdb
File Version : 5.1.2600.5698 (xpsp_sp3_gdr.081022-1932)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 5.1.2600.5698
ModLoad: 78520000 000a3000 C:\WINDOWS\WinSxS\x86_Microsoft.VC90.CRT_1fc8b3b9a1e18e3b_9.0.30729.4148_x-ww_d495ac4e\MSVCR90.dll (9.0.30729.4148) (-exported- Symbols Loaded)
Linked PDB Filename : msvcr90.i386.pdb
File Version : 9.00.30729.4148
Company Name : Microsoft Corporation
Product Name : Microsoft� Visual Studio� 2008
Product Version : 9.00.30729.4148
ModLoad: 76320000 0001d000 C:\WINDOWS\system32\IMM32.DLL (5.1.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : imm32.pdb
File Version : 5.1.2600.5512 (xpsp.080413-2105)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 5.1.2600.5512
ModLoad: 02560000 006fb000 C:\Program Files\ATI Stream\bin\x86\atiocl.dll (1.1.0.1) (-exported- Symbols Loaded)
Linked PDB Filename :
File Version : 1, 1, 0, 1
Company Name : Advanced Micro Devices Inc.
Product Name : OpenCL 1.1
Product Version : 1, 1, 0, 1
ModLoad: 5d3f0000 000a1000 C:\WINDOWS\system32\dbghelp.dll (5.1.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : dbghelp.pdb
File Version : 5.1.2600.5512 (xpsp.080413-2105)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 5.1.2600.5512
ModLoad: 77be0000 00058000 C:\WINDOWS\system32\msvcrt.dll (7.0.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : msvcrt.pdb
File Version : 7.0.2600.5512 (xpsp.080413-2111)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 7.0.2600.5512
ModLoad: 77bd0000 00008000 C:\WINDOWS\system32\VERSION.dll (5.1.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : version.pdb
File Version : 5.1.2600.5512 (xpsp.080413-2105)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 5.1.2600.5512
ModLoad: 69000000 00469000 C:\WINDOWS\system32\aticaldd.dll (6.14.10.838) (-exported- Symbols Loaded)
Linked PDB Filename : c:\workarea\8.78\drivers\cal\drivers\src\ddi\lib\build\2k\B_rel\aticaldd.pdb
File Version : 6.14.10.838
Company Name : Advanced Micro Devices Inc.
Product Name : ATI CAL DD
Product Version : 6.14.10.838
ModLoad: 5f070000 000cc000 C:\WINDOWS\system32\OPENGL32.dll (5.1.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : opengl32.pdb
File Version : 5.1.2600.5512 (xpsp.080413-0845)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 5.1.2600.5512
ModLoad: 6cef0000 00021000 C:\WINDOWS\system32\GLU32.dll (5.1.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : glu32.pdb
File Version : 5.1.2600.5512 (xpsp.080413-0845)
Company Name : Microsoft Corporation
Product Name : Syst�me d'exploitation Microsoft� Windows�
Product Version : 5.1.2600.5512
ModLoad: 736b0000 0004b000 C:\WINDOWS\system32\DDRAW.dll (5.3.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : ddraw.pdb
File Version : 5.03.2600.5512 (xpsp.080413-0845)
Company Name : Microsoft Corporation
Product Name : Syst�me d'exploitation Microsoft� Windows�
Product Version : 5.03.2600.5512
ModLoad: 73b10000 00006000 C:\WINDOWS\system32\DCIMAN32.dll (5.1.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : dciman32.pdb
File Version : 5.1.2600.5512 (xpsp.080413-2105)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 5.1.2600.5512
ModLoad: 05e00000 0002f000 C:\WINDOWS\system32\atiadlxx.dll (6.14.10.1054) (-exported- Symbols Loaded)
Linked PDB Filename : c:\workarea\8.78\drivers\adl\build\xp\B_rel\atiadlxx.pdb
File Version : 6.14.10.1054
Company Name : Advanced Micro Devices, Inc.
Product Name : ADL Component
Product Version : 6.14.10.1054
ModLoad: 778e0000 000f8000 C:\WINDOWS\system32\SETUPAPI.dll (5.1.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : setupapi.pdb
File Version : 5.1.2600.5512 (xpsp.080413-2111)
Company Name : Microsoft Corporation
Product Name : Syst�me d'exploitation Microsoft� Windows�
Product Version : 5.1.2600.5512
ModLoad: 78130000 0009b000 C:\WINDOWS\WinSxS\x86_Microsoft.VC80.CRT_1fc8b3b9a1e18e3b_8.0.50727.4053_x-ww_e6967989\MSVCR80.dll (8.0.50727.4053) (-exported- Symbols Loaded)
Linked PDB Filename : msvcr80.i386.pdb
File Version : 8.00.50727.4053
Company Name : Microsoft Corporation
Product Name : Microsoft� Visual Studio� 2005
Product Version : 8.00.50727.4053
ModLoad: 76be0000 0002e000 C:\WINDOWS\system32\WINTRUST.dll (5.131.2600.5922) (-exported- Symbols Loaded)
Linked PDB Filename : wintrust.pdb
File Version : 5.131.2600.5922 (xpsp_sp3_gdr.091223-1907)
Company Name : Microsoft Corporation
Product Name : Syst�me d'exploitation Microsoft� Windows�
Product Version : 5.131.2600.5922
ModLoad: 779e0000 00097000 C:\WINDOWS\system32\CRYPT32.dll (5.131.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : crypt32.pdb
File Version : 5.131.2600.5512 (xpsp.080413-2113)
Company Name : Microsoft Corporation
Product Name : Syst�me d'exploitation Microsoft� Windows�
Product Version : 5.131.2600.5512
ModLoad: 77a80000 00012000 C:\WINDOWS\system32\MSASN1.dll (5.1.2600.5875) (-exported- Symbols Loaded)
Linked PDB Filename : msasn1.pdb
File Version : 5.1.2600.5875 (xpsp_sp3_gdr.090904-1413)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 5.1.2600.5875
