WARNING: This website is obsolete! Please follow this link to get to the new Albert@Home website!
Running on ATI |
Message boards :
Problems and Bug Reports :
Running on ATI
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · Next
Author | Message |
---|---|
zombie67 [MM] Send message Joined: 10 Oct 06 Posts: 130 Credit: 30,924,459 RAC: 0 |
About CPU usage and run times. Q9550 w/ 5870 , 11.11 I set BOINC to use only 3 of 4 cores. Running no GPU task, total CPU usage was ~80%. Run an albert ATI task and CPU usage jumped to 100%. Then set BOINC to use only 2 of 4 cores. Running no GPU task, total CPU usage was ~53%. Run an albert ATI task and CPU usage jumped to ~82%. Summary: This ATI app, on this hardware, needs (or at least uses) ~29% of the total CPU power, or slightly more than a full core. I am going to let it run a full task with BOINC using only two cores, then using three cores, and will report on run times. Edit: One more data point. BOINCmanager says this task is using 0.15 CPU + 1.00 GPU. I am not sure where that number comes from, but it clearly way off the mark. Dublin, California Team: SETI.USA |
Oliver Behnke Volunteer moderator Project administrator Project developer Send message Joined: 4 Sep 07 Posts: 130 Credit: 8,545,955 RAC: 0 |
BOINCmanager says this task is using 0.15 CPU + 1.00 GPU. I am not sure where that number comes from, but it clearly way off the mark. Thanks for the info. These settings are set by us and they are realistic. Please note that there's a known bug in AMD's Catalyst drivers that can cause 100% CPU load (single core) when there are more than one GPUs available. This bug shouldn't apply in your case since you have only one GPU as far as I can tell. Or is there another one not reported by BOINC? You may test Catalyst 11.6 (although our app's recommended drivers are 11.7 and later) to see whether it's in fact the driver as some have reported that this issue can also affect single GPU setups... Cheers, Oliver |
zombie67 [MM] Send message Joined: 10 Oct 06 Posts: 130 Credit: 30,924,459 RAC: 0 |
There is only one CPU in that machine. FWIW, run times with only 2 of 4 cores crunching are about the same as 3 of 4 cores. No noticeable difference. Below are some examples. But running with all 4 cores, the time has shot way up. I don't have any examples to share yet. And at this rate, it looks like it will take about 12-13 hours to complete. So maybe tomorrow I will have an example to share. I think this means the task is CPU-starved. With 50%: http://albert.phys.uwm.edu/result.php?resultid=45270 Run time 4,650.91 CPU time 4,331.19 http://albert.phys.uwm.edu/result.php?resultid=48300 Run time 5,166.38 CPU time 4,776.22 (FWIW, validated against CUDA) With 75%: http://albert.phys.uwm.edu/result.php?resultid=45059 Run time 5,165.22 CPU time 4,847.52 http://albert.phys.uwm.edu/result.php?resultid=45079 Run time 5,594.42 CPU time 4,749.09 Edit: One other note: For CUDA apps, the CPU time is about 25% of the run time. For these ATI tasks, it is about 95%. Dublin, California Team: SETI.USA |
TRuEQ & TuVaLu Send message Joined: 11 Sep 06 Posts: 75 Credit: 615,315 RAC: 0 |
Thank you Oliver for the info. And zoombie for the comparissons. Mayby I should have tried with only this project running so that all CPU is free. I think one should be able to use the non-used cores to other projects and the 70% of the used core to other projects as well. I hope you will solve this in future releases. I am in pause here until 6.13.xx is boinc recomended version. |
pragmatic prancing periodic problem child, left Send message Joined: 26 Jan 05 Posts: 1639 Credit: 70,000 RAC: 0 |
Updated to 11.7 Catalysts. Now I'm using a full CPU core, whereas before it would use 10-20%. It's good I set BOINC to only use 2 cores, that way I have one left for non-BOINC stuff (2 core plus HT). Jord. BOINC FAQ Service They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake. |
Oliver Behnke Volunteer moderator Project administrator Project developer Send message Joined: 4 Sep 07 Posts: 130 Credit: 8,545,955 RAC: 0 |
Updated to 11.7 Catalysts. Now I'm using a full CPU core, whereas before it would use 10-20% See, that's what I said. It's a pitty but AMD's drivers are just a big PITA... The colleagues at SETI have similar issues and maybe we'll try a previous SDK/driver version (most likely 2.4/11.4) but that comes with its own set of issues. There's no universal solution right now. Oliver |
