Deprecated: trim(): Passing null to parameter #1 ($string) of type string is deprecated in /disks/centurion/b/carolyn/b/home/boincadm/projects/beta/html/inc/boinc_db.inc on line 147
MultiBeam V7 for CPU and ATi GPUs, rev390

MultiBeam V7 for CPU and ATi GPUs, rev390

Message boards : SETI@home Enhanced : MultiBeam V7 for CPU and ATi GPUs, rev390
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 41553 - Posted: 24 Nov 2011, 20:24:12 UTC
Last modified: 24 Nov 2011, 20:24:31 UTC

Here the link to binaries: http://files.mail.ru/6D6Q8X
It's required update to fix data processing bug found in AKv8 code base.
NV version will follow.

To i5, i7 and later Intel processors owners:
Test if SSSE3 build is faster indeed than SSE2 one.
Our internal tests show that SSSE3 build is faster for Core2 family but slower vs SSE2 for latest Intel processor generations.

Archive includes special build for low-end ATi HD4xxx GPUs as well.
News about SETI opt app releases: https://twitter.com/Raistmer
ID: 41553 · Report as offensive
Father Ambrose
Volunteer tester

Send message
Joined: 1 May 07
Posts: 556
Credit: 6,470,846
RAC: 0
United Kingdom
Message 41557 - Posted: 25 Nov 2011, 9:26:35 UTC
Last modified: 25 Nov 2011, 9:28:51 UTC

Good morning Claggy Raistmer.

Archive includes special build for low-end ATi HD4xxx GPUs as well.



Have just downloaded and installed r390 it appears it does not like latest AMD ATI cat 11.11. Will have to downgrade driver later.

I am away to Freemans for heart check up any advice on possible drivers cat and sdk before I get back and start in earnest with latest r390.

Michael
ID: 41557 · Report as offensive
LadyL
Volunteer tester
Avatar

Send message
Joined: 14 Sep 11
Posts: 129
Credit: 379,042
RAC: 0
Message 41558 - Posted: 25 Nov 2011, 9:39:43 UTC
Last modified: 25 Nov 2011, 10:23:21 UTC

Quoting from upcoming main release notes:

We now offer the application in three different variants, depending on card:
LHD4K - for owners of HD4xxx GPUs with max workgroup size of 128 (see below)
plain - for other HD4xxx owners and those with HD5xxx and HD6xxx cards who
suffer from driver incompatibilities.
HD5 - for most if not all owners of HD5xxx GPUs and HD6xxx GPUs.

To check workgroupsize run CLinfo, downloadable (at present) from
{http://boinc.berkeley.edu/dl/clinfo.zip}
If you have problems running the HD5 flavour (anything from reduced performance
to BSOD), you should run the plain variant.

The recommended apps (r521 for Astropulse, r390 for MultiBeam) need OpenCL
support from the ATI driver - for recent driver versions, look for a driver
*with* APP or OpenCL listed. For compatible drivers see
http://support.amd.com/us/gpudownload/Pages/index.aspx
For cards, see
http://developer.amd.com/gpu/AMDAPPSDK/pages/DriverCompatibility.aspx
On most systems GPU-Z (see above) will tell you if your setup supports OpenCL.
Minimum required driver: Something that is OpenCL capable. You can check with
GPU-Z (see above).
Catalyst 10.7b with SDK
Catalyst 10.10 Accelerated Parallel Processing (APP) Technology Edition
The following drivers are known to be incompatible with the app:
11.10 and 11.11.

On some drivers the app shows increased CPU usage.
ID: 41558 · Report as offensive
Father Ambrose
Volunteer tester

Send message
Joined: 1 May 07
Posts: 556
Credit: 6,470,846
RAC: 0
United Kingdom
Message 41560 - Posted: 25 Nov 2011, 9:58:17 UTC

Copy of my post 41437 under r365.




What driver are you using as the problem started after removing cat 11,6

Present driver and card CAL ATI Radeon HD 4600 series (R730) (1024MB) driver: 1.4.1546

Have tried most still the same. Each cat was uninstalled and PC restarted and/or started cold. I would like to continue but run out of ideas

Cat 11.2 through 11.9

SDK v2.2 through 2.5

GPU_MAX_HEAP_SIZE 256 through 512.

period_iterations_num 1 through 10 also 20 through 100.{in 10s}

The only constant is at progress.

Reaches 0.006 locks up at (between 30 45 seconds blank screen and comes back with progress 0.010.

Continues to run with no further increase in progress
MB7_win_x86_SSE3_openCL_ATi_r365_LHD4K.exe continues to run at 25% [0ne core at 100%] even after suspension of WU and sometimes after closing BOINC.

