Issues with ATI GPU Astropulse

Message boards : Number crunching : Issues with ATI GPU Astropulse
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
[AF>HFR>HOP!]SH4 Origon X

Send message
Joined: 10 Jun 03
Posts: 20
Credit: 66,986,761
RAC: 0
France
Message 1056458 - Posted: 16 Dec 2010, 0:20:53 UTC
Last modified: 16 Dec 2010, 0:24:59 UTC

Hello , i'm having some issues with ATI computing on Seti, every try on WU solves in errors, here is detailed
informations about my problem:

-Windows XP SP3 up to date
-BOINC 6.10.58
-ATI Stream SDK 2.2
-Catalyst 10.10
-Q6600
-HD4870 1go
-CPU client: AK V8 SSSE3x

Errors:
09/12/2010 16:32:04 SETI@home Starting ap_22ap10af_B4_P0_00121_20101209_17193.wu_0
09/12/2010 16:32:04 SETI@home Starting task ap_22ap10af_B4_P0_00121_20101209_17193.wu_0 using astropulse_v505 version 506
09/12/2010 16:32:05 SETI@home Task ap_22ap10af_B4_P0_00121_20101209_17193.wu_0 exited with zero status but no 'finished' file
09/12/2010 16:32:05 SETI@home If this happens repeatedly you may need to reset the project.
09/12/2010 16:32:05 SETI@home Restarting task ap_22ap10af_B4_P0_00121_20101209_17193.wu_0 using astropulse_v505 version 506
09/12/2010 16:32:06 SETI@home Task ap_22ap10af_B4_P0_00121_20101209_17193.wu_0 exited with zero status but no 'finished' file
09/12/2010 16:32:06 SETI@home If this happens repeatedly you may need to reset the project.
09/12/2010 16:32:06 SETI@home Restarting task ap_22ap10af_B4_P0_00121_20101209_17193.wu_0 using astropulse_v505 version 506
09/12/2010 16:32:07 SETI@home Task ap_22ap10af_B4_P0_00121_20101209_17193.wu_0 exited with zero status but no 'finished' file
09/12/2010 16:32:07 SETI@home If this happens repeatedly you may need to reset the project.
09/12/2010 16:32:07 SETI@home Restarting task ap_22ap10af_B4_P0_00121_20101209_17193.wu_0 using astropulse_v505 version 506
09/12/2010 16:32:09 SETI@home Task ap_22ap10af_B4_P0_00121_20101209_17193.wu_0 exited with zero status but no 'finished' file
09/12/2010 16:32:09 SETI@home If this happens repeatedly you may need to reset the project.
It goes like this for about 150 lines

And then:
09/12/2010 16:33:57 SETI@home Restarting task ap_22ap10af_B4_P0_00121_20101209_17193.wu_0 using astropulse_v505 version 506
09/12/2010 16:33:58 SETI@home Computation for task ap_22ap10af_B4_P0_00121_20101209_17193.wu_0 finished
09/12/2010 16:33:58 SETI@home Output file ap_22ap10af_B4_P0_00121_20101209_17193.wu_0_0 for task ap_22ap10af_B4_P0_00121_20101209_17193.wu_0 absent

WorkUnit end in error ...
Error in calculation
Client state calculation error
State at the exit : -226 (0xffffffffffffff1e)

The errors are the same even with Full OpenCL client (456) or OpenCL/Brook+ (r449) with or without overclocking.
The hybrid client CPU/GPU Brook+ (r453) is working fine by the way.

Thanks in advance o/
ID: 1056458 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1056530 - Posted: 16 Dec 2010, 5:44:41 UTC - in response to Message 1056458.  

Have a read of AstroPulse for ATI GPUs released, and see if setting some lower -ffa_block and -ffa_block_fetch parameter's help,

Claggy
ID: 1056530 · Report as offensive
[AF>HFR>HOP!]SH4 Origon X

Send message
Joined: 10 Jun 03
Posts: 20
Credit: 66,986,761
RAC: 0
France
Message 1056568 - Posted: 16 Dec 2010, 7:50:24 UTC

Thanks !

How am I supposed to do to add the variables ?

"<cmdline>-ffa_block_fetch 2048</cmdline>
<cmdline>-ffa_block 8192</cmdline>"

or

"<cmdline>-ffa_block_fetch 512 -ffa_block 2048</cmdline>"

or something else ? I'm not at ease with codes.
I tryed 1024/4096 and 512/2048 but the things are not going better.

ID: 1056568 · Report as offensive
Aker

Send message
Joined: 2 Nov 01
Posts: 24
Credit: 2,030,727
RAC: 0
United States
Message 1056571 - Posted: 16 Dec 2010, 8:19:11 UTC - in response to Message 1056568.  

Second way, all in one line works. If you check the results of one of the returned wus it will note that you've overidden the default.


The error in your reported tasks seems to be:
"Error: Building Program (clBuildProgram): -11
Internal error: Compilation init failed."

