Forest Right #5 part 2

Forest Right #5 part 2

Description

To render on CPU
Next 400 frames.

Message boards : Comments and discussion : 3203

Author Message
Profile DoctorNow
Project donor
Avatar
Send message
Joined: 11 Apr 05
Posts: 392
Credit: 2,168,338
RAC: 3
Message 14800 - Posted: 20 Oct 2016, 8:07:43 UTC
Last modified: 20 Oct 2016, 8:08:12 UTC

Well, that's a bit strange.
This session needs 9 GB as stated on the status page, but some of my tasks on my 16 GB-memory host go into "waiting for memory"-state after some minutes, although there are "only" about 29% of the memory occupied by other stuff, so there should be plenty to use it...
Also HDs have enough GB free, so that's not it either.
More strange, my other host with "only" 12 GB is doing the wus without problems, wus indeed run with 9 GB usage as task manager says.
____________
Life is Science, and Science rules. To the universe and beyond
Proud member of BOINC@Heidelberg
My BOINC-Stats

Profile DoctorNow
Project donor
Avatar
Send message
Joined: 11 Apr 05
Posts: 392
Credit: 2,168,338
RAC: 3
Message 14801 - Posted: 20 Oct 2016, 18:34:44 UTC
Last modified: 20 Oct 2016, 18:36:28 UTC

While I couldn't figure out why my Win 7-host with 16 GB doesn't run these tasks (though there is enough memory & space) but the Vista host with 12 GB does, the returned results catched my eye.
Not a single one which was reported got credit, they all were marked invalid!
Please have a look at the tasks below, I hope anyone can tell me what could have been wrong with them.
10219105
10218845
10218547
10218611
10218594
It is a bit suspicious to me that most of the time I had wingmen with Win 10 and these got all validated. Could it be that there is a validation mismatch between Win 7 and Win 10 or something like this?
____________
Life is Science, and Science rules. To the universe and beyond
Proud member of BOINC@Heidelberg
My BOINC-Stats

Profile Janus
Volunteer moderator
Project administrator
Avatar
Send message
Joined: 16 Jun 04
Posts: 4487
Credit: 2,094,806
RAC: 0
Message 14812 - Posted: 23 Oct 2016, 9:47:30 UTC
Last modified: 23 Oct 2016, 10:30:05 UTC

The "Waiting for memory"-bit is typically caused by BOINC preferences where "Use % of memory while in machine in use" is set to less than 100% of what it can use while not in use. Apart from that I can't come up with any other good idea as to why it would do that.

I'll pull the data on one of your workunits to check it out in a few hours

[Edit:] Looks like a fair amount of difference in the alpha channel around the top of the trees along with random "fireflies" of difference in colour throughout the remaining image. Seems like the scene is simply not fully deterministic. The validator unfortunately doesn't know better (yet, there has been some work on that, this is a common issue with Cycles renders and the reason why we still consider it experimental here) than to reject your result and accept two other ones that randomly happen to be much closer together.

Profile DoctorNow
Project donor
Avatar
Send message
Joined: 11 Apr 05
Posts: 392
Credit: 2,168,338
RAC: 3
Message 14813 - Posted: 23 Oct 2016, 13:30:17 UTC - in response to Message 14812.

The "Waiting for memory"-bit is typically caused by BOINC preferences where "Use % of memory while in machine in use" is set to less than 100% of what it can use while not in use.

Ouch, I totally forgot to check these settings, thought I had it up to 100%, but it actually was only at 50%, that explains it of course... sorry and thanks. ;-)
____________
Life is Science, and Science rules. To the universe and beyond
Proud member of BOINC@Heidelberg
My BOINC-Stats

Profile ChameleonScales
Avatar
Send message
Joined: 2 Apr 16
Posts: 143
Credit: 1,120
RAC: 0
Message 14816 - Posted: 24 Oct 2016, 1:29:37 UTC - in response to Message 14813.
Last modified: 24 Oct 2016, 1:53:51 UTC

Thanks and glad you figured it DoctorNow.

"difference in colour throughout the remaining image" ? Could this by any chance be the same kind of issue as this ?
I doubt it is since in your case you say the alpha channel is also altered, but I really wonder what's causing the differences in this.
____________
https://caetano-veyssieres.com

Profile Janus
Volunteer moderator
Project administrator
Avatar
Send message
Joined: 16 Jun 04
Posts: 4487
Credit: 2,094,806
RAC: 0
Message 14819 - Posted: 25 Oct 2016, 16:58:05 UTC - in response to Message 14816.

Could this by any chance be the same kind of issue as this ?

No it looks more like a kind of halo around the leaves and just a handful of random single pixels strewn around on the remaining image. You only notice it if you already know where to look. It is almost like there is some kind of backlight and just that light source is rendered ever so slightly different every time a frame is rendered.


Post to thread

Message boards : Comments and discussion : 3203