Downloading tasks for offline re-check

Message boards : Number crunching : Downloading tasks for offline re-check
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1991251 - Posted: 24 Apr 2019, 7:27:03 UTC
Last modified: 24 Apr 2019, 7:27:51 UTC

Recently I looked through my inconclusives list and found this task: blc35_2bit_guppi_58406_02921_HIP116398_0037.24560.818.22.45.57.vlar

http://setiathome.berkeley.edu/workunit.php?wuid=3445488506

My host result (Ti1050, x41p_V0.97, Cuda 9.00 special)
Spike count: 7
Autocorr count: 0
Pulse count: 8
Triplet count: 3
Gaussian count: 0

wingman result (Iris Pro, SoG SSSE3ux OS X 64bit Build 3321 (Apple OS))
Spike count: 7
Autocorr count: 0
Pulse count: 7
Triplet count: 3
Gaussian count: 0

difference in Pulse so I wanted to download this task until third host give own version. But Excel's macro (SETI general fanout sheet, created by Richard long time ago) gives wrong link like http://boinc2.ssl.berkeley.edu/sah/download_fanout/294/blc35
Am I doing smth wrong with macro or it requires update to new version or ? (Mine has @2015 update:
Added support for 'Guppi' (SaH v8) tasks/workunits
@ as last history update).
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1991251 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1991256 - Posted: 24 Apr 2019, 7:58:46 UTC - in response to Message 1991251.  

Try http://boinc2.ssl.berkeley.edu/sah/download_fanout/25e/blc35_2bit_guppi_58406_02921_HIP116398_0037.24560.818.22.45.57.vlar

My copy has release notes:

2016 update:
At some stage I switched to a different MD5 algorithm, which seems to have broken Open/Libre Office compatibility. I can't remember when or why, so I've reverted.
Also tweaked to cope with a different format of blc 'Guppi' filenames.
I'll email you a copy.
ID: 1991256 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1991257 - Posted: 24 Apr 2019, 8:00:51 UTC - in response to Message 1991251.  
Last modified: 24 Apr 2019, 8:03:40 UTC

I'm using the one from here, http://lunatics.kwsn.info/index.php?action=downloads;sa=view;down=415 It works for me in OSX using LibreOffice Vanilla and in Linux using LibreOffice Calc with the Macro Security set to Low. It gives the link;
http://boinc2.ssl.berkeley.edu/sah/download_fanout/25e/blc35_2bit_guppi_58406_02921_HIP116398_0037.24560.818.22.45.57.vlar which appears to be correct.

You need to update the CUDA App to V0.98b1 which gives much better results with the Arecibo work, along with being a little faster with the BLC tasks.
ID: 1991257 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1991259 - Posted: 24 Apr 2019, 8:07:48 UTC - in response to Message 1991257.  

I'm using the one from here, http://lunatics.kwsn.info/index.php?action=downloads;sa=view;down=415 It works for me in OSX using LibreOffice Vanilla and in Linux using LibreOffice Calc with the Macro Security set to Low. It gives the link;
http://boinc2.ssl.berkeley.edu/sah/download_fanout/25e/blc35_2bit_guppi_58406_02921_HIP116398_0037.24560.818.22.45.57.vlar which appears to be correct.

You need to update the CUDA App to V0.98b1 which give much better results with the Arecibo work, along with being a little faster with the BLC tasks.
Yes, the Lunatics download has a September 2016 datestamp and that same additional release note. It should be good. Thanks, that's saved me an email :-)
ID: 1991259 · 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 1991264 - Posted: 24 Apr 2019, 8:31:33 UTC
Last modified: 24 Apr 2019, 8:33:43 UTC

Thanks! I downloaded BOINC.7z from link in another thread so will do update. And wlll update macro also.

Update: macro from Lunatics download works fine, thanks again.
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1991264 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1991268 - Posted: 24 Apr 2019, 9:30:40 UTC - in response to Message 1991251.  

