Just installed 6.5.0 from Developer Versions for BOINC Manager. Immediately my boxes load new work and his with hight prio mode works now in normal mode. The new version use on my boxes 0.05 CPU instead 0.9. Don't know, if now the GPU task will take more time, will see in 6:20 hours on my GTX280. More later ...

---

Looks like the developer has found the philosopher's stone with the 6.5.0. I see a CPU usage of 2 to 4% and 4+1 WUs running. After a runtime of real 40 minutes the task has worn out 6.8%. This looks like a runtime of 9:48 h. It's a very rough estimate, but it looks fine for me.

---

Hmm, well ... did not work for me. I get the not avail for your type of computer message.

Hmm, well ... did not work for me. I get the not avail for your type of computer message.

I've installed now on 4 PCs. 3 of them running under Vista 64 bit (GTX280, 2xGTX260², 8800GT) with the 64 bit version, one on XP 32 bit (GTX260²). All are running fine. Try to reset the project, but only if you have no WU running. Maybe your DCF is messed up.

---

Yeah same here for me too. Only have 1 WU running and have to go to work. Guess I'll do a reset when I get home. Too bad my 280's are going to sit idle in a couple of hours with no work sent!!

Cheers

Hmm, well ... did not work for me. I get the not avail for your type of computer message.

I've installed now on 4 PCs. 3 of them running under Vista 64 bit (GTX280, 2xGTX260², 8800GT) with the 64 bit version, one on XP 32 bit (GTX260²). All are running fine. Try to reset the project, but only if you have no WU running. Maybe your DCF is messed up.

I did a project reset ... nope ...

So, i have the "correct" version of BOINC installed done a project reset ... nada ...

Detach, re-attach ... nope ...

Something else is going on ... what it is .... who knows.

I saw something akin to this on another project and it had to do with a bug in the feeder. A minor change to the feeder was made and it eased the problem though the underlying cause was never fixed (in *MY* opinion because it would have required a change to the design of the feeder, besides I was the one making the critique, always a fatal flaw).

Anyway ... I guess I will just fiddle with it and monitor to see if it ever gets fixed.

........... doesn't make a jot of difference for me.

I feel I've tried everything that is suggested or known to try and get my client - on multiple hosts - working the way it used to up until recently. However, nothing currently is working properly or relaibly - I do think the application/client environment has escaped from the control of the project team.

P.

Maybe this can help: I have under Advanced, Preferences the additional work buffer set to 2 days and the Connect about on 0.02 days.

---

Looks like the developer has found the philosopher's stone with the 6.5.0. I see a CPU usage of 2 to 4% and 4+1 WUs running. After a runtime of real 40 minutes the task has worn out 6.8%. This looks like a runtime of 9:48 h. It's a very rough estimate, but it looks fine for me.

Which OS are you using? Under XP SP3 I have still the high CPU-usage as before. Any remarks?

Which OS are you using? Under XP SP3 I have still the high CPU-usage as before. Any remarks?

Sorry for the late answer, needed some time to collect ;)

8800GT ........ XP 32 bit, AMD X2 6000+ 3 GHz, 16-20% CPU usage, (0.03% CPU 1 CUDA)
8800GT OC .. Vista Ultimate 64 bit, Phenom 9850 BE 2.5 GHz, 7-8% CPU usage, (0.05% CPU 1 CUDA)
GTX260² ....... Vista Business 64 bit, Phenom 9850 BE 2.5 GHz, 2-4% CPU usage, (0.05% CPU 1 CUDA)
GTX260² ....... Suse Linux 64 bit, Intel Q9550 2.83 ghZ, 1-2% CPU usage, (0.03% CPU 1 CUDA) - still 6.4.2!!!
GTX260² ....... XP Pro 32 bit, Phenom 9850 BE 2.5 GHz, 8-10% CPU usage, (0.05% CPU 1 CUDA)
GTX280 ........ Vista Ultimate 64 bit, Phenom 9950 BE 2.6 GHz, 2-4% CPU usage, (0.05% CPU 1 CUDA)

---

Thank you for your answer. I have nearly the same system then your first. 6000 X2 @ 3000 MHz and an 8800 GT.