ModLoad: 76c40000 00028000 C:\WINDOWS\system32\IMAGEHLP.dll (5.1.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : imagehlp.pdb
File Version : 5.1.2600.5512 (xpsp.080413-2105)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 5.1.2600.5512
ModLoad: 77b50000 00022000 C:\WINDOWS\system32\Apphelp.dll (5.1.2600.5512) (-exported- Symbols Loaded)
Linked PDB Filename : apphelp.pdb
File Version : 5.1.2600.5512 (xpsp.080413-2105)
Company Name : Microsoft Corporation
Product Name : Microsoft� Windows� Operating System
Product Version : 5.1.2600.5512
ModLoad: 07d80000 00115000 F:\BOINC\dbghelp.dll (6.8.4.0) (-exported- Symbols Loaded)
Linked PDB Filename : dbghelp.pdb
File Version : 6.8.0004.0 (debuggers(dbg).070515-1751)
Company Name : Microsoft Corporation
Product Name : Debugging Tools for Windows(R)
Product Version : 6.8.0004.0
ModLoad: 06cc0000 00048000 F:\BOINC\symsrv.dll (6.8.4.0) (-exported- Symbols Loaded)
Linked PDB Filename : symsrv.pdb
File Version : 6.8.0004.0 (debuggers(dbg).070515-1751)
Company Name : Microsoft Corporation
Product Name : Debugging Tools for Windows(R)
Product Version : 6.8.0004.0
ModLoad: 06d10000 0003b000 F:\BOINC\srcsrv.dll (6.8.4.0) (-exported- Symbols Loaded)
Linked PDB Filename : srcsrv.pdb
File Version : 6.8.0004.0 (debuggers(dbg).070515-1751)
Company Name : Microsoft Corporation
Product Name : Debugging Tools for Windows(R)
Product Version : 6.8.0004.0
*** Dump of the Process Statistics: ***
- I/O Operations Counters -
Read: 35, Write: 0, Other 516
- I/O Transfers Counters -
Read: 0, Write: 13583, Other 0
- Paged Pool Usage -
QuotaPagedPoolUsage: 44064, QuotaPeakPagedPoolUsage: 45636
QuotaNonPagedPoolUsage: 4384, QuotaPeakNonPagedPoolUsage: 4568
- Virtual Memory Usage -
VirtualSize: 109805568, PeakVirtualSize: 110333952
- Pagefile Usage -
PagefileUsage: 55939072, PeakPagefileUsage: 63606784
- Working Set Size -
WorkingSetSize: 58585088, PeakWorkingSetSize: 58585088, PageFaultCount: 20860
*** Dump of thread ID 3444 (state: Waiting): ***
- Information -
Status: Wait Reason: UserRequest, , Kernel Time: 0.000000, User Time: 0.000000, Wait Time: 105695168.000000
- Unhandled Exception Record -
Reason: Breakpoint Encountered (0x80000003) at address 0x7C91120E
- Registers -
eax=00000000 ebx=00000000 ecx=00000001 edx=0000613c esi=625f6d68 edi=735c6c61
eip=7c91120e esp=00b1fb90 ebp=00b1ff9c
cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00000246
- Callstack -
ChildEBP RetAddr Args to Child
00b1ff9c 0041e119 00000000 00b1ffb4 00b1ffb4 0041e079 ntdll!DbgBreakPoint+0x0
00b1ffac 0041e079 00b1ffec 7c80b729 00000000 735c6c61 primegrid_tpsieve_1.35_windows_!+0x0
00b1ffb4 7c80b729 00000000 735c6c61 625f6d68 00000000 primegrid_tpsieve_1.35_windows_!+0x0
00b1ffec 00000000 0041e060 00000000 00000000 017b0000 kernel32!GetModuleFileNameA+0x0
*** Dump of thread ID 2192 (state: Waiting): ***
- Information -
Status: Wait Reason: UserRequest, , Kernel Time: 625000.000000, User Time: 114062496.000000, Wait Time: 105695168.000000
- Registers -
eax=00000000 ebx=00000001 ecx=712f401b edx=00000000 esi=00000094 edi=00000000
eip=7c91e514 esp=0012fd70 ebp=0012fdd4
cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00000246
- Callstack -
ChildEBP RetAddr Args to Child
0012fdd4 7c802542 00000094 ffffffff 00000000 02dd44f8 ntdll!KiFastSystemCallRet+0x0
0012fde8 025abc4f 00000094 ffffffff 02dffde4 02dffde4 kernel32!WaitForSingleObject+0x0
0012fdfc 025abaf7 02dffde4 02dffdd8 02dd6290 00000001 atiocl!clGetSamplerInfo+0x0
0012fe18 025a606f 00000000 00434b34 02dffde4 0012fe50 atiocl!clGetSamplerInfo+0x0
0012fe34 02566fea 06bf2060 00002800 00002800 000ab9d6 atiocl!clGetSamplerInfo+0x0
00000000 00000000 00000000 00000000 00000000 00000000 atiocl!clWaitForEvents+0x0
*** Dump of thread ID 5060 (state: Waiting): ***
- Information -
Status: Wait Reason: UserRequest, , Kernel Time: 201718752.000000, User Time: 238593744.000000, Wait Time: 105695168.000000
- Registers -
eax=00000001 ebx=00000038 ecx=0603f904 edx=7c91e514 esi=00000020 edi=00007306
eip=7c91e514 esp=0603f8d4 ebp=0603fb48
cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00000202
- Callstack -
ChildEBP RetAddr Args to Child
0603fb48 69119980 d601263e 00007306 00000038 0603fbb8 ntdll!KiFastSystemCallRet+0x0
0603fc04 6910d228 05b62c70 00000000 00000000 00000400 aticaldd!calddiGetExport+0x0
00000000 00000000 00000000 00000000 00000000 00000000 aticaldd!calddiGetExport+0x0
*** Debug Message Dump ****
*** Foreground Window Data ***
Window Name :
Window Class :
Window Process ID: 0
Window Thread ID : 0
Exiting...
</stderr_txt>
]]>
I'm running the official application that just started to get work, but it used to work fine with the app_info version ...
Any ideas ? |
|
|
|
Third WU ended in the same state ... I'm going to suspend this project until I hear some news ... |
|
|
Ken_g6 Volunteer developer
 Send message
Joined: 4 Jul 06 Posts: 915 ID: 3110 Credit: 183,164,814 RAC: 0
                        
|
I'm not sure if this will help, but I need to ask it sometime. Have you tried installing the .NET 3.5 runtime?
____________
|
|
|
|
Same problem on mine, And yes .net 3.5 is installed.
everything works with app_info
Steve
what about .net version 4 - anyone running that? |
|
|
Ken_g6 Volunteer developer
 Send message
Joined: 4 Jul 06 Posts: 915 ID: 3110 Credit: 183,164,814 RAC: 0
                        
|
No, .NET 4 wouldn't matter. .NET 3.5 only matters because I compiled the client with VC++ Express 2008. I've since learned that I might be able to compile it statically; but I'd need a fresh install of Windows on that machine first.
In any case, this leaves "Maximum elapsed time exceeded". Which probably means the maximum elapsed time has been set too low.
____________
|
|
|
|
Same problem on mine, And yes .net 3.5 is installed.
everything works with app_info
Steve
what about .net version 4 - anyone running that?
Same here, last update to the 3.5 .net framework was on october 15th, and the application works with the app_info I used before ...
That's probably a project setting issue.
Also I'm wondering : is my 5770 supposed to be faster than a 9800 GTX+ ? Because I'm currently seeing similar times (30 minutes per WU) on both, and on other projects that have both ATI and NV applications, the ATI are usually way faster ... |
|
|
Scott Brown Volunteer moderator Project administrator Volunteer tester Project scientist
 Send message
Joined: 17 Oct 05 Posts: 2165 ID: 1178 Credit: 8,777,295,508 RAC: 0
                                     
