We have all heard of Redshift as a powerful fully GPU-accelerated biased renderer. This means the graphics card (GPU) is the driving factor for its rendering performance. There is no such thing as an overpowered GPU for Redshift; and by setting up multiple GPUs in a single system, Redshift does render faster. However, how well does rendering speed scale across multiple GPUs in Redshift, have you wondered? Let us get to the point and deep dive into how Redshift takes advantage of multi-RTX 4090 Configurations.
Maxon has expanded Redshift’s high-performance rendering support via Nvidia (CUDA) to Apple (Metal), and most recently via AMD (HIP) technology (Redshift 3.5.15 brings the public beta for AMD GPU support). Among those, Nvidia (CUDA) is still the best technology offering the best rendering performance in Redshift. Amongst NVIDIA’s graphics cards, the GeForce RTX 4090 is the most powerful and suitable GPU for Redshift rendering. Why is RTX 4090? Let’s dig deeper into how Redshift utilizes GPU to find the reason.
There are two aspects of a GPU that influence rendering capabilities in Redshift. They are video memory (VRAM) and the raw speed of the GPU itself.
Redshift stores scene data, such as geometry and textures, on the VRAM when rendering the scene. Accordingly, if the amount of VRAM is not enough, you will not be able to render complicated scenes as efficiently and quickly. RTX 4090 with a decent amount of VRAM (24GB) can meet almost all Redshift’s projects that contain simple to intermediate to complex and even very complicated scenes.
GeForce GPU tends to have good raw performance and RTX 4090 has the best raw performance amongst NVIDIA Geforce GPUs. RTX 4090 is powered by the ultra-efficient NVIDIA Ada Lovelace architecture. Thanks to its third-generation RT Cores, fourth-generation Tensor Cores, an eighth-generation NVIDIA Dual AV1 Encoder, and 24GB of Micron G6X memory capable of reaching 1TB/s bandwidth, RTX 4090 brings a massive boost in rendering performance for Redshift.
RTX 4090 beats other GeForce 40 Series GPUs (RTX 4080 & RTX 4070Ti) and GeForce 30 Series GPUs (RTX 3090Ti, RTX 3090, RTX 3080Ti, RTX 3080, etc) to be the top-performing GPU for Redshift.
Redshift 3.5.12 Benchmark Results (Source: Puget Systems)
Compared to the other two GPUs in the GeForce RTX 40 Series, which are RTX 4080 and RTX 4070Ti, the RTX 4090 wins regarding both the amount of VRAM and rendering performance. It is 22% and 55% faster, respectively than RTX 4080 and RTX 4070Ti. The results are more impressive when compared to its predecessors, GeForce RTX 30 Series. With the same amount of 24GB VRAM, RTX 4090 gives a faster render speed than RTX 3090Ti and RTX 3090 with 64% and 71%, respectively.
Note: We will not compare GeForce RTX 4090 with the Quadro GPU. While Quadro comes with larger amounts of VRAM, it costs more for the same level of raw performance as Geforce GPU.
All right, it’s enough for a single RTX 4090 in Redshift. Now, let’s get to the right point of this article: How well Redshift takes advantage of multi-RTX 4090 Configurations.
Putting Redshift 3.5.09 (RTX enabled) to the test, we are looking at scaling from one to seven RTX 4090s in a single workstation. Let’s see how well the rendering speed scales across multi-RTX 4090s.
The graph below shows the raw Redshift benchmark render times (in seconds) with 1, 2, 3, 4, 5, 6, and 7 of the GeForce RTX 4090 24GB VRAM:
Redshift 3.5.09 Benchmark Results (Source: Puget Systems)
Or another way to look at it. The following graph shows how adding GPUs increases rendering performance – shown as a percentage (%) compared to the speed of a single card:
Redshift 3.5.09 rendering performance scales across multi-RTX 4090 configurations (Source: Puget Systems)
Going from one card to two, we see an impressive gain in rendering performance, an 86% increase. From two cards to three, and three cards to four, both result in a >75% increase in performance. The result in a 4x RTX 4090 configuration performs about 3.4x faster than a single RTX 4090, still an impressive increase in performance over a single RTX 4090. After that, however, the gains drop off relatively. We only look at about a 48% performance gain from four cards to five and a 42% gain from five cards to six. The gains drop off sharply from six cards to seven. We are only looking at about a 24% performance gain for this seventh additional RTX 4090.
At the very peak, with 7x RTX 4090 GPUs, we notice exactly a 4.56x increase in performance over a single GPU. That is not terrible by any means, but it shows that RTX 4090 performance in Redshift scales as additional cards are added, but not very well. It is not perfect, or linear, scaling – there are some levels of diminishing returns as more cards are added to a single system.
Redshift differs from many other rendering benchmarks in that it times how long it takes to render a single scene rather than returning how many samples per second the system can handle. The problem with this method is that as hardware becomes faster and faster, the test takes less and less time to complete. This can have a direct impact on the measurement (especially because Redshift only reports results in whole seconds rather than fractions), but it can also cause things like scene load time to account for a bigger portion of the result rather than isolating the rendering portion.
As a result, it appears that the current Redshift benchmark is rather misleading when it comes to really powerful configurations like this. The performance scaling was suitable for up to four RTX 4090 GPUs, but the benefit per GPU started diminishing after that. We still ended up with 4.5x the maximum performance of a single RTX 4090, but we feel that real Redshift users rendering huge scenes would notice a larger advantage in most circumstances than what this benchmark is currently capable of proving.
It is a fact that Redshift is increasingly evolving with more and more powerful features. Creative individuals and studios all want to make more creative projects with complex scenes and therefore need super-powerful hardware resources to meet their rendering demand. Sadly, even though you have owned a strong GPU such as RTX 3090 or RTX 4090, there are still frames that take inevitably tens of minutes to hours to render. Not to mention, it is just a single frame in a sequence of frames that make up a creative project. At this time, the power of a powerful GPU is not enough. However, a multi-GPU system is sure to shine in these cases since Redshift actually scales quite impressively with multiple cards.
iRender gonna help you speed up Redshift rendering with the most multi-RTX 4090 machines on the market.
iRender provides high-configuration single and multi-GPU servers, specifically 1/2/4/6/8x RTX 4090 & RTX 3090. Powered by two of the most single-core performance CPUs for Redshift which are AMD Ryzen™ Threadripper™ PRO 3955WX @ 3.9 – 4.2GHz & AMD Ryzen™ Threadripper™ PRO 5975WX @ 3.6 – 4.5GHz, 256GB RAM and 2T NVMe SSD storage, all our servers can handle almost all levels of complexity of Redshift projects.
iRender gives you, all 3D artists, an affordable answer to unleash your creativity with the beast RTX 4090 from just 8.2 USD/hour. We are proud to be the only render farm where you can install any software and plugins of any version that serves your project. You will be given complete control over the servers and use them as your local computers.
Let’s check out our test videos about Redshift rendering performance on our multi-RTX 4090 servers.