Any other thoughts(also tried with and without flops, count, and stderr)

I am at present running off Raistmers WU.s. sorry for the errors.

{EDIT} Host is dual boot tried some WU's under Vista home prem.

Just possible starting point.(still waiting for Ambulance car)

MB7_win_x86_SSE3_openCL_ATi_LHD4K_r390.exe


Michael
ID: 41560 · Report as offensive
Profile Mike
Volunteer tester
Avatar

Send message
Joined: 16 Jun 05
Posts: 2531
Credit: 1,074,556
RAC: 0
Germany
Message 41561 - Posted: 25 Nov 2011, 10:23:00 UTC

Did you make sure you removed all left overs ?
Specially after 11.6 its very important to get rid of all *dll and *ocl files.

If everything fails you can try Cat 10.12.
Thats what i´m still using on my Vist 64 machine without any problems.

Mike

With each crime and every kindness we birth our future.
ID: 41561 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 29 May 06
Posts: 1037
Credit: 8,440,339
RAC: 0
United Kingdom
Message 41563 - Posted: 25 Nov 2011, 10:37:09 UTC - in response to Message 41561.  
Last modified: 25 Nov 2011, 10:46:32 UTC

Michael might need to delete any amdocl[32|64].dll as well, after he's uninstalled Cat 11.10

Kernel compiler problems in 11.10 - and cannot get rid of it

Claggy
ID: 41563 · Report as offensive
TRuEQ & TuVaLu
Volunteer tester
Avatar

Send message
Joined: 28 Jan 11
Posts: 619
Credit: 2,580,051
RAC: 0
Sweden
Message 41564 - Posted: 25 Nov 2011, 12:39:41 UTC

1'st task complete.

http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=10114904


period iterations 1 instances 1 -hp

Running driver 11.9 catalyst suite on ATI 5850 ; core 870Mhz mem 1100mhz

No lag at all and GPU load is 95%+ all the time.

Worked nice with first task.

:)

//TQ
TRuEQ & TuVaLu
ID: 41564 · Report as offensive
TRuEQ & TuVaLu
Volunteer tester
Avatar

Send message
Joined: 28 Jan 11
Posts: 619
Credit: 2,580,051
RAC: 0
Sweden
Message 41565 - Posted: 25 Nov 2011, 13:43:22 UTC

Several Tasks don now with same config as below except for gpu mem set to 1200

I notice a short period of lag(GUI freeze) when the 1 paket is finnished and a new pakets start.

I have no idea if it is the task the finnishes or the task that starts or both when they switch to the new tsak...

I can live with it though....


//TQ
TRuEQ & TuVaLu
ID: 41565 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 41566 - Posted: 25 Nov 2011, 18:04:51 UTC - in response to Message 41565.  

Several Tasks don now with same config as below except for gpu mem set to 1200

I notice a short period of lag(GUI freeze) when the 1 paket is finnished and a new pakets start.

I have no idea if it is the task the finnishes or the task that starts or both when they switch to the new tsak...

I can live with it though....


//TQ

removal of -hp switch will reduce lag
News about SETI opt app releases: https://twitter.com/Raistmer
ID: 41566 · Report as offensive
TRuEQ & TuVaLu
Volunteer tester
Avatar

Send message
Joined: 28 Jan 11
Posts: 619
Credit: 2,580,051
RAC: 0
Sweden
Message 41567 - Posted: 25 Nov 2011, 22:20:27 UTC - in response to Message 41566.  

Yep, the -hp switch made it dissapear.
I thought the run time should increase without the -hp
I can't really see any increase in run times so I'll keep running without -hp

//TQ
TRuEQ & TuVaLu
ID: 41567 · Report as offensive
Father Ambrose
Volunteer tester

Send message
Joined: 1 May 07
Posts: 556
Credit: 6,470,846
RAC: 0
United Kingdom
Message 41574 - Posted: 28 Nov 2011, 14:16:29 UTC - in response to Message 41560.  
Last modified: 28 Nov 2011, 14:17:06 UTC

Copy of my post 41437 under r365.




What driver are you using as the problem started after removing cat 11,6

Present driver and card CAL ATI Radeon HD 4600 series (R730) (1024MB) driver: 1.4.1546

Have tried most still the same. Each cat was uninstalled and PC restarted and/or started cold. I would like to continue but run out of ideas

Cat 11.2 through 11.9

SDK v2.2 through 2.5

GPU_MAX_HEAP_SIZE 256 through 512.

period_iterations_num 1 through 10 also 20 through 100.{in 10s}

The only constant is at progress.

