Anyone Have An app_info For Poem?

[nanoprobe]

I know it's a long list but only 7 of those are in service and of those only 5 are running at this time. For whatever reason boinc keeps track of every machine that's ever run on your network. I play around a lot with different OSs and setups and each one registers as a new device on the network.

[Duke of Buckingham]

Yes I know nanoprobe, I was joking.

Every computer is on the database for at least one year after the last crunch credits, I think.

I had a lot also but they were in fact only 3 and now they are only 2.

Joking Duke

[Mike029]

Fixed and Thank You CH!!!

[Duke of Buckingham]

I would try to take the flops (<flops>2.1e10</flops>) and expect the system to choose the default. I know I am not the best help but have you tried that yet? I had once troubles with that. I don't even know very well what that line does.

[Christopher Herr]

Mike, had this twice as well on my single PCIe slot Dell box, as soon as I took any AMD/ATI card out and put any of my GTX in instead, BUT, and that is the point, left the AMD/ATI drivers installed.
For some stupid reason, NVidia drivers refuse to work, when there is an AMD/ATI driver installed for the exact same slot. Competition BS from NVidia to hassle AMD owners, I guess.
From the messages output you were so wise to post I conclude there is no other graphics card in this system and from my recent poor experience with NVidia drivers regarding their competitor I conclude this must be it. Use DriverFusion (successor of DriverSweeper) to clean any AMD/ATI remnants directly after stock uninstall as well, then reboot!

After this procedure, my GTX 480/680 always purred along happily...

Hope this guessing in relative dark of mine helps!

Oh and reduce the number of parallel POEM tasks in your app_info to four, I suggest, since one needs 256 MB and you have 1024 MB; had the same with my GTX 480 (1536 MB) and seven instead of six, instantly the average runtime jumped a lot higher than sixteen percent relative to six. IIRC it was around 23 percent higher relatively! Reason as far as I could determine: the data with seven at once was copied out from RAM and into the VRAM and vice versa all the time, as the memory controller was double as active relative to six at once...
For the same reason watch the VRAM clocks! When the driver has reset the VRAM clock is at lowest possible stock and the average runtime leaps in an unwanted direction. At least that was the case twice with one of my GTX 680s I had kicked a little too hard too far.

[Christopher Herr]

I would try to take the flops (<flops>2.1e10</flops>) and expect the system to choose the default. I know I am not the best help but have you tried that yet? I had once troubles with that. I don't even know very well what that line does.

Ric, that line tells the server one running instance of this application has the performance in FLOPS given in that line under the settings chosen, e.g. for coprocessor or number of tasks in parallel and CPU.
Then for correct estimation of the average runtime of a single task the server has to calculate the Duration Correction Factor (DCF) for every single result uploaded and report this to your computer.
This line in fact is superfluous, you can erase it, but perhaps only after you have downloaded a two days' worth of cached work!
In the following you will get standardized WUs with default settings from the server and the DCF has to be adapted to a very small number, for GPU computing at least.

[Mike029]

Mike, had this twice as well on my single PCIe slot Dell box, as soon as I took any AMD/ATI card out and put any of my GTX in instead, BUT, and that is the point, left the AMD/ATI drivers installed.
For some stupid reason, NVidia drivers refuse to work, when there is an AMD/ATI driver installed for the exact same slot. Competition BS from NVidia to hassle AMD owners, I guess.
From the messages output you were so wise to post I conclude there is no other graphics card in this system and from my recent poor experience with NVidia drivers regarding their competitor I conclude this must be it. Use DriverFusion (successor of DriverSweeper) to clean any AMD/ATI remnants directly after stock uninstall as well, then reboot!

After this procedure, my GTX 480/680 always purred along happily...

Hope this guessing in relative dark of mine helps!

Oh and reduce the number of parallel POEM tasks in your app_info to four, I suggest, since one needs 256 MB and you have 1024 MB; had the same with my GTX 480 (1536 MB) and seven instead of six, instantly the average runtime jumped a lot higher than sixteen percent relative to six. IIRC it was around 23 percent higher relatively! Reason as far as I could determine: the data with seven at once was copied out from RAM and into the VRAM and vice versa all the time, as the memory controller was double as active relative to six at once...
For the same reason watch the VRAM clocks! When the driver has reset the VRAM clock is at lowest possible stock and the average runtime leaps in an unwanted direction. At least that was the case twice with one of my GTX 680s I had kicked a little too hard too far.

That did it!!! The AMD driver were causing the problem. Funny that there was no problem when the AMD card we installed. I removed the card and sent it in for repair, did not want to uninstall the drivers as I'm going to put the card back in when it is returned.

Thank you CH you are a genius.. Or I'm a dummy. Anyway the problem is fixed and my 560 Ti is crunching away at Poem.

[Christopher Herr]

Thank you CH you are a genius.. Or I'm a dummy. Anyway the problem is fixed and my 560 Ti is crunching away at Poem.

Now come on, Mike, this has neither anything to do with you being a dummy nor me being a genius! It has to do with having seen that crap and thinking "WTF, am I stupid?" also at that time...
Glad it did help, though, great for the long-term POEM quest!

Now that you mention it, I THINK I had seen that crap happening also when taking out the 7970 for RMA due to unnatural fan noise problems (which could have been lost by the "great" DHL on its way to the retailer, argh).
So it MIGHT be only working when there is actually an AMD card installed in the respective rig.
Anyway, it hope it is only a strange bug/glitch/whatever in the most recent WHQL NVidia driver and not updated, yet stupid policy of NVidia to suppress a competitor.
I had been ripping out AMD/ATI cards out of my single x16 PCIe slot Dell box all the time and installed the GTX 480 instead with both driver brands installed all the time...
Err, we will have to wait and see.

[nanoprobe]

OT: I was wondering who that was that recently ran over me at POEM. Nice to meet you CH. Just about the time I get dusted off STE\/E will be here.

[Duke of Buckingham]

Ric, that line tells the server one running instance of this application has the performance in FLOPS given in that line under the settings chosen, e.g. for coprocessor or number of tasks in parallel and CPU.
Then for correct estimation of the average runtime of a single task the server has to calculate the Duration Correction Factor (DCF) for every single result uploaded and report this to your computer.
This line in fact is superfluous, you can erase it, but perhaps only after you have downloaded a two days' worth of cached work!
In the following you will get standardized WUs with default settings from the server and the DCF has to be adapted to a very small number, for GPU computing at least.

Thanks CH that explains all. It happened maybe two months ago and GPU was waiting for something when I decided to clean that line. The GPU started working immediately. I didn't know why.