Page 20 of 24

Re: Different Works in Progress

Posted: Sun Aug 25, 2019 12:04 am
by FarbigeWelt
Laser through Mask
Laser through Mask
There are eight cubes with material Null and two different volume settings.
The cubes overlap each other.
I expect of material Null to have invisible cube faces but obviously there are visible cube faces.
The lower cubes contain small tetrahedrons, actually large enough to be good visible.
The tetrahedrons have material Glass with homogenous volume. The tetrahedrons are the reason for the additional scattering of the lower cubes and also for the green shapes due to light passing through the holes of the mask.

VRAM 10240 of 8176 MB

Posted: Mon Aug 26, 2019 11:17 pm
by FarbigeWelt
What is unbelievable here?
VRAM 10240 of 8176 MB
VRAM 10240 of 8176 MB
Object: cat-scaled.stl, Cat

Re: Different Works in Progress

Posted: Tue Aug 27, 2019 4:58 am
by lacilaci
40 copies (near 20M polys) wont even reach 2GB VRAM
cat.jpg
if by any means something else is eating up your ram then the answer would be instancing
cat2.jpg
But it's probably some bug, seems that it is not present in 2.8 version though..

Re: Different Works in Progress

Posted: Tue Aug 27, 2019 8:52 am
by FarbigeWelt
lacilaci wrote: Tue Aug 27, 2019 4:58 am 40 copies (near 20M polys) wont even reach 2GB VRAM
cat.jpg

if by any means something else is eating up your ram then the answer would be instancing
cat2.jpg

But it's probably some bug, seems that it is not present in 2.8 version though..
Add particles hair on a simple large plane. Use particle childrens in render (100). Set particles to 100‘000. Set different colors for root and tip. Reduce hair resolution to 3. Start openCL render. Wonder about export time of more than 4 minutes and have a look at VRAM used as soon as render starts visible but if you get an error message instead of an image reduce particles to 10’000.

Re: Different Works in Progress

Posted: Tue Aug 27, 2019 9:05 am
by lacilaci
FarbigeWelt wrote: Tue Aug 27, 2019 8:52 am
lacilaci wrote: Tue Aug 27, 2019 4:58 am 40 copies (near 20M polys) wont even reach 2GB VRAM
cat.jpg

if by any means something else is eating up your ram then the answer would be instancing
cat2.jpg

But it's probably some bug, seems that it is not present in 2.8 version though..
Add particles hair on a simple large plane. Use particle childrens in render (100). Set particles to 100‘000. Set different colors for root and tip. Reduce hair resolution to 3. Start openCL render. Wonder about export time of more than 4 minutes and have a look at VRAM used as soon as render starts visible but if you get an error message instead of an image reduce particles to 10’000.
Oh I understand now, so the particles "ate" your vram...

Re: Different Works in Progress

Posted: Tue Aug 27, 2019 10:03 am
by FarbigeWelt
lacilaci wrote: Tue Aug 27, 2019 9:05 am
Oh I understand now, so the particles "ate" your vram...
Exactly. The question is, why ate they more than available and render still run.

Re: Different Works in Progress

Posted: Tue Aug 27, 2019 10:05 am
by alpistinho
FarbigeWelt wrote: Tue Aug 27, 2019 10:03 am Exactly. The question is, why ate they more than available and render still run.
If the stats are accurate, swap most likely. The performance would probably suffer pretty badly, though

Re: Different Works in Progress

Posted: Tue Aug 27, 2019 10:13 am
by FarbigeWelt
alpistinho wrote: Tue Aug 27, 2019 10:05 am
FarbigeWelt wrote: Tue Aug 27, 2019 10:03 am Exactly. The question is, why ate they more than available and render still run.
If the stats are accurate, swap most likely. The performance would probably suffer pretty badly, though
VRAM-RAM-swapping? Never observed with my old graphic cards. In much earlier LuxCoreRender versions not enough VRAM lead to crash. Six to eight weeks ago crash had been fixed, an error message appears instead. I guess the stats are wrong.

Re: Different Works in Progress

Posted: Tue Aug 27, 2019 11:00 am
by Dade
FarbigeWelt wrote: Tue Aug 27, 2019 10:13 am
alpistinho wrote: Tue Aug 27, 2019 10:05 am
FarbigeWelt wrote: Tue Aug 27, 2019 10:03 am Exactly. The question is, why ate they more than available and render still run.
If the stats are accurate, swap most likely. The performance would probably suffer pretty badly, though
VRAM-RAM-swapping? Never observed with my old graphic cards. In much earlier LuxCoreRender versions not enough VRAM lead to crash. Six to eight weeks ago crash had been fixed, an error message appears instead. I guess the stats are wrong.
AMD Vega 56/64 supports swapping GPU RAM over CPU RAM so you can use more than available GPU physical RAM. I assume newer AMD GPUs have that feature too.

AMD Vega 56/64 supports swapping GPU RAM over CPU RAM

Posted: Tue Aug 27, 2019 6:25 pm
by FarbigeWelt
Dade wrote: Tue Aug 27, 2019 11:00 am
FarbigeWelt wrote: Tue Aug 27, 2019 10:13 am
VRAM-RAM-swapping? Never observed with my old graphic cards. In much earlier LuxCoreRender versions not enough VRAM lead to crash. Six to eight weeks ago crash had been fixed, an error message appears instead. I guess the stats are wrong.
AMD Vega 56/64 supports swapping GPU RAM over CPU RAM so you can use more than available GPU physical RAM. I assume newer AMD GPUs have that feature too.
Since both AMD Hawaii (R9 290X, R9 390X) died this summer and the first replacement is an AMD 5700 XT, waiting for its ordered one egg twin, with Navi GPU, swapping was my first thought.

:D :D :D Thank you much for the confirmation that swapping is possible and the stats are okay. This opens some new ways to be discovered. LuxCoreRender is a fantastic software surprising me every week further times. :D :D :D