Hi Marcus,
Rieche, Marcus wrote:
>
> I have made some tests. The same light and data etc. gives different
> Results.
>
> I data is based on one node. So I did a splitting. And based on that and
> the server count I get different FPS. The best result in cluster is with
> Sort Last and Pipeline Composer with
> 47 FPS. But on a single computer with no cluster window, but with the
> window I used before as client window I get 71 FPS.
If you can run the mode at 71 fps, why do you want to use SortLast? ;)
But seriously: SortLast has fairly noticeable constant overhead in the
form of reading the depth buffer, sending it over the network and
composing the partial depth buffers. But this overhead is independent of
scene complexity, so the main domain for SortLast is do support massive
models which can not be rendered on a single node, for which that
constant overhead is insignificant.
> I checked the traffic. It is about 20-30MB/s in 1GBit network. There is
> enough bandwidth. But maybe not enough speed?!?
The buffer read/send/compose are not pipelined right now, so for a small
scene and a small cluster like yours you will not see the network as the
bottleneck, so I'm not too surprised to see that result. Which graphics
hardware are you running on? I haven't run benchmarks on this kind of
configuration to see if there is some other problem, so YMMV.
> What can I do to improve the performance of SortLast?
Use a bigger model! :)
Hope it helps
Dirk
-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Opensg-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensg-users
