render speed of slave

[CHF]

Dear Ono and anyone else!

I could not find any information on that issue, but it seems to me, that my slave (quad 2.4) has the same speed as my master (duo 3.0). When rendering directly from indigo not as a slave on the quad, it runs at double the speed. Anyone else made the same notice of the phenomena? Or is that related to networking? Don't know if it fits the "bugs and requests".

CHF

[Zom-B]

I just have a quite similar strange situation where a Intel P4 2.4 Ghz (one core!) gets 4ksps while a core2duo with 2x1.8Ghz does 19k. BOTH computers are slaves here! The P4 maybe only has a older network card (10/100)...
I know a P4 is older etc. but such a big difference seems strange!!

[eman7613]

out of curiosity of corelation, both of your masters have less cores then your slaves - are your slaves running all CPU threads?

EDIT: nm, i miss read zomb's = but zmob - p4 is slow, and has many specific deficiencies that kill performance in random situations (well not random, specific ones realy), including being incredibly slow at shifting. Any core2 thought should be about 2x the speed of its p4 GHz counterpart using a single thread, the remaining 3k diffrence could easily be one of the p4s shortcomming poping up.

[CHF]

I just use the normal process: Start up the render master, turn on network render, start the render, while having the slave running. The render slave connects and picks up the data and starts rendering, using all 4 cores, but not being faster than its master.

[OnoSendai]

It may be delayed when uploading the frame to the master. Try setting the frame_upload_period to a greater value (e.g. once every 600 seconds / 10 mins or so).

[CHF]

Dear Ono,
thanks for the reply. I set the slave to 600sec. Speed of slave (123k) is still allmost the same as the master (121k).
Mh. I will check back what happens, when the quad is the master....

Best

CHF

[Zom-B]

I have some network rendering going on here now and it seems that the Samples per second in the network Tab seems to be way different from the stuff that you can read in the render client window!
I also miss some information about the actual Master samples per second (not only the sum of all!)...

BTW: the clients window reacts strange if you try to make the first fiel wider.... the scrollbar just skips from beginning to the end.... no real scrolling

**EDIT** it also seems that the transfered samples are really less then rendered by the slave!
A connected Mac Pro dual Qaud Xenon with 8 x 2.8Ghz does Approx. samples per second: 741111
But my Master tells me that he only sends 422k and my full master samples per second are around 550k (having total of 3 clients + MacPro + master)

Something smells fishy here!!