| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Speed issues Boinc vs UD |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 6
|
|
| Author |
|
|
biini
Senior Cruncher Finland Joined: Jan 25, 2007 Post Count: 334 Status: Offline Project Badges: 
|
Hi
----------------------------------------I did only a quick search so this might be one of those stupid questions already answered a hundred of times... Anyway, here it goes: I tested Boinc client last year and it felt kinda slow compated to UD. So I removed it from all my computers and reinstalled UD. Recently we all were forced to move to the Boinc and abandon UD. I did some calculations to compare the speed of the clients and found out that my first impression could've been correct. Here's the results: desktop: boing average result in: 7,12h UD average result in: 4,23h laptop1: boinc averate result in: 11,8h UD(1) average result in: 9,42h UD(2) average result in: 9,83h (I had reinstalled the client before on this computer) laptop2: boinc average result in: 8h UD average result in: 4,82h Does anyone have idea (or knowledge :) what causes this high speed-differences? I have/had cpu-usage set to max on both clients. Cheers! ---------------------------------------- [Edit 1 times, last edit by biini at Mar 11, 2008 11:40:57 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
It is impossible to do a direct comparison like that. Different work is available for each client, and work unit / batch sizes vary considerably. So, your results are useless, I'm afraid.
However, both clients use the same science applications, so under fair conditions they perform identically. BOINC has the advantage, though, because it can do twice as much work on dual core computers. UD is old, useless, and gone. Welcome to the wonderful world of BOINC! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello biini,
The work is done by the science application. Therefore the difference between UD and BOINC is only a few seconds. Did you do your benchmarks on a Pentium 4? BOINC runs multiple threads for multiple cores unless you tell it otherwise. If you run 2 threads simultaneously on a hyperthreading Pentium 4 (discouraged), then each thread will run much more slowly and your computer will use more memory, perhaps spilling over into slow virtual memory. Lawrence |
||
|
|
biini
Senior Cruncher Finland Joined: Jan 25, 2007 Post Count: 334 Status: Offline Project Badges:
|
Hi! Thanks for you replies.
----------------------------------------I have no P4. Only AMD XP3200+, PIII, and Amd Turion 64. I have chosen only to the same tasks with Boinc as I had with UD (see the badges). The values I posted are averages from hundreds and thousands of work units, so I think it should be comparable. You can see that laptop1 with two different UD installations gave close to each other results, but after I switched to boinc you can see the drastic speed drop in results. Could be I'm missing something. Cheers! ---------------------------------------- [Edit 2 times, last edit by biini at Mar 11, 2008 12:26:58 PM] |
||
|
|
biini
Senior Cruncher Finland Joined: Jan 25, 2007 Post Count: 334 Status: Offline Project Badges:
|
I recall same kind of behaviour in SETI@home clients back in 90's. The command-line version performing 10-20% faster than the windows client with fancy GUI and graphics. Anyhow, not propably anyway related to this issue :) Cheers!
-------------------------------------------------------------------------------- [Edit 1 times, last edit by biini at Mar 11, 2008 12:35:35 PM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
hi biini,
----------------------------------------As you can see from the below graphs the length of jobs both for FAAH and HPF2 have changed quite a bit with HPF2 in fact having had 50% added, never as long as now. What is evident is that since the more pushed conversion from UD to BOINC, the project average points per hour are on the rise by the principle of more quality hours, more points (though many are not interested in points, they are a mean of performance measurement). The undulation are phenomena related to work changes, and slow machines catching up with faster machines and the way work is completed during the weekly cycle. Important is to look at the longer trend.... smack in the face is the FAAH change in the 10 days of conversion push, whilst HPF2 has been going on for longer. Full BOINC projects run 72-76 points per hour and given the jobs are identical in science, gives me the idea that BOINC is more efficient, particular when it comes to buffering work. Whilst UD with servers are down grinds to a halt.... we've seen a few protracted cases in the last few months. Think WCG need to talk to the FAAH (and HPF) project owners to highlight the UD > BOINC conversion also. Oh and please visit the Start Here forum if not done already for a Migration FAQ and feature comparison.... too many bells to recount here. ttyl
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
|