Message boards :
Number crunching :
El Capitan and BOINC 7.6.12 jump in performance
Message board moderation
Author | Message |
---|---|
Marek Majewski Send message Joined: 26 Nov 00 Posts: 31 Credit: 18,604,752 RAC: 147 |
I have noticed a significant jump in averages right after I updated my two MACs to El Capitan OS X 10.11. Now, it may be because of the new OS (it is visibly faster with other applications), or it maybe because of the new BOINC client (7.6.12) I had to install post OS update, since my older version of BOINC refused to run under 10.11. It also maybe due to some other factors that I may be not aware of (different work-units, etc), but the jump is rather significant - almost double - on both host averages and user averages. I run BOINC on two MACs Intel i7 quad core CPUs, and did not change the CPU or time allotment, which I tuned so it does not spin the CPU fans past 2,200 RPM. The CPU utilization, temps and fan RPMs are pretty much the same as before, but the increase in averages is mind blowing, really. Since I do not believe one can achieve such and incredible - mind you, almost 100% - increase in efficiency by optimizing mature code, I wonder what may be causing that increase? Any ideas? * Wisdom is ability to get important things approximately right most of the time. * Breakthrough is seeing what everyone else has seen, but thinking what no one else has thought. * Cogito, ergo sum. |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
You have changed from using the Stock 7.00 Seti v7 app to the Optimised Seti v7 7.07 ssse3, sse4.1 and avx apps, these apps are significantly faster. Setiahome Applications Claggy |
Marek Majewski Send message Joined: 26 Nov 00 Posts: 31 Credit: 18,604,752 RAC: 147 |
I see - this must be it. Why would anybody run the less-efficient version of the client? -mm- * Wisdom is ability to get important things approximately right most of the time. * Breakthrough is seeing what everyone else has seen, but thinking what no one else has thought. * Cogito, ergo sum. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Yes, you can check it by comparing Run Times at similar Angle Ranges. http://setiathome.berkeley.edu/result.php?resultid=4433919040 Application version: SETI@home v7 v7.00 WU true angle range is : 0.422131 Run time: 2 hours 44 min 17 sec CPU time: 2 hours 34 min 43 sec http://setiathome.berkeley.edu/result.php?resultid=4435187664 Application version: SETI@home v7 v7.07 (avx) WU true angle range is : 0.422084 Run time: 1 hours 12 min 35 sec CPU time: 1 hours 9 min 7 sec So, at that Angle Range My App is over Twice as fast as the Old App on Your machine. It does vary with different machines and Angle Ranges, but it's usually almost Twice as Fast. As for El Capitan, well, on My machine it slowed the SETI Apps down so much I went back to Mavericks. Yosemite slowed it down a little and El Capitan even more. On the Machine that compiled those New Mac CPU Apps, Mavericks gives the best ATI & CPU times. I can't run the nVidia 750Ti in Mavericks, but Yosemite was better than El Capitan with my PC 750Ti. I moved the GTX 750Ti to a Linux machine for now. |
Marek Majewski Send message Joined: 26 Nov 00 Posts: 31 Credit: 18,604,752 RAC: 147 |
Thanks for the explanation. It makes sense, although I do not understand what the "angle range" is. Where can I read up on it? Thanks. -mm- * Wisdom is ability to get important things approximately right most of the time. * Breakthrough is seeing what everyone else has seen, but thinking what no one else has thought. * Cogito, ergo sum. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
It's pretty self explanatory. Each WU contains a portion of the Sky in Angle Ranges. A high AR means the telescope was moving across the Sky at a higher than normal rate of change. The Normal AR is when the telescope is stationary and the change in Angles is caused by the Earth's rotation. Low ARs happen when the telescope is moving to stay fixed on a certain part of the Sky. Similar ARs usually take about the same amount of time to process, hence when comparing Run Times you should compare similar ARs. If you preform an Advanced Search using WU true angle range you will find many posts on the subject. |