|
Ran some fine earlier, but all mine after Rytis fixed the CPU load problem (where the app ran at more than .5 CPUs making a dual GPU machine run 1 CPU idle) I get the exceeded maximum runtime error. Looks like the fix broke something else...
____________
141941*2^4299438-1 is prime!
|
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2649 ID: 1 Credit: 26,363,112 RAC: 0
                    
|
Limit has been changed and will affect new WUs (we have to work through the current batch before).
____________
|
|
|
|
Great news, thanks ! :) |
|
|
|
3 so far for me ending in maximum time exceeded. How long till the old wu's are done ( estimated ) ? Thanks, Jack |
|
|
|
I'm still downloading wu with time extimates completely wrong... 4 minutes for boinc manager while really it takes one order of magnitude more (40 minutes) so the lock (boinc exit status: -177) ...
Maybe the flops counter is still wrong?? |
|
|
|
I think we'll need to work through heaps upon heaps of resends of resends of resends... It would be faster and simpler just to release a new app version with parameters set to right values.
I'm wasting ~20 min for each WU that starts on my HD4870 and bombs out @ ~75% done. I hope we'll get at least some kind of compensation for all that wasted time and energy.
BR,
____________
|
|
|
|
I have a ti HD3870 and use it mainly for MilkyWay@Home, as it supports double precision and can crunch a WU MilkyWAy faster than my GTX260.
For some reason PrimeGrid can not see the device and all my WUs for PrimeGrid end up in error. A 3870 has 320 Unified Shaders, but they do not show up in CPU-Z. Main problem with the 3870 seems to be that it does not support OpenCL.
____________
|
|
|
|
I have a ti HD3870 and use it mainly for MilkyWay@Home, as it supports double precision and can crunch a WU MilkyWAy faster than my GTX260.
For some reason PrimeGrid can not see the device and all my WUs for PrimeGrid end up in error. A 3870 has 320 Unified Shaders, but they do not show up in CPU-Z. Main problem with the 3870 seems to be that it does not support OpenCL.
OpenCL is supported onyl from HD4xxx.
Also, double precision here doesn't matter anything. |
|
|
|
I think we'll need to work through heaps upon heaps of resends of resends of resends... It would be faster and simpler just to release a new app version with parameters set to right values.
I'm wasting ~20 min for each WU that starts on my HD4870 and bombs out @ ~75% done. I hope we'll get at least some kind of compensation for all that wasted time and energy.
BR,
I still think that they do not have fixed the problem. Also, it's not an app problem, but a value problem: they estimated too few flops required for these wu and so boinc manager kills them when they last too long. It's a simple parameter server side to modify, but i'm not sure they're tackling the problem correctly |
|
|
|
OpenCL is supported onyl from HD4xxx.
Also, double precision here doesn't matter anything.
I know OpenCL is only fully supported from the HD4xxx series. I mention the double precision and MilkyWay@Home to prove that a HD3870 is a quite capable cruncher, and that a CAL+/Stream application could get my card to work for PrimeGrid as well. Even other Ati cards might benefit from a CAL+/Stream application.
If support of OpenCL is a prerequisite, a HD3870 should not get WUs.
____________
|
|
|
|
OpenCL is supported onyl from HD4xxx.
Also, double precision here doesn't matter anything.
I know OpenCL is only fully supported from the HD4xxx series. I mention the double precision and MilkyWay@Home to prove that a HD3870 is a quite capable cruncher, and that a CAL+/Stream application could get my card to work for PrimeGrid as well. Even other Ati cards might benefit from a CAL+/Stream application.
If support of OpenCL is a prerequisite, a HD3870 should not get WUs.
unfortunately, OpenCL has some strict requirements on memory structure and so it's not portable on old architectures
OpenCL was choosen and not Brook+ because OpenCL is "future-proof", while Brook+ has been abandoned.
CAL, on the other way, is the subsystem for both OpenCL and Brook+ but is much complex to deal with directly! And maybe they're using some CAL functions that do not exist on older HD 3xxx
Stream is the name of the software SDK that ATi (AMD) releases to developer |
|
|
|
I'm also surprised by the performances of my 5770 vs 9800 GTX+ : they both need about 30 minutes to complete thier PPS WUs ...
Is this application unable to use all SPs in ATI chips ? |
|
|
|
I'm also surprised by the performances of my 5770 vs 9800 GTX+ : they both need about 30 minutes to complete thier PPS WUs ...
Is this application unable to use all SPs in ATI chips ?
simply put, the cuda version has been developed more. The people behind the ati version have left. Let's hope someone will bring on the development. In the meantime, this app is better than nothing! |
|
|
Scott Brown Volunteer moderator Project administrator Volunteer tester Project scientist
 Send message
Joined: 17 Oct 05 Posts: 2165 ID: 1178 Credit: 8,777,295,508 RAC: 0
                                     
