Mac OS Sierra

Message boards : Number crunching : Mac OS Sierra
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 . . . 9 · Next

AuthorMessage
W3Perl Project Donor
Volunteer tester

Send message
Joined: 29 Apr 99
Posts: 251
Credit: 3,696,783,867
RAC: 12,606
France
Message 1824169 - Posted: 14 Oct 2016, 9:44:11 UTC

Would you be interested in a Mac with GeForce 320M to run beta ?
It's currently running Lion but I can upgrade it to Sierra if needed.
Quite old Mac but still running :)
ID: 1824169 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1824182 - Posted: 14 Oct 2016, 11:26:24 UTC - in response to Message 1824169.  

Yes, that GPU would be helpful. Currently there aren't any Pre-Fermi GPUs running the CUDA42 App. The current OS would be fine as CUDA doesn't work in Yosemite and above running Pre-Fermi GPUs. The highest OSX that can run CUDA on a Pre-Fermi GPU is Mavericks (Darwin 13.x). For those Pre-Fermi cards in Mavericks, and other cards in Mountain Lion & Lion, you should install this driver; NVIDIA CUDA 5.5 for MAC
The App hasn't been tested in Snow Leopard, but, it should work in Snow Leopard using this driver; NVIDIA CUDA 4.2 for MAC
If you compare the times of my Pre-Fermi Cards running CUDA42 with similar cards running the Mac OpenCL App, the CUDA42 App should be quite a bit faster on the Older cards. Unfortunately, Beta has already deleted the tasks run back in August on my older cards, so, there aren't any completed tasks to show.
ID: 1824182 · Report as offensive
W3Perl Project Donor
Volunteer tester

Send message
Joined: 29 Apr 99
Posts: 251
Credit: 3,696,783,867
RAC: 12,606
France
Message 1824194 - Posted: 14 Oct 2016, 12:52:49 UTC - in response to Message 1824182.  

Up and running....but no CPU tasks have been downloaded, only GPU.

https://setiweb.ssl.berkeley.edu/beta/results.php?hostid=80460
ID: 1824194 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1824206 - Posted: 14 Oct 2016, 13:56:48 UTC - in response to Message 1824194.  

It seems you also don't have the CUDA Driver installed. This is what it should show with driver 5.5.47 installed;
NVIDIA GeForce 320M (256MB) driver: 3290.07 OpenCL: 1.0
Without the CUDA driver you will not be sent any CUDA tasks.

Also, the block on OpenCL Apps is only for Macs with El Capitan & Sierra, therefore, your Mac will be sent OpenCL Apps which could take some time to finish on that machine. Next you will probably be sent the other OpenCL App, it could be days before you are sent the CUDA App...once you get the driver installed. After the driver is installed it will also show up in the Event Log on Startup.
Did you get any warnings when you tried to install the CUDA Driver? Once the driver is installed you will see a CUDA preference in System Preferences, do you see one of those?
In my Lion system, which is where the CUDA42 App was compiled, I have CUDA-Z installed, it gives you CUDA Info and will tell you if CUDA is working. You have to use an older version with Lion, I'm using version CUDA-Z 0.7.184. The complete list is here, CUDA-Z
If you get CUDA working, I'd suggest Aborting ALL those OpenCL Tasks. Else it might be some while before you are sent a CUDA task. I had to Abort around 130 OpenCL tasks on my Mac before it sent a CUDA task. Since the Block was appiled last month you don't have that problem in El Capitan & Sierra.
ID: 1824206 · Report as offensive
gregjsmith
Volunteer tester

Send message
Joined: 29 Dec 99
Posts: 19
Credit: 13,241,489
RAC: 5
United States
Message 1824211 - Posted: 14 Oct 2016, 14:27:05 UTC - in response to Message 1824130.  

Ah I get it. I have the beta running on a MacBook Pro. I will get it on the other machines when I get a chance.

Did you modify the URL as described above?
Site is this:
http://setiweb.ssl.berkeley.edu/beta/

When you "Add project" Copy/Paste this in the URL field:
http://setiweb.ssl.berkeley.edu/beta/


It sounds the same as in this post, [edit]...I went in through "Add Projects" by modifying the URL for the Seti@Home option in the list to the Beta url. This time it worked.
ID: 1824211 · Report as offensive
W3Perl Project Donor
Volunteer tester

Send message
Joined: 29 Apr 99
Posts: 251
Credit: 3,696,783,867
RAC: 12,606
France
Message 1824268 - Posted: 14 Oct 2016, 19:26:24 UTC - in response to Message 1824206.  

