Message boards :
Number crunching :
Open Beta test: SoG for NVidia, Lunatics v0.45 - Beta6 (RC again)
Message board moderation
Previous · 1 . . . 11 · 12 · 13 · 14 · 15 · 16 · 17 . . . 32 · Next
Author | Message |
---|---|
I3APR Send message Joined: 23 Apr 16 Posts: 99 Credit: 70,717,488 RAC: 0 |
Someone must have read my posts...I was able to download v0.45 40 minutes ago, and the first results seems promising. Thank you A. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14679 Credit: 200,643,578 RAC: 874 |
As the title says. Beta test only, try at your own risk. And a further update to Beta4, to include r3500 where available - that's the SoG app for NVidia cards, and the ATi app for discrete GPUs (ATi_APU not updated this time). For NVidia SoG, please note that Raistmer has changed the name of the command-line tuning file from mb_cmdline_win_x86_SSE3_OpenCL_NV.txt to mb_cmdline_win_x86_SSE3_OpenCL_NV_SoG.txt I think he expects the optimal values to change too, so please refer to the support thread for details, don't simply blind-copy the numbers from the old file to the new file. This installer will also place an updated ReadMe_MultiBeam_OpenCL.txt file in your docs folder, which may help. Apologies to anyone who tried to install ATi (non-HD5) r3430 to a 64-bit machine last time - that's what Betas are for. I think I've fixed the typo this time... As before, I'll leave the Beta3 files on OneDrive for a day or two just in case - check which file you want to download. If all goes well, I'll deprecate the old version in due course. Download url remains: https://onedrive.live.com/redir?resid=8D83BF774A4A86F5!979&authkey=!AK7t-36P9nAec2Y&ithint=folder%2cexe |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14679 Credit: 200,643,578 RAC: 874 |
I've just received some updated documentation, so I'm going to remove Beta4 temporarily (shouldn't be more than 10 minutes) and replace it. OK, updated. I've named the replacement documentation file ReadMe_MultiBeam_OpenCL_NV_SoG.txt to avoid clashes with the generic MB OpenCL file delivered with other versions of the applications - we'll tidy that up before the final release. The major changes seem to be in these suggested templates for various classes of NV cards, at the foot of the document (but in respect of the first 'known issues' comment, note that we are recommending a minimum driver version of 347.09 for SoG): ============= NV specific info ================ Known issues: - With NV drivers past 267.xx GPU usage can be low if CPU fully used with another loads. App performance can be increased by using -cpu_lock switch in this case, and CPU time savings are possible with -use_sleep switch. Suggested command line switches: Entry Level cards NV x20 x30 x40 -sbs 128 -spike_fft_thresh 2048 -tune 1 2 1 16 (*requires testing) If you experience screen lags or driver restarts use -sbs 128 -period_iterations_num 250 -no_defaults_scaling Mid range cards x50 x60 -sbs 192 -spike_fft_thresh 2048 -tune 1 64 1 4 (*requires testing) Super clocked x50TI / x60TI -sbs 256 -spike_fft_thresh 2048 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 32 -oclfft_tune_cw 32 (*requires testing) High end cards x8x x80TI Titan / Titan Z -sbs 384 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64 (*requires testing) =============================================== |
Mike Send message Joined: 17 Feb 01 Posts: 34365 Credit: 79,922,639 RAC: 80 |
I strongly suggest to update for all using AMD GPU`s. Some bug fixes has been made as well as some nice improvements. You can also download beta 4 from my website now. SETI@home v8 (anonymous platform, ATI GPU) Number of tasks completed 35598 Max tasks per day 6098 Number of tasks today 131 Consecutive valid tasks 3166 Average processing rate 157.97 GFLOPS Average turnaround time 0.56 day Not one single invalid since 4 weeks. With each crime and every kindness we birth our future. |
qbit Send message Joined: 19 Sep 04 Posts: 630 Credit: 6,868,528 RAC: 0 |
That's strange. Mine is named "mb_cmdline_win_x86_SSE3_OpenCL_NV.txt" and still it works. |
Mike Send message Joined: 17 Feb 01 Posts: 34365 Credit: 79,922,639 RAC: 80 |
This is for stand alone package only for the moment. My version has the SOG extension but in the hurry it seems Richard didn`t notice or simply hasn`t implement it for now. Its still beta. The WoW event is starting so he wanted to get it ready. With each crime and every kindness we birth our future. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14679 Credit: 200,643,578 RAC: 874 |