|
I'm also surprised by the performances of my 5770 vs 9800 GTX+ : they both need about 30 minutes to complete thier PPS WUs ...
Is this application unable to use all SPs in ATI chips ?
simply put, the cuda version has been developed more. The people behind the ati version have left. Let's hope someone will bring on the development. In the meantime, this app is better than nothing!
No exactly...the ATI app is OpenCL. If the NVidia cards only used OpenCL app versions, then a similar slowdown would occur. Conversely, if the ATI app used the Brook/CAL (i.e., the ATI native language or equivalent to CUDA), then the ATI cards would be faster. They don't, because the developer of the app could not do so (which is not surprising...any app at all is impressive since he does not even have an ATI GPU!).
____________
141941*2^4299438-1 is prime!
|
|
|
|
I'm also surprised by the performances of my 5770 vs 9800 GTX+ : they both need about 30 minutes to complete thier PPS WUs ...
Is this application unable to use all SPs in ATI chips ?
simply put, the cuda version has been developed more. The people behind the ati version have left. Let's hope someone will bring on the development. In the meantime, this app is better than nothing!
No exactly...the ATI app is OpenCL. If the NVidia cards only used OpenCL app versions, then a similar slowdown would occur. Conversely, if the ATI app used the Brook/CAL (i.e., the ATI native language or equivalent to CUDA), then the ATI cards would be faster. They don't, because the developer of the app could not do so (which is not surprising...any app at all is impressive since he does not even have an ATI GPU!).
this is absolutely wrong, I'm sorry if i'll be a bit rude.
OpenCL is not defective by design, in fact is the ONLY language recommended by ATi to write GPU software! Brook+ is an older language based on Brook that has been discontinued and is not updated anymore. CAL is the layer "under the hood", really very difficult to use (only simple algorithm like those used in collatz or milkyway@home can be deployed with this language).
The problem here is that the ATi app, written in OpenCL, is not developed extensively as the nvidia one, written in CUDA.
You can reach similar performances on nvidia writing a good OpenCL app, but unfortunately the same app would not be good performing on ATi, because you have to write custom kernels per brand if you want to get maximum performance. This is the problem that we have here: OpenCL kernels used in ATi app are tailored to the characteristics of nvidia gpus, because they come from the porting of the CUDA app.
I hope I'm clear... |
|
|
Vato Volunteer tester
 Send message
Joined: 2 Feb 08 Posts: 785 ID: 18447 Credit: 263,436,450 RAC: 0
                     
