Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Topics - agentdark45

Pages: [1]
1
I was reading about the following and was wondering if this applies to Corona/3ds max in any way:

https://www.reddit.com/r/Amd/comments/e24jqp/extremetech_how_to_bypass_matlabs_cripple_amd_cpu/

Cliff notes:

Anything that uses the Intel Math Kernel Library (https://en.wikipedia.org/wiki/Math_Kernel_Library) uses a discriminative CPU Dispatcher that does not use efficient codepath according to SIMD support by the CPU, but based on the result of a vendor string query. If the CPU is from AMD, the MKL does not use SSE3-SSE4 or AVX1/2 extensions but falls back to SSE1 no matter whether the AMD CPU supports more efficient SIMD extensions like AVX2 or not.

When fixed, the results are claimed to be: "AMD’s performance improves by 1.32x – 1.37x overall. Individual test gains are sometimes much larger. Obviously these results are much worse for Intel, changing what looked like a narrow victory over the 3960X and a good showing against the 3970X into an all-out loss."

Devs can you weigh in on this?

2
AMD tweeted out a link earlier about their new Ryzen optimisation/performance analysis tool: https://developer.amd.com/amd-uprof/?sf215410082=1

I was wondering whether this was already being used by the dev team, and if not would it be useful in squeezing out more performance from the new Zen 2 Ryzens/upcoming Threadrippers? Seeing as the new chip architecture is drastically different from anything that's come before it, would this be something worth looking into?

The reason I bring this up is due to the lower than expected performance of the current gen Threadrippers with Corona (compare the corona bench scores to similar scoring Xeons, then compare those chips Cinebench score - something is amiss). However this could also be due to the already noted windows scheduling/NUMA issues...

Some insight here would be appreciated!

3
Hi all,

We require advanced 3D modelling and texturing, organic sculpting and general 3D interior skills with the ability to carefully follow mark-ups and other briefing instructions accurately.

Attached are examples of the complex furniture that needs to be modeled. These will go into a collaborative 3DS max scene.

It is likely that the first task will be to model and texture furniture and not be working in the scene itself. Please reply if you are able to meet this standard of modelling such items.
 
Please let us know your costs for individual furniture modelling and texturing or if you would rather cost on a rate basis.


Project Info:
6 no. interior visuals (working with base scenes provided by us, final renders done by us or by you depending on skill level).
Modelled in 3DS max, rendered using Corona
Start date: immediately
Deadline: 13th April (TBC) but future work will be required on a regular basis

4
https://blogs.nvidia.com/blog/2017/05/10/ai-for-ray-tracing/


Around the 2 minute mark in the video above.

Very interesting happenings! I wonder what else neural nets could speed up in the rendering process?

5
General CG discussion / F-Storm LUT capture?
« on: 2017-10-09, 15:23:36 »
Hi guys, just wondering if anyone has any experience with F-Storm here?

I'm part of the F-Storm Facebook group and am constantly impressed by how well tone-mapped some of the images are, however I can't get my hands on the software at the moment to do any experimenting with (it appears to not be for sale).

My question is whether someone who access to the engine is able to capture whatever they are doing with their default tonemapping as a LUT for use with the Corona VFB? The highlights, midtone contrast, colours and shadows are just spot on in almost everything I've seen (calling on Dubcat!).

Some examples here:

https://evermotion.org/vbulletin/showthread.php?116783-Visualization-FStorm-renders-collection

This isn't a knock to Corona in any way (I'll be sticking with it for the foreseeable future) but there's definitely some tonemapping magic going on in F-Storm that has me extremely intrigued.

6
Work in Progress/Tests / Atomizer concept renders
« on: 2016-12-23, 02:38:16 »
Hi everyone, just posting up some renders of an atomizer design (for vaping/advanced e-cigarette use) I came up with as a personal project. Everything is modeled to scale (width of the tank is 25mm). I'm currently in talks with a company to get this put into production after posting this on reddit's e-cigarette sub-reddit. The metal was done using the new triplanar mapping feature which is awesome btw!

7
Not sure if this has been posted before, but this just popped up on my facebook feed. Looks very interesting!

Technical stuff is linked in the paper below:

http://hexus.net/tech/news/software/94882-researchers-devise-100x-faster-way-render-realistic-metallics/

http://cseweb.ucsd.edu/~ravir/glints.pdf

8
http://hexus.net/tech/news/cpu/90458-amd-zen-based-opteron-chips-detailed-cern-slides/

With a 40% IPC improvement and individual cores with hyperthreading AMD might finally be back in the game! Anyone ever own an Althon chip back in the day?

I've never looked at Opteron chips before but anyone care to make a speculative guess on how much a 32 core Opteron would cost? I'm currently running a 12 Core Xeon, but if the new Zen chips live up to expectations I would have no issue switching to AMD.

Also, would network rendering work with machines running completely different CPU's?


9
Hi all, I've recently built a new 12-core Xeon v3 system and I had a few questions:

1. Do I need to configure Corona / Max in any way to obtain optimal rendering speed with this many cores or is everything good to go?
2. I've noticed IR renders are a lot more noisy /don't clean up as well than non-IR renders for the same amount of time. Is this something to do with how IR uses the CPU cores or just the way the IR algorithm works?
3. Will there be any issues using my old 3930k 6-core system as a render slave considering it's a different type of CPU and uses DDR3 ram? (new build uses DDR4)

The old system is out of commission until I get my new graphics cards, so I haven't been able to test DR, but I thought I'd ask ahead to time / to save myself any headaches.

