Call for testers - low-end ATi HD4xxx, part 2
Message boards :
Number crunching :
Call for testers - low-end ATi HD4xxx, part 2
Message board moderation
| Author | Message |
|---|---|
 Raistmer  Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 
 |
There is new build (thanks to Urs Echternacht who implemented corresponding oclFFT fixes) for low-end HD4xxx GPUs (with max workgroup size of 128 instead of 256). It's already proved to work on usual GPUs. If someone with corresponding (HD4xxx low-end GPU, OpenCL-supported, 128workitems per workgroup) hardware want to test it, please, PM me. ID: 1148524 ·  |
Floyd  Send message Joined: 19 May 11 Posts: 524 Credit: 1,870,625 RAC: 0 
|
There is new build (thanks to Urs Echternacht who implemented corresponding oclFFT fixes) for low-end HD4xxx GPUs (with max workgroup size of 128 instead of 256). I could have helped before I bought the 460SE... My MB only has 1 X16 slot and 2 PCI slots... Sorry ! ID: 1149171 · |
|
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 
|
I could have helped before I bought the 460SE... My MB only has 1 X16 slot and 2 PCI slots... [offtopic] You could fill up your mobo with GT430 or HD5450 PCI cards.. ;-) ('Message 1120861') [/offtopic] - Best regards! - Sutaru Tsureku, team seti.international founder. - Optimize your PC for higher RAC. - SETI@home needs your help. - 
ID: 1149384 · |
|
archae86 Send message Joined: 31 Aug 99 Posts: 909 Credit: 1,582,816 RAC: 0
|
If someone with corresponding (HD4xxx low-end GPU, OpenCL-supported, 128workitems per workgroup) hardware want to test it, please, PM me.To be on-topic, I offered this host as a test platform. The CPU side has been running Einstein for months, while the 4550 ATI GPU has been running Raistmer's Astropulse application. In offline testing it passed running Knabench tests, and it has been running production MultiBeam since 23:06 September 4 UTC. Only two of the new MB results have validated, but most of the others have a quorum partner with a history of several day reply time. Some salient host characteristics: Windows 7 64-bit E5620 CPU (basically a Nehalem on 32nm, overclocked at 3.4 GHz) with 12G RAM GIGABYTE GV-R455D3-512I Radeon HD 4550 512MB 64-bit DDR3 graphics card ID: 1149401 · |
|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121
|
Cause it proven to work on at last one target GPU I start wider public beta (i.e. use it on your own risk and don't forget to remove it if wrong results will be produced). Link for downloading: http://files.mail.ru/HJVMPG Corresponding app_info section resembles one for usual ATi MB release. App can be used both on SETI main and on SETI beta cause it's V7-compatible and V6 back compatible one. Don't forget to list link on host where you plan to use it in this thread. ID: 1149663 · |
|
archae86 Send message Joined: 31 Aug 99 Posts: 909 Credit: 1,582,816 RAC: 0
|
With the recent protracted period in which the scheduler declined to send work to anonymous platform ATI GPUs (at least I think that is what Claggy discovered and got DA to fix) I've not been running this much, but with the fix in place I got 80+ results to work on yesterday, and already have seen again a problem I think I saw before the shutdown. My description is vague, and I can't provide much evidence. This post is more an alert and request for comment than a useful bug report. The basic bad behavior is failure to make progress. It seems that it may be associated with restarts. The one I started last night was a non-shortie which had a good start --about 20% reported complete before I went to bed, with perhaps 1.5 hours elapsed. First thing this morning, I found this had apparently run all night, was reporting a total of about 8 hours of elapsed computation, but had a completion report in the very low 20s percent. I chose to abort it. Possibly as a result, the stderr listing is essentially empty of useful information save reporting that I aborted it. I think I saw something like this more than once before the Great Shortage. Aside from being a very low-end card (4550) my host has the preference setting "Suspend GPU work while computer is in use?" activated, and as this is my primary personal use computer and I currently allow GPU work to resume after one minute of inactivity there are sometimes very many restarts. This seemed