|
Actually, Scott is much closer to the truth in this particular situation, though there is some merit in saying that the original GPU code was for CUDA.
____________
|
|
|
Scott Brown Volunteer moderator Project administrator Volunteer tester Project scientist
 Send message
Joined: 17 Oct 05 Posts: 2165 ID: 1178 Credit: 8,777,295,508 RAC: 0
                                     
|
this is absolutely wrong, I'm sorry if i'll be a bit rude.
OpenCL is not defective by design, in fact is the ONLY language recommended by ATi to write GPU software! Brook+ is an older language based on Brook that has been discontinued and is not updated anymore. CAL is the layer "under the hood", really very difficult to use (only simple algorithm like those used in collatz or milkyway@home can be deployed with this language).
The problem here is that the ATi app, written in OpenCL, is not developed extensively as the nvidia one, written in CUDA.
You can reach similar performances on nvidia writing a good OpenCL app, but unfortunately the same app would not be good performing on ATi, because you have to write custom kernels per brand if you want to get maximum performance. This is the problem that we have here: OpenCL kernels used in ATi app are tailored to the characteristics of nvidia gpus, because they come from the porting of the CUDA app.
I hope I'm clear...
You are not rude since I do not see an insult or anything of that ilk in your response.
As for your belief that OpenCL apps can be as efficient as as CUDA or CAL, theoretically you are correct. However, the reality is that one would have to code OpenCL with perfect efficiency to match CAL. That is because OpenCL is built on top of CAL (see here for example) for ATI (and on top of CUDA for NVidia). It is not perfectly analogous to the difference between C and assembly, but the analogy is useful. That is, though C code can be written very well, it is not as efficient as the raw assembly code (and similarly, the latter--like CAL as you note--is more difficult to write).
____________
141941*2^4299438-1 is prime!
|
|
|
|
12/12/2010 2:22:24 PM PrimeGrid Aborting task pps_sr2sieve_5119963_0: exceeded elapsed time limit 27057.583038
... |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2649 ID: 1 Credit: 26,363,112 RAC: 0
                    
|
12/12/2010 2:22:24 PM PrimeGrid Aborting task pps_sr2sieve_5119963_0: exceeded elapsed time limit 27057.583038
...
Aurimai, dirbam prie to - naujai sukurtos užduotys turėtų nebeturėti šios problemos, tiesiog užtruks šiek tiek, kol iš serverio išsivalys senosios.
Malonu matyti lietuvius :)
____________
|
|
|
|
Ryti, daug maloniau matyti lietuvius kitoje puseje - Tavo atliktas darbas daug naudingesnis nei aukojanciuju kompiuteriu laika.
Tik viena pastaba:
http://www.primegrid.com/workunit.php?wuid=146115678 - rodo, kad jis sukurtas 9 Dec 2010 16:35:13 UTC
Tuo tarpu jau gruodzio 6 diena rasei:
Limit has been changed and will affect new WUs
As cia nelabai ka nutuokiu, taciau tai kelia itarima :) |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2649 ID: 1 Credit: 26,363,112 RAC: 0
                    
|
Iki 6 dienos lūžo absoliuÄiai visos užduotys, tada buvo pakeista ir daugeliui vartotojų susitvarkÄ—, taÄiau daliai vistiek iÅ¡liko problema. Kas bjauriausia, kad NVidia turÄ—tojams viskas tiesiog veikia, o ATI - nesugaudau... Vistik vakar padariau dar vienÄ… pakeitimÄ…, todÄ—l tikiuosi, kad dabar jau bus viskas gerai. AiÅ¡ku, vienintelis bÅ«das yra tiesiog pabandyti - pats neturiu ATI plokÅ¡tÄ—s...
____________
|
|
|
|
Sugavau http://www.primegrid.com/workunit.php?wuid=146364436
11 Dec 2010 19:46:37 UTC
Tikiuosi, jis buvo pagamintas jau po pakeitimo. Tokiu atveju pranesiu apie rezultatus.
Beje, idomus pastebejimas. Kai paleidi Collatz ant GPU, jis labai minimaliai trukdo darbui kompiuteriu, o nuo DNETC ir Prime, net klaviatura rasant raides atsilieka :) |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2649 ID: 1 Credit: 26,363,112 RAC: 0
                    
