E-Cycles - March Update with cleaner viewport and denoise addon included

commercial

(Komposthaufen) #62

I downloaded the RTX build and tested the same scene, but it’s significantly slower on my GTX 980 (maybe it’s to old).

RTX E-Cycles Time: 30:25.40 Peak Memory: 655.69M


The result is nearly the same to E-Cycles, only a few fireflies are different.

@bliblubli Is this Build using Optix?


(mathieu) #63

no optix, it’s just that Nvidia fixed cuda 10, so it now supports older cards with Cycles. So I compiled with it. The 1080Ti is 6 to 12% faster with it. Sad it does the contrary on older generation. I’ll continue to provide both then, 9.1 and 10.0 cuda kernels.


#64

I don’t want to prove nothing.
For me, it’s as simple as the fact that of 24h cycle the bright part is the day and the dark is the night — visual difference, while same values are used (especially with one and the same engine) generally mean something is wrong.
The last example (ayreon2) show it even more.

Trying to sell the same “tech” to any other render engine & developers should be ready for a war :wink:


(mathieu) #65

Do you mean it’s bad because you have less noise for the same sample count so it changes the image? Try 2.78 and latest 2.78 and compare images of the same render, you can get same differences due to changes in the russian roulette by Brecht. Or do you think Cycles should keep it’s noise level and noise pattern forever so that those image comparison stays black?


#66

There’s nothing to think about, because nothing was really explained.
To me the difference in brightness is just too much. Simple as that.
ie. There were similar cases with Corona development (sampler) that introduced somewhat inconsistent results (but were way less prominent) yet most were considered as bugs and got fixed.


(mathieu) #67

I worked with a company making renderings for Zaha Hadid architects and they found the renderings done with those patches good enough. So at least until this level, it pleases clients. Maybe there is higher.

And I spoke with one official cycles dev. He has the code and their is interest to get the tech I use here (which comes from a research paper and which accounts for the changed noise pattern) in master at some point. You can then discuss about these point during review. As I said. Everything in here will be submitted for review after a year.


(mathieu) #68

Happy new year everyone,

you can now get only one month of updates if you want to.
And people joining during the first week get the 30% off for the first month, but also the following ones if they stay: https://gumroad.com/l/vkTeQ/66cx403


(BloQi) #69

Hello,

I bought to try it out and it is really cool, works great with a 1080Ti
I don’t know if it has been mentionned before (if so, I’m sorry), but there’s no default addons with it ? Like node wrangler, etc.


#70

I copied the addons from the official buildbots, it works.
The speedup is indeed insane, but the most cool part is the auto tile size. I found it so cumbersome to benchmark every scene for the optimal size. Why not simply remove the UI part @bliblubli? It would make some room.


(mathieu) #71

Thanks for reporting. I added the contrib addons, you can find the new build is on the product page.

@janbauer ok, it will be available in the next release.


(cactus1138) #72

Hi
Its there a way to switch off thsi automatic tile size ?
In my case with this feature it render way longer than standard build on GPU+CPU with 16/16 tiles


(mathieu) #73

Hi,

yes it’s possible, will be added in the next build. However, you should really check if CPU+GPU is faster than GPU alone. Only if your CPU is much faster than your GPU will it bring something (like a GTX 1050 on a threadripper).

Can you please report your times on BMW with CPU+GPU and with GPU only? It will be 16x16 anyway on this scene. https://download.blender.org/demo/test/BMW27_2.blend.zip


(beep) #74

Is it possible to get OSX builds?


(mathieu) #75

If there is enough demand, yes. If at least 5 people want to take the year option, I would invest in a small mac to do the builds. So if you find 4 other people, you get it.


(mathieu) #76

The linux build with an option to deactivate auto tile size is already up, the win version comes tomorrow.


(mathieu) #77

The windows build is uploaded. I recommend to check if it’s really faster with CPU. I’ll be happy to see your comparison with and without CPU at the tile size you prefer.


(cactus1138) #78

I have render bmw27_gpu

Blender 2.79.6
GPU+CPU /tiles 16/16 - the fastest setup
time 2:20

E-cycles
GPU+CPU/tiles 16/16
time 1:28

So it looks like a nice speedup , I will chck it with other files

I have i7-3930K, GTX 1060, win10


(Lissanro Rayen) #79

I have tested render speed in BMW scene. With vanilla Blender it was 209 seconds, with E-Cycles it is down to 119 seconds (1.75x faster).

Classroom - 627 seconds with vanilla Blender and 446 seconds with E-Cycles (1.4x faster).

Impressive speed-ups. For rendering I have used CPU and overclocked GTX 680 (1200 MHz GPU and 7100 MHz memory).

Also, support is good, all issues (specific to E-Cycles build) I have found and reported were quickly fixed.


Work faster - Create your own Blender, with new modifiers and faster rendering
(mathieu) #80

Good to see it also gives good speedup using CPU. Just to know, could you render the same scene but with only the GPU?


(mathieu) #81

You’re welcome :slight_smile: