Message boards :
Number crunching :
Strange 'To Completion' numbers
Message board moderation
Author | Message |
---|---|
The_bestest Send message Joined: 7 Oct 06 Posts: 36 Credit: 82,706,887 RAC: 79 |
Went out for a few minutes this morning, and when I cam back, I'm getting some rather strange entries in the 'To Completion' column in my Tasks screen. Instead of getting the usual completion times of anywhere from 30 minutes to say three hours, I'm now showing completion times of 30 hours plus. Haven't done the reboot thing yet, but that's next. Pc is running fine (Quad Core QX6700 Extreme, OC'ed slightly to 3GHz, 2GB RAM). CPU usage appears normal, no strange tasks running that I can see. Anyone else running into this? Thanks in advance for any thoughts. |
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
Go to "your account", then "computers on this account - view", then click on the "computer ID" of the host in question, and see what values you have for the following: % of time BOINC client is running 93.8113 % While BOINC running, % of time work is allowed 99.9929 % Average CPU efficiency 0.990161 Result duration correction factor 0.44414 (these are from my host) |
The_bestest Send message Joined: 7 Oct 06 Posts: 36 Credit: 82,706,887 RAC: 79 |
Hmmmmmm. Something doesn't seem right here. Here's the info for my Quad-Core % of time SETI is Running 95.3612 % of time work is allowed 53.1876 (huh?) Average CPU efficiency 0.911759 Result duration correction factor 0.28845 Compared to my AMD X64 Dual Core 4800 % of time SETI is Running 82.1284 % of time work is allowed 93.2581 Average CPU efficiency 0.944237 Result duration correction factor 0.30093 Both machines sit running SETI pretty much 24x7, although the Quad is the machine I use for an hour or so in the morning, and maybe an hour in the evening after work. That % of time work is allowed entry for the quad is just not right. Will look at settings |
SMW Send message Joined: 16 May 99 Posts: 22 Credit: 29,285,238 RAC: 16 |
Went out for a few minutes this morning, and when I cam back, I'm getting some rather strange entries in the 'To Completion' column in my Tasks screen. I have had the same thing on both the Macintosh and the PC platforms. I have found that the actual time is then far less and works just fine. I have also seen where as the units are processed, the amount of time goes up rather than down. In all instances I find that the real time for units is reasonable and within norms, getting reported back with no problems. "It is better to be hated for what you are then to be loved for what you are not" - Andre Gide (1869-1951) |
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
There is a formula used to determine the "to comp" field. I remember that the project supplied "fpops_estimate", and "RDCF" is used, but I can't remember if benchmarks play a role. Your RDCF seems OK, and since fpops est is supplied with the wu, I'm guessing that the AR of the wu being run is leading to a large fpops est, in relation to the average wu. However, If benchmarks are used and something like "heat" is triggering a frequency reduction, then benchmarks would be cut roughly in 1/2. I see your benchmarks are: Measured floating point speed 2848.78 million ops/sec Measured integer speed 3743.35 million ops/sec as last reported. I wonder if that's high or low for your machine. |
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
I've found this in the wiki: The Scheduling Server(initial duration estimate) estimates the amount of time a Work Unit will take to complete with the formula (number of flops)/(flops per second)+(number of iops)/(iops per second). So, benchmarks are a part of it. |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
I've found this in the wiki: That's the calculation the Scheduler uses when responding to a work request and determines how many tasks get sent. The "To completion" time in BOINC Manager of unstarted work uses the rsc_fpops_est which the splitter generated, the p_fpops Whetstone benchmark of the host, and the Duration Correction Factor (DCF). Given p_fpops of 2848.78 million and DCF of 0.28845, the largest "To completion" estimate (for angle range 0.2259) should be just over 8 hours. My guess is the host has run benchmarks which have not been reported yet and got a much lower Whetstone score. Rerunning the benchmarks would probably fix it. Joe |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
I've found this in the wiki: The other possibiolity is that the machine has just completed an evercrunch and the DCF is now very large. BOINC WIKI |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.