Well, my CPU says this;
04:23:41 (9949): Can't set up shared mem: -1. Will run in standalone mode.
Build features: SETI8 Non-graphics FFTW FFTOUT JSPF AVX2 64bit 
 System: Linux  x86_64  Kernel: 4.15.0-47-generic
 CPU   : Intel(R) Core(TM) i5-6600 CPU @ 3.30GHz
 4 core(s), Speed :  3627.385 MHz
 L1 : 64 KB, Cache : 6144 KB
 Features : FPU TSC PAE APIC MTRR MMX SSE  SSE2 HT PNI SSSE3 SSE4_1 SSE4_2 AVX  AVX2  

ar=0.012812  NumCfft=110467  NumGauss=0  NumPulse=40881363840  NumTriplet=53846729888
In v_BaseLineSmooth: NumDataPoints=1048576, BoxCarLength=8192, NumPointsInChunk=32768
Linux optimized setiathome_v8 application
Version info: AVX2xjf (Intel, Core 2-optimized v8-nographics) V5.13 by Alex Kan
AVX2xjf Linux64 Build 4008 , Ported by : Raistmer, JDWhale, Urs Echternacht

Work Unit Info:
...............
Credit multiplier is :  2.85
WU true angle range is :  0.012812
Pulse: peak=7.932369, time=45.9, period=22.91, d_freq=4055342465.12, score=1.04, chirp=-1.2693, fft_len=2k
D:	threshold 1.00288; unscaled peak power: 1.038439 exceeds threshold for 3.546%
Spike: peak=24.94693, time=65.86, d_freq=4055334632.17, chirp=-14.592, fft_len=64k
Spike: peak=24.99146, time=65.86, d_freq=4055334632.18, chirp=-14.597, fft_len=64k
Spike: peak=24.34843, time=65.86, d_freq=4055334632.14, chirp=-14.642, fft_len=64k
Spike: peak=25.58207, time=65.86, d_freq=4055334632.16, chirp=-14.647, fft_len=64k
Spike: peak=25.65295, time=65.86, d_freq=4055334632.17, chirp=-14.653, fft_len=64k
Spike: peak=24.56381, time=65.86, d_freq=4055334632.19, chirp=-14.658, fft_len=64k
Spike: peak=24.5657, time=90.91, d_freq=4055338159.59, chirp=-18.359, fft_len=16k
Pulse: peak=2.01093, time=45.86, period=3.551, d_freq=4055338232.75, score=1.02, chirp=-23.498, fft_len=1024 
D:	threshold 0.1781281; unscaled peak power: 0.180496 exceeds threshold for 1.329%
Pulse: peak=3.400584, time=45.84, period=7.393, d_freq=4055336939.6, score=1.019, chirp=28.984, fft_len=512 
D:	threshold 0.1365628; unscaled peak power: 0.1385716 exceeds threshold for 1.471%
Pulse: peak=0.5805106, time=45.81, period=0.4299, d_freq=4055345165.31, score=1.02, chirp=-39.299, fft_len=32 
D:	threshold 0.003034519; unscaled peak power: 0.003056987 exceeds threshold for 0.7404%
Pulse: peak=6.718574, time=45.9, period=17.09, d_freq=4055337178.72, score=1.062, chirp=-43.516, fft_len=2k
D:	threshold 0.9294373; unscaled peak power: 0.9793674 exceeds threshold for 5.372%
Triplet: peak=10.4096, time=34.52, period=24.57, d_freq=4055341574.12, chirp=44.54, fft_len=256 
Pulse: peak=2.039099, time=45.86, period=3.031, d_freq=4055335979.15, score=1.04, chirp=-54.12, fft_len=1024 
D:	threshold 0.1741923; unscaled peak power: 0.1787721 exceeds threshold for 2.629%
Triplet: peak=11.95138, time=38.41, period=31.54, d_freq=4055335763.87, chirp=65.664, fft_len=512 
Pulse: peak=6.295819, time=45.99, period=13.33, d_freq=4055341469.72, score=1.185, chirp=-82.14, fft_len=4k
D:	threshold 1.687322; unscaled peak power: 1.950637 exceeds threshold for 15.61%
Pulse: peak=2.451421, time=45.82, period=3.964, d_freq=4055339481.91, score=1.029, chirp=82.53, fft_len=128 
D:	threshold 0.02654426; unscaled peak power: 0.02708057 exceeds threshold for 2.02%
Triplet: peak=11.42224, time=55.87, period=12.78, d_freq=4055336027.07, chirp=-94.975, fft_len=128 

