It’s my thought exactly but we can wait another year and the introduction price will most likely still be pretty high for high end card. Unless Nvidia really surprise us as with 1080Ti but I doubt it. Then the software we’ll need some time for optimization so it’ll really can be a pretty long wait and I have an opportunity for a good deal now. So I’m debating at the moment.
I answered some points in your original thread if you want. I’ll also update the bench results asap including the 1080Ti, 2080Ti, 2080 (should be about like 2070super) and 2070.
All 2019 users wanting to benefit from the special upgrade price again can send me a mail (answer any mail from Gumroad about E-Cycles) or the “message creator” button on the right panel of the Blender market page and I’ll send you your coupons.
PS: Please use the same account/email you used to buy 2019 if you have multiple ones.
PS2: there may be a Mac version after all
I installed a 2080 a month ago and it’s fantastic. Rendering in Arnold, Cycles, and working in Substance Painter(also rendering in Iray) is WAY faster than my CPU(which is an i9)
Also, texture baking is so much faster. All in all, the time saved is life-changing
Thank you. I appreciate real life experience. I’ve been hearing a lot of positives about RTX in regards to 3D graphics. The software support is getting better so I finally see a point in considering the upgrade if the price of hardware is right.
I installed a 2080 a month ago and it’s fantastic.
By the way, what you upgraded from? What was the previous card?
I was running a GTX 1080, before upgrading to Dual RTX 2080ti founder editions. Even a single RTX 2080ti was a huge difference from the 1080! The RTX couple with E-cycle was the best decision for far! I’m a modeler by trade but want to get into mograph in the future and rendering in general so a fast GPU was a better fit.
In the beginning I was skeptical about E-Cycles rendering speed but I took a chance and bought it , it lives up to the hype and already paid for itself in some side gig rendering logo like items.
Thanks. Great to hear from expirience that it does make sense to upgrade from 1080 Ti to 2080 Ti performance wise.
If you can swing getting 2 or 3 GPUs, get 2070 Supers. I have three EVGA 2070 Super Hybrids. At around $550 each, that is just a little more than a single 2080Ti. But even just one 2070 Super is pretty formidable with E-Cycles RTX.
I was on a 1060.
Consider me asking for a 2.82 Mac build.
E-Cycles 2.82 for Mac is now available E-Cycles builds based on stable Blender releases in 2020 will now also get a mac build.
Please spread the word everywhere you can. If enough people join, I may restart the weekly builds too for Mac!
Made with E-Cycles by @Luan3dr:
E-Cycles can give you more detailed images with 8x less samples than the already awesome OIDN node from Blender 2.81. If you don’t believe, look at this comparison by Marvin Luebke. And it works with any device, so you can use it with OpenCL and CPU rendering
Faster render, more details and one click setup
That’s a great comparison, I love how the timber floor detail is still consistent after denoise in E-Cycles.
Is there any progress in implementing NVIDIA® OptiX™ AI-Accelerated Denoiser? We are planning to buy a lot of new generation of RTX card. Since we run things on server so we really want to make use of VRAM more than RAM.
NVIDIA® OptiX™ AI-Accelerated Denoiser is already in E-Cycles 2.82 and 2.83 builds since December. It’s the one used in the viewport and is available in the pass system like the old NLM denoiser.
But E-Cycles AI denoiser is of better quality still in all the tests I and other artists have done (see my previous post for example), so it’s the best choice for final render. If you find a case where OptiX denoiser is better, you are welcome to show it
Very nice comparisson! I would like to know if it’s possible the E-Cycles settings for that render. Just to compare with more information like light paths, AO bounces, etc.
Viewport OptiX denoiser is amazing! It really makes a difference! Great job.
I have a question regarding out of core memory rendering. So if we want full speed whole scene has to fit GPU memory, right? I was rendering 2K res scene and had to limit textures to 2048MB, but when I set resolution to 4K, everything was fine to about 50% of the render, then the CUDA error happened.
OptiX CUDA error CUDA_ERROR_ILLEGAL_ADDRESS in cuMemcpyHtoDAsync(launch_params_ptr + offsetof(KernelParams, tile), &wtile, sizeof(wtile), cuda_stream[thread_index]), line 753OptiX CUDA error CUDA_ERROR_ILLEGAL_ADDRESS in cuStreamSynchronize(cuda_stream[thread_index]), line 2576
So when the render resolution rises, there is even more hunger for VRAM?
Is there any bypass solution other than rendering with CPU or buying quadro?
(btw. I have 2x 2080Ti).
And I was wondering what does “Transparent Shadows” option do exactly? I know what is it used for, but what does the code do in the background? Do I even need light output node in my glass shader?
You can ask it’s original author here.