I was watching BOINC Manager as the 4 remaining running tasks I had were finishing. I'm running an Intel i7 with 4 cpus and an ATI 7820 GPU (with from what I gather has 10 cores). One of the tasks was an Astropulse 505 v5.05. The others were SETI@homeEnhanced v6.03. I was following the Windows Resource Monitor to see how the CPUs responded. The Astropulse task finished first. Apparently it was running on CPU 0 since that processor activity essentially went to 0. The 3 remaining processors seemed to switch nicely to carry extra load for the SETI@homeEnhanced v6.03 tasks, but CPU 0 was essentially not running. I wonder if Astropulse is not releasing some resources or something to make that core available. I have no idea what's going on with the GPU. I just learned OpenCL a few days ago.

Can you shut down and re-start the BOINC Manager then post the opening messages?

Cheers.

---

This is what came from the event log, most of it is failed uploads and downloads., Forget about the ATI 5700 thing. I had to hump the display driver to get OpenCL to recognize the GPU. Bottom line is that even without the GPU things don't seem quite right. I can provide additional information if you point me in the right direction.

9/30/2011 9:24:44 PM | | Starting BOINC client version 6.12.34 for windows_x86_64
9/30/2011 9:24:44 PM | | log flags: file_xfer, sched_ops, task
9/30/2011 9:24:44 PM | | Libraries: libcurl/7.21.6 OpenSSL/1.0.0d zlib/1.2.5
9/30/2011 9:24:44 PM | | Data directory: C:\ProgramData\BOINC
9/30/2011 9:24:44 PM | | Running under account dlrichter
9/30/2011 9:24:44 PM | | Processor: 4 GenuineIntel Intel(R) Core(TM) i7 CPU Q 840 @ 1.87GHz [Family 6 Model 30 Stepping 5]
9/30/2011 9:24:44 PM | | Processor: 256.00 KB cache
9/30/2011 9:24:44 PM | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 syscall nx lm vmx smx tm2 popcnt pbe
9/30/2011 9:24:44 PM | | OS: Microsoft Windows 7: Ultimate x64 Edition, Service Pack 1, (06.01.7601.00)
9/30/2011 9:24:44 PM | | Memory: 7.93 GB physical, 15.85 GB virtual
9/30/2011 9:24:44 PM | | Disk: 237.70 GB total, 111.76 GB free
9/30/2011 9:24:44 PM | | Local time is UTC -7 hours
9/30/2011 9:24:44 PM | | ATI GPU 0: ATI Radeon HD 5700 series (Juniper) (CAL version 1.4.1523, 1024MB, 1120 GFLOPS peak)
9/30/2011 9:24:44 PM | SETI@home | URL http://setiathome.berkeley.edu/; Computer ID 6176855; resource share 100
9/30/2011 9:24:44 PM | SETI@home | General prefs: from SETI@home (last modified 21-Sep-2011 07:38:08)
9/30/2011 9:24:44 PM | SETI@home | Host location: none
9/30/2011 9:24:44 PM | SETI@home | General prefs: using your defaults
9/30/2011 9:24:44 PM | | Reading preferences override file
9/30/2011 9:24:44 PM | | Preferences:
9/30/2011 9:24:44 PM | | max memory usage when active: 7304.55MB
9/30/2011 9:24:44 PM | | max memory usage when idle: 8116.17MB
9/30/2011 9:24:44 PM | | max disk usage: 100.00GB
9/30/2011 9:24:44 PM | | (to change preferences, visit the web site of an attached project, or select Preferences in the Manager)
9/30/2011 9:24:44 PM | | Not using a proxy
9/30/2011 9:24:45 PM | SETI@home | Started upload of ap_10jl11aa_B5_P0_00037_20110915_02322.wu_2_0
9/30/2011 9:24:45 PM | SETI@home | Started upload of ap_25se10af_B6_P1_00356_20110831_19539.wu_4_0
9/30/2011 9:24:45 PM | SETI@home | Started download of 21jn11ac.20139.11524.3.10.228
9/30/2011 9:24:45 PM | SETI@home | Started download of 21jn11ac.20162.11524.4.10.202
9/30/2011 9:25:06 PM | SETI@home | Temporarily failed upload of ap_10jl11aa_B5_P0_00037_20110915_02322.wu_2_0: connect() failed
9/30/2011 9:25:06 PM | SETI@home | Backing off 13 min 34 sec on upload of ap_10jl11aa_B5_P0_00037_20110915_02322.wu_2_0
9/30/2011 9:25:06 PM | SETI@home | Temporarily failed upload of ap_25se10af_B6_P1_00356_20110831_19539.wu_4_0: connect() failed
9/30/2011 9:25:06 PM | SETI@home | Backing off 12 min 39 sec on upload of ap_25se10af_B6_P1_00356_20110831_19539.wu_4_0
9/30/2011 9:25:06 PM | SETI@home | Started upload of ap_15my10aa_B0_P0_00063_20110901_26236.wu_2_0
9/30/2011 9:25:06 PM | SETI@home | Started upload of 23jn11aa.31205.21744.13.10.129_0_0
9/30/2011 9:25:14 PM | | Project communication failed: attempting access to reference site
9/30/2011 9:25:16 PM | | Internet access OK - project servers may be temporarily down.
9/30/2011 9:25:28 PM | SETI@home | Temporarily failed upload of ap_15my10aa_B0_P0_00063_20110901_26236.wu_2_0: connect() failed
9/30/2011 9:25:28 PM | SETI@home | Backing off 13 min 3 sec on upload of ap_15my10aa_B0_P0_00063_20110901_26236.wu_2_0
9/30/2011 9:25:28 PM | SETI@home | Temporarily failed upload of 23jn11aa.31205.21744.13.10.129_0_0: connect() failed
9/30/2011 9:25:28 PM | SETI@home | Backing off 39 min 17 sec on upload of 23jn11aa.31205.21744.13.10.129_0_0
9/30/2011 9:25:28 PM | SETI@home | Temporarily failed download of 21jn11ac.20139.11524.3.10.228: connect() failed
9/30/2011 9:25:28 PM | SETI@home | Backing off 10 hr 24 min 51 sec on download of 21jn11ac.20139.11524.3.10.228
9/30/2011 9:25:28 PM | SETI@home | Temporarily failed download of 21jn11ac.20162.11524.4.10.202: connect() failed
9/30/2011 9:25:28 PM | SETI@home | Backing off 9 hr 21 min 18 sec on download of 21jn11ac.20162.11524.4.10.202
9/30/2011 9:25:28 PM | SETI@home | Started download of 21jn11ac.20139.11524.3.10.234
9/30/2011 9:25:28 PM | SETI@home | Started download of 21jn11ac.20139.11524.3.10.239
9/30/2011 9:25:31 PM | | Project communication failed: attempting access to reference site
9/30/2011 9:25:32 PM | | Internet access OK - project servers may be temporarily down.