zombie67 [MM] Send message Joined: 10 Oct 06 Posts: 130 Credit: 30,924,459 RAC: 0 |
But running with all 4 cores, the time has shot way up. I don't have any examples to share yet. And at this rate, it looks like it will take about 12-13 hours to complete. So maybe tomorrow I will have an example to share. I think this means the task is CPU-starved. For the record, after about 6 hours, the whole system locked up and required a hard reboot. Upon reboot, the work for the ATI task was not lost. That's good. But I could not wait any longer (our team is joining another challenge today, which requires 6.12.xx or less). So I changed the CPU usage down to 50% to get it completed quickly. I won't have an example to show for a week or so. But I think we all know the behavior well enough now. With 11.11, the ATI tasks require a whole free core. Dublin, California Team: SETI.USA |
TRuEQ & TuVaLu Send message Joined: 11 Sep 06 Posts: 75 Credit: 615,315 RAC: 0 |
Ok, now i figure out how to run albert on my system. I use a dual-core with ATI 5850 and driver Catalyst 11.9 When I have 1 free core for albert the GPU load is 65-75% , 99% of the time. When I change CPU priority between low and high I can't see any difference in performance. The core is dedicated to albert when the app needs it. The CPU load is between 15-25% of the free core. progress meter shows 29% done in 29minutes and it looks very stable. Thanks to you who brought the free core information to this thread. |
Tex1954 Send message Joined: 1 Dec 11 Posts: 10 Credit: 25,500 RAC: 0 |
I have two systems I am testing on. 1) AMD955BE 3.7GHz with HD6990 930MHz. 2) Intel i7-950 4.259GHz with DUAL GTX560 Ti SC 900MHz Performance of the 1.19 tasks on the Nvidia GTX560 CUDA32 is about seven (7) times faster than on the ATI HD6990 OpenCL. I just wonder why it runs slow on the HD6990 since the HD6990 is 7 times faster than the GTX560 on Milkyway tasks. Also, I have to run the 955BE at 50% CPU's to allow two cores to service the HD6990... otherwise it chocks. This is similar to Moo! Wrapper. Very heavy CPU use... one full core per GPU. You can see the difference in CPU times on the two systems... ATI uses a ton more CPU... 8-) http://albert.phys.uwm.edu/results.php?userid=333355&offset=0&show_names=0&state=2&appid= |
Tex1954 Send message Joined: 1 Dec 11 Posts: 10 Credit: 25,500 RAC: 0 |
Boo Boo, I meant to say Performance of the 1.19 tasks on the Nvidia GTX560 CUDA32 is about two (2) times faster and about sever (7) times less CPU than on the ATI HD6990 OpenCL. 8-) |
Jon Fox Send message Joined: 9 Feb 05 Posts: 7 Credit: 138,252 RAC: 0 |
Received a large set of WUs on 26 November, all failed with "Error while downloading" http://albert.phys.uwm.edu/host_sched_logs/1/1301 or "Error while computing" http://albert.phys.uwm.edu/host_sched_logs/1/1301 Core Client Startup Event Log:
Tue Nov 29 20:26:26 2011 | | This a development version of BOINC and may not function properly Tue Nov 29 20:26:26 2011 | | log flags: file_xfer, sched_ops, task Tue Nov 29 20:26:26 2011 | | Libraries: libcurl/7.21.7 OpenSSL/0.9.7l zlib/1.2.5 c-ares/1.7.4 Tue Nov 29 20:26:26 2011 | | Data directory: /Library/Application Support/BOINC Data Tue Nov 29 20:26:26 2011 | | Processor: 4 GenuineIntel Intel(R) Core(TM) i5-2500S CPU @ 2.70GHz [x86 Family 6 Model 42 Stepping 7] Tue Nov 29 20:26:26 2011 | | Processor features: FPU VME DE PSE TSC MSR PAE MCE CX8 APIC SEP MTRR PGE MCA CMOV PAT PSE36 CLFSH DS ACPI MMX FXSR SSE SSE2 SS HTT TM PBE SSE3 PCLMULQDQ DTES64 MON DSCPL VMX SMX EST TM2 SSSE3 CX16 TPR PDCM SSE4.1 SSE4.2 xAPIC POPCNT AES PCID XSAVE OSXSAVE Tue Nov 29 20:26:26 2011 | | OS: Mac OS X 10.7.2 (Darwin 11.2.0) Tue Nov 29 20:26:26 2011 | | Memory: 4.00 GB physical, 906.57 GB virtual Tue Nov 29 20:26:26 2011 | | Disk: 930.71 GB total, 906.32 GB free Tue Nov 29 20:26:26 2011 | | Local time is UTC -6 hours Tue Nov 29 20:26:26 2011 | | OpenCL: ATI GPU 0: ATI Radeon HD 6770M (driver version 1.0, device version OpenCL 1.1, 512MB) Tue Nov 29 20:26:26 2011 | | ATI GPU is OpenCL-capable
|
Oliver Behnke Volunteer moderator Project administrator Project developer Send message Joined: 4 Sep 07 Posts: 130 Credit: 8,545,955 RAC: 0 |
I just wonder why it runs slow on the HD6990 since the HD6990 is 7 times faster than the GTX560 on Milkyway tasks. Simple, the implementation is not yet fully optimized for AMD's GPU architecture.