Thanks!

10
I need help / I need a glow plugin that works with Corona
« on: 2014-06-25, 14:09:20 »
Hi everyone, I've got a scene with ~ 3000 tiny led lights (corona light mtl, all one object) that need to glow quite brightly, rather than being just white dots. The problem is that they are so small the regular mask / photoshop method isn't working. Is there some sort of plugin I can use in 3ds max that enables a glow / glare effect on the light emitting material?

I can get the glow effect on the led's in keyshot really well...but I'd rather not have to composite anything - it would be great to do it all out of 3ds max...especially as this is going to be animated.

Many thanks!

11
Resolved bugs / vfb saving issues
« on: 2014-06-04, 15:47:41 »
I'm getting some weird issues saving files from the vfb in the A7 preview build. The saved filed only contains a region (with a black background) that doesn't correspond to the location of the 3ds max region. It also is affecting VFB+ and outputting the wrong region to it. I can copy the image from the vfb fine however.

Not sure what's going on or if I've messed up a setting somewhere.

12
Hi everyone, I'm in the early stages of working on a few concept renders of a hotel lobby for a client and the main feature is that there are several thousand glass rods that go through the ceiling of the building into the lobby, which will illuminate the interior with natural daylight (as well as the rods glowing from the outside of the building at night due to the internal artificial light being transported through the glass rods).

I'm having some trouble getting the glass rods to "glow" or transport the light without making them look matte (by messing with the glossiness parameter). At the moment they look far too transparent. I realise this is probably a nightmare scenario for any renderer...but any tips would be really really helpful! The glass material in the attached render is a standard one (ior 1.6 for both refl/refract, slow option not ticked).

The client wants this as a white render for now, which I assume will also make this tricky as there's not much to refract / reflect apart from a range of grey hues...

Many thanks


13
Hi all,

I thought I would create this thread to discuss how one might apply real photographic techniques to our renders as seen in the following link: http://fstoppers.com/architectural-real-estate-photos-how-to-1

I'm at the point where I can confidently say I've got a good grasp of how to set up a 3D scene to emulate a physically correct "real world" environment (i.e sun + sky + interior lights)...but this got me thinking. If I were a photographer, I wouldn't just walk into a room, snap a shot with my SLR and go home - I would spend quite some time setting up strobes, reflectors, merging exposures e.t.c to control light in a very specific way. This wouldn't be fakery with Corona, it would be physically simulating real world pro photographic lighting techniques

This is something that I haven't ventured into yet in the 3D world as I haven't come from a photographic background, purely c.g, and secondly I haven't a clue where to begin learning about strobe placement, fill lights, when / if they are needed e.t.c.



Take the above image for example, I would have lit this scene with an HDRI environment to simulate sunlight + sky and corona lights for the chandelier lights and the fireplace, hit render, tweak the exposure / light compression and presto a physically correct image (oh it it were only this easy). However, according to the above article the image was made up of 8 exposures with sotfboxes, directional lights, a speed-light and a ceiling bounce light, then merged in photoshop - a whole other layer of lighting on top of my would be "physical" approach.

I am confident that Corona would work excellently with these techniques as I've never seen another render engine manage light quite so well.

Discuss!

14
Hi guys I've been having some trouble with IES / down lighting recently. When adding in IES lights things seem to take 3-8x longer to render due to all of the additional noise.

The 1st attachment shows the scene lit with just an HDRI and an emitter material for the ceiling light, render time ~ 15 mins.
The 2nd attachment has all of the IES lights enabled (photometric light type in max with Corona shadows ticked), render time ~ 40 mins
The 3rd attachment uses a very dark HDRI and all of the IES lights enabled. This one took around 50 mins, you can see the extent of the noise...

All renders are using the same exposure with basic render settings (HD cache, adaptive: 4, latest daily build) One thing that did strike me as odd is that I had to crank the lm values of the lights to 35000.0 else they were too dark.  Exposure is set to 2.75, highlight compression is set to 4.

All glass is set to hybrid mode, the pool is a single plane set to onesided (solid), but with "casts shadows" unticked. Trees are opacity mapped and the RGB of all white materials are set to 180.

Also I'm rendering this on an overclocked 3930k with 32gb ram. I completed the Corona benchmark scene in about 3 minutes 20.

I don't particularly need to use IES lights, I just need some down/spot lights - so anything that can shave some render time off will be a massive help.

Thanks for all suggestions / tips!

15
Just thought I'd post up a bug that I've been experiencing when rendering at higher resolutions (eg 2100x1600px+).

The VFB works great for low res shots, but as soon as I render high res shots and try to zoom in /maximize the VFB while rendering, 3ds max claims an error has occurred and force closes. The image will render file if I don't touch the VFB at all. If it's any help the VFB tends to get garbled sometimes when I minimize Max, and the only way to see my render is to maximize the VFB window again.

Running a very stable machine (3930k with 32gb ram). Also using the Alpha build 5 posted on the forums.

Thanks for taking the time to read this.

Pages: [1]