I'm using BOINC v6.4.1. Did a project reset after my last WU finished this AM. BOINC then downloaded 2 new WUs. Two pieces of good news:

1) The estimated times are much more accurate (only about 20% high, which is fine with me).
2) CPU time has fallen from 13-18% CUDA usage on my quad down to 0%. 5 processes running, 4 at 25% each, CUDA at 0%.

This is very good news as the CPU usage problem in Windows (32bit XP SP3) seems to have been fixed somehow.
The WU seems to be progressing normally percentage-wise.

The changes you're seeing are caused by the new windows app.

MrS

---

Scanning for our furry friends since Jan 2002

Phoneman1 - is 6.3.21 actually working for you?

6.3.21 is working for me. Never gave me any grief after it was worked out that I needed to set GPUGRID to 1000 share and not run too many CPU projects. I got in a little trouble when I put three CPU projects on the machine, but GPUGRID @ 1000 + one CPU @ 100 with 3+ days queue and I never run out of work.

Everyone I know who tried 6.4.anything has regretted it, sorry to say.

The changes you're seeing are caused by the new windows app.

Saw that after my post. Thanks for the heads up though.

fractal, I was doing fine with v6.4.1 until I saw the recommendation on the front page to upgrade to v6.4.5. After going to v6.4.5 nothing worked right.
Reinstalling v6.4.1 fixed some of the problems and resetting the project fixed the rest.

BTW, the project front page is still saying to go to v6.4.5. Shouldn't that be changed?

is it 6.4.2 or 6.4.5?

I am on 6.4.5 and working a 6.58 job. It now refuses to get new work. The new box that I started tonight will not get work either.

Linux os/9600GSO.

No memory leak:D

mike

Try to suspend the other projects to get work.

MrS

---

Scanning for our furry friends since Jan 2002

I did that and had also tried when GPU was the only project. Tried a reset also with no luck. I'll let it run and see if it will pick some thing up.

mike

It now refuses to get new work.

do you mean BOINC is requesting 0 seconds of new work or the dreaded "molecular dynamics for cell processor not available for your system" message?

MrS

---

Scanning for our furry friends since Jan 2002

It now refuses to get new work.

do you mean BOINC is requesting 0 seconds of new work or the dreaded "molecular dynamics for cell processor not available for your system" message?

MrS

It is asking for a bunch but getting the red letters:(

mike

I got that too ...

Try increasing your queue size too (if it is small like mine was at 0.1 days the default) ...

I checked when I got up, it had not received any new work. I did a manual update and got 2 wu's for each box,

I have 2 happy gpu's now:D

mike

I had 6.4.2 running with minor problems, i.e. I had to suspend all CPU projects to make it request new WUs.. but I got them.

Today I got the message "no work sent, cell not available for you" again, whithout any changes on my side. Project reset didn't help, the server has work, setting nrpc=0 didn't help and sending a double request (like Kokomiko suggested) didn't help. Too bad I'm out of house now until tomorrow.

Maybe the problem solved itself until then. Otherwise: any suggestions I should try?

MrS

---

Scanning for our furry friends since Jan 2002

I had 6.4.2 running with minor problems, i.e. I had to suspend all CPU projects to make it request new WUs.. but I got them.

Today I got the message "no work sent, cell not available for you" again, whithout any changes on my side. Project reset didn't help, the server has work, setting nrpc=0 didn't help and sending a double request (like Kokomiko suggested) didn't help. Too bad I'm out of house now until tomorrow.

Maybe the problem solved itself until then. Otherwise: any suggestions I should try?

MrS

It sounds like someone changed something on the server or maybe the server process is stuck.
Work doesn't seem to be going out this AM at least to NVidia cards. See this thread:

http://www.gpugrid.net/forum_thread.php?id=594

Now we're up to 978 WUs ready to send on the server status page.

I had 6.4.2 running with minor problems, i.e. I had to suspend all CPU projects to make it request new WUs.. but I got them.

Today I got the message "no work sent, cell not available for you" again, whithout any changes on my side. Project reset didn't help, the server has work, setting nrpc=0 didn't help and sending a double request (like Kokomiko suggested) didn't help. Too bad I'm out of house now until tomorrow.

Maybe the problem solved itself until then. Otherwise: any suggestions I should try?

MrS

It sounds like someone changed something on the server or maybe the server process is stuck.
Work doesn't seem to be going out this AM at least to NVidia cards. See this thread:

http://www.gpugrid.net/forum_thread.php?id=594

Now we're up to 978 WUs ready to send on the server status page.

You can never tell anything on the server status due to lumping GPU & PS3 work together.

This project has turned into a nightmare to run the last few weeks....one problem after another ARGHHHHHH!

This is strange. I had the same problem described above (Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.) I tried a project reset, suspending all other projects, setting a larger queue, reinstalling BOINC (i have 6.4.5), no luck.

I was just browsing the forum and thought to try again, and lo and behold, i got a new workunit. Are you sure the server queue isn't full of PS3 units?

I also just received WUs!

I have now also 2 boxes without work. They got no work in the last 24 hours and so this boxes are now idle. Is there really no new work available? So far I got work by manually update, but in the last 24 hours both boxes got every time the same message ... no work for your type of computer ... and the boxes can't call automatically for work, then after this error they go in the 24 hours loop. I will shut down now this 2 boxes and try tomorrow again, no need to run this PCs without a chance for work ...

This boxes are effected:
Linux 65 bit GTX260² 15924
Vista 64 bit GTX260² 17745

---

After trying for an hour I finally got WUs on this both boxes. Looks like really no work on the server was available.

---

- "No work sent, molecular dynamics not available for you" again
- 2 different hosts
- Correct BM, correct driver, correct temperature at home, even weather outside is correct (if some1 would like to ask :P) etc.
- Nothing helps (nrpc, DCF, suspending other projects - but request is fine...)
- 1 box got some tasks in queue, 2nd box got nothing in queue (I`m going to try reset project after finishing these 2 WU`s which I`m crunching now, but I don`t believe it will help).

And probably I`m one of not so many who are taking this situation absolutely easy :). I understand this situation. I could understand this situation even for whole next year if... If something what I asked earlier could become soon.

Please. Please change `per-CPU limit` from 1 to 3 and give us possibility to download more than 1-2 WU`s `at once` (5-6 will be OK). After these changes I`ll do not care even if server will go down for 24h. I want to have enough WU`s in stock (queue) to let me crunch at least 24h if some problems will occur. I want to have possibility to DL enough WU`s even if possibility to DL new tasks can occur once per day. And pls change this nonsense that BM is `deferring communication` for next 24h if can`t reach work. That`s really biggest `bunkum` :).

And then I`ll can sleep well.

Mby more of you, crunchers, could sign this supplication? :)
Or mby you disagree?



PS. I`m curious how GPUGrid is working `under` VM. I think I`ll install 3-4 OS`s on each host, just to be sure that I always have enough WU`s. And I`ll switch between them if necessary. Or I`ll install Ubuntu inside Win (by using this option with `filepartition`).

---

Agreed none of the usual "tricks" seem to force a download of work at the moment - but I am lucky enough to have a queue of 2 waiting to go on my machine with the GTX260 card.

On the machine with the 8600GTS card I have more work units than I can process before their deadline. With this machine trebling the number of wu available to download would just mean I would have to abort more wus every few days, unless the deadline was revised accordingly.

Ideally, the number of wus downloaded should be limited to capability of the card in relation to the deadline; now where have I heard that before...:-)

Looks like some people are getting work though, I guess because other people are aborting wus or wus are passing their deadline and are being re-issued.

Phoneman1