Large Boinc WUs

SunRedRX7

Member

​

Joined

Apr 26, 2003

Location

Kenmore, NY

• Jul 12, 2004

• #1

Anyone else getting the larger WUs for BOINC?

They look like they'll take roughly 6-7 times as long to complete(based on "to completion" column), not that it matters as credit is given based on their formula that also takes into account time.

I wonder if these are the units they were talking about sending out based on how powerful a machine is. Probably wouldn't want a work unit that takes 20hrs on an XP2500 to go to a Pentium2.

 

SkyHook

Senior Village Idiot

​

Joined

Jan 24, 2002

Location

York, Pennsylvania

• Jul 12, 2004

• #2

NOPE! Has nothing to do with the makeup of the WU. They simply inflated the projected completion time so that when your client requests XXXXsecs. of work, they can meet that request with fewer WUs thereby lessening the load on their distribution system. This has fairly well been documented on the messageboards on the BOINC site, which of course is unavailable right now.

My experience has shown that these WUs process in the same timeframe as the more normally estimated ones.

SkyHook

 

OP

SunRedRX7

Member

​

Joined

Apr 26, 2003

Location

Kenmore, NY

• Jul 12, 2004

• Thread Starter
• #3

Well maybe I'll teach them a thing or 2 and set my cache 6-7 times higher

Thanks for the info.

 

S

SCREAMING EAGLE

Member

​

Joined

Apr 3, 2004

• Jul 12, 2004

• #4

what a cheap trick. btw under prefrences there are percentages of time connected ,time host is active, ect. Would they use these numbers as part of there formula for how many wu's to send ya?

 

SkyHook

Senior Village Idiot

​

Joined

Jan 24, 2002

Location

York, Pennsylvania

• Jul 12, 2004

• #5

SCREAMING EAGLE said:

what a cheap trick. btw under prefrences there are percentages of time connected ,time host is active, ect. Would they use these numbers as part of there formula for how many wu's to send ya?

Click to expand...

The formula, if in fact one exists, is like a lot of other things about BOINC right now, steeped in mystery and intrigue. Berkeley has IMHO been anything but worthcoming with actual factual info about many aspects of this project. Common sense would lead one to believe that the things you brought up certainly should have a bearing on how much work an individual would receive, but I can't remember any official statement backing up that assumption.

SkyHook

 

S

SCREAMING EAGLE

Member

​

Joined

Apr 3, 2004

• Jul 12, 2004

• #6

Well I think that this is the case. The machine that I leave connected cache fills no problem , while my other will only get one or two no matter how I set the buffer.

 

orthus

Member

​

Joined

Mar 6, 2004

Location

Just east of St Louis, MO

• Jul 12, 2004

• #7

It's really insane. On the fastest of my machines, it will only cache 2-3 of the "28 hour" WUs. However, on a slower machine it always has a cache of like 8-9 of these WUs. Totally baffling

 

G

glasszon

Member

​

Joined

Feb 28, 2004

• Jul 14, 2004

• #8

My celeron computer is even worse, boinc estimate the computer will take about 4 hours to complete each WU and it ends up taking 13 hours on average.

So now A LOT of cache on my celeron computer is over the report date and I can't do anything about it.

 

TC

Senior Seti Addict

​

Joined

Jan 15, 2001

Location

Denver, CO

• Jul 15, 2004

• #9

Yeah their benchmark is a mystery to me. I've got a 1.6GHz pentium M laptop that benches about double what my 2.4GHz athlon 64 gets. Then I have a 3.4GHz P4 with HT that gets a lower score than a 1.8GHz athlon xp. And the funny thing is the score can double or halve each time I run it.