It seems you also don't have the CUDA Driver installed. This is what it should show with driver 5.5.47 installed;
NVIDIA GeForce 320M (256MB) driver: 3290.07 OpenCL: 1.0
Without the CUDA driver you will not be sent any CUDA tasks.


CUDA driver have been installed, I've downloaded the file from
http://www.nvidia.com/object/macosx-cuda-5.5.47-driver.html
and I had no error on install.


Also, the block on OpenCL Apps is only for Macs with El Capitan & Sierra, therefore, your Mac will be sent OpenCL Apps which could take some time to finish on that machine. Next you will probably be sent the other OpenCL App, it could be days before you are sent the CUDA App...once you get the driver installed. After the driver is installed it will also show up in the Event Log on Startup.
Did you get any warnings when you tried to install the CUDA Driver? Once the driver is installed you will see a CUDA preference in System Preferences, do you see one of those?
In my Lion system, which is where the CUDA42 App was compiled, I have CUDA-Z installed, it gives you CUDA Info and will tell you if CUDA is working. You have to use an older version with Lion, I'm using version CUDA-Z 0.7.184. The complete list is here, CUDA-Z
If you get CUDA working, I'd suggest Aborting ALL those OpenCL Tasks. Else it might be some while before you are sent a CUDA task. I had to Abort around 130 OpenCL tasks on my Mac before it sent a CUDA task. Since the Block was appiled last month you don't have that problem in El Capitan & Sierra.


Computer is at work so I will check CUDA on monday.
ID: 1824268 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1824483 - Posted: 15 Oct 2016, 17:01:40 UTC - in response to Message 1824211.  

Ah I get it. I have the beta running on a MacBook Pro. I will get it on the other machines when I get a chance.

Just a little reminder about how nVidia Drivers are released. Usually, after Apple releases a new OS to the Public nVidia will release a New driver for that OS if needed. It usually takes a couple days after the Apple Public release before nVidia releases the Drivers that support that new OS. It's been that way for quite a while now, and nVidia users know to wait a few days for the new nVidia driver before updating the Apple OS.

It would appear the Beta Darwin 16.1 release does Not work very well with the existing nVidia CUDA driver. nVidia probably will not release an updated CUDA driver for Darwin 16.1 until a couple of days after the Public release. For this reason I would suggest not using a Beta OS with the CUDA driver. If you look back through the CUDA driver Updates at nVidia it would seem that lately just about every new point release of OSX has required a New CUDA driver. This also means if you are using Darwin 16.0 with a CUDA driver released for Darwin 15.x it probably will not work very well.
ID: 1824483 · Report as offensive
W3Perl Project Donor
Volunteer tester

Send message
Joined: 29 Apr 99
Posts: 251
Credit: 3,696,783,867
RAC: 12,606
France
Message 1824500 - Posted: 15 Oct 2016, 17:58:30 UTC - in response to Message 1824206.  

OK...running now with CUDA driver. I just need to reboot the laptop.
ID: 1824500 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1824529 - Posted: 15 Oct 2016, 20:03:59 UTC - in response to Message 1824500.  
Last modified: 15 Oct 2016, 20:57:48 UTC

Yes, it's showing the CUDA driver now. Judging by your cache of OpenCL tasks, it could be most of a day before those are finished. Looking at a faster machine, that's still running the OpenCL tasks, that Arecibo VLAR could take over 8 hours; Run time: 7 hours 26 min 41 sec
Those BLC7s are good for around 5 hours a piece, Run time: 4 hours 43 min 13 sec, the others are over an hour or 2 each.
Abort, Abort, Abort. I'd keep aborting them until it sends a CUDA task and then grab as many CUDAs as you can as fast as you can. After it sends a number of CUDA tasks, it will go back to sending OpenCL tasks...again. When it starts sending OpenCL tasks again, set it to No New Tasks and then repeat the procedure when the CUDA tasks are finished.
ID: 1824529 · Report as offensive
gregjsmith
Volunteer tester

Send message
Joined: 29 Dec 99
Posts: 19
Credit: 13,241,489
RAC: 5
United States
Message 1824766 - Posted: 16 Oct 2016, 21:37:10 UTC - in response to Message 1824483.  

Is there a difference to S@H from the hardware drivers and the CUDA drivers? My hackintoshes all have a control panel for the hardware and one for CUDA. My MacBook Pro only has the CUDA driver control panel.

I don't even know if I can install the Nvida control panel on a MacBook Pro.


Just a little reminder about how nVidia Drivers are released. Usually, after Apple releases a new OS to the Public nVidia will release a New driver for that OS if needed. It usually takes a couple days after the Apple Public release before nVidia releases the Drivers that support that new OS. It's been that way for quite a while now, and nVidia users know to wait a few days for the new nVidia driver before updating the Apple OS.