to work OK for many months of running Raistmer's ATI Astropulse application. Possibly this Multibeam application is less robust for restarts, or maybe restarts have nothing to do with the problem. Is this sort of behavior a known issue? Is it possible that I'm seeing something for which I should try to capture more useful information than this vague bug report? If so, are there any suggestions of ways I could instrument the problem or otherwise provide a more useful report? One last thing--before the the Great Shortage I believe I saw one of these for which I chose to let it restart yet again, and it ran on to successful completion. If that was a correct observation, and yet it was the same problem, then the "hung up" state may not necessarily be persistent across restarts for an in-process WU. ID: 1159429 · |
|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121
|
Actually latest MB builds have "improved" shutdown protection (it bases on communication with BOINC about shutdown in general)... there are 2 not nessesary connected issues you reported: 1) app "hung" w/o making progress. 2) no stderr captured. What drivers do you use? (OpenCL 1.0 ATI-Stream-v2.3 (451) - what Cat version is it?) I found Cat 11.2 most stable for now for my own hardware. Besides of this recommendation I have too little info to suggest anything, try to get more statistics... And please, when you will see hang (w/o making progress) again, please, capture stderr.txt from corresponding slot directory by hands before aborting task. That way we get more info perhaps about what happens. ID: 1159431 · |
|
archae86 Send message Joined: 31 Aug 99 Posts: 909 Credit: 1,582,816 RAC: 0
|
What drivers do you use? (OpenCL 1.0 ATI-Stream-v2.3 (451) - what Cat version is it?)Catalyst Control Center reports Catalyst Version = 11.5 on my host I found Cat 11.2 most stable for now for my own hardware.I checked the driver availability page, and see they have a September 28 released version 11.9. Would it make sense for me to try that after (if) see a repetition of this behavior on the theory that other users may have the most recent release, or would it be more diagnostic for me to find and install 11.2? I do see that you posted on the Lunatics site some unfavorable performance results from 11.9 in the presence of a busy CPU. As I keep my CPUs fully busy with Einstein, I imagine I'd get a performance hit. I found a legacy drivers page for Windows XP-suitable Catalyst, but not yet found for Windows 7 64bit or Vista. when you will see hang (w/o making progress) again, please, capture stderr.txt from corresponding slot directory by hands before aborting task. That way we get more info perhaps about what happens. I shall do so. ID: 1159452 · |
|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121
|
YEs, new drivers have own issues, I'd better try downgrading to 11.2 first (or 11.4 as some reported this version as last one w/o CPU-consuming bugs). ID: 1159455 · |
Mike   Send message Joined: 17 Feb 01 Posts: 34257 Credit: 79,922,639 RAC: 80
|
ID: 1159460 · |
|
Floyd Send message Joined: 19 May 11 Posts: 524 Credit: 1,870,625 RAC: 0
|
Ok I have my old computer working , I put my old ati Video card in it , ati HD-4670 , that I had in my new computer till I switched it for the Nvidia GTX 460SE . This ati card is only 128 bit , so what programs do I need to install to get this computer crunching ok... Do I need Lunatics 38 for cpu ? ati ? Link to computer number. http://setiathome.berkeley.edu/show_host_detail.php?hostid=6193616 Computer number , Computer 6193616 
ID: 1159824 · |
|
Mike Send message Joined: 17 Feb 01 Posts: 34257 Credit: 79,922,639 RAC: 80
|
You can run the installer yes, but to get your ATI working you need to download this http://files.mail.ru/HJVMPG app. Copy those files into boinc data directory. Edit appinfo renaming Ati app entry. rename <name>MB_6.10_win_x86_SSE3_OpenCL_ATI_r177.exe</name> to <name>MB7_win_x86_SSE3_OpenCL_ATi_r365_LHD4K.exe</name> With each crime and every kindness we birth our future. ID: 1159832 · |
|
Floyd Send message Joined: 19 May 11 Posts: 524 Credit: 1,870,625 RAC: 0
|
You can run the installer yes, but to get your ATI working you need to download I downloaded a rar file , Just don't have way to open it on this computer. Can I get it as a exe file for windows ?