Reaches 0.006 locks up at (between 30 45 seconds blank screen and comes back with progress 0.010.

Continues to run with no further increase in progress
MB7_win_x86_SSE3_openCL_ATi_r365_LHD4K.exe continues to run at 25% [0ne core at 100%] even after suspension of WU and sometimes after closing BOINC.

Any other thoughts(also tried with and without flops, count, and stderr)

I am at present running off Raistmers WU.s. sorry for the errors.

{EDIT} Host is dual boot tried some WU's under Vista home prem.

Just possible starting point.(still waiting for Ambulance car)

MB7_win_x86_SSE3_openCL_ATi_LHD4K_r390.exe


Michael


Good Afternoon,

Tried a few moor things as well as all those before.

Last cat run was your 10.12 with sdk v2.3 and sdk 2.4, beginning to think it could be the card. Has anyone got a 46xx too run?

CPU version of r390 running on both hosts no problems so far

HKEY_LOCAL_MACHINE\\SYSTEM\\CURRENTCONTROLSET\\CONTROL\\SESSION MANAGER\\ENVIRONMENT - GPU_MAX_HEAP_SIZE = 50% thro 100% 512_1024 M

HKEY_LOCAL_MACHINE\\SYSTEM\\CURRENTCONTROLSET\\CONTROL\\GraphicsDrivers – TdrDelay = 60 thro 128
HKEY_LOCAL_MACHINE\\SYSTEM\\CURRENTCONTROLSET\\CONTROL\\GraphicsDrivers - TdrDdiDelay = 128 thro 256

Reset time too blank screen down too 31 seconds.

No ATI problems on moo milkyway ect TV play back and games okay.

Michael.
ID: 41574 · Report as offensive
Urs Echternacht
Volunteer tester
Avatar

Send message
Joined: 18 Jan 06
Posts: 1038
Credit: 18,734,730
RAC: 0
Germany
Message 41575 - Posted: 28 Nov 2011, 14:29:56 UTC - in response to Message 41574.  
Last modified: 28 Nov 2011, 14:31:38 UTC


Last cat run was your 10.12 with sdk v2.3 and sdk 2.4, beginning to think it could be the card. Has anyone got a 46xx too run?

Sorry, not on windows. But works ok on linux with driver <= Cat 11.9

It is a driver problem and already reported to AMD.
_\|/_
U r s
ID: 41575 · Report as offensive
Father Ambrose
Volunteer tester

Send message
Joined: 1 May 07
Posts: 556
Credit: 6,470,846
RAC: 0
United Kingdom
Message 41577 - Posted: 28 Nov 2011, 14:46:40 UTC - in response to Message 41575.  

Many thanks Urs just thinking out loud.

Will continue with CPU.

Michael
ID: 41577 · Report as offensive
Fred J. Verster
Volunteer tester

Send message
Joined: 3 May 10
Posts: 88
Credit: 1,594,385
RAC: 0
Netherlands
Message 41599 - Posted: 9 Dec 2011, 1:01:20 UTC - in response to Message 41577.  

Many thanks Urs just thinking out loud.

Will continue with CPU.

Michael


Another possebillity, running the 2 5870 GPUs with BROOK version, if it's available?
And I could try cat.11.9, which is the only, I haven't tried, yet!
I'll be back and report.

ID: 41599 · Report as offensive
Exzypher
Volunteer tester

Send message
Joined: 27 Jul 11
Posts: 11
Credit: 204,026
RAC: 0
United States
Message 41610 - Posted: 10 Dec 2011, 19:05:52 UTC - in response to Message 41599.  
Last modified: 10 Dec 2011, 19:06:51 UTC

Have the work units changed? I am getting run times of ~8,000 seconds (r390) vesus ~4,000 seconds (r365).


HD 4870
AMD Catalyst 11.9
Windows 7 X 64


Also, I have no diea what all the hysetria is over the catalyst drivers, I have never once seen the 100% CPU bug. Mine is always at 0-1%.

For what it is worth, in the last week (before I started r390), I downloaded BOINC 7.0.2 as it was required for Einstein beta (Albert@home). WU's under 6.13.12 all errored out immediately:

<core_client_version>6.13.12</core_client_version>
<![CDATA[
<message>
couldn't start Can't link app version file: 2
</message>
]]>
ID: 41610 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 29 May 06
Posts: 1037
Credit: 8,440,339
RAC: 0
United Kingdom
Message 41611 - Posted: 10 Dec 2011, 19:23:27 UTC - in response to Message 41610.  

Have the work units changed? I am getting run times of ~8,000 seconds (r390) vesus ~4,000 seconds (r365).


