Message boards :
Number crunching :
I've Built a Couple OSX CUDA Apps...
Message board moderation
Previous · 1 . . . 54 · 55 · 56 · 57 · 58 · Next
Author | Message |
---|---|
TimeLord04 Send message Joined: 9 Mar 06 Posts: 21140 Credit: 33,933,039 RAC: 23 |
Frankly, it's impossible to trash your work by adding that one line to the CPU section. So, you must have done something else besides adding that one API line. I didn't even get a chance to copy over the Line you posted here for me... I moved in the r3711 CPU exe and Updated the Appinfo.xml File with the r3711 CPU "8.03" info, (first Suspending BOINC and Exiting), reinstalled BOINC, and hoped to see CPU and GPU crunching... Well, I got the CPU Units almost immediately, and they were crunching fine. The GPU Units started to "halt" with the "Waiting...." Msg. When I removed the r3711 exe and Appinfo Lines, (again first Suspending BOINC), then reinstalling BOINC again, ALL WORK disappeared from Tasks' List. :-( TL [EDIT:] I think, actually, I'll end up NOT changing the Command Line. Instead, I've decided to move the 1050 with it's sister, (another 1050), over to Prometheus. I'll then be pulling the 760 from Prometheus and keep it as a spare/backup GPU. So, I won't need to modify the MacPro Command Line after all. Thanks, though; you're a great help to me and everyone else. :-) TL TimeLord04 Have TARDIS, will travel... Come along K-9! Join Calm Chaos |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Now you're saying it was the GPU tasks 'waiting to run'? Before all you mentioned was the CPU tasks. Kinda hard to follow what you were doing. What did you do? From here it sounds like you went from Stock to Anonymous platform, downloaded some tasks, then went back to Stock. Yes, going from Anonymous platform to Stock will trash ALL your existing Tasks, that's been known for quite some time and not a secret. The usual reason a GPU task will not start is there is something in the Commandline it doesn't like. The fix is to remove any offending Commands, not remove the app_info. I've found some GPUs don't work with the -oclfft_tune_gr command, that's why it's not in the cmdline I posted earlier. If your cmdline has -oclfft_tune_gr in it, that would be my first guess. Try the GPU cmdline without -oclfft_tune_gr, or, maybe you should just stick with Stock. |
TimeLord04 Send message Joined: 9 Mar 06 Posts: 21140 Credit: 33,933,039 RAC: 23 |
Now you're saying it was the GPU tasks 'waiting to run'? Before all you mentioned was the CPU tasks. Kinda hard to follow what you were doing. No, I DID say the GPU Tasks were getting the "Waiting" Msg... The CPU Tasks started hogging the crunching time away from the GPU Tasks. (Eight CPU and One GPU Task are SUPPOSED to crunch at one time... Or, they did on the Stock App on the Daedalus Profile.) On the Andromeda Profile with the Anonymous Platform, the GPU Tasks were ALL getting the "Waiting" Msg after Installing the r3711 App and Appinfo.xml Info. Anyway, I've Restored the Original Andromeda Profile and Anonymous Platform Apps and Info... WITHOUT CPU Crunching, the Andromeda Profile Crunches fine, one Unit at a time - as you said that it would when on the Hackintosh-Andromeda. I just thought I'd be able to Crunch CPU, (now that I have the available Threads to do so), on the MacPro 5,1... (This is why I hate messing around with Anonymous Platform UNLESS it's completely done through Lunatics! Unfortunately, Lunatics IS NOT available on Mac.) I will just leave it crunching GPU on the r3709 App and Info files. As stated, I plan, (now), to move the 1050 to Old Prometheus, (on the new PC Platform), with a sister 1050 for TWO GTX-1050 Cards that will utilize Lunatics. When I make the CPU changeover, I will install the one good 750TI SC into the MacPro. So, as stated in my last post, I won't change the Command Line, as