Product News

Publish Date : 2024/03/27

NVIDIA RTX 5880 Ada Performance


▲ NVIDIA RTX 5880 Ada Generation

NVIDIA RTX™ 5880 Ada Generation GPU is built on the all-new NVIDIA Ada Lovelace architecture, using a 4nm manufacturing process, and equipped with 48GB of GDDR6 memory. These specifications enable this graphics card to be applied to various professional workloads and creative processes, including but not limited to: 3D scene construction, GPU rendering, smooth real-time interaction, GPU-accelerated computing, high-resolution visualization, and currently popular AI projects.

1. Test Environment

2. List of test software

3. Graphics card specifications

4. Software Test Results and Analysis

4.1 Compute Performance: CUDA-Z

CUDA-Z, similar to CPU-Z and GPU-Z, provides a visual display of some performance information about the graphics card.

▲ Screenshot of RTX 5880 Ada Test Results

▲Test Results Chart

Single-precision performance is an important indicator for measuring the computational performance of graphics cards, with many calculations being performed using single precision. From the test results, the single-precision performance of the RTX 5880 Ada reaches 75T, while the RTX 5000 Ada is 66.8T. In comparison, the single-precision performance of the RTX 5880 Ada has increased by 12%.

In terms of double-precision performance, the performance of the RTX 5880 Ada has also increased by 14% compared to the RTX 5000 Ada. In other precision calculation performance aspects, the performance of the RTX 5880 Ada has also improved.

4.2 Graphics Performance: SPECviewperf 2020 V3

SPECviewperf 2020 V3 is primarily used to evaluate the professional graphics performance of graphics cards. It includes performance tests for common software such as 3ds Max, Maya, Catia, SNX, SolidWorks, Creo, as well as medical and energy simulation performance tests. It scores the performance based on the speed of operation in software scenes, and finally generates relative performance of graphics cards.

▲ Screenshot of RTX 5880 Ada Test Results

▲ Test Results Chart

From the test results of SPECviewperf 2020 V3, it can be seen that the graphics interaction performance of RTX 5880 Ada has improved compared to RTX 5000 Ada. The performance has increased by 10% in SNX, 26% in SolidWorks, and 28% in Maya.

4.3 Real-time Rendering Performance

4.3.1 Blender Cycles

Blender is a free and open-source 3D graphics software that offers a range of solutions for animation short film production, including modeling, animation, materials, rendering, audio processing, and video editing. The Cycles renderer is the GPU renderer supported by Blender. In this Blender real-time rendering test, the Cycles renderer and Optix technology were chosen for real-time rendering, comparing the time needed for different GPU graphics cards to render a frame in real-time.

▲ Screenshot of the RTX 5880 Ada Rendering Scene

▲ Test Results Chart

From the test results, rendering performance of RTX 5880 Ada in Blender has improved by 50% compared to RTX 5000 Ada.

4.3.2 NVIDIA Omniverse USD Composer

Omniverse™ is a collaborative design platform, and NVIDIA Omniverse USD Composer (formerly known as "Create") is a foundational application within the Omniverse platform used for constructing large-scale 3D worlds and composite scenes based on OpenUSD (Universal Scene Description) workflows. Omniverse USD Composer supports collaborative building of 3D scenes online by multiple users simultaneously, integrates with a rich DCC ecosystem, and also provides GPU-accelerated real-time rendering with realistic ray tracing.

▲ Screenshot of the RTX 5880 Ada Rendering Scene

The "Restaurant" scene is a large 3D scene included with the Omniverse USD Composer software, approximately 30GB in size. In this test, real-time rendering will be performed based on the "Restaurant" scene, comparing the frame rates of real-time rendering on different graphics cards.

▲Test Results Chart

From the real-time ray tracing rendering in Omniverse USD Composer, it can be observed that when DLSS frame generation is disabled, the real-time rendering frame rate of RTX 5880 Ada has increased by 33% compared to RTX 5000 Ada. Additionally, with DLSS frame generation enabled, RTX 5880 Ada can achieve a real-time rendering frame rate of 32 frames at 4K resolution, representing a 52% increase compared to RTX 5000 Ada.

4.3.3 Unreal Engine

Unreal Engine is a comprehensive game development platform aimed at next-generation consoles and PCs, providing game developers with a wealth of core technologies, data generation tools, and foundational support they need. In this test, the official animation assets from UE 5.3 were used for real-time rendering, comparing rendering performance based on frame rates.

▲Test Results Chart

From the test results, the real-time rendering performance of RTX 5880 Ada in UE has improved by 13% compared to RTX 5000 Ada.

4.4 Offline rendering performance

Offline rendering is widely used in media entertainment, design for manufacturing, and architectural industries. Currently, many renderers support GPU rendering, and the speed of GPU rendering depends on the calculation speed of the graphics card.