How should the fourth CPU (core) be occupied if only three tasks remain?

Gruß,
Gundolf

I think the fourth CPU core could be used to heat the house. More seriously, the Astropulse task finished first and CPU core 0 activity went from 100% to near 0%. As the 3 SETI@home Enhanced tasks finished one by one, the workload seemed to distribute itself over cores 1,2 and 3 yet core 0 remained unused. All tasks that had completed the compute phase were stuck in the upload phase while the server was down. The SETI@home tasks displayed the behavior I would expect, namely using the CPU resources as they became available. Cores 1,2 and 3 stayed at near 100% usage as the SETI@home tasks wound down from 3 to 2 to 1 as they finished computing and went into the upload stage. I hope I'm making myself clear. I'm just a bit curious about this.[/code]

Makes sense to my English head as well :-)

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

yes 1 core per WU. and it isn't an opencl issue since that is only used on the GPU WU's for ATI and newer nVidia cards

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

My understanding is that OpenCL has the concept of compute devices and should be able to use CPUs as well as GPUs and anything else that could be considered a compute device. But Astropulse working on a single thread starts to make sense. The SETI@home seemed to possibly be multi-threaded. I'm pretty sure the WUs spread across the remaining 3 cores as they finished, so when there was only one left to execute there was still lots of activity on those 3 cores. I didn't have a long time to observe it though.