HD 4870
AMD Catalyst 11.9
Windows 7 X 64


Also, I have no diea what all the hysetria is over the catalyst drivers, I have never once seen the 100% CPU bug. Mine is always at 0-1%.

For what it is worth, in the last week (before I started r390), I downloaded BOINC 7.0.2 as it was required for Einstein beta (Albert@home). WU's under 6.13.12 all errored out immediately:

<core_client_version>6.13.12</core_client_version>
<![CDATA[
<message>
couldn't start Can't link app version file: 2
</message>
]]>

No, the Workunits haven't changed, there is still a mixture of Angle Range Wu's here, which take a different amount of time to do,

I don't believe Cat 11.9 has the High CPU usage Bug, which is what you're running now, before that, in September you were running Cat 11.6,
and you seem to have missed running Cat 11.7 and 11.8 which did cause high CPU usage,

Claggy
ID: 41611 · Report as offensive
Exzypher
Volunteer tester

Send message
Joined: 27 Jul 11
Posts: 11
Credit: 204,026
RAC: 0
United States
Message 41612 - Posted: 11 Dec 2011, 2:33:21 UTC - in response to Message 41611.  

I don't believe Cat 11.9 has the High CPU usage Bug, which is what you're running now, before that, in September you were running Cat 11.6,
and you seem to have missed running Cat 11.7 and 11.8 which did cause high CPU usage,

Claggy



Ah, my mistake then. I could have sworn I read people saying >11.4 had the CPU bug. Well, in any case, thanks :)
ID: 41612 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 41614 - Posted: 11 Dec 2011, 20:25:01 UTC

High CPU usage bug was fixed only for single GPU config and immediately incorrect computations were introduced.
Look AMD forums for more info, there are quite hot discussion about quality of recent AMD drivers...
In short, I would recommend 11.2 still as single stable working driver with low CPU usage on my own config (Q9450 + HD6950 under Vista x86).

P.S.
For further reading:
http://forums.amd.com/forum/messageview.cfm?catid=390&threadid=156704&enterthread=y

http://forums.amd.com/forum/messageview.cfm?catid=390&threadid=153211&enterthread=y

http://forums.amd.com/forum/messageview.cfm?catid=390&threadid=156623&enterthread=y

http://forums.amd.com/forum/messageview.cfm?catid=390&threadid=156203&enterthread=y

http://forums.amd.com/forum/messageview.cfm?catid=390&threadid=155039&enterthread=y

And topic in progress:
http://forums.amd.com/forum/messageview.cfm?catid=390&threadid=155591&enterthread=y

For now there is no sense to discuss CPU usage cause with new driver app can't recive correct results at all.

News about SETI opt app releases: https://twitter.com/Raistmer
ID: 41614 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 41615 - Posted: 11 Dec 2011, 20:34:50 UTC - in response to Message 41610.  



Also, I have no diea what all the hysetria is over the catalyst drivers, I have never once seen the 100% CPU bug. Mine is always at 0-1%.


Name: ATI RV770
Vendor: Advanced Micro Devices, Inc.
Driver version: CAL 1.4.1546
Version: OpenCL 1.0 AMD-APP-SDK-v2.5 (767.1)

It's from stderr of one of your tasks.
Usually problems beging when driver has (VM) in its name. Your one has not.

Also, don't forget that even if one downloads same binary package for HD4xxx/HD5xxx /HD6xxx GPUs it absolutely doesn't mean that driver binaries used for these GPU families also the same.
All these families have diferent architectures. So, If for example HD6xxx + Cat 11.9 has problems and HD4xxx + Cat 11.9 doesn't - it doesn;t mean that all who use HD6xxx should forget about their issues as imaginary ones. It just means that HD4xxx drivers are good and HD6xxx ones are broken, nothing more.

News about SETI opt app releases: https://twitter.com/Raistmer
ID: 41615 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 41616 - Posted: 11 Dec 2011, 20:38:36 UTC - in response to Message 41599.  


Another possebillity, running the 2 5870 GPUs with BROOK version, if it's available?

I would not consider it as real alternative.
Brook-based app performance too bad vs OpenCL one so better to find way to run OpenCL on OpenCL-compatible device.
Moreover, all this has nothing with this particular thread, it's about MultiBeam app, not about AstroPulse app who has both Brook and MultiBeam versions.
Please, stay on topic, there is too much to search through it already...


News about SETI opt app releases: https://twitter.com/Raistmer
ID: 41616 · Report as offensive

Message boards : SETI@home Enhanced : MultiBeam V7 for CPU and ATi GPUs, rev390


 
©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.