4.4.1 V-Ray Benchmark

V-Ray Benchmark is a free standalone rendering speed testing software used to test the offline rendering speed of computers. It can quickly and easily test two rendering engines:

  • V-Ray GPU CUDA - CUDA rendering mode test
  • V-Ray GPU RTX - RTX rendering mode test

 

It includes 3 specific test scenes to test the speed of different V-Ray rendering engines and record the final scores.

▲ Screenshot of RTX 5880 Ada Test Results

▲ Test Results Chart

In the CUDA rendering test, the performance of RTX 5880 Ada has improved by 33% compared to RTX 5000 Ada. In the RTX rendering test, the performance of RTX 5880 Ada has improved by 48% compared to RTX 5000 Ada.

4.4.2 Octane Benchmark

Octane (OC) is a frequently used GPU renderer, with plug-ins available for many familiar software among users. In this comparison, we conducted tests using OC's Benchmark software and recorded the final scores.

▲ Screenshot of RTX 5880 Ada Test Results

▲ Test Results Chart

In the OC Benchmark test, using the RTX rendering mode, RTX 5880 Ada shows a 41% performance improvement compared to RTX 5000 Ada. Without using the RTX rendering mode test, RTX 5880 Ada shows a 27% performance improvement compared to RTX 5000 Ada.

4.4.3 Blender

Blender's GPU renderer Cycles supports two rendering modes—CUDA and Optix. In this article we will test the offline rendering acceleration of different GPUs for these two rendering modes and compare the time needed for offline rendering.

▲ Screenshot of the RTX 5880 Ada Testing Process

▲Test Results Chart

In the Blender Cycles test, using Optix rendering, the performance of RTX 5880 Ada has improved by 39% compared to RTX 5000 Ada. When using CUDA rendering, RTX 5880 Ada has improved by 29% compared to RTX 5000 Ada.

4.4.4 Keyshot

KeyShot is an interactive ray tracing and global illumination 3D rendering and animation software that supports both GPU and CPU rendering. This test will focus on KeyShot GPU rendering acceleration, comparing the time taken by different graphics cards to render the same image by exporting an 8K image through GPU rendering.

▲ Test Results Chart

From the test results, it can be seen that the offline rendering performance of RTX 5880 Ada in KeyShot has improved by 20% compared to RTX 5000 Ada.

4.5 AI Performance

4.5.1 Stable Diffusion

Stable Diffusion is a deep learning text-to-image model released by the startup company Stability AI. It is primarily used to generate detailed images based on textual descriptions. This article uses Stable Diffusion for AI image generation, testing and comparing the speed of different graphics cards in generating images based on text under the same settings.

▲ Screenshot of RTX 5880 Ada Test Results

▲ Test Results Chart

In the test of Stable Diffusion's text-to-image generation, it can be observed that the AIGC graphic generation capability of RTX 5880 Ada has improved by 22% compared to RTX 5000 Ada.

4.5.2 AI Benchmark

AI Benchmark is a hardware performance benchmarking suite used to evaluate artificial intelligence algorithms. It includes multiple test modules covering various fields such as computer vision, natural language processing, and more.

By running AI Benchmark, users can test the performance of different hardware platforms. This test will run AI Benchmark on different graphics cards to compare the final scores of the benchmark test.

▲ Test Results Chart

From the test results, it can be seen that the performance of RTX 5880 Ada has improved by approximately 20% in inferencing, training, and overall AI scores compared to RTX 5000 Ada

5. Conclusion

▲Official Rendering Image of NVIDIA RTX 5880 Ada Generation

The NVIDIA RTX 5880 Ada architecture GPU is currently Ultra High-end professional graphics card in the NVIDIA RTX GPU lineup. With its 48GB of ultra-large memory, it can support larger Batch_Size, allowing more training data to be loaded into memory at once, resulting in better performance in the field of AI.

From the test data in this article:

。 In terms of AI performance, the performance of the RTX 5880 Ada has improved by about 20% compared to the RTX 5000 Ada.

。 In professional graphics, the RTX 5880 Ada has seen a performance improvement of approximately 10% to 30% compared to the RTX 5000 Ada.

。 In rendering, the improvement with the RTX 5880 Ada is more pronounced. In CUDA rendering mode, there's about a 30% improvement with the RTX 5880 Ada compared to the RTX 5000 Ada; in ray tracing rendering, there's a 40% to 50% improvement with the RTX 5880 Ada compared to the RTX 5000 Ada.

Overall, the NVIDIA RTX 5880 Ada architecture GPU will be the top choice for high-end graphic applications, GPU rendering, and AI applications combined with visualization projects in the foreseeable future.