the 750 will be utilized on the Mac. [EDIT:] OK - In rereading my Post about the "Waiting" Msg, I could have been more clear in how I typed it out... Sorry, the "Waiting" Msg was ONLY being received AFTER Installing the r3711 App and Info to start Crunching CPU WITH GPU Tasks. Apparently more coaxing is needed on the Anonymous Platform to get that to happen. As stated in this post, above, the original Daedalus Profile on the Mac crunched 9 Total Units, 8 CPU and 1 GPU at a time. I thought I could do the same with the Anonymous Platform App. TL TimeLord04 Have TARDIS, will travel... Come along K-9! Join Calm Chaos |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Whatever floats your boat TL. Meanwhile, I'll go back to watching one of My Macs Pass this one remaining Windows Box. Another 2k or so and there won't be any Windows to look through. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
TimeLord04: OK - I think my appconfig.xml file screwed over my initial work on the new Prometheus, Computer ID: 7440554. I initially received 8 to 10 WUs and ALL immediately received Compute Errors... You know, IF you have one of those appconfig.xml files in your Mac, that could very well be what's causing Apps to go into waiting to run mode. Try removing the file and see how it works. Personally, I don't use an appconfig.xml file due to it causing problems with running tasks when running other tasks outside the Project. I find it easier to just use the previous methods of setting task limits. |
TimeLord04 Send message Joined: 9 Mar 06 Posts: 21140 Credit: 33,933,039 RAC: 23 |
TimeLord04: OK - I think my appconfig.xml file screwed over my initial work on the new Prometheus, Computer ID: 7440554. I initially received 8 to 10 WUs and ALL immediately received Compute Errors... Haven't had an app_config.xml on the MacPro 5,1... ONLY had the app_info.xml that was originally created on the Hackintosh-Original Andromeda. Thanks, though. :-) TL TimeLord04 Have TARDIS, will travel... Come along K-9! Join Calm Chaos |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Well, other than that, I don't know of any reason why you are having that problem. I'm not having that problem, and there is at least 1 other Mac Pro running 3 GPUs without it as well, https://setiathome.berkeley.edu/show_host_detail.php?hostid=8424399 I've run My Mac in High Sierra with the same 3 GPUs as in Sierra without any trouble running CPU tasks. The only other problem I can think of is having the Mixed GPUs and running OpenCL. BOINC is still screwed with device numbers in OpenCL. It works fine in CUDA, as long as you have a GPU with over 2 GBs of vRam to run the Screen. In High Sierra, BOINC doesn't give the Pre-Fermi Warning you get in previous Systems, instead it just doesn't list the GPUs as having OpenCL. I.E. in HS, the two GPUs listed below aren't even listed as having OpenCL, only the 1050 Ti would be listed in the OpenCL section, Mon Jul 23 13:16:20 2018 | | Starting BOINC client version 7.10.3 for x86_64-apple-darwin Mon Jul 23 13:16:21 2018 | | NVIDIA GPU 1: GeForce GTX 1050 cannot be used for CUDA or OpenCL computation with CUDA driver 6.5 or later Mon Jul 23 13:16:21 2018 | | NVIDIA GPU 2: GeForce GTX 950 cannot be used for CUDA or OpenCL computation with CUDA driver 6.5 or later Mon Jul 23 13:16:21 2018 | | CUDA: NVIDIA GPU 0: GeForce GTX 1050 Ti (driver version 9.1.178, CUDA version 9.1, compute capability 6.1, 4096MB, 3398MB available, 2255 GFLOPS peak) Mon Jul 23 13:16:21 2018 | | CUDA: NVIDIA GPU 1: GeForce GTX 950 (driver version 9.1.178, CUDA version 9.1, compute capability 5.2, 2048MB, 1999MB available, 2022 GFLOPS peak) Mon Jul 23 13:16:21 2018 | | CUDA: NVIDIA GPU 2: GeForce GTX 1050 (driver version 9.1.178, CUDA version 9.1, compute capability 6.1, 2048MB, 1992MB available, 1976 GFLOPS peak) Mon Jul 23 13:16:21 2018 | | OpenCL: NVIDIA GPU 0: GeForce GTX 1050 Ti (driver version 10.18.5 378.05.05.25f10, device version OpenCL 1.2, 4096MB, 3398MB available, 2255 GFLOPS peak) Mon Jul 23 13:16:21 2018 | | OpenCL: NVIDIA GPU 1: GeForce GTX 1050 (driver version 10.18.5 378.05.05.25f10, device version OpenCL 1.2, 2048MB, 2048MB available, 619 GFLOPS peak) Mon Jul 23 13:16:21 2018 | | OpenCL: NVIDIA GPU 2: GeForce GTX 950 (driver version 10.18.5 378.05.05.25f10, device version OpenCL 1.2, 2048MB, 2048MB available, 622 GFLOPS peak) Mon Jul 23 13:16:21 2018 | | OpenCL CPU: Intel(R) Xeon(R) CPU E5472 @ 3.00GHz (OpenCL driver vendor: Apple, driver version 1.1, device version OpenCL 1.2) Mon Jul 23 13:16:26 2018 | | OS: Mac OS X 10.12.6 (Darwin 16.7.0) Again, No problem running 3-4 CUDA tasks on My Macs. If I wanted to run OpenCL right now, only the 1050 Ti would work in BOINC. Maybe you should try the latest BOINC and see how that works, for Me, CUDA is working just fine. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
BTW, in High Sierra it looks as though; Tue Aug 21 05:54:04 2018 | | Starting BOINC client version 7.10.3 for x86_64-apple-darwinWhere'd my GPUs go? Only One GPU in OpenCL. But, there are 3 there...really. CUDA Special Now works in High Sierra. Truly the Fastest Mac App; CUDA_Special-v0.97-Sierra+ |
TimeLord04 Send message Joined: 9 Mar 06 Posts: 21140 Credit: 33,933,039 RAC: 23 |
BTW, in High Sierra it looks as though; Possibly a 'stupid' question, I have the NVIDIA CUDA Driver "8.0.71" currently installed on the 5,1 - do I need to Upgrade that, and to what? I ask 'cause for a long time it was mentioned NOT to go above the 8.0.71 Driver. (Thanks in advance.) TL TimeLord04 Have TARDIS, will travel... Come along K-9! Join Calm Chaos |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Well, CUDA 8 is for El Capitan. For Sierra and High Sierra you need to use the Highest CUDA version. Just keep Updating it in CUDA Preferences until it says Nothing Newer. There is still the problem with the 2GB GPUs. Don't connect them to a Monitor, use a GPU with at least 3GB of RAM, such as my 1050Ti. Or connect the monitor to an ATI GPU. The 2GB NV GPUs work fine as long as they are not connected to a Monitor. See this, CUDA version 9.2, compute capability 6.1, 4096MB, 3550MB available That's how much vRAM OSX uses Just to run the monitor. That's a pretty big chunk out of a 2 GB GPU. |
TimeLord04 Send message Joined: 9 Mar 06 Posts: 21140 Credit: 33,933,039 RAC: 23 |
Well, CUDA 8 is for El Capitan. For Sierra and High Sierra you need to use the Highest CUDA version. Just keep Updating it in CUDA Preferences until it says Nothing Newer. Thanks, system is in the shop right now for Drive Cloning for the Win7 Pro x64 Drive. TechTool Pro indicated that the SMART Check was FAILING on that Drive... OF COURSE it's a cheap ass Seagate Drive that isn't even worth being a paperweight! Drive is being cloned to a Samsung 860 EVO 1TB SSD. I've NEVER had a GOOD Seagate Drive. (Even going back to the MFM Connor days.) TL TimeLord04 Have TARDIS, will travel... Come along K-9! Join Calm Chaos |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
An Updated version Now Available. Anyone running the Special App in Sierra or High Sierra needs to update; CUDA_Special-v0.97b1-Sierra+ |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Well...... Further Testing has found there is a Problem with CUDA versions above 0.97 in MacOS. Most seriously in High Sierra where all Pulses are missed with 0.97b1 & b2. A newly compiled version of 0.97 is now at Crunchers Anonymous. Anyone running the CUDA Special App in Sierra or High Sierra Should update to this new App. Same place as always; CUDA_Special-v0.97-Sierra+ |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Was going to try your new cuda app until I realized I had too old a GPU in that mac, hahaha... I refuse to buy a new laptop.... |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