Best spike: peak=25.65295, time=65.86, d_freq=4055334632.17, chirp=-14.653, fft_len=64k
Best autocorr: peak=17.65161, time=62.99, delay=4.3583, d_freq=4055338164.71, chirp=-26.17, fft_len=128k
Best gaussian: peak=0, mean=0, ChiSq=0, time=-2.124e+11, d_freq=0,
	score=-12, null_hyp=0, chirp=0, fft_len=0 
Best pulse: peak=6.295819, time=45.99, period=13.33, d_freq=4055341469.72, score=1.185, chirp=-82.14, fft_len=4k
Best triplet: peak=11.95138, time=38.41, period=31.54, d_freq=4055335763.87, chirp=65.664, fft_len=512 
Spike count:    7
Autocorr count: 0
Pulse count:    8
Triplet count:  3
Gaussian count: 0

Time cpu in use since last restart: 2724.8 seconds
05:10:37 (9949): called boinc_finish(0)

But, you really don't need to waste all that time. I just check the other Host and see how it's working, as in,
All tasks for computer 7857610
State: All (170) · In progress (8) · Validation pending (67) · Validation inconclusive (43) · Valid (52) · Invalid (0) · Error (0) Application: All (170) · AstroPulse v7 (1) · SETI@home v8 (169)
That's really all you need to do. In the case of the Old Windows CUDA Apps, if the other run-time is Lower than My CUDA Time I don't even bother checking it, since that's an impossibility. Other times a quick check will show the other Host has a Very High Inconclusive or Invalid rate... that's all I need to see.
If you removed all the obviously Bad Wingmen from the CUDA results, the Inconclusive rate would be Very low.

Recently I looked through my inconclusives list and found this task: blc35_2bit_guppi_58406_02921_HIP116398_0037.24560.818.22.45.57.vlar

http://setiathome.berkeley.edu/workunit.php?wuid=3445488506

My host result (Ti1050, x41p_V0.97, Cuda 9.00 special)
Spike count: 7
Autocorr count: 0
Pulse count: 8
Triplet count: 3
Gaussian count: 0

wingman result (Iris Pro, SoG SSSE3ux OS X 64bit Build 3321 (Apple OS))
Spike count: 7
Autocorr count: 0
Pulse count: 7
Triplet count: 3
Gaussian count: 0

difference in Pulse so I wanted to download this task until third host give own version. But Excel's macro (SETI general fanout sheet, created by Richard long time ago) gives wrong link like http://boinc2.ssl.berkeley.edu/sah/download_fanout/294/blc35
Am I doing smth wrong with macro or it requires update to new version or ? (Mine has @2015 update:
Added support for 'Guppi' (SaH v8) tasks/workunits
@ as last history update).
ID: 1991268 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1991271 - Posted: 24 Apr 2019, 10:09:42 UTC - in response to Message 1991268.  
Last modified: 24 Apr 2019, 10:11:34 UTC

But, you really don't need to waste all that time. I just check the other Host and see how it's working, as in,

SETI@home v8 v8.00 (opencl_intel_gpu_sah)
x86_64-apple-darwin
Yes, I think we've identified that one as a problem.

I'd be more worried by

Workunit 3445027226
http://boinc2.ssl.berkeley.edu/sah/download_fanout/3c/blc35_2bit_guppi_58406_01258_HIP116971_0032.20297.409.22.45.69.vlar

The cuda50 app is normally OK, and should be fine on a GTX 560, but 3 / 1 / 14 / 2 is very different from 1 / 1 / 6 / 2

Mind you, computer 8706539 is a pretty dodgy wingmate, too.
ID: 1991271 · 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 1991274 - Posted: 24 Apr 2019, 10:44:59 UTC

Thanks, host-checking before task downloading and rerun will save some time indeed.
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1991274 · 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 1991275 - Posted: 24 Apr 2019, 10:46:00 UTC - in response to Message 1991271.  

But, you really don't need to waste all that time. I just check the other Host and see how it's working, as in,

SETI@home v8 v8.00 (opencl_intel_gpu_sah)
x86_64-apple-darwin
Yes, I think we've identified that one as a problem.