My understanding is that OpenCL has the concept of compute devices and should be able to use CPUs as well as GPUs and anything else that could be considered a compute device. But Astropulse working on a single thread starts to make sense. The SETI@home seemed to possibly be multi-threaded. I'm pretty sure the WUs spread across the remaining 3 cores as they finished, so when there was only one left to execute there was still lots of activity on those 3 cores. I didn't have a long time to observe it though.

Your computers are using the stock S@H applications. Which are not OpenCL based nor multi-threaded. Also there is not currently a stock S@H application for ATI video cards. BOINC, the application that hosts S@H, will run several instances of the applications to use all of your CPU cores if you set it to do so.

To use ATI GPUs with S@H you would need to use the optimized applications from Lunatics.

When you say ATI 7820 I am guessing that is an ATI FirePro M7820 in a HP EliteBook 8740w. I am not sure if they have released a driver for the FirePro or FireGL cards, but currently there is not a driver for the Mobile Radeon GPUs that will allow the optimized ATI application to work.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

Yes it's an ATI FirePro M7820 in a Dell box. I got the driver from the AMD website after I was trying to get some OpenCL code to work and got an error about a version mismatch. When I used the Dell driver OpenCL could see my CPUs, but not the GPU. Turns out other people had run into the same problem. The solution is to get the Radeon HD 5700 driver available at AMD/ATI. Now my OpenCL program sees everything; 4 CPU cores and 10 GPU cores. I'll check out Lunatics. What you said seems to make some sense out of what I was seeing. Thanks much!!!

Doug

Wow! I got the optimized software for the ATI GPU from Lunatics and got a 2.5 times speed increase over the generic stuff for S@H Enhanced. WU that used to take 5.5 hours are now taking slightly over 2 hours. I am impressed. Thanks for pointing me in the right direction. It will be interesting to see what it does when it hits an Astropulse task.

Doug

just remember that your laptop is going to generate a lot of heat. You should consider a nice cooler to keep it from frying on you

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

Wow! I got the optimized software for the ATI GPU from Lunatics and got a 2.5 times speed increase over the generic stuff for S@H Enhanced. WU that used to take 5.5 hours are now taking slightly over 2 hours. I am impressed. Thanks for pointing me in the right direction. It will be interesting to see what it does when it hits an Astropulse task.

Doug

Did you use the installer ?

You just running on CPUs not GPU yet.

---

With each crime and every kindness we birth our future.

Wow! I got the optimized software for the ATI GPU from Lunatics and got a 2.5 times speed increase over the generic stuff for S@H Enhanced. WU that used to take 5.5 hours are now taking slightly over 2 hours. I am impressed. Thanks for pointing me in the right direction. It will be interesting to see what it does when it hits an Astropulse task.

Doug

Did you use the installer ?

You just running on CPUs not GPU yet.

I can't seem to get GPU work for my HD5770 eithier, if i have Nvidia GPU work fetch enabled i can get Cuda work without too much problem,
But if i disable Nvidia GPU work fetch (as well as CPU work fetch) and just do ATI GPU work fetches, i don't receive any ATI work at all, if i snip a Wu out of my client_state.xml and try and get it resent to my ATI nothing happens,
i keep getting Message from server: No tasks sent:

02/10/2011 11:01:20 SETI@home [sched_op_debug] Starting scheduler request
02/10/2011 11:01:20 SETI@home [wfd] request: 191196.19 sec CPU (0.00 sec, 0.00) NVIDIA GPU (0.00 sec, 0.00) ATI GPU (191196.19 sec, 1.00)
02/10/2011 11:01:20 SETI@home Sending scheduler request: To fetch work.
02/10/2011 11:01:20 SETI@home Reporting 1 completed tasks, requesting new tasks for GPU
02/10/2011 11:01:20 SETI@home [sched_op_debug] CPU work request: 0.00 seconds; 0.00 CPUs
02/10/2011 11:01:20 SETI@home [sched_op_debug] NVIDIA GPU work request: 0.00 seconds; 0.00 GPUs
02/10/2011 11:01:20 SETI@home [sched_op_debug] ATI GPU work request: 191196.19 seconds; 1.00 GPUs
02/10/2011 11:02:09 SETI@home Scheduler request completed: got 0 new tasks
02/10/2011 11:02:09 SETI@home [sched_op_debug] Server version 613
02/10/2011 11:02:09 SETI@home Message from server: No tasks sent
02/10/2011 11:02:09 SETI@home Project requested delay of 303 seconds
02/10/2011 11:02:09 SETI@home [sched_op_debug] handle_scheduler_reply(): got ack for result 24ap11ah.7385.7434.12.10.163_1
02/10/2011 11:02:09 SETI@home [wfd] backing off ATI GPU 59 sec
02/10/2011 11:02:09 SETI@home [sched_op_debug] Deferring communication for 5 min 3 sec
02/10/2011 11:02:09 SETI@home [sched_op_debug] Reason: requested by project
02/10/2011 11:02:09 [wfd] Request work fetch: RPC complete

Claggy

Very weird.
Just checked client state.
ATI backoff was at 4800 sec.

02.10.2011 17:02:29 SETI@home [work_fetch] ATI GPU: fetch share 0.00 LTD 0.00 backoff dt 4800.00 int 4800.00 (comm deferred)

Resheduled a few units to check behaviour.

---

With each crime and every kindness we birth our future.

I just used the Lunatics installer and ...

used the SSSE3x radio button (since I have an Intel i7). I wasn't sure if I should use the SSE4.1 or not. I haven't searched that one out yet.

went to the next page and enabled processing for AstroPulse v5.05 (ATI) and MultiBeam (ATI). I used the ap_5.06_win_x86_SSE2_OpenCL_ATI_r521 radio button for AstroPulse and the MB_6.10_win_SSE3_ATI_r177 radio button for MultiBeam. That decision was purely based off the descriptions in the installer. I didn't get a whole lot out of the documents I could find on the Lunatics website to point me in the right direction. Part of my problem is that I had to swap out my Dell provided ATI driver for my ATI FirePro M7820 because I couldn't get the ATI OpenCL sdk to recognize my GPU. Some others had a similar problem and fixed it by getting a different driver from ATI. It now says that I have an ATI Mobility Radeon HD 5800 Series, OpenCL can now see my GPU and everything is hunky-dory I'm not sure about my MultiBeam decision, but it seems to be running like gangbusters. It does seem that since my computer thinks I have a HD 5800 I should use MB_6.10_win_SSE3_ATI_HD5_r177, but I went the more conservative route since I'm just kind of blindly doing this. Make sense? If I hadn't started to program the damn computer and using calls like clGetDeviceIDs and clGetDeviceInfo I would have never figured this stuff out. Well, that and a lot of informed people who know a lot more than me, have been through this and post. There must be something out there like CPU-Z for GPUs. I'll probably be brain dead by the time I really understand all of this. I am having fun with OpenCL now though.

GPU-Z: http://www.techpowerup.com/gpuz/
GPU Caps Viewer: http://www.ozone3d.net/gpu_caps_viewer/

---

I watched again as the WUs wound down, distributing work among the idled cores, now that I'm using the Lunatics apps. It behaved exactly as I expected which would explain its 2.5 times speed increase even though I have yet to hit the GPU as far as I can tell. I can't wait to get a ATI GPU task to see how it does. Thanks again to everyone for helping me understand this. I've learned a lot.

Doug

That's interesting for the first scenario. I've been told but haven't seen it on Linux machines, a single-threaded (or single-CPU) application will be assigned one core/CPU at the start, and will stay on that core/CPU until it finishes or gets interrupted by the CPU scheduling function in the kernel.

Windows will balance the load across all the available cores. 4 cores with a single-threaded task will result in all four cores having ~25% load unless you use task manager to manually set the affinity for that particular process.

I don't know why you experienced the completely idle core issue with the stock setup. I thought the OS kernel was the one that made the scheduling decisions like that, but changing from stock to Lunatics seems to have "fixed" that weird behavior.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

I think Bill Gates had something to do with it. I'm going back to Linux as soon as I can get my internet phone (MagicJack) to work on it or buy a second computer. I've been waiting 2 years.