That's not a problem. Since you already have a Real Mac, making one of your old PCs a CUDA Special App Beast is a breeze. Much easier than making one without a Real Mac. Now that the problem has been identified as a driver conflict, it's easy to avoid it. Run the screen on a GPU not crunching or just Turn the Screen Off and run it with the Classic Linux screensaver, a Black Screen. Works for me. Hopefully my next Fake Mac will be able to use the iGPU to run the screen and I won't have to worry about it. Or just run the old zi3v in Sierra, the old Apps work fine in Sierra. It will just be Two times as fast as Windows instead of Three times as fast ;-) It appears everything was stuck in Orlando for today, the ram arrived there a few hours after the board. The Heat sink was supposed to arrive today, but didn't. I suspect it too is in Orlando. Hopefully everything will arrive here tomorrow. Look at that, the ram is in Tampa now, surely the rest is here too; 9/21/2018 - Friday 4:03 pm Arrived at FedEx location: FEDEX SMARTPOST TAMPA, FL The Schedule has been updated as well; Scheduled delivery: Saturday 9/22/2018 by end of day ;-) |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Now I'm convinced FedEx is just messing with me. It was a little strange sending the Ram on a road trip around Ohio, but, that wasn't the end of it. After the Ram arrived in Tampa, the jokers sent it BACK to Orlando. You can't make this stuff up. They also changed the Schedule back to Monday once it arrived back in Orlando; 9/22/2018 - Saturday 1:06 am Arrived at FedEx location ORLANDO, FL 9/21/2018 - Friday 6:03 pm Departed FedEx location FEDEX SMARTPOST TAMPA, FL 9/21/2018 - Friday 4:03 pm Arrived at FedEx location FEDEX SMARTPOST TAMPA, FL 9/21/2018 - Friday 8:17 am In transit BELLE ISLE, FL Belle Isle is South Orlando. There is still time to get it here Sat, just put it back on the Truck and it will be back in Tampa in Two Hours. WTH? Why did they send it back to Orlando??? The Post Office isn't much better. The Heat Sink was supposed to arrive on Friday...it didn't. This is all they are saying, notice they still don't give a location; September 21, 2018 In Transit to Next Facility Your package will arrive later than expected, but is still on its way. It is currently in transit to the next facility. The last I heard, the board left Orlando headed (I Hope) for Tampa; 9/21/2018 - Friday 6:42 pm Departed FedEx location ORLANDO, FL I'm about 70 Google Miles from Orlando. Tampa isn't much further from Orlando... |
Brent Norman Send message Joined: 1 Dec 99 Posts: 2786 Credit: 685,657,289 RAC: 835 |
Flight got turned back because of weather likely. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
I kinda doubt they would fly something from Orlando to Tampa, it's around 80 to 90 miles, depending what side of town(s) you're talking. According to the tracking, it was in Tampa for two hours before they sent it back. |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
I've seen that when they don't want to deliver it before it's scheduled. ie.. you didn't pay for a Saturday delivery but a weekday one so they turn the package around and send it back to the central location and wait to reship it out the night before original delivery. Used to live in Orlando, over by the airport. Long time ago. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Twilight Zone... They are sending the ram Back to Tampa after being in Orlando for less than two hours; 9/22/2018 - Saturday 2:59 am Departed FedEx location ORLANDO, FL 9/22/2018 - Saturday 1:06 am Arrived at FedEx location ORLANDO, FL I went to Bootcamp in Orlando, back when the Navy had Bootcamp in Orlando, early '72. |
©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.