[pluMmet]

Maybe somone could elaborate on how you take separate low quality images and merge them into a singular, superior image. ......definitely not clear on that.

Well the seed number is different...Very roughly it means that the noise is different on each image meaning that the clearity is different. So it adds the areas of clearity together.

There will be however a sweet spot of how many cpus are needed and more then that will be a waste.

[ivox3]

Okay, .....now that makes a little sense. ......

[Mihai]

There will be however a sweet spot of how many cpus are needed and more then that will be a waste.

Big assumption, how do you know it works like that? I think you can merge as many mxi's as you want. Think of it not in terms of pixels, but "layers" of calculated values that you can stack on top of each other, which are then in turn transformed to rgb space.

[Leonardo]

I don't want to hurt anybody's feeling... but for some odd reason I feel that this news... are bad news.

I'm not talking about... the fabulous render farm that we are going to need so speed is no longer an isse. Or the fact that we will need more license....

I'm talking about the "e-mail"= I get the feeling RC6 is not coming anytime soon and when it comes, it will be another RC1!
No mac, no linux, no way to use it... unless you're tom or Mverta. Unrealistic e-mails by Maya, that CAN'T duplicate with the same software.
It happened with Beta
It happened with RC1,2,3, and 4
It will happen again

EDIT- For the record... this are only "my predictions"
Just like this one
http://www.maxwellrender.com/forum/view ... light=beta
oh.. but I have some good news.... I save a sh*t load of time and money by switching to V-ray

[paxreid]

There will be however a sweet spot of how many cpus are needed and more then that will be a waste.

Big assumption, how do you know it works like that? I think you can merge as many mxi's as you want. Think of it not in terms of pixels, but "layers" of calculated values that you can stack on top of each other, which are then in turn transformed to rgb space.

Again..I think it is a brilliant solution, but how many bought into maxwell thinking that in order to get a good image, in hours instead of days, would have to buy a license for 8-12 cores + purchase 3-5 additional dual core computers to get acceptable rendering times. The assumption by NL that the 'problem is solved' by the customers shelling out an additional 4k-6k to make a rendering farm is a slap in the face. Very dissapointing.

[Fabius]

If i remember correctly Victor, after the defacement of RC1, RC2, RC3, etc. told us in a thread that "they will consider a way of refund" for Maxwell customers.

"The more machines you use, the faster and the more accurate the results"

So, why don't give us at least a 1 cpu licence?

This could be a great way to stabilize the histeric feeling in the forum (to wait something every friday...)

[pluMmet]

There will be however a sweet spot of how many cpus are needed and more then that will be a waste.

Big assumption, how do you know it works like that? I think you can merge as many mxi's as you want. Think of it not in terms of pixels, but "layers" of calculated values that you can stack on top of each other, which are then in turn transformed to rgb space.

Well lets see....If 1,000 cpus reach level 1.25 each how would such an image look if compiled?

I think my assumption is correct but your right it is an assumption.

Though my assumption is the opposite of this...I said too many would be a waste. In co-op mode this makes sense. If 20 cpus can reach level 13 in a hour and compile into a perfect image then more cpus won't matter.

Again if 1,000 cpus reach level 1.25 in 3 min. you'ld get mud as your final image. There is a CPU/level sweet spot.

[Kabe]

Now, if you were combining a series of noisy renders
to get a nicer result (different seed values suggestion)
how would the program tell the data from the noise?
Chaos math?

Nope, simple averaging. To clear images by averaging is quite common in
digital imaging, be it in scanning, time lapse video capture, photography.

This feature is relatively cheap, because all you need is the current state
from your network render nodes when the noise is random for each image.

The noise in the result get's decreased because it's normally distributed.

Kabe

[aitraaz]

Well, if you take every fifth letter and run it through some pre-Benedictine, post albigensian crusade cryptographic cipher systems you come up with this:

sixth tuesday lunar cycle 16 polar equation northwest unbiased strata beacon

I mean, I think they were just trying to send everyone a periodic update. I wouldn't read too much into it, i mean, i hope it doesnt mean another release is that far off...

[dutch_designer]

Well, if you take every fifth letter and run it through some pre-Benedictine, post albigensian crusade cryptographic cipher systems you come up with this:

sixth tuesday lunar cycle 16 polar equation northwest unbiased strata beacon

I mean, I think they were just trying to send everyone a periodic update. I wouldn't read too much into it, i mean, i hope it doesnt mean another release is that far off...

Well I read in another thread Mike mentioned another round of testing is currently going on after the playtime they had a while ago... meaning that we'll probably see a new release soon.
Of course, nobody knows how long the testing will take eh?

[Mihai]

Again if 1,000 cpus reach level 1.25 in 3 min. you'ld get mud as your final image. There is a CPU/level sweet spot.

I don't think so, not with Maxwell's system. You would get a clean image if you combine those 1000 mxi's, even if they individually reach a low sampling level.

[goncalo]

One thing i didn´t understand...

They gonna change the number of licenses to 10 ????

Because Maxwell still so slow and they can solve that problem...

Is that it?

[jdp]

no email here yet...

[Frances]

I don't think so, not with Maxwell's system. You would get a clean image if you combine those 1000 mxi's, even if they individually reach a low sampling level.

Yes, I think that's right. Theoretically, the samples represent discreet portions of the image that have been completed. Those samples (presumably) would not be further refined.

That's my take on it anyway. But I am often wrong and sometimes wronger.

[Yog]

Again if 1,000 cpus reach level 1.25 in 3 min. you'ld get mud as your final image. There is a CPU/level sweet spot.

Heck, you guys are getting renders in 3 minutes ?

The last two scenes I tried to pass through it, voxolisation on one took 20 mintues, and the other took 40 mintues to finish voxolisation
Test renders my A**

OK, so they were rather large scenes, but I was getting rather nice final images from other renderers in the same time