|
Šitas turėtų būti jau geras (t.y. labai tikiuosi).
DÄ—l Collatz - jie mažiau efektyviai iÅ¡naudoja GPU, todÄ—l lieka laiko Windowsams pieÅ¡ti GUI. Na o PG duoda daug didesnÄ™ apkrovÄ… :) AÅ¡ pats CUDA užduotis leidžiu tik tada, kai nenaudoju kompiuterio, nes jauÄiasi stabdymas, ypaÄ Å¾iÅ«rint video. Galima bÅ«tų dirbtinai lÄ—tinti skaiÄiavimus, bet Äia vÄ—lgi bjauru, nes skirtingose plokÅ¡tÄ—se reikia tai daryti skirtingai...
____________
|
|
|
|
12/12/2010 5:49:20 PM PrimeGrid Aborting task pps_sr2sieve_5149793_0: exceeded elapsed time limit 9199.578233
Ups... Gal laika ne i ta puse pasukei? :) |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2649 ID: 1 Credit: 26,363,112 RAC: 0
                    
|
PanaÅ¡u į tai :) PakeiÄiau į kitÄ… pusÄ™.
____________
|
|
|
|
?
sry i dont speak your language but for know i got the first Wu's with errors too.
I dont crunch work that waist my time, so for the moment i switch over to whatever.
http://www.primegrid.com/workunit.php?wuid=146350418
http://www.primegrid.com/workunit.php?wuid=145648528
http://www.primegrid.com/workunit.php?wuid=146097102
http://www.primegrid.com/workunit.php?wuid=146370100
http://www.primegrid.com/workunit.php?wuid=146370095
http://www.primegrid.com/workunit.php?wuid=146370073
http://www.primegrid.com/workunit.php?wuid=145901072
http://www.primegrid.com/workunit.php?wuid=145858022
____________
Public Energy -Crunch da Power- |
|
|
|
Jo, pabandysiu sugauti nauja uzduoti, bet kol kas dar siulo senus.
Cia kaip supratau pavadinime pps_sr2sieve_5166359 kuo didesnis skaicius, tuo naujesni WU? |
|
|
|
I hope that's the way it works with the higher number being newer. The WU's don't error at the start, they run the full 2800-2900 seconds finish and abort. I aborted a bunch and am hoping the higher number WU's will complete. Will these be gone in time for the challenge? |
|
|
|
don't know what you changed, but now it seems to work, at least on my host, without app_info!!
(on my host, i manually modified duration_correction_factor back in the days so maybe this also helped if someone other is still having problems) |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2649 ID: 1 Credit: 26,363,112 RAC: 0
                    
|
I believe (or more like, hope) that it should work now :)
____________
|
|
|
|
looking not so good for me;
http://www.primegrid.com/results.php?hostid=173800&offset=0&show_names=0&state=4&appid=
Do i had to switch away? i dont want to crunch for nothing, thats not my fault.
____________
Public Energy -Crunch da Power- |
|
|
rroonnaalldd Volunteer developer Volunteer tester
 Send message
Joined: 3 Jul 09 Posts: 1213 ID: 42893 Credit: 34,634,263 RAC: 0
                 
|
looking not so good for me;
http://www.primegrid.com/results.php?hostid=173800&offset=0&show_names=0&state=4&appid=
Do i had to switch away? i dont want to crunch for nothing, thats not my fault.
Maybe your boinc client 6.12.8 causes this problems...
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
Not having the best of times with some work units, some seem to like NVIDIA more than my ATI...
http://www.primegrid.com/workunit.php?wuid=145385494
http://www.primegrid.com/workunit.php?wuid=145662009
http://www.primegrid.com/workunit.php?wuid=146350165
http://www.primegrid.com/workunit.php?wuid=146368463
http://www.primegrid.com/workunit.php?wuid=146368440
http://www.primegrid.com/workunit.php?wuid=146365471
You can see that my host is 86532
I'm not worried about the above failures but I hope the problems go away.
Peter
____________
35 x 2^3587843+1 is prime! |
|
|
|
Ah Ronald bitte, das ist doch Unfug.
Auf mehreren anderen Rechner hat diese Wu auch Probleme und ist "wertlos".
Was mich ärgert ist, das ich die voll durchgerechnet habe und nicht einen credit dafür bekomme.
In der Zeit hätte ich 100 MW Wu's rechnen können.
____________
Public Energy -Crunch da Power- |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2649 ID: 1 Credit: 26,363,112 RAC: 0
                    
