[Opensim-users] PhyFPS

[Karen Palen]

>From the sim owner's viewpoint this boils down to "too many scripts" - how
many of them do really dumb things (like this) is at least a second order
analysis!

This is the case for a really comprehensive "benchmark" exercise of the kind
that Tom's hardware does on CPUs, however that kind of effort would  totally
occupy the OpenSimulator effort! The manpower simply does not exist, and if
it did even *I* could think of 1000+ better places to use the resource!

A US$10M "donation"/"investment" (in manpower NOT "green $!") could change
everything of course ... That works out to around 30 people full time for a
year with overhead.

Karen

On Thu, Sep 23, 2010 at 4:51 PM, Teravus Ovares <teravus at gmail.com> wrote:

> Actually, the reason that I mentioned scripts is because I've seen some
> scripts fire tiny interval timers that move things around rapidly in the
> scene, set things physical, then phantom..    non-phantom,....  move the
> object around again..   etc.
>
> Get a group of those going and you're giving the physics plugin a run for
> it's money and that can lead to an increase in the amount of time it takes
> to process a frame..   which will lead to a lower frame rate.
>
> Regards
>
> Teravus
>
>
> On Thu, Sep 23, 2010 at 5:46 PM, Serendipity Seraph <sseraph at me.com>wrote:
>
>>
>> On Sep 22, 2010, at 9:39 PM, Karen Palen wrote:
>>
>> I can add one more datum point to this, I have the latest Diva distro
>> (V0.7.0.2) standalone running on an Athlon triple core 3X 445 with very
>> light cpu load up to 5 users - the most I have tried.  I documented this in
>> the wiki running on my Intel I7 machine, but the MUCH lower performance low
>> end Athlon (by about 200X!) is giving very similar results. The Athlon was
>> simply chosen as the cheapest CPU on th e market last month (US$64) and has
>> proven to be quite adequate for 10 regions.
>>
>> The major limit seems to be the IO bandwidth of my consumer cable
>> connection. As noted previously, this dramatically increases CPU usage as
>> the IO "maxes out".
>>
>> As Teravus commented, scripted objects and things like collisions make a
>> far greater difference than the basic region maintenenace.
>>
>>
>> Why would most scripts make that much difference unless they are still all
>> getting a huge amount of memory?  Is the script engine that bad?  Does it
>> have to be?    Since not that many things are usually moving about to
>> collide it seems a bit odd that that would so much difference either.   How
>> are these things validated?
>>
>> - seren
>>
>>
>> Try it out for yourself on:
>> http://mars-simulator.hobby-site.org:9000/wifi
>>
>> Karen
>>
>>
>> On Wed, Sep 22, 2010 at 1:05 PM, Oliver Guerino <oguerino at gmail.com>wrote:
>>
>>> Hi,
>>>
>>> I have the OpenSim version 0.6.9 installed on a server QuadCore Xeon 2.0
>>> with 12GB of RAM. When you have more than 20 avatars PhyFPS is connected
>>> to around 35fps out why this might be happening?
>>>
>>> Oliver
>>>
>>> _______________________________________________
>>> Opensim-users mailing list
>>> Opensim-users at lists.berlios.de
>>> https://lists.berlios.de/mailman/listinfo/opensim-users
>>>
>>>
>> _______________________________________________
>> Opensim-users mailing list
>> Opensim-users at lists.berlios.de
>> https://lists.berlios.de/mailman/listinfo/opensim-users
>>
>>
>>
>> _______________________________________________
>> Opensim-users mailing list
>> Opensim-users at lists.berlios.de
>> https://lists.berlios.de/mailman/listinfo/opensim-users
>>
>>
>
> _______________________________________________
> Opensim-users mailing list
> Opensim-users at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/opensim-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://opensimulator.org/pipermail/opensim-users/attachments/20100923/d359bf2e/attachment.html>