I'd be more worried by

Workunit 3445027226
http://boinc2.ssl.berkeley.edu/sah/download_fanout/3c/blc35_2bit_guppi_58406_01258_HIP116971_0032.20297.409.22.45.69.vlar

The cuda50 app is normally OK, and should be fine on a GTX 560, but 3 / 1 / 14 / 2 is very different from 1 / 1 / 6 / 2

Mind you, computer 8706539 is a pretty dodgy wingmate, too.

Downloaded, thanks.
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1991275 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1991277 - Posted: 24 Apr 2019, 10:56:04 UTC - in response to Message 1991271.  
Last modified: 24 Apr 2019, 11:12:22 UTC

Jason had a 560Ti. The way I remember it, the card was notorious for producing a 30% inconclusive rate out of the box . Jason found the factory clock was set too high and the core voltage needed to be increased to compensate. Unfortunately, convincing someone to increase the voltage on their card meets with little success, so, they usually produce a higher number of Inconclusive/Invalid results.

Just a few other Apples from my basket,
https://setiathome.berkeley.edu/results.php?hostid=8653444
https://setiathome.berkeley.edu/results.php?hostid=5857582
https://setiathome.berkeley.edu/results.php?hostid=8323253
https://setiathome.berkeley.edu/results.php?hostid=8696806
https://setiathome.berkeley.edu/results.php?hostid=8377304
https://setiathome.berkeley.edu/results.php?hostid=8687146
https://setiathome.berkeley.edu/results.php?hostid=7146191
https://setiathome.berkeley.edu/results.php?hostid=8190220
https://setiathome.berkeley.edu/results.php?hostid=7539509
https://setiathome.berkeley.edu/results.php?hostid=8148397
https://setiathome.berkeley.edu/results.php?hostid=8299073
https://setiathome.berkeley.edu/results.php?hostid=6066861
https://setiathome.berkeley.edu/results.php?hostid=8414601
https://setiathome.berkeley.edu/results.php?hostid=8456013
https://setiathome.berkeley.edu/results.php?hostid=8616853

Now, compare those to one of my OSX machines running the CUDA App;
All tasks for computer 8097309
State: All (2192) · In progress (331) · Validation pending (1001) · Validation inconclusive (27) · Valid (827) · Invalid (1) · Error (5) Application: All (2192) · AstroPulse v7 (8) · SETI@home v8 (2184)

Again, remove all those obviously Bad Wingmen from my host and the Numbers are Much Better.
ID: 1991277 · Report as offensive
Profile -= Vyper =-
Volunteer tester
Avatar

Send message
Joined: 5 Sep 99
Posts: 1652
Credit: 1,065,191,981
RAC: 2,537
Sweden
Message 1991284 - Posted: 24 Apr 2019, 13:18:38 UTC - in response to Message 1991277.  

Jason had a 560Ti. The way I remember it, the card was notorious for producing a 30% inconclusive rate out of the box .


Yes
You're correct, i had one of those 560Tis aswell, it was faulty at stock, but i didn't increase the voltage, i took a bios editor and cloned my bios, and then lowered the core frequency about 50Mhz or so. Then it produced valid results and thus i could edit and reflash the bios with failsafe values instead.. :)

_________________________________________________________________________
Addicted to SETI crunching!
Founder of GPU Users Group
ID: 1991284 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1991552 - Posted: 26 Apr 2019, 11:29:07 UTC - in response to Message 1991275.  
Last modified: 26 Apr 2019, 12:03:45 UTC

But, you really don't need to waste all that time. I just check the other Host and see how it's working, as in,

