GPU Rendering - is it as great as
people say? Let’s test it today and answer this question! For this,
I will be using Chaos V-Ray GPU. Rendering is a complex task and it is one
of the biggest challenges to computers to calculate this. In general, the machine
can handle this process in two ways. A computer can process renders via CPU or
GPU. V-Ray GPU also gives the possibility to use a hybrid-rendering mode that
utilizes both - CPU and GPU devices taking benefits from all available
hardware on a specific workstation. As GPU hardware is developing quickly,
it increases performance at a lower cost. This is the main reason why
it makes GPU Rendering so popular. So let me present my results on two
machines using different rendering methods. Firstly, I will show you the main
specifics of both machines so you have a clear understanding of what is going on. The first machine is my PC with Intel Core
i9-7960X and Nvidia GEFORCE GTX 1070 Ti - FE. I have 2 times 32 GB RAM on
the computer, so in total 64GB. The second machine is a laptop
MSI Creator 15. Here, big thanks to NVIDIA Studio for being a partner of
this video and sending me this laptop. The laptop has Intel Core i7-10875H and
NVIDIA GeForce RTX 3070. With a 2 times increase in rendering performance as well
as a 2 times increase in graphics memory compared to the Series 20, working with very
large 3D models is faster than ever before. Besides, it has 32GB RAM with the
possibility to increase up to 64 GB. What I really like about it is
that you can use fingerprint or face recognition to unlock the laptop. Another cool option is a touch screen that
makes the user experience much better! Plus, I really like the big TouchPad that makes
working without a mouse way more comfortable. With that being said, let me show
you what rendering options we have. So with V-Ray, we can use
both CPU and GPU Rendering. After installation, we only
need to change the renderer. Anyway, if you want to render with CPU, you need
to choose V-Ray, in my case version 6 Hotfix 3. While if we want to use GPU Rendering, we
will choose V-Ray GPU, for me also version 6. Now, for GPU Rendering, V-Ray
offers a couple of options. We can render with CUDA engine or RTX, plus we
can select a GPU device for denoising as well. Let’s start with the default engine - CUDA. It offers hybrid rendering that
uses both CPU and GPU devices. You can choose only GPU, in the case of
my machine it is NVIDIA GeForce RTX 3070. Or only CPU. We can also choose RTX as an engine. And here,
we can use only a GPU device for rendering. It is worth mentioning that if you have more than
one GPU available, it will be listed here as well. Similarly, for the CUDA engine. Plus, as mentioned before, we can additionally
choose GPU for denoising if we wish. Now, I will show you my results on both machines. Here are my input parameters. I used the Living Room scene
from our Visualization Training for testing. Link to the course in the corner. I rendered the images with a
size of 4000 by 2000 pixels. Also, I set the noise threshold equal to 0.005. For the CPU Rendering, I rendered with
Bucket as well as Progressive Image Sampler, I chose the Brute Force as the primary
engine to be able to use the lightmix. For the GPU Rendering, I chose the
default settings and I rendered with Bucket and Progressive Image Sampler
as well to test all the options. Here, you can see my render time results. First of all, when we look at all the options
Bucket Image Sampler is a winner, however, this is not a super large difference, in this case, most
of the time, it is just a couple of minutes saved. But here remember: sometimes, a few minutes make a
difference, for example, if you render animation. Anyway, I will leave only
the times with the Bucket Image sampler to make the comparison more obvious. So you can see that when I
use the CPU Rendering method, the results are better on my computer which
makes sense as I have a stronger CPU there. But still, less than 2 hours for an image
of this quality is a really good result! Ok, but let’s go further… when I use the
CUDA engine, utilizing only a graphic card, I get almost 2 times better results
on the laptop. It makes perfect sense, as my graphic card on PC is not
as good for today’s standards. But when I used a hybrid rendering, the laptop
is still a winner, but the difference is pretty small, less than a minute! It is because my
PC makes up for speed with a good processor. With the last option with the
CUDA engine, using only the CPU, the PC wins this battle unquestionably.
PC was more than 2 times faster than the laptop. The reason for this is obviously
the stronger processor in my computer. And finally, the results from another engine
from GPU Rendering - RTX. Here, a clear winner is a laptop. It rendered the image only for less
than 19 minutes! This is a great result, isn’t it? Here you can see that the NVIDIA
GeForce RTX 3070 makes a real job! To sum up, on my PC, the fastest method of
GPU Rendering is almost two times faster than CPU Rendering. While on the laptop
this difference is more than 5 times! Comparing only the rendering time, it
is obvious to choose GPU Rendering on both machines, however, it makes
much more sense on the laptop. Now, render times are definitely not
the only things to consider. To make an honest comparison, we need
to compare the outputs as well. So I saved everything in the history panel.
First of all, you can see that there is no difference in the output when rendering
by different engines using the V-Ray GPU. It doesn’t matter if I choose CUDA
or RTX, the image looks the same. However, it is not the case, if we compare the image rendered by CPU
Rendering and GPU Rendering methods. On the LHS I have the visualization from the
CPU Rendering, and on the RHS - from GPU. You can notice that reflections
look slightly different, Materials look different, we can see
quite a big difference in the fabrics. You can see it, especially on the sofa. As I read on the Chaos website, because
V-Ray GPU is a very different engine, the bump map itself can look a
bit different compared to V-Ray. And it seems like a case here. Also, the light works differently, the
light spreads differently in the room. I think the biggest difference
is in the environmental lighting. Let me show you the light pass for this. Yeap, here we can see a clear difference. It looks like the contrast between
light and shadow is more prominent in CPU Rendering which makes
the details more visible. I will show you also the reflection path so
you can see the differences more clearly. You can notice dissimilarities in small elements like the decor on the
table as mentioned previously. Or in the standing lamp. But also on the wall material for example. Now, a couple of important notes! The image was originally created in V-Ray,
and to make a fair comparison, I would need to create this visualization from scratch in V-Ray
GPU and adjust materials, lighting, etc. in this. On the other hand, it will be hard to compare
differences if we have various parameters. Another thing to be aware of is that not
all the features are supported by V-Ray GPU. You can find all the information
on the Chaos website. So everything that has this
green icon works in V-Ray GPU, while the red icon presents
non-supported features. Some options have additional notes
that indicate some limitations. huge difference, especially, in exterior
images where there are a lot of plants. Anyway, most of the features are supported and
with each upgrade, there are more possibilities so I’m sure the Chaos didn’t say their last word and
there will be tons of improvements in the future. To sum up, GPU Rendering has
definitely enormous potential. The render times are very good, and I can
imagine that with the use of the NVIDIA GeForce RTX 4090 which is the best in the market
right now, these times will be ridiculously fast! So let’s answer the question: when it
pays off to use GPU Rendering in V-Ray? So first of all, you need to
have a graphics card from NVIDIA. But not whatever graphic card in the market, you need a good one. Great parameters of this
are a must if you want to take real advantage. Another thing that speaks for the
GPU option is the fact that you can utilize multiple GPU devices which increases
significantly its potential to render faster. And of course, taking into consideration the
economic aspect, GPU Rendering is a choice to go. Also, for the quality I’m looking for in
images, GPU Rendering will work fine for interiors but I’m not sure about exteriors
because of the lack of the translucency mode. So if you want to make sure
GPU Rendering is for you, I think testing on your machine
will be the best option if possible. As you can see how much time you can save and if it is worth compromising it with the lack
of some features that are not supported. As GPU Rendering is getting more and more
popular and it develops really quickly, for my next computer, I will definitely
invest in a very strong GPU or maybe even two!