That's odd. I did notice, and thought I'd replaced the file itself and changed the AIstub (that's why I mentioned it). I'll check and repair later if needed. |
Mike Send message Joined: 17 Feb 01 Posts: 34365 Credit: 79,922,639 RAC: 80 |
That's odd. I did notice, and thought I'd replaced the file itself and changed the AIstub (that's why I mentioned it). I'll check and repair later if needed. Yes, its in the folder. Installed it to testfiles folder. Both the Aistub and the comandline text file has the SOG extension. With each crime and every kindness we birth our future. |
qbit Send message Joined: 19 Sep 04 Posts: 630 Credit: 6,868,528 RAC: 0 |
Anyway, I think it was the right decision to release the installer with r3500. Raistmer did a good job on this one, it's fast, stable and puts (much) less stress on the CPU compared to the previous version (r3472 IIRC). |
The_Matrix Send message Joined: 17 Nov 03 Posts: 414 Credit: 5,827,850 RAC: 0 |
Why is there not any Readme.txt file in this package ? I think that´s bad for crunchers they don´t know that they have to configure their txt-files to get best performance. They all have to wend to the Lunatics Page to get the ReadMe´s. Even when they know. C:\ProgramData\BOINC\projects\setiathome.berkeley.edu\oldApp_backup\docs Are they there ? It might be a "Programs Menü Option" to install them as a icon in menues ? |
Mike Send message Joined: 17 Feb 01 Posts: 34365 Credit: 79,922,639 RAC: 80 |
Why is there not any Readme.txt file in this package ? I think that´s bad for crunchers they don´t know that they have to configure their txt-files to get best performance. On my test install yes. Thats why i made some changes in the last minute. With each crime and every kindness we birth our future. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14679 Credit: 200,643,578 RAC: 874 |
Why is there not any Readme.txt file in this package ? Because it's still a Beta. Full public releases get an announcement in this sticky thread, and an associated discussion thread pre-populated with a full copy of the ReadMe file. But we won't move on to that stage until we know the final wrinkles have been ironed out of the application. (Could somebody translate that final idiom into colloquial German, please?) |
rob smith Send message Joined: 7 Mar 03 Posts: 22508 Credit: 416,307,556 RAC: 380 |
First impressions (no figures) With no special tuning, just whatever the installer put in place: Mouse and keyboard are less "soggy", slightly lower CPU usage, and maybe a little bit quicker. Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
They all have to wend to the Lunatics Page to get the ReadMe´s. Even when they know. You look at the old docs Try to find ReadMe_MultiBeam_OpenCL_NV_SoG.txt in: C:\ProgramData\BOINC\projects\setiathome.berkeley.edu\docs  - ALF - "Find out what you don't do well ..... then don't do it!" :)  |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
And a further update to Beta4, to include r3500 where available - that's the SoG app for NVidia cards, and the ATi app for discrete GPUs (ATi_APU not updated this time). I have a few questions. I tested Lunatics_Win32_v0.45_Beta4-for-SoG_setup.exe with TestMode=1 I run it two times extracting files to: TestFiles - Lunatics_Win32_v0.45_Beta4-for-SoG_setup - ATi_HD5_r3500 TestFiles - Lunatics_Win32_v0.45_Beta4-for-SoG_setup - NV_SoG_r3500 Â Â Â 1) In the case of ATi_HD5_r3500 there are 2 older files extracted but not used (not in MB8_win_x86_SSE2_OpenCL_ATi_HD5.aistub): MultiBeam_Kernels_r3430.cl MB8_win_x86_SSE2_OpenCL_ATi_HD5_r3430.exe I wonder why they are still in the installer and why they are extracted? Â Â Â 2) In both ATi_HD5_r3500 and NV_SoG_r3500 cases the same file is extracted: MultiBeam_Kernels_r3500.cl - 197 438 bytes, date 29.07.2016 But looking in Lunatics_Win32_v0.45_Beta4-for-SoG_setup.exe by 7-Zip I see two files MultiBeam_Kernels_r3500.cl - the second is: MultiBeam_Kernels_r3500.cl - 199 678 bytes, date 31.07.2016 I extracted the file by 7-Zip and there are some differences/additions to the newer file (I don't know if they are "ready to use" or are some tests). Is the file dated 31.07.2016 used for any selection in the installer? Â Â Â 3) In MB8_win_x86_SSE2_OpenCL_ATi_HD5.aistub I see this: <app_version> <app_name>setiathome_v8</app_name> <version_num>812</version_num> <api_version>7.5.0</api_version> Is <api_version>7.5.0 compatible with BOINC 6.10.58 Â Â - ALF - "Find out what you don't do well ..... then don't do it!" :) Â |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14679 Credit: 200,643,578 RAC: 874 |
