Hmmm...a man of few words.

Since it's posted, it must be safe...

____________


Consciousness: That annoying time between naps......

Experience is a wonderful thing: it enables you to recognize a mistake every time you repeat it.

It is also out for Linux, x86 and 64.

Wow, that looks a lot better than 6.3.14 and the 32bit 6.3.18 to me...
Now without setting cpus or needing to suspend most tasks, it runs 2+1 tasks on a Dualcore with 1 GPU...
I suspended the units, restarted the client etc, it keeps the same number running

Much better now :)

Well it seems to be working better. 4 CPU tasks plus 1 CUDA.

However it did something odd when I first started....it says it generated a new CPID for the PC and trashed the old CUDA WU's in the cache. The WU status briefly changed to "Cancelled by Project" or something similar and then it downloaded a new set of CUDA WU's.

The box is this one: http://www.gpugrid.net/show_host_detail.php?hostid=10619

It shows 3 client aborted WU's, followed by 4 new ones. I've never seen that behaviour before when upgrading a version of Boinc. However, it does seem to be running correctly.
____________

I don't see any difference from 6.3.17. Runs fine on Win64 (so far...). I have been running 6.3.17 fine with 2+1 running, without making any changes in client settings.

2 cpu + 1 gpu task on an AMD X2, just like 6.3.17.

With the GPU using 16% of cpu cycles, just like 6.3.17.

I did not lose any GPU tasks (1 running, 1 ready to run) after the upgrade.

____________


Consciousness: That annoying time between naps......

Experience is a wonderful thing: it enables you to recognize a mistake every time you repeat it.

Now the 2nd box is upgraded and this time no lost WU's. Odd....

Not to worry: it looks good with no pre-empted (waiting) Cuda or CPU WU's so far
____________

Installed on 5 Box's with no Problems or lost Wus, upon startup all 5 Box's showed 5 Wu's running with a CUDA Wu being 1 of them. But now after 10-15 Minutes only 1 Box still shows 5 Wu's runnning with the other4 showing 4 Wu's running with a CUDA Wu being 1 of them ...

Installed on 6 machines now: across Linux, Vista32 and XP32. Only the first box had a strange event loosing some WU's on re-start, all others were ok.

I have the same symptoms as reported by ETA and PoorBoy - occasional switching of a CPU task to waiting, but it seems to settle down after a while. The Cuda task seems unaffected.
____________

I did two windows x86 installs. I only had a few minutes to observe them though as it was the end of the work day.

One seemed to be fine running max of 2 cpus + 1 cuda.

The other did too, but within a minute it was not. I quit it, checked all config and preferences files, all looked ok with nothing to cause that. I turned on some debug flags and restarted it, it ran fine while watching it a few minutes. scratched head. then turned of debug as I did not want to leave it on all night and within 1 minute it was only running 1 cpu + 1 cuda. It is not perfect.

I don't have time to fully debug right now, next day or so or maybe the weekend.

However on quick test I was not able to get some of the bad errors that were in 6.3.16 or 6.3.17 so maybe those are gone.

I've installed 2 on Vista 64, one on XP 64. ALl are starting with 5 tasks (4+1). When I stop the running projekt and restart them, only 4 tasks are running (3+1). After a while 5 tasks are running again. Then I shorten the switch time between the application (Preferences, switch time between the application) and I could watch a strange behavior. 4 tasks running, 5 tasks running, 4 tasks running, can't find a system behind this.
____________

Finding the same scenario everyone else is with sometimes 4 total tasks running and sometimes 5 but the gpu always runs now...no more stalls so that is good. I decided to just go with the 5 ncpus solution for now until its fixed. At least I will get maximum utilization that way ;)

running fine here on server 2008 x64. The lag in the eye-candy animations is greatly reduced, which is nice.

I think I found out why it was smother, the GPU app was taking up less than 1% of the CPU. I set it to realtime priority and its now 20% of 4 cores.

Wow, that looks a lot better than 6.3.14 and the 32bit 6.3.18 to me...
Now without setting cpus or needing to suspend most tasks, it runs 2+1 tasks on a Dualcore with 1 GPU...
I suspended the units, restarted the client etc, it keeps the same number running

Much better now :)

My quad is running 3 CPU + 1 CUDA. I have no ncpus set. Should it be lie this? Or should it be 4+1? I think this is set on the project side, right?
____________
Reno, NV
Team: SETI.USA

6.3.17 sometimes also processed only 2 tasks, CPU+GPU WUs. I'll take a look at 6.3.19 now and compare.