Please see my previous posts: that's most likely due to a known bug in the AMD driver. Oliver |
TRuEQ & TuVaLu Send message Joined: 11 Sep 06 Posts: 75 Credit: 615,315 RAC: 0 |
Tex1954 What driver are you running on the ATI?? They say that there is a bug in newer ATI drivers that consumes 100%CPU. Look at my tasks. http://albert.phys.uwm.edu/results.php?hostid=1353 I have a CPU load between 15-25% I have an ATI 5850 and use Catalyst 11.9 Parhapse you can try the 11.9 drivers and it will work with the 69xx as well. I think it is worth a try. Nice card btw, and it looks as you say that the card has 2GPU which needs 2CPU cores to feed it. |
Tex1954 Send message Joined: 1 Dec 11 Posts: 10 Credit: 25,500 RAC: 0 |
I just wonder why it runs slow on the HD6990 since the HD6990 is 7 times faster than the GTX560 on Milkyway tasks. Thank you. I didn't know about an AMD bug since I only received the 6990 recently and it's the first ATI/AMD GPU product I have ever had. Started with 3DFX Voodoo's back when and been Nvidia since... figured I would try the fastest card in the world to see how it would crunch... And BTW, the new AMD FX processors don't crunch well at all... at least the FX-8120 stinks... http://www.overclock.net/t/1143670/fx-8120-boinc-benchmarks-real-world-stuff 8-) |
Tex1954 Send message Joined: 1 Dec 11 Posts: 10 Credit: 25,500 RAC: 0 |
Tex1954 I have the 11.9 drivers as well running on Win7-64b... I don't know about that bug, but it does in fact require 2 cores out of 4 to keep it fed... 8-) |
TRuEQ & TuVaLu Send message Joined: 11 Sep 06 Posts: 75 Credit: 615,315 RAC: 0 |
tex 1954 Hmm... You are using 11.9 the same as me.. I use 5xxx you 6xxx I use 1gpu you 2gpu I use 32bit you 64bit My guess here is that ATI drivers and multiGPU together with OpenCl makes the CPU-core goes 100% Albert uses 2feeding cores to supply your 2GPU's whatever the CPU load is here at albert it still requires 1 free CPU to feed each GPU. I guess if you try to run Milkyway and Collatz your CPU usage is much lower. I hope ATI can work with OpenCl programmers to fix this (bug). |
Tex1954 Send message Joined: 1 Dec 11 Posts: 10 Credit: 25,500 RAC: 0 |
tex 1954 That is exactly correct and true for the HD6990. In contrast, the Nvidia CUDA tasks use very little CPU resources. ATI/AMD seems a bit behind CUDA in their API's for sure. Let's just hope they progress more quickly and perhaps get their stream API working more easily. OpenCL is maybe nice for games or whatever, but linking to generalized libraries vs. optimized has to be part of the problem; especially considering the massive 2Gig memory available to each GPU on the HD6990. It certainly seems something is lacking in whatever SDK is being used... :) |
TRuEQ & TuVaLu Send message Joined: 11 Sep 06 Posts: 75 Credit: 615,315 RAC: 0 |
[quote]tex 1954 Albert requires as Oliver says here: http://albert.phys.uwm.edu/forum_thread.php?id=8838 in another thread. For instance SETI BETA openCl can use older drivers like 11.2-11.4 with SDK 2.4 and they get around this CPU bug problem. Here we just have to wait for someone somewhere to fix this with newer drivers And: I don't think CUDA app here uses OpenCL yet..... That's why there is a great performance difference as of now, maybe not in the future.... |
Tex1954 Send message Joined: 1 Dec 11 Posts: 10 Credit: 25,500 RAC: 0 |
Well, if OpenCL applied to Nvidia GPU's slows it down, I am against it! LOL! But, I have no real idea what is really required... However, I suspect we all are on the same track and wish the same goals... that being the most efficient and speedy code that can be achieved. I'm happy to be able to crunch Einstein tasks on ATI at all! That's a great leap ahead already! All good on the developers so far! 8-) |
[SETI.USA]Tank_Master Send message Joined: 22 Jan 05 Posts: 17 Credit: 301,403 RAC: 0 |
I'm using the ATI 11.11 drivers and the 2.5SDK installed, on my i7 2600K @ 4.7GHz, i typically see 3-4% of the total CPU usage going to the albert openCL app... I have a ATI 6970 and running windows 8 x64. My typical GPU load is around 80% |