ID: 1159835 · |
Richard Haselgrove  Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 
 |
You can run the installer yes, but to get your ATI working you need to download RAR is a common, but proprietary, archive format for Windows. The standard tool, WinRAR, has to be paid for, and unlike the other standard compression format, ZIP, Windows can't extract the archived files itself. The best tool to use to get the files out of the archive is the free, open-source 7-zip ID: 1159837 · |
|
Floyd Send message Joined: 19 May 11 Posts: 524 Credit: 1,870,625 RAC: 0
|
You can run the installer yes, but to get your ATI working you need to download Ok I did this as best as I understood... Got error when I restarted Boinc... 10/7/2011 2:19:13 PM | | ATI GPU 0: ATI Radeon HD 4600 series (R730) (CAL version 1.4.1016, 1024MB, 512 GFLOPS peak) 10/7/2011 2:19:13 PM | SETI@home | Found app_info.xml; using anonymous platform 10/7/2011 2:19:13 PM | SETI@home | File referenced in app_info.xml does not exist: MB7_win_x86_SSE3_OpenCL_ATi_r365_LHD4K.exe 10/7/2011 2:19:13 PM | SETI@home | File referenced in app_info.xml does not exist: MultiBeam_Kernels_r177.cl 10/7/2011 2:19:13 PM | SETI@home | [error] State file error: missing application file MB_6.10_win_SSE3_ATI_r177.exe10/7/2011 2:19:13 PM | SETI@home | URL http://setiathome.berkeley.edu/; Computer ID 6193616; resource share 200 10/7/2011 2:19:13 PM | SETI@home | General prefs: from SETI@home (last modified 07-Oct-2011 12:36:15) 10/7/2011 2:19:13 PM | SETI@home | Computer location: home
ID: 1159894 · |
|
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4
|
You can run the installer yes, but to get your ATI working you need to download You're got to change each entry in the app_info for MB_6.10_win_SSE3_ATI_r177.exe to MB7_win_x86_SSE3_OpenCL_ATi_r365_LHD4K.exe and each entry for MultiBeam_Kernels_r177.cl to MultiBeam_Kernels_r365.cl Claggy ID: 1159902 · |
|
Floyd Send message Joined: 19 May 11 Posts: 524 Credit: 1,870,625 RAC: 0
|
You can run the installer yes, but to get your ATI working you need to download Ok... Maybe I got it right this time , No more notices from Boinc now... Hopefully no invalids either... ! ... :-) Sad that the est time to crunch is over 20 hours for a vlar ( all 6.03's )is about 22 hours... SAD ! and the one ati WU I had is gone now anyway...
ID: 1159910 · |
|
Floyd Send message Joined: 19 May 11 Posts: 524 Credit: 1,870,625 RAC: 0
|
Ok... Bad news... The four ati wu's all had computation errors , two was vlars and the others wasn't. So I still have a problem with this !
ID: 1159925 · |
|
Treasurer Send message Joined: 13 Dec 05 Posts: 109 Credit: 1,569,762 RAC: 0
|
Ok... Bad news... The four ati wu's all had computation errors , two was vlars and the others wasn't. I tried the app on this machine (HD4650). Same result: error after couple of seconds. Tried to change iterations_num, still errors. ID: 1159929 · |
|
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4
|
Ok... Bad news... The four ati wu's all had computation errors , two was vlars and the others wasn't. You have got APP drivers installed, or the SDK, haven't you? Edit: now you're reported your errored tasks, i see that it does report finding an OpenCL platform. Claggy ID: 1159932 · |
©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.