Marek Majewski Send message Joined: 26 Nov 00 Posts: 31 Credit: 18,604,752 RAC: 147 |
Now I understand the Angle Range - thanks for your time. * Wisdom is ability to get important things approximately right most of the time. * Breakthrough is seeing what everyone else has seen, but thinking what no one else has thought. * Cogito, ergo sum. |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
El Capitan appears to be much more stable than Yosemite. Looks like Apple learned from it's mistake in rushing a OS out to the public So far (fingers crossed) it hasn't crashed. Ran a few work units here, now over on Beta running all the apps to see how it handles. |
Marek Majewski Send message Joined: 26 Nov 00 Posts: 31 Credit: 18,604,752 RAC: 147 |
El Capitan appears to be much more stable than Yosemite. I have never had a problem with Yosemite. I kept it running for as long after a boot as I could, and was only limited by an occasional re-start necessary as a part of an upgrade or install of something, easily getting 30 days of uptime in between reboots. I see the same happening with El Capitan. Could it be you had had a pesky piece of so software screwing your Yosemite instance? -mm- * Wisdom is ability to get important things approximately right most of the time. * Breakthrough is seeing what everyone else has seen, but thinking what no one else has thought. * Cogito, ergo sum. |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Doubtful. 2 separate MacBooks. Both subject to frequent crashes. 1 was barebones. This after a complete removal of all software and install of clean OS and only BOINC and seti. Yet it continued to frequently crash. Problem with Yosemite, it appeared to not want you to use your computer all day. It acted more like a novelty program, rather than a robust OS for a daily driver. No wonder there were waves of people at the apple stores reverting back to earlier OS. At least with El Capitan, it looks like they took the time to test it under real life usage to make sure it works. As for pesky piece of software. Any software other than the basic that comes with a mac is considered pesky. Unrealistic in this day and age. Don't get me wrong, I love Mac. Use it for my daily driver. Been a user since the days of Performa. Even had the apple newton. But Yosemite was a rushed job and apple knew it. Lets see how El Capitan holds up. Zalster |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
My Last day with El Capitan I walked in to find my Mac had rebooted after suffering what appears to have been a nVidia Driver crash with the GTX 750Ti. I've already seen 2 different nVidia Web Drivers since the release, I think I'll wait for the next one. With Yosemite the 750 has racked up 7833 Consecutive valid CUDA tasks without an Error or Invalid, and I don't want to lose that over new flakey video drivers. I didn't have any problems with Yosemite running 3 video cards and 4 CPU tasks, it was just a little slower than Mavericks on My machine. Unfortunately my Yosemite turned into El Capitan as I didn't have a spare partition for yet another system...oh well. |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
With the laptops I haven't had to worry about updating the drivers. So I'm safe on that aspect. Still running cuda 6.0.51 I back up both Maverick versions of the HD and now the El Capitan just to be safe |
Marek Majewski Send message Joined: 26 Nov 00 Posts: 31 Credit: 18,604,752 RAC: 147 |