SETI@home v8 v8.00 (opencl_intel_gpu_sah)
x86_64-apple-darwin
Yes, I think we've identified that one as a problem...
Perhaps it would help to remind those concerned with the Intel App about the Other problems with OpenCL on Macs since the 2015.4 "Update". The Darwin 15.4 Update Broke OpenCL on Macs for nVidia in both Multibeam and AstroPulse. Since that Update Raistmer's code will not produce a working NV Multibeam App, and works extremely slowly with the existing nVidia AstroPulse. So you see, it's not just the Intel OpenCL App that doesn't work well, in fact, the Intel build works Very Well when run on an AMD or NV GPU, it just Fails on an Intel GPU. The Current Mac nVidia MultiBeam App on SETI Main is in fact, wait for it, the Intel Build! The NV GPUs are running the Intel Build because since 15.4 the NV Build works slower than a CPU, which basically means the GPU isn't being used in the NV GPU Mac Build.
So you see, it's Not just a problem with the Intel OpenCL App, not only does the Intel Build Not work on Intels, but, the NV Build doesn't work on NVs either. About the only OpenCL Build that works semi-correctly is the ATI/AMD Build.
ID: 1991552 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1991580 - Posted: 26 Apr 2019, 16:07:37 UTC - in response to Message 1991552.  

Very interesting post TBar. I know you've said you have informed Jeff and Eric about the problems with those apps and they have done nothing about it. How many more voices will it take for them to pull those apps from circulation? Substitute the AMD OpenCL app for the Intel gpus seems the solution. I know that the most active development lately has been on the AMD Vulkan platform which has cross-platform capability for OpenCL as its design constraint.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1991580 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1991582 - Posted: 26 Apr 2019, 18:08:00 UTC - in response to Message 1991580.  

I don't think Jeff was involved in the SETI Apps, mainly just Raistmer & Eric. Anyone at BETA back then could have been reading the comments, as I'm sure Richard was. That was a while back, I just wanted to refresh peoples memory about the other Mac OpenCL Apps as I've had to do about Non-Replaceable Mac Drivers. The problem right now is there isn't an App that works correctly on the Intel GPUs, there are working Apps for AMD & nVidia, the Intel App actually works nicely on the nVidia GPUs.
ID: 1991582 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1991600 - Posted: 26 Apr 2019, 21:51:42 UTC - in response to Message 1991582.  
Last modified: 26 Apr 2019, 21:53:04 UTC

Yes, I am interested: I'd like to explore exactly what Apple did in 2015.4, but I've got visitors this weekend: I'll do a fuller reply on Sunday pm or Monday.

I have a sneaking suspicion that the existing Apple Apps may have been compiled by Charlie Fenton, who appears to be a fully fledged Apple developer and has been keeping the BOINC client and Manager up to date for Apples through all the various OS X changes. But he'll just have done quick app builds for SETI as a favour, without fully studying the underlying subleties. Charlie wants to move away from BOINC (or possibly retire - I don't know how old he is), and he's just hanging on for care and maintenance until a replacement comes forward. Adverts have been placed, and one potential candidate has put themselves forward: more may appear. I'd suggest we watch the situation and when BOINC has a new Apple developer, and he or she is in too deep to back out... - then we can ask ever so gently if they could build for SETI as well.

But we'll need to know what went wrong in 2015, and thus know what we're asking them to do.
ID: 1991600 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1991617 - Posted: 26 Apr 2019, 23:29:50 UTC - in response to Message 1991600.  
Last modified: 26 Apr 2019, 23:53:22 UTC

Hmmmm, the amount of Memory Booster needed to fix that post would probably pose a health risk. Perhaps you should ask Raistmer where he was getting all those Mac Apps he passed on to Eric. I think Charlie maybe the one providing the CPU App, which is coming from the seti-boinc folder. All the other Mac Apps are from Raistmer's sah_v7_opt folder and provided by either Urs or Myself. Every time I try to compile something from the seti-boinc folder it fails on the ScreenSaver, so, I gave up on that years ago. I do hope the new BOINC Mac person knows their main function will be to fix a broken ScreenSaver every few months when an 'update' breaks it, hopefully they are one of the ones that still likes screensavers. As for the current Mac Apps, all of them except the AP, CPU, and Intel App came from my Mac.
Darwin starts late in the Year, that means the problem began early 2016. Adobe would know what happened as they and others had to rebuild their Mac Apps due to the change in OpenCL that started with Darwin 15.0. It took until 15.4 to totally break the SETI NV Apps, the AMD App actually improved.
ID: 1991617 · Report as offensive
Profile Unixchick Project Donor
Avatar

Send message
Joined: 5 Mar 12
Posts: 815
Credit: 2,361,516
RAC: 22
United States
Message 1991638 - Posted: 27 Apr 2019, 0:53:57 UTC

As a mac person (at this time), I'm interested in this discussion of the code base for MacOS, in part to make sure that I only return good data. I currently only crunch on the CPU, not on my GPU due to things I've read on the boards here. It is also a neat programming problem to think about. I'd love to see this in a separated possibly pinned thread so that it can support anyone who decides to tackle this issue. If I ever get to having some spare time (hahaha), I think seti programming would be a cool project to look at.