Thanks for checking - I'll grab some coffee and head downstairs to investigate. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14679 Credit: 200,643,578 RAC: 874 |
I have a few questions. And here are the answers. I wonder why they are still in the installer and why they are extracted? They're extracted because they're still in the installer - the process is "deploy all files found in the stated folder". And they were still in the folder because I forgot to delete them while updating - finger fumble. Slight additional payload to download, no other penalty. Now deleted from build system, won't appear in next version. Is the file dated 31.07.2016 used for any selection in the installer? I downloaded a fresh copy of all three r3500 applications from Raistmer's mail.ru dropbox, and repeated your comparisons on a different machine. Same conclusion - the kernels for ATi_HD5 and NV_SoG are identical, the kernels for ATi (base version) are newer and larger. I don't know why - I'll draw it to Raistmer's attention (his department, not mine) - I think it simply confirms that development work is ongoing, and none of these apps is finally certified for prime time yet. Which is why this is still a Beta. The {newer, larger} version will certainly be deployed if the ATi base version app is selected via the UI. The installer logic processes ATi before NV, and the duplicate file handler is set at "SetOverwrite ifnewer". So the SoG version will also use the {newer, larger} version if it is installed in the same installer run with the ATi base version: if SoG is selected on its own, or in conjunction with ATi_HD5, the older version will be the only one deployed. That feels all wrong, but again it's Raistmer's department: I'll remind him that if he updates a file, he has to increment the version number too. Is <api_version>7.5.0 compatible with BOINC 6.10.58Â We had a long discussion about this at SETI Beta recently, because TBar tripped over a missing back-compatibility in Jason's X-branch CUDA code. Basically, the problem is that there was a (very slow, very gradual) transition between two different ways for the BOINC client to tell the science application which of the multiple GPUs in the computer BOINC has set aside for a newly-starting science task. BOINC v6: by command line, BOINC v7 (starting with v7.1.0): via an init_data.xml file. The two systems ran in parallel for some four years, but the command line version was finally disabled two years ago because it was incompatible with a Bitcoin Utopia application that they couldn't modify (yes, really. That's how complicated compatibility gets in this world. Welcome to what used to be known as 'DLL Hell'.) But I'm pretty certain that Raistmer uses a "new enough" API these days, and handles the fallback to ancient clients OK. v6.10.58 was released 01-Jul-2010 11:55, so it's over six years old now, and it doesn't contain proper native OpenCL detection and scheduling: people shouldn't really be trying to use it with Raistmer's apps anyway. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14679 Credit: 200,643,578 RAC: 874 |
Some people have reported difficulty choosing the right download file from OneDrive. This is the view I use: Select it this way: Hope that helps. |
Mark Stevenson Send message Joined: 8 Sep 11 Posts: 1736 Credit: 174,899,165 RAC: 91 |
On the other OneDrive view you can always right click and select details to see what the full name is Life is what you make of it :-) When i'm good i'm very good , but when i'm bad i'm shi#eloads better ;-) In't I " buttercups " p.m.s.l at authoritie !!;-) |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
For me your images don't show, I changed your links: (I used the same view yesterday, but sorted by date) Some people have reported difficulty choosing the right download file from OneDrive. Â - ALF - "Find out what you don't do well ..... then don't do it!" :) Â |
©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.