My experience is very different, and I am not trying to be a fanboy. I approach technology from a pragmatic and practical standpoint: if it works, I like it, if it does not, I don't. Three Macs: Mac Mini (late 2012) MacBook Pro (late 2013) MacBook Air (late 2013) All solid as can be, under Yosemite, or El Capitan. They all have maxed out RAM (16GB, 16GB, 8GB) and all run from SSDs, and stay up until I shut them down or reboot. Pro and Mini are loaded with various stuff: external USB-3 and Thunderbolt drives, twin monitors, Office, VMware Fusion, OS X Server, countless other programs and utilities, i.e.: Mini has 380 processes and 1872 threads and the Pro has 292 processes and 1397 threads running currently, both on over 10 days of uptime at the moment. I use them all day long for work and fun, and for crunching in the background, but I limit crunching to two CPUs per machine, since I like to keep the cooling fans around 2,200 RPMs. Before I had dozens of Windows machines, and they all exhibited the stability and grace of a heard of hippos trying to dance Swan Lake. But as they say: your MPG may vary... -mm- * Wisdom is ability to get important things approximately right most of the time. * Breakthrough is seeing what everyone else has seen, but thinking what no one else has thought. * Cogito, ergo sum. |
Woodgie Send message Joined: 6 Dec 99 Posts: 134 Credit: 89,630,417 RAC: 55 |
I was wondering why my RAC was higher than normal. OK, shall have a look-see and post anything interesting I find for no reason other than I like the sound of my own voi... keyboard clicking. ~W |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Hmmm...Suddenly I'm reminded of Hundreds of Errors generated by iMacs with the ATI HD4 cards. It is possible to add a single line to a cc_config.xml file and avoid all those Errors. If the Host doesn't have a cc_config.xml file you can create one by duplicating an existing .xml file, then changing the file name to cc_config.xml and using TextEdit to replace the contents with; <cc_config> <log_flags> </log_flags> <options> <save_stats_days>365</save_stats_days> <ignore_ati_dev>0</ignore_ati_dev> </options> </cc_config> That should solve it, http://boinc.berkeley.edu/wiki/Client_configuration In other news it appears nVidia is thinking about making their Web Drivers available to All Macs with nVidia cards instead of just the older MacPros. For months now the following line has been included in the Release Notes; http://www.nvidia.com/download/driverResults.aspx/92192/en-us Yes, it's here too; Operating System: Mac OS X Yosemite 10.10.5 http://www.nvidia.com/download/driverResults.aspx/89538/en-us |
Thanar Send message Joined: 14 May 99 Posts: 50 Credit: 3,223,553 RAC: 0 |
So what you're all saying is that the new faster SSE versions of the OS X client came out about the same time as we upgraded to El Capitan? Is that correct? Regarding El Capitan experience, I would say I am quite pleased. I'm on an 27" iMac, Late 2009, on 24/7, with record uptime of almost 3 months and was lately experiencing frequent freezing issues under Mavericks (basically I believe it was some major process crashing, since I had no interaction whatsoever, although audio for one kept on playing). El Capitan has been running solid since updating. On a sidenote, since I updated to El Capitan and reinstalled BOINC, I have been having errors when the project is utilizing the GPU (Radeon HD 4850). I guess -TBar- it's the errors you're talking about in you post; however, I thought HD4850 had openCL support, and don't remember such frequent errors in SETI@home in the past. What could be the culprit? I also note that even when running 4x SSE tasks at 100% CPU utilization, CPU temp stays at 67°C and fan RPMs at lowest 1800. That's interesting... Any comments on that? |
Urs Echternacht Send message Joined: 15 May 99 Posts: 692 Credit: 135,197,781 RAC: 211 |
So what you're all saying is that the new faster SSE versions of the OS X client came out about the same time as we upgraded to El Capitan? Is that correct? Could you try some other BOINC version (latest), to test if that avoids the SIGPIPE errors, which seem to happen before the app is starting to crunch ? _\|/_ U r s |
Urs Echternacht Send message Joined: 15 May 99 Posts: 692 Credit: 135,197,781 RAC: 211 |
