So Im noticing just how much more memory it takes to use Adaptive Subdivision versus Adaptive QMC. My question is has anyone figured out what settings you need with the adaptive QMC to equal the quality of adaptive subdivision at say 0,3?
well…qmc has different setting for its final result. For example qmc will produce a sharper image then adaptive. But it will require more samples for the less noisy picture then adaptive. Because adaptive will produce always soft result. I found that to get better results qith qmc you need to work with qmc sampler. I believe that some where in the instructions to vray it was stated that the two are interconnected. What I found was that buy decreasing the noise threshold from 0.005 (default) to 0.004 and increasing the globals subdivision multipler by 4, (default 1) the result is crisp but with no noise.
Also, I found that when rendering complex geometry (millions of details) adaptive seems to take much longer time then amc. Infact I dont use adaptive anymore.
Anyways it my take on it.
Cheers
thanks for the info. So what do you use for your AA settings then using the qmc?
usually 1/4 and default qmc sampler with mitchell natravalli antialiaser. But sometimes I use soften filter with radius of 4.0 (to blend things like adaptive)
Ditto 8)
Mine is defaulted to adaptive QMC as well. That’s also advised in the manual when using lots of glossy effects. Once you get used to it, it is very reliable and predictable and your QMC sampler will become your best friend 
To get similar render quality you will need to increase your settings though either global or local. If you change it global, you might want to consider to lower the subdiv. settings for irriadiance, since it affects all subdivisions (except lightcache)
I also use QMC all the time.
AA set to 1/4, QMC settings noise th = 0.001 and local glossy subdivs around 30, same or even higher for area shadows.
noise th 0.001 is very high quality, but rendertimes are lot higher than with0.005 of course
Im wondering if Vray’s qmc is the same as brazil’s? would be nice if it would.
wow…ill have to post some results…but I had to really jack up the AA settings with adaptive qmc to get even close to the adaptive subdivision. Of course, like stated adjusting the global qmc settings will improve that…will have to try that out.
okay this is getting crazy. Heres a test with adaptive subdivision 0,3. QMC settings are default!
Heres same image with adaptive qmc 1,4. Ive gone way overboard with qmc settings. .001 noise threshold and 15.0 global subdivs!.
Load those two up in ram player and scrub them back and forth and you can really see the difference. Ive even tried jacking up AA to like 6,15 and it still doesn’t look as good as standard settings with adaptive subdivision AA. Now the render times are alot faster, and of course memory is MUCH lower which is nice, but the quality stinks.
ive even tried taking adaptive amount to .7 and global subdivs to a crazy 30.0 and still looks like utter shite compared to adaptive subdivision…
can we test this scene?
well, I can’t post it for everyone to download, but I could up it to ftp tomorrow. (cant upload to ftp at home, my modem quits working when i do for some reason) Email you the download location if youll take a look.
yeah lets do that
okay anyone can try this file:
http://myweb.cableone.net/amylenhart/stuff/glossytesting.max
have issues getting good smoothing of glossies. Adaptive subdivision just looks so much better with same material glossy subdivision amount.
Try 30 subdivs for the glossy material, and 0.001 for noise th in the QMC rollout, global subs=1.
Don’t be fooled by the global multiplier, here’s how it works:
local subs=8
global=15
result: 8*8*15=960
local=30
global=1
result: 30*30*1=900
So 15 global is not really that much, and multiplying global with 2 is not the same as multiplying local subs with 2.
I always use local=30 and global=1, with 0.001 noise th and glossies are perfect. I’m on max 5 so I can’t open your file…
In a test like yours, adaptive subdivision is probabely faster for the same quality, but in a scene with lots of glossies, area lights, displacement, trees, etc… qmc AA is definately faster.
Here’s my test:
Materials glossy 30 subdivs, adaptive sub AA 0/3, qmc noiseth=0.005
Materials glossy 30 subdivs, adapt QMC AA 1/4, qmc noiseth=0.001 (=900 samples in total)
Materials glossy 8 subdivs, adapt QMC AA 1/4, qmc noiseth=0.001, global subdivs=14 (=896 samples in total)
The adaptive subdiv test is a little bit better, but adjusting the material subdivs to 35 would solve this. Already in this test, adapt QMC is faster for the same quality, imagine what it would do in a more complex scene! It definately rocks
Percy,
just tested your file. Try for example min samples = 16 and global subdiv. = 16 and noise treshold = 0.001 and you wil get the same quality with faster rendertime. (AA adaptive QMC settings default)
thanks for the help. The thing is, I kinda knew if I upped the material glossy subdivs it could smooth them out, its just that it didn’t seem necessary if I used adaptive subdivision. Even default values of 8 looked great. Seemed strange to me. But I can see Ill just have to go at it that way. 
Thanks again.
There’s something about AA in the helpfiles which explains why adapt QMC or even fixed is faster than adapt sub with lot’s of glossies or other blurry parameters (MB, area shadows, …). I think the reason you need to up the material subdivs is related to that.