I'd try and see if changing driver vers does anything. Can't hurt ;)

http://sites.amd.com/us/game/downloads/Pages/radeon_xp-32.aspx

Uninstall Stream sdk and any drivers you have now and install the APP package.


ID: 1056571 · Report as offensive
[AF>HFR>HOP!]SH4 Origon X

Send message
Joined: 10 Jun 03
Posts: 20
Credit: 66,986,761
RAC: 0
France
Message 1056579 - Posted: 16 Dec 2010, 9:35:57 UTC - in response to Message 1056571.  

Done. APP Package 10.12 installed.
Still the same error with <cmdline>-ffa_block_fetch 1024 -ffa_block 4096</cmdline> or <cmdline>-ffa_block_fetch 512 -ffa_block 2048</cmdline>.
ID: 1056579 · Report as offensive
Aker

Send message
Joined: 2 Nov 01
Posts: 24
Credit: 2,030,727
RAC: 0
United States
Message 1056617 - Posted: 16 Dec 2010, 15:07:59 UTC - in response to Message 1056579.  

I don't think tweaking the settings with a compile error like that will help.

Does your app_info entry look like..

<app>
<name>astropulse_v505</name>
</app>
<file_info>
        <name>ap_5.06_win_x86_SSE2_OpenCL_ATI_r456.exe</name>
        <executable/>
    </file_info>
<file_info>
    <name>AstroPulse_Kernels.cl</name>
    <executable/>
</file_info>
    <app_version>
        <app_name>astropulse_v505</app_name>
        <version_num>506</version_num>
	  <flops>16000000000</flops>
<avg_ncpus>0.01</avg_ncpus>
<max_ncpus>0.01</max_ncpus>
                <plan_class>ati13ati</plan_class>
                <cmdline>-ffa_block 4096 -ffa_block_fetch 1024</cmdline>
          <coproc>
              <type>ATI</type>
        <count>1</count>
          </coproc>
        <file_ref>
            <file_name>ap_5.06_win_x86_SSE2_OpenCL_ATI_r456.exe</file_name>
            <main_program/>                           
        </file_ref>
<file_ref>
    <file_name>AstroPulse_Kernels.cl</file_name>
    <copy_file/>
</file_ref>
    </app_version>

ID: 1056617 · Report as offensive
[AF>HFR>HOP!]SH4 Origon X

Send message
Joined: 10 Jun 03
Posts: 20
Credit: 66,986,761
RAC: 0
France
Message 1056666 - Posted: 16 Dec 2010, 17:01:43 UTC - in response to Message 1056617.  
Last modified: 16 Dec 2010, 17:04:47 UTC

Mine had <flops>16000000000</flops> a little lower and with different number (like on lunatics page). The only difference, and I just tryed your App_Info... guess what ? same error :/


____________
ID: 1056666 · Report as offensive
Profile skildude
Avatar

Send message
Joined: 4 Oct 00
Posts: 9541
Credit: 50,759,529
RAC: 60
Yemen
Message 1056678 - Posted: 16 Dec 2010, 17:30:20 UTC - in response to Message 1056666.  
Last modified: 16 Dec 2010, 17:37:17 UTC

Just to clarify about installing the New drivers. Did you uninstall the old ones before installing the new drivers. The ATI/AP and MB apps are very sensitive to this. I found this out the hard way.


In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope
ID: 1056678 · Report as offensive
[AF>HFR>HOP!]SH4 Origon X

Send message
Joined: 10 Jun 03
Posts: 20
Credit: 66,986,761
RAC: 0
France
Message 1056683 - Posted: 16 Dec 2010, 17:36:44 UTC - in response to Message 1056678.  
Last modified: 16 Dec 2010, 17:37:34 UTC

I did complete uninstall, reboot + pass of driver cleaner + manual search of ATI folders + CCleaner pass and reboot to install 10.12.
And thanks again to all for your replies :)
ID: 1056683 · Report as offensive
Profile skildude
Avatar

Send message
Joined: 4 Oct 00
Posts: 9541
Credit: 50,759,529
RAC: 60
Yemen
Message 1056685 - Posted: 16 Dec 2010, 17:39:55 UTC - in response to Message 1056683.  
Last modified: 16 Dec 2010, 17:40:13 UTC


for some hosts with only HD4xxx GPU installed additional environment variable may be needed: GPU_MAX_HEAP_SIZE=256
To set it go My computer->Properties->Advanced->Environment variables (path can be slightly different in different windows versions). But before setting such variable run application and check stderr.
If you see line:
Global memory size:
with value >128MB - you don't need this environment variable
]


have you taken care of that


In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope
ID: 1056685 · Report as offensive
[AF>HFR>HOP!]SH4 Origon X

Send message
Joined: 10 Jun 03
Posts: 20
Credit: 66,986,761
RAC: 0
France
Message 1056687 - Posted: 16 Dec 2010, 17:47:12 UTC - in response to Message 1056685.  
Last modified: 16 Dec 2010, 17:54:35 UTC

I have two values in every sterr out:
Global memory size: 1073741824
Global memory size: 536870912

Should be good no ?
This problem is so weird i'm sure about nothing now xD
ID: 1056687 · Report as offensive
Profile skildude
Avatar

Send message
Joined: 4 Oct 00
Posts: 9541
Credit: 50,759,529
RAC: 60
Yemen
Message 1056701 - Posted: 16 Dec 2010, 18:25:02 UTC - in response to Message 1056687.  