I have running one CPDN as parallel task. The CPU limited to 50 % for maximum performance of GPU-Grid. Then I have one Core CPDN (50 %) and one core GPU-Grid (35-40%). Testing Linux (Ubuntu) I have 2-3% on the GPU-Grid-core.

Have you change any preferences on your first machine to lower the GPU-Grid CPU-usage under Win XP?

Have you change any preferences on your first machine to lower the GPU-Grid CPU-usage under Win XP?

Maybe the driver can be the cause. I use on the Windows 32 bit XP box the 178.13_geforce_winxp_32bit_international_whql - driver.

---

The task I downloaded yesterday under 6.4.2 started its run today under 6.5.0. on my machine with the GTX260 card. After 3.5 hours I think GPU is running a little bit slower than previously, it is now showing 32.86% done; under earlier versions of Boinc I think it would have been in excess of 35% done by now. I'll have to see if it runs any longer than 11 hours to complete the job to know for sure.

However, the GPU and the video ram temperatures are down by 5C and I am not hearing the fastest fan speed kicking in at all today!

Phoneman1

Looks like the same here. The fast cards need fast service of the CPU, so I'm sure, the 6.5.0 is slower then the other versions. If you prefer faster GPU calculation, you can change the value on:

Advanced
Preferences
Processor usage
On multiprocessor system, use at most XX% of the processors.

For XX% set 99% and you get full speed for your GPU, but one core less of your CPU for other BOINC projects.

---

After all the issues I've had recently with 6.4.2, I figured I might as well try this after Kokomiko said it worked for him.

I had 3 GPUGrid WU's pending running, with estimated times of 200+ hours (my 8800GT's take about 12.5 hours on the shorter WU's). After installing 6.5.0, the 3 pending WU's showed now < 15 hours each for estimated time. Immediately, BOINC connected and downloaded another WU (so I have a max of 4 on my twin 8800GT's). Two are now running, neither are High Priority. It's still very early on, but I noticed a dramatic difference in several things:

1) estimated times are back lower without any project reset, etc.
2) automatically download a new WU without having to suspend other project or manually connect.
3) no longer running High Priority

We'll see how it goes moving forward, but from an instant glance 6.5.0 works much better for me than 6.4.2 (or any other previous version) in relation to the problems we've been seeing.

The runtime in real time for my first 2435.94 credit WU on BM 6.5.0 on a GTX280 is 10:54h, the approximate elapsed time for the WU is 18576.841 s (=5:09:37h). Looks like this time is no longer the real runtime of the whole WU, only the CPU time. 10:54h is way too long for a GTX280, so this version gives away a lot of the power of the GPU. The temperature was the whole time 5°C lower then with the 6.4.2. I will change the processor usage to 99% to get the full power of my card for GPUGrid and waivers the additional core for CPU crunching.

---

We should come out with an optimized Windows version soon.

gdf

DCF is beginning to drift upward, now over 7 on the machine with GTX260 card after 24+ hours running with 6.5.0. On the 8600GTS machine over 2 after similar time.

Also both machines have stopped getting automatic downloads. I've had to suspend all processing and edit the client_state.xml files on both machines again. The nrpc_failures values on both were non-zero. I don't know what is causing that at the moment. It can't be the Boinc installer because I haven't run it since I installed 6.5.0 a day ago and I did get some automatic downloads on 6.5.0 yesterday.

The files downloaded since correcting the file are showing longer likely completion times in BoincView (due to the dcf value) and the ones on the GTX260 machineare likely to run in high priority mode as the estimated run time exceeds the deadline by nearly 4 days!

Phoneman1

BOINC 6.5.0 has now been running unattended for about 16 hours. It has completed two WU's and fetched two more, without any intervention. My DCF remains about 0.97. As far as I can report so far, it is working as expected.

For me 6.5.0 works fine. I could suggest this version, but I don|t want to be against GDF who suggested 6.4.2 - but that|s the only one reason :).

---