Puget Systems print logo
Read this article at https://www.pugetsystems.com/guides/1780
Article Thumbnail

Best Workstation PC for Redshift 3D (Spring 2020)

Written on June 10, 2020 by William George


Here at Puget Systems, specifically in the Labs department, most of the content that we write falls into one of two categories: either long, fairly in-depth articles looking at the performance of various PC components in a given application or recommended system pages, with multiple configuration options aimed at different budgets and lots of supporting data drawn from the aforementioned articles. For some readers all of that is information overload, though, and they just want a quick question answered: what is the best computer for my needs? We’re going to try answering that question more directly in a new series of short articles, like this one.

Today we are looking at Redshift, a GPU-based ray tracing engine from Maxon. As applications go, it is fairly straightforward in its needs: NVIDIA graphics cards, and lots of them. The more cards you can pack into a system, and the more powerful each individual card is, the faster your rendering will be!

Good PC Configuration for Redshift

We'll start off with a good, solid workstation design for Redshift - hopefully without breaking the bank. This is also a more compact computer, so if you don't have a lot of desk space it is a great choice. Within the limits of a smaller case we can fit up to two video cards, and with how well Redshift scales as you go to higher-end GPUs we are going for the top-end of the GeForce line: a pair of RTX 2080 Ti 11GB cards. If you needed to save some money, these could be scaled back... but lower models will have less VRAM, which can impact performance in complex scenes. However, Redshift is able to use "out of core" memory - so not all of the data related to what you are rendering has to fit in the GPUs' memory. For that reason, this configuration features a bit more system memory that is strictly necessary. As for the CPU, it just needs to be fast - with the number of cores not being a big factor for GPU rendering.

CPU Intel Core i7 9700K 8-core
Video Cards 2 x NVIDIA GeForce RTX 2080 Ti
Drives 1TB NVMe & 2TB SATA SSDs

Better PC Configuration for Redshift

Our most popular workstations for Redshift bump the number of video cards up from two to four. That is the most dual-slot video cards that can fit in a tower chassis, and effectively doubles performance of the system compared to the "good" configuration above by using the same model of video cards. That goal of having four GPUs drives the choice of a larger motherboard, chassis, and power supply. With so many expensive components in the system, bumping the hardware warranty up to three years better protects your investment.

CPU Intel Xeon W-2235 6-core
Video Cards 4 x NVIDIA GeForce RTX 2080 Ti
Drives 1TB NVMe & 2TB SATA SSDs

Best PC Configuration for Redshift

As of the date of this publication, the most potent single video card from NVIDIA is the Quadro RTX 8000 with a massive 48GB of onboard VRAM. If you are working with such enormous projects that these video cards are worthwhile, though, we should probably add in more RAM for any out-of-core needs... but the CPU doesn't need to change. For storing the larger scenes and texture data that would require this kind of hardware there needs to be an increase the size of the SSDs in the system too.

CPU Intel Xeon W-2235 6-core
Video Cards 4 x NVIDIA Quadro RTX 8000
Drives 1TB NVMe & 4TB SATA SSDs

That is the most powerful traditional, tower workstation that can be built for Redshift right now - but there are other ways to move beyond even this powerhouse of a PC. For example, if this were going to be a server-style system that was accessed remotely, you could go with a rackmount chassis that would offer more PCI-Express slots for video cards; we have options along those lines with support for up to eight GPUs (the maximum that Redshift supports per session). Or you could use PCI-Express expander cards to external enclosures in order to connect more video cards into the computer, though doing so reduces the bandwidth to each card (which will reduce performance). Redshift even supports network rendering, so you could split out processing across several systems in an office.

Additional Resources

If you want to know more about Redshift performance, we have published several articles over the years looking at how individual CPUs compare as well as how well this software scales across multiple cards. Or if you want to configure a computer more tailored to your specific needs, including the option of a rackmount-style server mentioned above, we have a few recommended systems tailored to running Redshift that might interest you. And if you aren’t sure what you need, or if your workflow includes multiple applications, please feel free to call or email our consultants to get a more personalized configuration.

Looking for a Redshift Workstation?

Puget Systems offers a range of poweful and reliable systems that are tailor-made for your unique workflow.

Configure a System!

Labs Consultation Service

Our Labs team is available to provide in-depth hardware recommendations based on your workflow.

Find Out More!
Tags: Render, Rendering, Workstation, PC Workstation, PC, GPU Acceleration, Redshift
Erkan Özgür Yılmaz

Nice article, but something caught my eyes. Normally with Redshift, it is recommended that the system RAM to be bigger than twice the total VRAM in your system. So if you put 4x 2080 Ti's on the system the minimum system RAM should be 11 x 4 x 2 = 88 GB or 96 GB which is the closest physically possible config. Yor first config conforms to that, but your second and third configs needs to be updated accordingly. I'll add the source for that info from Redshift developers later on.

Posted on 2020-06-19 08:15:33

Yeah, I've seen that recommendation as well - though I don't recall where - but I have yet to see anywhere near that much physical system memory being used during actual rendering. Maybe I should reach out to the developers and see if they have any insight on why that recommendation exists, I'd love to better understand :-)

Posted on 2020-06-19 17:38:57

I wouldn't be surprised if it was a "rule of thumb" years ago that is no longer relevant, but managed to stick around. Just like how some applications recommend "4GB of RAM per CPU core". It doesn't actually mean that the app will use 4GB per core, but back when you at most had 4 CPU cores, that was a quick and easy way to give people a RAM recommendation based on their system specs.

If anything, I would imagine that the relationship between VRAM and system RAM should be linear. IE, if with 4GB of VRAM you should have 8GB of system RAM, then with 8GB of VRAM you should need 12GB of system RAM. Both is just an flat increase of 4GB, rather than multiplicative.

Posted on 2020-06-19 18:48:56
Erkan Özgür Yılmaz

I remember that I had lockups and crashes of the render process over and over again with my 4x 1080 Ti and 64 GB of RAM which are solved by increasing the RAM to 128 GB. But I don't know if this proves anything, it could simply be a coincidence or a special case. Oh and anytime I do comp with Fusion alongside rendering something in behind, whenever the comp cache reaches a certain amount of RAM, like 20 GB or so, my renders start crashing. This happens even if I disable OpenCL or set it to use CPU in Fusion. So I set the maximum cache size to 4 GB to be able to do comp while rendering with Redshift.

Posted on 2020-06-25 20:17:41

I went ahead and changed the RAM, just to be on the safe side, but if you can send me the source you have for that info (I know I've read it before too, I just can't recall where) then I'd like to do some more digging and see if it is still an accurate need that Redshift has or if it is outdated. Thanks! :)

Posted on 2020-06-22 18:21:52
Erkan Özgür Yılmaz

Hey William, I found the source:


But it is not exactly saying what I recall it was saying. Here:


We recommend having at least twice as much memory as the largest GPUs installed on the system. I.e. if the system is using one or more TitanX 12GB, the system should have at least 24GB of RAM.

If you’ll be rendering multiple frames at once (as explained in the previous section), the memory should be multiplied accordingly. I.e. if rendering 1 frame needs 16GB, rendering two frames simultaneously will need roughly 32GB.

To recap:
If you’ll be installing multiple GPUs per computer, add plenty of CPU RAM".

Posted on 2020-06-22 22:42:33

Good find! That makes a lot more sense to me. Based on this, I am going to move back to the previous RAM amounts - which will all suffice for rendering up to two frames at a time, per those guidelines. I could see wanting to do that to get a little more efficiency out of the systems than just having all four GPUs working on a single frame, in some situations.

Posted on 2020-06-24 20:58:11