I noticed in a few other threads that people are setting up their clients to cache wu's. My client usually has 2 wu at any time, and it will download another when it is about to finnish one.

How can I set it to cache more wu's?

---

See section at bottom of page:

http://www.pperry.f2s.com/boinc-start.htm

regards,

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

> How can I set it to cache more wu's?

Hello Mjolinir!
It cannot be done manually.

The Boinc client calculates with the crunching speed of your computer (benchmarks), the size of the WUs (dependent of the project, CPDN real big, P@H quiet small) and the time you set in your general preferences at 'Connect to network about every (determines size of work cache; maximum 10 days)'

If you increase this number, you get a bigger cache (if there is any WU on the server).

But as the projects do not seem to calculate the crunching of the others, and these are the 'general preferences' for all projects, the possibility of losing WUs via deadline passing increases, especially with P@H (deadline 3-7 days).

I usually connect every day, so I put my time to 1.5 days, to have a bit cache for weekends or such. But with CPDN, the possibility of running out of work is real small ;-)

---

Gruesse vom Saenger

For questions about Boinc look in the BOINC-Wiki

Thank you very much, not sure how I missed that =)

---

> I noticed in a few other threads that people are setting up their clients to
> cache wu's. My client usually has 2 wu at any time, and it will download
> another when it is about to finnish one.
>
> How can I set it to cache more wu's?
>

To increase the cache size goto your account and under "Network" select "General Preferances" then change "Connect to network about every 0.01 days".

I am running a 2.4G machine and mine is set for 0.05 days. This dnlds aprox 8 work units since I complete 1 in about 4.5 hours.

Doing this will also change the Average turnaround time for the PC's that you have so dont make the number too big so you dont go over the 14 day max. turn around time per work unit.

Good luck.

---

> I am running a 2.4G machine and mine is set for 0.05 days. This dnlds aprox 8
> work units since I complete 1 in about 4.5 hours.

whats a 2.4G machine (2.4Ghz?)

if its 2.4Ghz do u do stuff at the same time coz 4.5 hours is a long time ive got a xp3200+(2.2Ghz) and that does a WU in about 2.5 hours

---

<BR>AMD XP3200+
<img src="http://boinc.mundayweb.com/seti2/stats.php?userID=2327&amp;trans=off"><img src="http://petrus.homeftp.org/bws/counter_big.php?id=7828479">
<a href="http://homepage.ntlworld.com/paulandrew.odell/">MY SITE!</a>

> Thank you very much, not sure how I missed that =)


Hey! There is a lot more to learn ... come read with me .... :)

---

> > I am running a 2.4G machine and mine is set for 0.05 days. This dnlds
> aprox 8
> > work units since I complete 1 in about 4.5 hours.
>
> whats a 2.4G machine (2.4Ghz?)
>
> if its 2.4Ghz do u do stuff at the same time coz 4.5 hours is a long time ive
> got a xp3200+(2.2Ghz) and that does a WU in about 2.5 hours
>
I'm not sure, but if that PC thats takes 4.4hours (the 2.4Ghz) is pentium with HT then your processign 2 WU at the same time, and I not quite sure but... that might take sometime to do the both WU on a single processor (2 virtually)
I'm not an Intel Pro becuase I havent used one since a quite while ago, so I might be wrong.

---

> > I am running a 2.4G machine and mine is set for 0.05 days. This dnlds
> aprox 8
> > work units since I complete 1 in about 4.5 hours.
>
> whats a 2.4G machine (2.4Ghz?)
>
> if its 2.4Ghz do u do stuff at the same time coz 4.5 hours is a long time ive
> got a xp3200+(2.2Ghz) and that does a WU in about 2.5 hours
>
A few years ago Intel screwed up the numbers...before that the numbers meant the speed of the chip. Now they do not. AMD tried staying with the old system but lost SOOO much market share because their chips were "percieved" as being slower they finally had to go along with the new numbeing scheme. BUT being AMD they figured out how to do soem things faster than an equal speed Intel chip, hence the + on the end. An Intel 3200 will run most applications at about the same speed as a 3200+ AMD chip, but there a FEW things that AMD does run faster.
I am sure that Intel chips can do somethings faster than AMD but have not figured out a way to market that yet.

These are AMD numbers:
Model Number Frequency L2 Cache Packaging
4000+ 2.4 GHz 1 MB 939-pin
3800+ 2.4GHz 512KB 939-pin
3700+ 2.4GHz 1MB 754-pin
3500+ 2.2GHz 512KB 939-pin
3400+ 2.4GHz 512KB 754-pin
3200+ 2.0GHz 1MB 754-pin
3200+ 2.0GHz 512KB 939-pin
3000+ 2.0GHz 512KB 754-pin
3000+ 1.8GHz 512KB 939-pin
I am sure Intel numbers are similar.

---

> I'm not sure, but if that PC thats takes 4.4hours (the 2.4Ghz) is pentium with HT then your processign 2 WU at the same time, and I not quite sure but...
> that might take sometime to do the both WU on a single processor (2
> virtually)
> I'm not an Intel Pro becuase I havent used one since a quite while ago, so I
> might be wrong.
>
You're right. My P4@3.2GHz HT processes 2 WU's at the same time. That's because it have 2 CPU's (virtually). Both of them gives 100% CPU load. That could be changed in general preferences by decreasing "On multiprocessors, use at most:" to 1. Didn't try, but I guess it will decrease a CPU load to 50%.

---

<img src="http://petrus.homeftp.org/bws/counter_big.php?id=7828882" /> <img src="http://boinc.mundayweb.com/one/stats.php/userID:553/prj:2/trans:off/.png" />

> > I'm not sure, but if that PC thats takes 4.4hours (the 2.4Ghz) is pentium
> with HT then your processign 2 WU at the same time, and I not quite sure
> but...
> > that might take sometime to do the both WU on a single processor (2
> > virtually)
> > I'm not an Intel Pro becuase I havent used one since a quite while ago,
> so I
> > might be wrong.
> >
> You're right. My P4@3.2GHz HT processes 2 WU's at the same time. That's
> because it have 2 CPU's (virtually). Both of them gives 100% CPU load. That
> could be changed in general preferences by decreasing "On multiprocessors, use
> at most:" to 1. Didn't try, but I guess it will decrease a CPU load to 50%.
>
That is true, but you would still be using the non virtual processor 100% of the time. You just would not be using the virtual processor at all.

---

That's great. So if I want to speed up processing of an individual WU's, just have to decrease number of CPU's used. It still be working with a 100% CPU load. So, I can choose: do I want to crunch an individual WU more quickly, or do I want to do it slowly, but crunch 2 WU's at the same time, by using both of virtual CPU's.
Have to see what is more effective.

---

<img src="http://petrus.homeftp.org/bws/counter_big.php?id=7828882" /> <img src="http://boinc.mundayweb.com/one/stats.php/userID:553/prj:2/trans:off/.png" />

> That's great. So if I want to speed up processing of an individual WU's, just
> have to decrease number of CPU's used. It still be working with a 100% CPU
> load. So, I can choose: do I want to crunch an individual WU more quickly, or
> do I want to do it slowly, but crunch 2 WU's at the same time, by using both
> of virtual CPU's.
> Have to see what is more effective.

Having the two virtual processors is most effective. You get slower speed on each but higher throughput. So, over time, you get more done.

---