OctaneRender® for 3ds max® v2018.1 - 7.04

[acc24ex]

How do we activate texture compression, and does "use cpu memory" actually do something ..
.. also how does nvlink thing runs, I got 2 x2080ti nvlink is on the way - what should we keep in mind - currently working on a scene 10gb vram and testing how high it can go..

.. also why doesn't set as movable proxy pop up the model in the viewport and requires reload (not refresh)? did I change something in the prefs or is it a feature/bug?

[senorpablo]

How do we activate texture compression, and does "use cpu memory" actually do something ..
.. also how does nvlink thing runs, I got 2 x2080ti nvlink is on the way - what should we keep in mind - currently working on a scene 10gb vram and testing how high it can go..

.. also why doesn't set as movable proxy pop up the model in the viewport and requires reload (not refresh)? did I change something in the prefs or is it a feature/bug?

For texture compression, you have to go into every single map in your scene that you would like to be compressed, and select a compression option(BC1, BC3, BC7, etc.) from the Channel Format dropdown within RGB Image, Greyscale Image, etc.

It's gets a little ludicrous for large scenes with many textures, which naturally are precisely the scenes you need compression for. Especially since Octane introduces terrible lag in the Max material editor.

I suggested a while back that the compressed textures be saved out locally, and a global compression switch would load the compressed version(dds) of a texture if it existed, rather than regular versions(tga, jpg, png, etc.). That's how it's done for games.

[acc24ex]

damn, how counterintuitive - I looked into the manual and everywhere and checked the maps, no where it says "compress texture" ? why the hell would you name it "channel format" yeah I saw that it saves cache locally, but didn't know this was made like this with no global switch .. yeah maybe have a switch on textures, and a global switch to activate all or disable all seems very obvious way to do it .. also how does the compression affect the quality I'm using 8k displacement maps, so OK, if I knew this already I'd already compress them when I made them (and test for differences) ..

still, use CPU memory doesn't do anything to me - last time the scene hanged for half and hour or more when I decided to shut it down,,

Octane team, now we paid for the upgrade, at least make the obvious stuff obvious and fix those little crap things

BTW is nvlink working in max, anyone tested? I got one but my motherboard is crap, and I couldn't activate sli, what are the sli requiremnts - googling it I mostly get gamer reviews - also, is this maybe only available on win10 - I'm using win 7 - because of that crappy vram overhead thing from win10 - currently running a scene that's 10 gb large on 2x2080ti and I actually need all of this

[coilbook]

I agree with CPU memory. Still getting crashes out of memory even though CPU mem is on and before when it worked tons of CUDA out of memory errors even thought CPU mem was on.

[mbetke]

SOmewhere it said we don't need the use CPU Memory option anymore because octane will detect lack of vram and enable it internally. But nothing more heard of it....I hope they fix the lags and get more stability into editing materials while having interactive viewport open.

[acc24ex]

SOmewhere it said we don't need the use CPU Memory option anymore because octane will detect lack of vram and enable it internally. But nothing more heard of it....I hope they fix the lags and get more stability into editing materials while having interactive viewport open.

.. well I'm working on this scene with 10gb of vram - and I usually see it crashing when I go over - actually every time, nvlink is a bit buggy on my motherboard, waiting for a new one, maybe that might do something - in the meantime, yeah what's happening with cpu memory thing, and nvlink - how do we test if it's working apart from this https://www.pugetsystems.com/labs/suppo ... s-10-1266/
.. it says nvlink not operating - I'm using a 2010. asus ws mobo (with 32 pci lanes somehow on i7 950 cpu) .. impressive it still works, but nvlink doesn't like it - also on win7, also seems that you need a monitor plugged into one of the gpus for sli to turn on - and also makes the whole system sluggish even if you don't use anything on that screen ..

- I mean 2080ti works pretty fine on their own, but this new nvlink thing is probably gonna need more work

[mbetke]

I didn't bought a NVLink bridge but will in a few months. So I can't tell about its stability. But some people on facebook group got it working as far as I know.

[coilbook]

why do i need to use alpha shadows for cast/receive shadows to work. Alpha shadows slows rendering a lot but I need my objects not to cast shadows. Please fix this otoy

[acc24ex]

I didn't bought a NVLink bridge but will in a few months. So I can't tell about its stability. But some people on facebook group got it working as far as I know.

ok, so they managed to run nvlink on standalone build, also what is xb1 - x beta1? Really we should relax with those numbers - how about just add a +1 at the end, I don't care as long as I can keep up, so if my version is 6000 so I can know that 5600 is lower version, 2018 xb1 vs 2018.1 - 7.04.. I'm lost.. I'm guessing this is newer than that version as it was in december 2018.. and there's a .1 at the end..

viewtopic.php?f=33&t=69896#p353452

Anyhow, I used the nvlink test tool from puget systems link above - installed win 10 - and sli worked and the tool said nvlink was operational
- I then loaded a cube divided by 200x200x200 and did 10 then 20 copies.. etc.. after 40 or so 3dmax crashed - I even installed a fresh max 2019 - so nothing else was on it, just octane - did it again, with less cubes, again just crashed - so I'm guessing nvlink not stable, not workin - also this on a old asus p6t ws mobo - so it could be either of those ..
..can someone go a bit deeper into what's going on with nvlink - does it need pcie x8, v3.0, what are the minimum system requirements (I couldn't fin much on the web, does it need win 10 etc)

..just trying to work out can I rely on the nvlink vram shareing or should I stick to optimizing so the scene is under 10 gigs (or how much the win 10 overlay is)

BTW also didn't realize we need alpha shadows to cast/receive shadows, I do leave it on cause there's issues with stuff like cloth with transparency doesn't pass much light, and similar issues.. so I leave it on even though I don't really have any opacity masks..

[mbetke]

7.05 when? V4 got an update. Or I am just impatient?