Scheduler is an idiot

keputnam
keputnam
Joined: 18 Jan 05
Posts: 47
Credit: 84,110,314
RAC: 2

OK, re-booted and snoozed

OK, re-booted and snoozed BOINC. With System Idle Process at 97-98% ran Benchmarks again

These are the results

7/10/2013 2:11:04 PM | | Benchmark results:
7/10/2013 2:11:04 PM | | Number of CPUs: 2
7/10/2013 2:11:04 PM | | 824 floating point MIPS (Whetstone) per CPU
7/10/2013 2:11:04 PM | | 1357 integer MIPS (Dhrystone) per CPU

Certainly higher than the last set

Also, the current estimated total run-time after just less than 30 Hours of run-time is now 231 hours (301 remaining), so the total estimate is (finally) starting to come down

Large Software's PC Tune up ran when I booted so the registry should be clean and I have DiskKeeper set to run automatically. Saw it running two days ago, so I don't see much else I could to to tweak the system

Thanks for your help BTW. And for letting me vent a little


Miklos M.
Miklos M.
Joined: 3 Apr 05
Posts: 19
Credit: 1,735,494,997
RAC: 890,821

7/10/2013 5:19:25 PM |

7/10/2013 5:19:25 PM | Einstein@Home | No work is available for Gravitational Wave S6 LineVeto search (extended)
7/10/2013 5:19:25 PM | Einstein@Home | No work is available for Gamma-ray pulsar search #2
What is going on?

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,869
Credit: 113,203,975,416
RAC: 36,771,180

RE: 7/10/2013 5:19:25 PM |

Quote:
7/10/2013 5:19:25 PM | Einstein@Home | No work is available for Gravitational Wave S6 LineVeto search (extended)
7/10/2013 5:19:25 PM | Einstein@Home | No work is available for Gamma-ray pulsar search #2
What is going on?


The S6LV search has finished. If you check the server status page, you will see about 2,500 tasks still 'out there'. If any of those fail (some will) you could be lucky and get a resend but I don't like your chances :-).

The FGRP2 search hasn't finished but for some reason, the generation of new work is currently 'disabled' - again, check the status page. There are currently around 90K tasks in progress and quite a few will fail so if your client keeps asking, you'll probably get resends from time to time - for a while, anyway. Hopefully, work generation will be turned back on at some point soon.

Your other alternative for CPU work until the new GW search gets underway is BRP4 from the new Arecibo batch that Bernd announced recently. Check the Technical News.

Cheers,
Gary.

Jord
Joined: 26 Jan 05
Posts: 2,952
Credit: 5,779,100
RAC: 0

RE: 7/10/2013 2:11:04 PM |

Quote:
7/10/2013 2:11:04 PM | | Benchmark results:
7/10/2013 2:11:04 PM | | Number of CPUs: 2
7/10/2013 2:11:04 PM | | 824 floating point MIPS (Whetstone) per CPU
7/10/2013 2:11:04 PM | | 1357 integer MIPS (Dhrystone) per CPU


Something else must be running then, taking up the CPU (and throwing off the benchmark plus slowing down calculations). For comparisons, my old Pentium 4 3.00GHz, still in my list of computers:
[pre]Measured floating point speed 1637.91 million ops/sec
Measured integer speed 3193.48 million ops/sec
[/pre]

Heck, even the older P4 2.6GHz had better values than yours:
[pre]Measured floating point speed 1189.07 million ops/sec
Measured integer speed 2409.2 million ops/sec
[/pre]

Anything else using the CPU is in the way of the calculations. Especially completely useless things such as PC Tune Up and Diskkeeper. Either run those manually when BOINC isn't running, or set BOINC to run with one CPU core only (On multiprocessor systems, use 50% of the processors), so that those CPU swamping programs can use the other core. With luck, it'll be the virtual core.

By the way, the CPU benchmarks are also higher when you switch off hyper threading. But, the benchmarking isn't used to determine how many floating operations the project should add to a task sent to your BOINC, but to determine the maximum floating operations the CPU is capable of at the time of benchmarking.

No project sends out work with fpop estimate values specifically for your CPU. That other projects may work better, is then because their work suits your CPU better, or because BOINC has had several (at least 10) tasks run from this project, and adjusted the (earlier mentioned) duration correction factor, so that BOINC knows how long these tasks run and will ask for so much work accordingly.

keputnam
keputnam
Joined: 18 Jan 05
Posts: 47
Credit: 84,110,314
RAC: 2

Jord Don't know what to

Jord

Don't know what to tell you

That last set of Benchmarks was run after re-booting and holding down the shift key - bypassing the startup folder entirely

As I said, when I started the benchmarks, Task Manager was indicating 97-98% CPU used by the "System Idle Process" (and that is what it went back to after the benchmarks were done), so basically nothing else was running

I have no idea if it makes any difference, but this is an HP/Compaq DC5000

The BIOS is at the last one provided by HP, whatever that is, and I don't remember making any drastic settings in the BIOS that should affect CPU speed at all

I'll just follow Gary's advice and select the application with the shorter WUs here and let Einstein and the rest of my projects crunch along

Tried re-booting again, disconnected from the network and shutdown everything I could think of, including my AV (Lavasoft)

New benchmark figures are

7/10/2013 9:06:07 PM | | Number of CPUs: 2
7/10/2013 9:06:07 PM | | 1109 floating point MIPS (Whetstone) per CPU
7/10/2013 9:06:07 PM | | 1767 integer MIPS (Dhrystone) per CPU


DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1,364
Credit: 3,562,358,667
RAC: 0

If you haven't done so, I'd

If you haven't done so, I'd suggest running CPU-z to see how fast your CPU is actually running. Overheating and slowing down to avoid damage because the heat sink is clogged or the fan on it failed is a reasonably common cause for older computers to slow down.

http://www.cpuid.com/softwares/cpu-z.html

keputnam
keputnam
Joined: 18 Jan 05
Posts: 47
Credit: 84,110,314
RAC: 2

Dan, Using CPUz v1.41, I

Dan,

Using CPUz v1.41, I get a reported CPU speed or 2992.4 MHz

One thing that may be playing into this is that I also run T4T, which previously required BOINC to run under a logged on user rather than as a service. They use an Oracle VM to process their WUs

They have recently been pushing a new wrapper that runs headless. After checking on their Forum. I have re-installed BOINC as a service and will start logging out when not actually using it

With no excess crap running under my user ID, I *should* see at least some improvement in BOINC throughput


Claggy
Claggy
Joined: 29 Dec 06
Posts: 560
Credit: 2,699,403
RAC: 4

For comparison I got my PIII

For comparison I got my PIII 800 @896MHz to grab a Wu, it came in at 117 Hours in the end, against your P4 @3.00GHz at 198 hours.

All tasks for computer 933098

Claggy

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.