|
looking not so good for me;
http://www.primegrid.com/results.php?hostid=173800&offset=0&show_names=0&state=4&appid=
Do i had to switch away? i dont want to crunch for nothing, thats not my fault.
Maybe your boinc client 6.12.8 causes this problems...
I have checked your tasks on server - your client isn't uploading any files at all. Do you see any error messages in your client?
____________
|
|
|
|
Hi Rytis, no in my client are no errors pointed.
Curently MW seems to be down, so a few results can not be uploaded.
Thats all, the Primgrid Wu's are running and i see only in the Resultspage that they are invalid. So i had no other choice to abourt a hugh count of PG Wu's today.
Edit
There are new Wu that have a estimated runtime of ~5h, do they cause the problems?
____________
Public Energy -Crunch da Power- |
|
|
|
Mano ATI per beveik 9 valandas sukramte naujai sukurta uzduoti, taigi, lyg ir viskas gerai. Aciu. |
|
|
rroonnaalldd Volunteer developer Volunteer tester
 Send message
Joined: 3 Jul 09 Posts: 1213 ID: 42893 Credit: 34,634,263 RAC: 0
                 
|
I see some valid units on your host.
Are your ATIs running in crossfire-mode or as single-gpu?
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
The CUDA app is working as expected. However the ATI app is asking for .52 CPU/WU. That forces dual GPU boxes to waste a CPU core. Using an appropriate app_info.xml can bring the asked for CPU reservation down to .05 with no slowdown. Could you please adjust the default app to ask for less CPU so we can eliminate the need for the app_info.xml? Thanks!
|
|
|
|
Hello, I've got the same problem with errors in workunits.
For example this one: http://www.primegrid.com/workunit.php?wuid=146919177
As you can see, this WU generates errors on both ATI and NVIDIA boxes.
I've got two boxes with GF240 and one with ATI 5750, one GF240 running linux generates errors all the time, second GF240 running on Windows XP generates errors from time to time, and ATI 5750 running WIndows 7 generates errors all the time too.
What's wrong? Will it be fixed or should I just detach Primegrid and use some other project? |
|
|
|
I see some valid units on your host.
Are your ATIs running in crossfire-mode or as single-gpu?
I think you mean me, yes they run in CF mode, otherwise Boinc will only enable one card.
I can run different apps on this setup, PG + MW, PG + CC, CC + MW. Only Dnect reserve both cards for one Wu. Interessting, because it's a 5870 and a 5850. o-O
____________
Public Energy -Crunch da Power- |
|
|
|
Hello, I've got the same problem with errors in workunits.
For example this one: http://www.primegrid.com/workunit.php?wuid=146919177
As you can see, this WU generates errors on both ATI and NVIDIA boxes.
I've got two boxes with GF240 and one with ATI 5750, one GF240 running linux generates errors all the time, second GF240 running on Windows XP generates errors from time to time, and ATI 5750 running WIndows 7 generates errors all the time too.
What's wrong? Will it be fixed or should I just detach Primegrid and use some other project?
Your Boxes are hidden, no one can see anything.
____________
Public Energy -Crunch da Power- |
|
|
rroonnaalldd Volunteer developer Volunteer tester
 Send message
Joined: 3 Jul 09 Posts: 1213 ID: 42893 Credit: 34,634,263 RAC: 0
                 
|
Hidden or not, it would change nothing because: "Rytis" wrote: In preparation for the upcomming challenge, were only keeping results in the database for 4 hours after their validation. It will return back to normal once the challenge is done.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
The CUDA app is working as expected. However the ATI app is asking for .52 CPU/WU. That forces dual GPU boxes to waste a CPU core. Using an appropriate app_info.xml can bring the asked for CPU reservation down to .05 with no slowdown. Could you please adjust the default app to ask for less CPU so we can eliminate the need for the app_info.xml? Thanks!
Yes! Please, pretty please! It's bad enough, that ATI cards get beaten by nVidia here, so please, can I get at least one core back (there are 2 ATIs in this box).
BR
____________
|
|
|
|
For me too pls, two cores for free. ^^
____________
Public Energy -Crunch da Power- |
|
|