It would appear the Beta Darwin 16.1 release does Not work very well with the existing nVidia CUDA driver. nVidia probably will not release an updated CUDA driver for Darwin 16.1 until a couple of days after the Public release. For this reason I would suggest not using a Beta OS with the CUDA driver. If you look back through the CUDA driver Updates at nVidia it would seem that lately just about every new point release of OSX has required a New CUDA driver. This also means if you are using Darwin 16.0 with a CUDA driver released for Darwin 15.x it probably will not work very well.
ID: 1824766 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1824773 - Posted: 16 Oct 2016, 22:21:10 UTC - in response to Message 1824766.  

If you are referring to the NVIDIA Drive Manager Preference Pane, you get that by installing the nVidia Graphics driver. There really isn't any need to install it unless you are running a GPU that isn't supported by the Apple Graphics driver. You could install it on a Laptop by installing the driver, But, the Graphics driver is different for each system build and will only work with specific builds. Something similar to the CUDA Driver but much more strict. If the system build doesn't match, the Graphics driver will not install or disable itself if already installed. The current one is here, macOS Driver Release 367.15.10.05
Note the required system build, Graphics driver updated for macOS Sierra 10.12 (16A323). If your build number is something else, it will not work. A couple of days after the next system update that disables the Graphics driver nVidia will release a New Graphics driver, similar to the way they release CUDA Drivers.
ID: 1824773 · Report as offensive
[SG]HeIsBack
Volunteer tester

Send message
Joined: 25 Jun 14
Posts: 1
Credit: 8,694,112
RAC: 8
Germany
Message 1824853 - Posted: 17 Oct 2016, 7:50:57 UTC

I joined the beta too.

Took me some time to find the reason why my mac did not get any units for the GPU any more.

Well, due to problems with the current SETI nVidia OpenCL Apps and OSX there has been a ban on Hosts running Darwin 15.0 (El Capitan) and above. The Ban went into effect around Sept 6th. There are nVidia CUDA Apps at Beta that work with El Capitan & Sierra but very few Macs have bothered to do any testing. To test the CUDA Apps you must install the CUDA Driver. If you wish to see OSX nVidia Apps that work with El Capitan & Sierra you may want to install the latest CUDA driver and sign up at Beta, CUDA driver update to support macOS 10.12


Looks like it's running fine now.
ID: 1824853 · Report as offensive
W3Perl Project Donor
Volunteer tester

Send message
Joined: 29 Apr 99
Posts: 251
Credit: 3,696,783,867
RAC: 12,606
France
Message 1824866 - Posted: 17 Oct 2016, 9:28:31 UTC - in response to Message 1824773.  

Get some cuda42 wu but they fail on blc with :
"Postponed: CuFFT Plan Failure, temporary exit"
ID: 1824866 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1824872 - Posted: 17 Oct 2016, 10:11:48 UTC - in response to Message 1824866.  
Last modified: 17 Oct 2016, 10:13:40 UTC

Get some cuda42 wu but they fail on blc with :
"Postponed: CuFFT Plan Failure, temporary exit"

OK, that GPU has half the memory my 8800 GT has, but, it still should work. It could be a driver problem, 5.5.47 was the highest driver I could get to work in Lion. But it did work with my 2 cards in Lion. I would suggest trying a lower driver. Installing the OSX CUDA driver is very simple, you don't have to remove anything, all you have to do is run the installer. Try the next lower series and see if that works, http://www.nvidia.com/object/macosx-cuda-5.0.61-driver.html. It helps if you have CUDA-Z installed. CUDA-Z will let you know if CUDA is working Before you launch BOINC, so, you don't have to worry about warnings from BOINC if something isn't working correctly. I chose driver 5.5.47 because on my machine the two Pre-Fermi cards worked with that driver in all 3 OSX versions, Lion, Mountain Lion, and Mavericks. I suppose it's possible other scenarios could be different.
ID: 1824872 · Report as offensive
W3Perl Project Donor
Volunteer tester

Send message
Joined: 29 Apr 99
Posts: 251
Credit: 3,696,783,867
RAC: 12,606
France
Message 1824894 - Posted: 17 Oct 2016, 12:15:46 UTC - in response to Message 1824872.  