God damn it! Again I see only 2 tastks running, CPU+GPU... But it had honestly started 3 tasks. And after a while it became to process only 2 WUs...

OK, heres some good news. I may have captured the bug happening with some debug info. Below is part of the email I sent the developer team, along with the debug info. I hope with this they can figure it out. I had 7 debug flags set, so hopefully ever little change was recorded.

If everyone would just remain calm and patient, we will eventually get this worked out. Remember that this is all new to users and devlopers, it has only been in development part of this year, whereas boinc itself has over 4 years of development.

Host #1 still running fine, 2 CPU + 1 CUDA/CPU<1.

Host #2 which was not I discovered the CPU job after at least 2 hours was not running, even though it said it was. In windows task manager it showed 0 CPU usage, so for some reason I've not seen it was hung up.

I shut down client, set debug flags on and restarted, then left client alone, I think. It only resumed the 1 CPU job and 1 CUDA/CPU<1 at that time. **Wrong behavior*

A little while ago I looked over and it showed the screen running 1 CPU, 1 CUDA/CPU<1 and 1 CPU as waiting to run. Note several other users on PS3GRID forum report same, less than max CPU running along with the CUDA/CPU<1.

I saved all the debug info, only problem is I do not know how far back that occurred, but it has to be in there somewhere. Notice too the time shows 00:00:35, If I remember there other times this happened, it happened at that exact same time, around 34-35 seconds they go to waiting.

Attached is debug log, cc_config, global_pref, global_prefs_override, client_state screenshot of tasks and screenshot of resource shares. I hope this would be enough to figure this out. I'll look it over my self tonight when I get home, although I do not follow all the debug too well. Hope its not too much info. Oh and I remember to reduce the images to jpeg this time instead of large bitmaps.

I just started running CUDA on a quad core with two 8800 GTS's using 6.3.19. I get two CUDA units running but only two units on the quad core where I used to have four running.

I notice that in the first few seconds I get four quad core units running but when the two CUDA units start, two of the other units stop.

How do I get six units running at the same time?

More good news, A change has been checked in, next client may be perfect.

David 29 Oct 2008 client: the preemptability order was wrong

Now: // and we want the best replacement at the front,
New: // and we want the most preemptable task at the front

Have you noticed a performance decrease between leaving one CPU to manage the GPU task, or using it for other workunits (provided that you are using 6.3.19!!!)?

gdf

I have seen my times go up on a Q6600 with a gtx260 from 28000 sec to over 30000 sec. Slight increase to free up that other core. FYI


____________

GTX280


6.3.14 + 6.48
GPU=1

Time per step: 26.120 ms
Time per step: 27.412 ms

6.3.14 + 6.48
GPU<1

Time per step: 31.970 ms
Time per step: 49.019 ms
Time per step: 34.924 ms
Time per step: 35.736 ms

6.3.19 + 6.48
GPU<1
Time per step: 31.853 ms just one WU until now)


It`s just if some1 wants to know.
____________

I found something in stderr.out on one WU:

No heartbeat from core client for 30 sec - exiting

Should I be worried about it or "have a break, have a kitcat"?

same problems for me on vista x64

gpu is still starved of cpu time running 4 cpu tasks and 1 gpu. Changing to real time priority for the gpu didn't help

with 4 cpu and 1 gpu
http://www.gpugrid.net/result.php?resultid=98085

with just 1 gpu running and no other tasks
http://www.gpugrid.net/result.php?resultid=96758

Problems with vista32 - have to suspend other project to get new tasks.

Request for next version:
Have the GPU reserve/use 16% of CPU to have better performance.
eg.
quad core = 4 cpu task + 1 GPU (but GPU uses 16% of CPU)

the remaining 9% will be for other cpu task.

As of now, I set Boinc to use 3 cpu to have a better performance for GPU but losing 9% CPU doing nothing.

I don't think a setting like you suggest x-files 27 will work for everyone even though it might be perfect for you.For example one of my hosts uses approx 8% of CPU and another uses 0-2%.

Until they can develop a client that has a sliding scale on cpu usage applied to the gpu a set % will not work well. I don't think the developers are even close to this option.