Thanks for testing the latest (7.6.12) version of BOINC. Its still showing the same error. ... OpenCL Platform Name: Apple Number of devices: 1 Max compute units: 10 Max work group size: 1024 Max clock frequency: 503Mhz Max memory allocation: 134217728 Cache type: None Cache line size: 0 Cache size: 0 Global memory size: 402653184 Constant buffer size: 65536 Max number of constant args: 8 Local memory type: Scratchpad Local memory size: 16384 Queue properties: Out-of-Order: No Name: Radeon HD 4850 Vendor: AMD Driver version: 1.0 Version: OpenCL 1.0 Extensions: cl_APPLE_SetMemObjectDestructor cl_APPLE_ContextLoggingFunctions cl_APPLE_clut cl_APPLE_query_kernel_names cl_APPLE_gl_sharing cl_khr_gl_event SIGSEGV: segmentation violation Crashed executable name: astropulse_7.07_x86_64-apple-darwin__opencl_ati_mac Machine type Intel 80486 (64-bit executable) System version: Macintosh OS 10.11 build 15A284 Tue Oct 13 13:59:31 2015 atos cannot load symbols for the file astropulse_7.07_x86_64-apple-darwin__opencl_ati_mac for architecture x86_64. 0 astropulse_7.07_x86_64-apple-darwin__opencl_ati_mac 0x0000000101c09074 SIGPIPE: write on a pipe with no reader 1 astropulse_7.07_x86_64-apple-darwin__opencl_ati_mac 0x0000000101bfa396 SIGPIPE: write on a pipe with no reader 2 libsystem_platform.dylib 0x00007fff9a2f552a SIGPIPE: write on a pipe with no reader 3 ??? 0x0000000000000001 SIGPIPE: write on a pipe with no reader 4 CoreFoundation 0x00007fff8ddcdd18 SIGPIPE: write on a pipe with no reader 5 CoreFoundation 0x00007fff8ddcd80d SIGPIPE: write on a pipe with no reader 6 ATIRadeonX2000GLDriver 0x0000000102e22d16 SIGPIPE: write on a pipe with no reader 7 OpenCL 0x00007fff8f912754 SIGPIPE: write on a pipe with no reader 8 OpenCL 0x00007fff8f8ff3a6 SIGPIPE: write on a pipe with no reader 9 astropulse_7.07_x86_64-apple-darwin__opencl_ati_mac 0x0000000101ba72de SIGPIPE: write on a pipe with no reader 10 ??? 0x0000000000000001 ...Will check further ... _\|/_ U r s |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
So what you're all saying is that the new faster SSE versions of the OS X client came out about the same time as we upgraded to El Capitan? Is that correct? The Mac ATI Apps were released back on May 21st, http://setiathome.berkeley.edu/apps.php. AFAIK they have Never worked on the High Definition 4 cards, why you just noticed it is a mystery. There is a similar iMac on Beta, and as you can see it has never worked with the GPU Apps, http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=74464&offset=280 Whether they will ever work with the HD4 cards is unclear, best to ask Urs about that. I will say I have a PC version of the 4670 and it has Never worked with the SetiatHome MB apps, so I doubt that one will ever work. The 47/48xx PC cards did work but they were so slow the energy to credit ratio is prohibitive and the recent CPU Apps are almost as fast. Looking at your CPU times it would appear they are similar to other machines. Comparing the Run times against the Old App also shows the expected speed up; http://setiathome.berkeley.edu/result.php?resultid=4438903635 Application version: SETI@home v7 v7.07 (sse41) WU true angle range is : 0.010699 Run time: 2 hours 22 min 57 sec CPU time: 1 hours 59 min 55 sec http://setiathome.berkeley.edu/result.php?resultid=4438662118 Application version: SETI@home v7 v7.00 WU true angle range is : 0.012306 Run time: 5 hours 1 min 24 sec CPU time: 3 hours 24 min 8 sec I don't know how that i5 should run, but it appears to be working normally. You could look around and see how other similar machines are running, maybe even reduce it down to just 2 running cores and see how the Turbo boost works. |
Urs Echternacht Send message Joined: 15 May 99 Posts: 692 Credit: 135,197,781 RAC: 211 |
Mac OS X/64-bit Intel 7.08 (opencl_ati5zc_mac) 16 Oct 2015, 18:45:36 UTC 1,922 GigaFLOPS New Mac OS X GPU apps are released, but the problem with the "opencl_ati_mac" app version of MBv7 and ATI HD 4000 Series GPUs are not resolved. Sorry. Hopefully someone of you that have such hardware are willing to help to fix this ? add: i'll leave this call for help for a week here, before deprecating the "opencl_ati_mac" app. _\|/_ U r s |
©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.