[b-kandor]

Hi,

In coop, if I select the job and click display, it computes an image and then displays it. The tab for the coop image also shows the aggregate sl number.

Here is what I see:

4 machines at sl 10.89, 8.63, 8.37 and 11.00

But - the coop image tab shows aggregate sl of 10.17 - is this due to merge losses? If so, the loss seems very high, since I would technically have a higher sl with just my one machine?

Kandor

[Peter Cernak]

this could be due to the time consuming merging process in higher resolutions IMO. when computing of coop image starts the highest single SL is 10.17. when it finishes the highest single SL is already 11.


or maybe it's a problem rogerteddy described in his post and i have this problem as well [url]http://www.maxwellrender.com/forum/view ... art=15[url] .

it seems to me maxwell is assigning same seeds to different nodes in one job.

it shows success after the merging process but the resulting mxi has often lower SL then the highest single SL on a strongest machine.
if you try to merge manually all the cooperative.mxis from temp directories some of them aren't mergible due to same seed.

and as i did write in my earlier post - i can see two computers in ONE AND THE SAME job writing final mxis.
sometimes - for example with three machines P4, dual xeon, dual quad xeon - P4 is writing final mxi a split of a second later then dual quad xeon is writing its final mxi (already merged with dual xeon). the resulting SL of the job is in this case a SL of P4 node alone.

as i said earlier maybe it's a network setup problem (but i tried to share temp folders on the network and set all the permissions and it didn't help. but i'm no IT specialist).

this coop thing is really a great and unmatched idea and concept. when it works it really rocks. but now i can't use it for some reason.

P.

(mw 32bit on 6 pcs, one of them XPpro 64bit, others 32bit.)

if somebody could help i woud be thankfull.

[b-kandor]

I would really like to figure out why this happens?

On a new render with 4 machines, currently at:

13.07
11.53
11.3
13.95

The coop - display tab indicates an sl of 13.13. It just can't be accurate.

Also, the image display of the 13.95 is demonstrably better than the coop display.

[Peter Cernak]

yes, that's the case.

Probably machines with SL 13 merged their's mxis, while at the same time machines with SL 11 merged their's.

Then if you watch the merging process in the MR viewer, you would probably see "writing final mxi" on two nodes. That should not happen - you shoud get only one final mxi. And in most cases - because the slower machine writes the final mxi slower, their final mxi overwrites the final mxi with higher SL.


At least that's what i was able to figure out.