Is it two issues? One to get WUs to run more efficiently (faster with no errors) on Mac's special OpenCL , and two to get the screen saver to work and continue to work as Mac changes things ??

I know my old Mac (NV) did recently get an updated seti app though.
ID: 1991638 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1991651 - Posted: 27 Apr 2019, 2:03:34 UTC - in response to Message 1991638.  

Well, to understand things one needs to be aware of the current events. Apple is moving away from both OpenCL and CUDA to the, at this time, Closed Metal. Seeing as how nVidia views Metal as a direct competitor to CUDA I can see why they don't see any upside to spending resources on Metal. Both OpenCL and CUDA will still work, for a while, but clearly Apple will be steering toward Metal. There is an interesting post at NV, how accurate it is isn't known, ...just stick with macOS 10.13 High Sierra on the cMP 5,1, it’s somewhat stable and your Nvidia cards will work smooth on it for a few years yet.
Hmmm, High Sierra on a cMac Pro, yep, seems to work, https://setiathome.berkeley.edu/show_host_detail.php?hostid=6796479
The Good news is the last NV Toolkit for Mac includes support for the Turing cards, previous versions didn't.... maybe soon....
ID: 1991651 · Report as offensive
Profile TimeLord04
Volunteer tester
Avatar

Send message
Joined: 9 Mar 06
Posts: 21140
Credit: 33,933,039
RAC: 23
United States
Message 1991656 - Posted: 27 Apr 2019, 2:43:17 UTC - in response to Message 1991638.  
Last modified: 27 Apr 2019, 2:47:25 UTC

As a mac person (at this time), I'm interested in this discussion of the code base for MacOS, in part to make sure that I only return good data. I currently only crunch on the CPU, not on my GPU due to things I've read on the boards here. It is also a neat programming problem to think about. I'd love to see this in a separated possibly pinned thread so that it can support anyone who decides to tackle this issue. If I ever get to having some spare time (hahaha), I think seti programming would be a cool project to look at.

Is it two issues? One to get WUs to run more efficiently (faster with no errors) on Mac's special OpenCL , and two to get the screen saver to work and continue to work as Mac changes things ??

I know my old Mac (NV) did recently get an updated seti app though.

Speaking for myself, (as another Mac User), I started with TBar's CUDA42, then CUDA75, then OpenCL, and finally the new CUDA90 App. ALL Apps, (except the new CUDA90), were run on my old Hackintosh-Andromeda that died around April of last year. (Intel Quad Core Extreme - QX9650 3GHz, Gigabyte GA-EP45-UD3P MOBO, TWO EVGA GTX-750Ti SC 2GB GDDR5 VRAM Cards, with Legacy Installed Clover --> El Capitan 10.11.3, Updated to .5 and finally .6) The CUDA90 App was first used by me on Daedalus under the Andromeda BOINC ID, (Dual Xeon 2.4GHz, EVGA GTX-1050 2GB GDDR5 VRAM Card - PCI-e Slot 3, ATI HD5770 1GB VRAM Card - PCI-e Slot 1.) My first use of CUDA90 on MacOS was with TBar's v0.97b2 App.

On Original Hackintosh-Andromeda under El Capitan 10.11.3 I experienced HIGH numbers of Inconclusives with CUDA75 and the OpenCL App. I wasn't on CUDA42 very long when TBar instructed me to move to the CUDA75 App. Then when he released the Mac OpenCL App, he instructed me to move to that to help reduce the Inconclusives. He found that 10.11.3 onward seemed to "break" things with the Apps, and found this was Driver Related. I held off Updating to 10.11.4 for fear of making things worse, but did eventually Update to 10.11.5 and later 10.11.6. The OpenCL App was the BEST App on the Original Hackintosh. On 10.11.6, in addition to the Web Driver Issues, there also seemed to be CUDA Driver Issues, and TBar stated NOT to go above CUDA Driver 8.0.71 for El Capitan.

