1) Message boards : Server and website : GPU Results ready to send - number dwindling (Message 18764)
Posted 4962 days ago by Michael E. Malis
Thanks for the feedback. I have my CPU OCed to 3.33GHz with no load other than the usual background OS stuff. I also have had the Swan_Sync going for almost a month now. Haven't had many issues with failed WUs since I backed the GPUs down from 825 to 800 though. I guess it's Vista in combination with a slower CPU. I know when I overclocked my CPU my RAC jumped about 10% immediately after it had been leveled out and has been slowly climbing ever since but I'm as far as I can go on that since the bios doesn't go any further than 25% OC on the Rampage III.
2) Message boards : Server and website : GPU Results ready to send - number dwindling (Message 18758)
Posted 4963 days ago by Michael E. Malis
GPU Results ready to send: 329 and falling.
I'm a bit worried.


You and me both. Out of curiosity, how are you getting 5,500 more RAC than me? I've got both my GTX480s already overclocked to 800 MHz and 2 GHz for the memory. That's as high as they go without getting failed work units every other one.......at least on air anyways. Are you running yours even higher than that on liquid cooling or thermoelectric???
3) Message boards : Server and website : GPU Results ready to send - number dwindling (Message 18482)
Posted 4985 days ago by Michael E. Malis
GPU Results ready to send 2,075


NICE!!! Hopefully that will last a couple weeks and we can get another job in before it runs out.
4) Message boards : Server and website : GPU Results ready to send - number dwindling (Message 18441)
Posted 4988 days ago by Michael E. Malis
You're right about that. This kind of simulation practically needs infinite time to finish. But as far as I know this project, its tasks depend on a previous task, as every task we crunch covers 5ns (a fatty task covers 10ns) of the given chemical reaction. In that way this project can run out of work, but of course only temporarily.


Maybe they could setup different bins for nS times to be crunched. The default would be 5nS but they could have bins for 10, 15, 20, 25......all the way up to what is necessary for the entire simulation to meet whatever requirements they have for it. For the higher bins though they need an auto-reload for failed WUs (they actually should have that now!) so that if a computation error occurs, it just restarts from the last checkpoint that was saved to disk (BOINC default is 60 seconds). This way the most time anyone would lose on any given computation error is 59.999999 seconds of computing time instead of losing the entire WU, which is what happens currently.
5) Message boards : Server and website : GPU Results ready to send - number dwindling (Message 18434)
Posted 4989 days ago by Michael E. Malis
Got one "Fatty" WU.......now dry again. I don't understand how molecular biology simulations could EVER run out of work. It makes absolutely NO SENSE WHATSOEVER. There is enough calculation that needs to be done in this area to keep computers busy for at least the next 10,000 years (by today's computing technology capabilities).
6) Message boards : Server and website : GPU Results ready to send - number dwindling (Message 18404)
Posted 4990 days ago by Michael E. Malis
Id like to see at least a 24-hour cache just in case something goes wrong with the server. Currently if GPUGRID goes down for something, I'm dry in 5-6 hours with this ridiculous max queue of 4 WUs. This is the most restricted DC project out there when it comes to the size of the buffer.
7) Message boards : Frequently Asked Questions (FAQ) : FAQ - Best configurations for GPUGRID (Message 18259)
Posted 5011 days ago by Michael E. Malis
I'm not sure what the specific cause of it is but DO NOT USE THE SWAN_SYNC WITH PHYS-X DEDICATED TO THE CPU IF YOU ARE RUNNING VISTA 64-BIT. It causes Windows Explorer to continuously stop working and then restarts Explorer over and over in an infinite loop and is extremely difficult to maneuver to change the settings back because you only have about 8 seconds in-between Explorer crashes.


//