Get some cuda42 wu but they fail on blc with :
"Postponed: CuFFT Plan Failure, temporary exit"

OK, that GPU has half the memory my 8800 GT has, but, it still should work. It could be a driver problem, 5.5.47 was the highest driver I could get to work in Lion. But it did work with my 2 cards in Lion. I would suggest trying a lower driver. Installing the OSX CUDA driver is very simple, you don't have to remove anything, all you have to do is run the installer. Try the next lower series and see if that works, http://www.nvidia.com/object/macosx-cuda-5.0.61-driver.html. It helps if you have CUDA-Z installed. CUDA-Z will let you know if CUDA is working Before you launch BOINC, so, you don't have to worry about warnings from BOINC if something isn't working correctly. I chose driver 5.5.47 because on my machine the two Pre-Fermi cards worked with that driver in all 3 OSX versions, Lion, Mountain Lion, and Mavericks. I suppose it's possible other scenarios could be different.


Idem....same error. CUDA-Z installed and showing transfers (host to device and back). CUDA is in preference and update to 5.5.47 is available.
ID: 1824894 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1824915 - Posted: 17 Oct 2016, 13:55:02 UTC - in response to Message 1824894.  

Hmmm, that's interesting. I was just looking over my Host from when I ran the CUDA42 tests and about the only difference I see is I was using BOINC 7.4.36.
Have you tried running it in Standalone? You could download KWSN OSX MB Bench v2.01.07 and v8 reference workunit from Lunatics and see if it will run. Mine is running a BLC2 in Standalone right now, probably going to take another hour to finish on the GTS 250. It appears yours is having problems with a BLC7? After it finishes I'll try Beta and see if it still works the same as it did back in August.

If you could look at the stderr.txt in the Slots folder and post the Exact Error that might help.
I have a feeling it's going to say something about Memory...I suppose you've rebooted already.
ID: 1824915 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1824918 - Posted: 17 Oct 2016, 14:08:59 UTC - in response to Message 1824915.  
Last modified: 17 Oct 2016, 14:12:06 UTC

General gist of what's going on, may be analogous to what happened on Windows some years back. This entailed the switch from a physical VRAM model, to a virtualised one, followed by progressive bloat on CUFFT with each Cuda version. The directly allocated video memory space is pretty small ( iirc <80MiB), though CUFFT plans tend to intrinsically/internally bloat. In this sense (bigger picture) it makes sense to try build with the oldest supported Cuda toolkit as a starting point, especially for older hardware. For prefermi gear in particular, you should find Cuda 3.2 workable, with few if any bugs, coupled with a lower overall memory footprint.

With CUFFT an unknown variable, it's probable I will work toward providing an alternative that is more definite, though the millions spent on that library seem to be operating in line with the prefermi deprecations, probably extending to preKepler generations after Cuda 8. IOW, get more video memory, or use old toolkits.
[Cuda 3.2 and libraries contain raw forward compatible PTX, therefore should still run under new drivers]
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1824918 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1824929 - Posted: 17 Oct 2016, 15:19:26 UTC - in response to Message 1824918.  
Last modified: 17 Oct 2016, 15:44:59 UTC

I looked at 3.2, and decided not to go there. My 8800 GT & GTS 250 is working as usual in Darwin 11.4.2, so, it must be something to do with that low memory mobile GPU. The screen lag is horrible running 2 BLC7s, and the GTS 250 finished the Test BLC2 about a minute faster than the CPU. Not much advantage there, might as well just run the CPUs. Probably not a bad thing the lower end Low end cards won't run it. I'm going to finish the tasks it has and then go back to the newer cards.

Dang, the Screen Lag was soooo bad I couldn't even tell I was still running OpenCL tasks. Well, those are finally all gone, and the Screen Lag is not nearly as bad ;-) It only took 68 Abortions to get some CUDA42s.
Looking at Activity Monitor, it shows each CUDA42 task is using around 78MBs of memory. I have no idea what the cards are using...
ID: 1824929 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1824939 - Posted: 17 Oct 2016, 15:40:14 UTC - in response to Message 1824929.  

I looked at 3.2, and decided not to go there. My 8800 GT & GTS 250 is working as usual in Darwin 11.4.2, so, it must be something to do with that low memory mobile GPU. The screen lag is horrible running 2 BLC7s, and the GTS 250 finished the Test BLC2 about a minute faster than the CPU. Not much advantage there, might as well just run the CPUs. Probably not a bad thing the lower end Low end cards won't run it. I'm going to finish the tasks it has and then go back to the newer cards.


Yeah, well in my mind NV still wrote off that generation too early. The memory amount considerations are definitely part of the situation (especially Fermi's 64 bit capability). IMO The real pain will come with the plans to similarly deprecate anything Pre-Kepler. That'll leave a lot of actively crunching hardware cold.
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1824939 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1825018 - Posted: 17 Oct 2016, 21:21:09 UTC

Thank You hard working experts!!! (I do not have a MAC, but I appreciate Your work!)
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1825018 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 . . . 9 · Next

Message boards : Number crunching : Mac OS Sierra


 
©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.