Try setting the heap size to 256 and see if that works. It was another little thing that the 4XXX series tends to need to make it work properly


In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope
ID: 1056701 · Report as offensive
[AF>HFR>HOP!]SH4 Origon X

Send message
Joined: 10 Jun 03
Posts: 20
Credit: 66,986,761
RAC: 0
France
Message 1056710 - Posted: 16 Dec 2010, 19:31:34 UTC - in response to Message 1056701.  
Last modified: 16 Dec 2010, 19:36:36 UTC

If I get it good, I just have to add a new system variable named GPU_MAX_HEAP_SIZE with 256 as value ? Done, still same error :(
ID: 1056710 · Report as offensive
Profile skildude
Avatar

Send message
Joined: 4 Oct 00
Posts: 9541
Credit: 50,759,529
RAC: 60
Yemen
Message 1056732 - Posted: 16 Dec 2010, 20:39:55 UTC - in response to Message 1056710.  
Last modified: 16 Dec 2010, 20:52:08 UTC

have you restarted your PC after updating the drivers and the Heap max...
You might also give the Lunatics forums a try. Raistmer the genius that created the ATI AP GPU APP frequents there. He lives in Russia so don't expect him to be on all the time


In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope
ID: 1056732 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1056741 - Posted: 16 Dec 2010, 20:54:49 UTC

At least one link on stderr of failed task ?
Stderr (listed in result page) is the first place to look and this info still missed in this thread.
ID: 1056741 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1056746 - Posted: 16 Dec 2010, 21:03:51 UTC

Well...
"
Error: Building Program (clBuildProgram): -11
Internal error: Compilation init failed.
"
this error can't be healed with params tuning or memory env variable setting.

Last time I've seen such error there were old SDK DLLs left in app folder.
Make sure you have no ATI-related DLLs in SETI project folder.
Also make sure that slot with task contains CL file, same CL file should be in project folder too.
ID: 1056746 · Report as offensive
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 34258
Credit: 79,922,639
RAC: 80
Germany
Message 1056892 - Posted: 17 Dec 2010, 8:13:57 UTC
Last modified: 17 Dec 2010, 8:14:41 UTC

My 5850 was running all night.
When i woke up i see a message display driver restored in the task bar.

It also seems the card only crunched 5 out of 7 hours.

I crunched MW and collatz for 4 weeks without a problem.

Any ideas ?

Latest drivers installed.


With each crime and every kindness we birth our future.
ID: 1056892 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1056969 - Posted: 17 Dec 2010, 13:52:32 UTC - in response to Message 1056892.  

My 5850 was running all night.
When i woke up i see a message display driver restored in the task bar.

It also seems the card only crunched 5 out of 7 hours.

I crunched MW and collatz for 4 weeks without a problem.

Any ideas ?

Latest drivers installed.


Decrease params (fetch) values in app command line.
ID: 1056969 · Report as offensive
[AF>HFR>HOP!]SH4 Origon X

Send message
Joined: 10 Jun 03
Posts: 20
Credit: 66,986,761
RAC: 0
France
Message 1057023 - Posted: 17 Dec 2010, 16:11:29 UTC
Last modified: 17 Dec 2010, 16:27:42 UTC

> skildude: I did it for sure ;)

Hi Raistmer ! :D

First here is a complete sterr of the last error unit:
ap_24se10ab_B0_P1_00073_20101216_09714.wu_1

Second, no old or ati related DLL's in app folder (no DLL's at all when using r456), only AK v8 files in slots 0 to 3, and good files for your app in slot 4, no old DLL's here too.
AstroPulse_Kernels.cl is in app and slot 4 folders.
ID: 1057023 · Report as offensive
[AF>HFR>HOP!]SH4 Origon X

Send message
Joined: 10 Jun 03
Posts: 20
Credit: 66,986,761
RAC: 0
France
Message 1057109 - Posted: 17 Dec 2010, 18:19:40 UTC
Last modified: 17 Dec 2010, 18:35:45 UTC

I add a sterrout of an older unit when i was Cat 10.10 and SDK 2.2, maybe can help too.

ap_22ap10af_B3_P1_00253_20101209_03508.wu_1


Edit: I was wondering, I saw in the output that the client seems to be OpenCL 1.1, according to GPU-Z my GFX card is OpenCL 1.0 compatible, could it be the source of the problem ?
ID: 1057109 · Report as offensive
1 · 2 · Next

Message boards : Number crunching : Issues with ATI GPU Astropulse


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.