NOW, with New Hackintosh-Andromeda, (Intel i7 7700K 4.2GHz, Gigabyte GA-H270-HD3 MOBO, MacVidCards' GTX-1070 8GB GDDR5 VRAM Card, and secondary EVGA GTX-1050 2GB GDDR5 VRAM Card), and Daedalus restored to original BOINC ID Parameters, both my Systems are on CUDA90 v0.97 - (No Beta); things are working well as long as I follow the rule of keeping the monitors on BOTH Systems OFF while Crunching. (OH, and Daedalus received a new/used Mac GTX-970 4GB GDDR5 VRAM Card to replace the ATI HD-5770.) IF the monitors are accidentally left ON while Crunching, ROUTINELY I experience the Bug of "Best Pulse = 0" which results in HIGH numbers of Invalids. Currently, both Macs are on 10.13.6 - (17G6030), Web Driver 387.10.10.10.40.124, and CUDA Driver 418.105. Current Web Driver Issues include the STRIPPING of OpenCL and OpenGL on ALL secondary and tertiary Cards on both Systems.

Again, as long as I follow TBar's instructions, (took me several PMs and Public Message Posts in various Threads to get things straight), TBar has helped me get the MOST out of MacOS on both Systems. Unfortunately, due to California's HIGH and Exorbitant Electricity Rates, I now MUST keep Daedalus OFF. ONLY Hackintosh-Andromeda Crunches, and ONLY from 18:00 to 09:00 Pacific.

Don't be afraid to try out TBar's CUDA90 App. IF you need help getting things installed and tweaked to best fit your System, TBar is FANTASTIC at getting things ironed out. Since getting the 1070 and 1050 Cards working together, I seem to be able to hit and hold approx. 43K RAC with STEADY work flow. Just re-reached 38K+ on Andromeda and a Total of 39K with both Andromeda and my dad's Systems Crunching. Dad's System is Farragut, an i7 7700 3.6GHz System on a Gigabyte GA-H270-HD3 MOBO. Farragut may Crunch 2-3 Hours every two to three days. Farragut is on Win 10 Pro and uses the Lunatics' 0.45Beta 6 Optimized App - AVX on CPU and SoG on the Integrated Intel HD630 GPU. To keep the System running smoother, I choose NOT to Crunch on CPU on ANY of my Systems. Andromeda is my heavy Cruncher now.


TL
TimeLord04
Have TARDIS, will travel...
Come along K-9!
Join Calm Chaos
ID: 1991656 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1991674 - Posted: 27 Apr 2019, 8:51:25 UTC - in response to Message 1991638.  
Last modified: 27 Apr 2019, 9:24:25 UTC

As a mac person (at this time), I'm interested in this discussion of the code base for MacOS, in part to make sure that I only return good data. I currently only crunch on the CPU, not on my GPU due to things I've read on the boards here. It is also a neat programming problem to think about. I'd love to see this in a separated possibly pinned thread so that it can support anyone who decides to tackle this issue. If I ever get to having some spare time (hahaha), I think seti programming would be a cool project to look at.

Is it two issues? One to get WUs to run more efficiently (faster with no errors) on Mac's special OpenCL , and two to get the screen saver to work and continue to work as Mac changes things ??

I know my old Mac (NV) did recently get an updated seti app though.
The screensaver problem is in two parts.

1) Interacting with the operating system to display a screensaver - any screensaver - when the time is right. Charlie has been looking after that in the BOINC code with patches like #2726 - whoever takes over from him will have to take on that bit of it too.

2) updating the specifically SETI code to display specifically SETI data and graphics when the screensaver module chooses SETI for display.

And, as you say, the whole screensaver/graphics question is another issue from the basic accuracy and speed of the SETI search application.
ID: 1991674 · Report as offensive
1 · 2 · Next

Message boards : Number crunching : Downloading tasks for offline re-check


 
©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.