Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 

Quick Go »
No member browsing this thread
Thread Status: Active
Total posts in this thread: 8
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 2608 times and has 7 replies Next Thread
AgrFan
Senior Cruncher
USA
Joined: Apr 17, 2008
Post Count: 396
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Completion times different between similiar machines

I have three single core machines dedicated to this project running at stock speeds.

Specs:
1 - Intel P4 2.53 (northwood, SSE2), 512mb
2 - AMD Sempron64 2800+ (palermo, SSE2), 512mb
3 - AMD Sempron 2500+ (thoroughbred-b, SSE), 1gb

#1 and #2 are completing units in 8-9 hours .. #3 is completing units in 16-18 hours.

Is this project optimized for SSE2 instructions ... does that explain why #3 is taking twice as long as #1 and #2 to complete a unit?

Is there any information available that lists the cpu optimizations for each project?
----------------------------------------

  • i5-10400 (Comet Lake, 6C/12T) @ 2.9 GHz
  • i5-7400 (Kaby Lake, 4C/4T) @ 3.0 GHz
  • i5-4590 (Haswell, 4C/4T) @ 3.3 GHz
  • i5-3330 (Ivy Bridge, 4C/4T) @ 3.0 GHz

----------------------------------------
[Edit 1 times, last edit by AgrFan at May 14, 2009 1:30:53 AM]
[May 14, 2009 1:30:04 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: Completion times different between similiar machines

Hello AgrFan,
The projects do not use optimized code because the final results would differ subtly. This is only required by bit-for-bit comparison for validation so it may change for future projects as we develop more expertise in other types of validator functions.

For the moment you must look elsewhere.

What OS?

Lawrence
[May 14, 2009 2:28:28 AM]   Link   Report threatening or abusive post: please login first  Go to top 
nasher
Veteran Cruncher
USA
Joined: Dec 2, 2005
Post Count: 1423
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Completion times different between similiar machines

the main reason is every work unit is different
and every project is different

you will see some projects that 1 computer dose the work faster and on a different project the reverse is true

this is normal..
----------------------------------------

[May 14, 2009 4:54:33 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: Completion times different between similiar machines

Apart from the reasons mentioned above, his can also be caused by the following:

- Different OS.
- One computer has more running proccesses in the background.
- One computer sits idle all day, the other is being used.
- Your settings differ per machine (CPU%, in use when active)
- One of the machines use a heave screensaver that is taking all the CPU.
- One of the machines is running hot and is being limited to prevent overheating.
- One of the CPU's is already been damaged slightly in the past.
- A dodgy PSU that cannot deliver enough power to the machine.
- Your mom\brother is snoosing WCG when they use the machine.
- Your memory has a bit of a problem.

And i can think of atleast 50 more causes for such a discriptency. So take your pick. Only proper testing with different projects and or tools (stresstest\benchmark\voltagemeter) will be able to give you a solid final awnser.

I would probably start with checking the running processes and see what can be tweaked there. (the less is running in the background the more will be dedicated to WCG)
[May 14, 2009 6:56:35 AM]   Link   Report threatening or abusive post: please login first  Go to top 
AgrFan
Senior Cruncher
USA
Joined: Apr 17, 2008
Post Count: 396
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Completion times different between similiar machines

Hello Laurence,
#1 is running Ubuntu 7.10, #2 is running Windows XP Home SP3 and #3 is running Windows XP Home SP 2. The number of background tasks is pretty much identical on the two Windows boxes.

I suspect the difference has to do with the fact #3 is a older Socket-A board running with DDR PC2700 RAM. I'm also running AVG Antivirus on this box. I excluded the BOINC directory last night but that didn't seem to make much difference.

#1 is running at a faster clock speed and #2 is a Socket 754 board running with DDR PC3200 RAM.

The difference is about 30% at this point ...
----------------------------------------

  • i5-10400 (Comet Lake, 6C/12T) @ 2.9 GHz
  • i5-7400 (Kaby Lake, 4C/4T) @ 3.0 GHz
  • i5-4590 (Haswell, 4C/4T) @ 3.3 GHz
  • i5-3330 (Ivy Bridge, 4C/4T) @ 3.0 GHz

[May 16, 2009 12:45:58 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: Completion times different between similiar machines

I just switched to running the latest version of BOINC.

19/05/2009 07:21:58 Starting BOINC client version 6.6.28 for windows_intelx86
19/05/2009 07:21:58 log flags: task, file_xfer, sched_ops
19/05/2009 07:21:58 Libraries: libcurl/7.19.4 OpenSSL/0.9.8j zlib/1.2.3
19/05/2009 07:21:58 Data directory: C:\ProgramData\BOINC
19/05/2009 07:21:58 Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz [x86 Family 6 Model 15 Stepping 11]
19/05/2009 07:21:58 Processor features: fpu tsc pae nx sse sse2 pni mmx
19/05/2009 07:21:58 OS: Microsoft Windows Vista: Home Premium x86 Edition, (06.00.6000.00)
19/05/2009 07:21:58 Memory: 3.00 GB physical, 6.17 GB virtual
19/05/2009 07:21:58 Disk: 457.76 GB total, 363.72 GB free
19/05/2009 07:21:58 Local time is UTC +1 hours
19/05/2009 07:21:59 CUDA device: GeForce 8400 GS (driver version 18250, compute capability 1.1, 256MB, est. 5GFLOPS)
19/05/2009 07:21:59 Not using a proxy

So far all seems well, except that the 'To completion' time is wildly inaccurate for influenza WUs. As I'm getting mainly influenza WUs I assume they have priority for download just now. Press on regardless. smile
[May 19, 2009 7:03:57 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Sekerob
Ace Cruncher
Joined: Jul 24, 2005
Post Count: 20043
Status: Offline
Reply to this Post  Reply with Quote 
Re: Completion times different between similiar machines

Download priority is weight controlled and project mix selection dependent. If you make a choice of e.g. FLU and CEP you'd currently get allot more FLU. Also much depends on what's in the feeders. No message would be given if there were no work for X or Y project if at least one of the selected projects has work.

FLU projected run times wildly inaccurate? If you did a clean install of 6.6.28 then your client has to learn from scratch. A quick view on one machine for the last 15 returned shows a variance of actual CPU time between 4.91 and 6.76 hours. The project mean was yesterday 6.6 hours.

The server continuously computes a mean runtime so simplistically, the work units send out have a header that tells MY client it has a number of FPOPS so it will take my computer 5.84 hours on average (6.76+4.91)/2. i.e. faster than project mean. Non-Deterministic as these computations are, they could though each take more or less FPOPS as the practical measurement shows hence the variation observed.
----------------------------------------
WCG Global & Research > Make Proposal Help: Start Here!
Please help to make the Forums an enjoyable experience for All!
[May 19, 2009 7:37:28 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: Completion times different between similiar machines

Thanks for the info Sekerob.
[May 19, 2009 8:28:07 AM]   Link   Report threatening or abusive post: please login first  Go to top 
[ Jump to Last Post ]
Post new Thread