3 CPU + 1 GPU
25.859 ms
21980.556 s

4 CPU + 1 GPU
46.794 ms
39774.961 s

Having the GPU use only like 1% CPU clearly loses some performance...about 55%

It's difficult to judge times, but I'd say that with 4+1 i'm not loosing more than 4% compared to 3+1. In the case of 4+1 the GPU task gets between 6 and 20% of my quad, whereas 3+1 leads to 16 - 20% utilization.

MrS

I did set my Client to use only 1 CPU on my DualCore.

That way only one Seti WU and one GPUGrid WU are running.

Times are exactly as before (no wonder - the GPU WU gets one whole Core to mess with)
I did try one GPU-WU running together with 2 CPU WUs -- the running time seemed to be >10% longer than without - so I didn´t want to lose to much speed and decided to not test that any longer.


Cu KyleFL

it's the same over here - VISTAx64 and XP32 - boinc currently is not able to assign resources to projects and fools around. it's everything:

1st: running fewer CPU-tasks than possible
2nd: suspending GPU-tasks to run CPU and vice versa
3rd: not getting CPU-WUs (and running dry) because GPU-task filled the queue
4th: requesting GPU-tasks according to number of CPUs (who the hack got that idea?)

currently it's FUBAR!

frank.

I just did the Linux64 upgrade from Boinc..16 to ..19. And re-attached to the GPUgrid project site. Seems to be fine.

I did notice my (quad) Linux64 box running (only cpu tasks) for a few minutes prior to the upgrade. But it changed to 4+1 just before I performed the upgrade, and it is still running 4+1. If it has been switching back and forth, I have not seen an increase in my wu completion times (8800GS). The Linux64 GPU has been running a steady 1-2% cpu usage for Boinc..16 and now ..19.

I have not run out of werk for either GPU or CPU (yeti...).

____________


Consciousness: That annoying time between naps......

Experience is a wonderful thing: it enables you to recognize a mistake every time you repeat it.

I don't have any problems running out of work either, that's not the issue here. The issue for me anyway is I can't get the Client to run what I want it to run. No matter what I set it to it does what it wants to do.

I really only want to run 3 CPU Wu's plus 1 GPU Wu so I set it to <ncpus>4</ncpus> which should give me what I want. But when I look most of the time it's running 4 CPU Wu's & 1 GPU Wu on all my CUDA Capable Box's. So unless I'm putting the wrong command in the <ncpus></ncpus> line in the cc_config.xml file the Client isn't doing what it's told to do, Thats the issue.

My times for the GTX 260's I have vary by as much as 2-3 hr's because of this problem ... :)

Hang in there PoorBoy, This can't stay ALPHA forever...

I haven't been playing with the <ncpus> option. I normally just use the 'max cpu's' in the project profiles I am running (Home, werk,...). Set it to 3 (for my quad) and let 'er rip. That has let me run 3+1 or 4+1 whenever I want (as long as the developers don't mess with the server settings..again..lol).

And so far my Linux64 GPU box runs 1-2% cpu usage on 3+1 and 4+1, so I don't bother changing any more. I'll have to recheck my Win64 box again to see if that is any different.

____________


Consciousness: That annoying time between naps......

Experience is a wonderful thing: it enables you to recognize a mistake every time you repeat it.

I don't have any problems running out of work either, that's not the issue here. The issue for me anyway is I can't get the Client to run what I want it to run. No matter what I set it to it does what it wants to do.

I really only want to run 3 CPU Wu's plus 1 GPU Wu so I set it to <ncpus>4</ncpus> which should give me what I want. But when I look most of the time it's running 4 CPU Wu's & 1 GPU Wu on all my CUDA Capable Box's. So unless I'm putting the wrong command in the <ncpus></ncpus> line in the cc_config.xml file the Client isn't doing what it's told to do, Thats the issue.

My times for the GTX 260's I have vary by as much as 2-3 hr's because of this problem ... :)

I've tried that too, setting my ncpu's to 3, problem is thats all it run's sometimes is 3. Two Regular Wu's & 1 GPU Wu ...

Well, while we are waiting for version 6.3.21, I've decided to try to build my own. I've read the directions and begun downloading all the parts I need. Although I will have to do the actual build when I get home, later tonight or the weekend. This way I can begin testing the fixes when made before a release and see if they bugs are fixed.

Then I Checked the timeline and Rom, two minutes ago tagged a release for 6.3.21.

So a new version will be ready very very soon.

***** Note however it is untested. I do not know if it will work correctly yet. Use at you own risk. *****