Old Server GPU Beats RTX2070?

Old Server GPU Beats RTX2070?

AD

APIMart

APIMart

AI API, 99.9% SLA. Your AI, Always On.

AI API, 99.9% SLA. Your AI, Always On.

Try Now

Table of Contents:

1. Introduction
2. Specifications of the NVIDIA Tesla K80
3. Why the Tesla K80 uses Kepler architecture instead of Maxwell
4. Use cases of the Tesla K80
   • Computing and data science
   • 3D rendering and photogrammetry
5. Blender performance comparison with modern GPUs
6. Benchmarking with common Blender scenes
   • BMW scene
   • Classroom scene
   • Barbershop scene
7. Price-to-performance comparison
8. Considerations for using the Tesla K80
   • Cooling requirements
   • Need for a secondary graphics card
   • Compatibility with the motherboard
   • Power supply considerations
9. Conclusion
10. Future testing possibilities in different workloads

Article:

The NVIDIA Tesla K80: Unleashing the Power of Enterprise-Level GPU Computing

Introduction

The NVIDIA Tesla K80 is far from just another ordinary graphics processing unit (GPU). Released in November 2014, this enterprise-level graphics card boasts two GK210 GPUs that are essentially enhanced versions of the GTX 780 Ti GPUs, built upon the Kepler architecture. In a market populated with the latest and greatest GPUs, one might wonder why NVIDIA decided to opt for Kepler instead of the more recent Maxwell architecture. In this article, we will Delve into the specifications, use cases, and performance of the NVIDIA Tesla K80 in comparison to modern GPUs, particularly in the Context of Blender rendering.

Specifications of the NVIDIA Tesla K80

The Tesla K80 may belong to an older architecture, but its specifications are nothing to scoff at. Equipped with 2496 CUDA cores and 12 GB of GDDR5 ECC VRAM per GPU, this powerful card packs a total of nearly 5000 CUDA cores and 24 GB of ECC VRAM. However, despite its impressive specs, the Tesla K80 is not well-suited for gaming due to its lack of video outputs. This is because it was specifically designed for use in data centers, where outputting video is unnecessary. Furthermore, its absence of fans adds to its suitability for data centers as it eliminates any potential issues associated with fan noise or failure.

Why the Tesla K80 uses Kepler architecture instead of Maxwell

At the time of the Tesla K80's release, the Maxwell architecture-Based 900 series GPUs had also just hit the market. So why did NVIDIA opt for Kepler instead of the newer architecture? The answer lies in the significant problems that plagued the Maxwell architecture, making it less ideal for their use case. Rather than trying to overcome these challenges, NVIDIA chose to take AdVantage of the already-established and stable Kepler architecture, making it easier to upgrade and optimize. Although Kepler may be an older architecture, the Tesla K80's specifications compensate for any perceived shortcomings.

Use cases of the Tesla K80

While not ideal for gaming, the NVIDIA Tesla K80 thrives in computing-intensive applications that benefit from robust single or double precision performance and ample VRAM. One such application is 3D rendering, where the Tesla K80's impressive CUDA Core count and VRAM capacity allow for faster rendering times, especially when using Blender's tile-based rendering system. Other use cases include photogrammetry and data science, as these fields require substantial computational power for complex calculations and simulations.

Blender performance comparison with modern GPUs

To truly gauge the value of the Tesla K80 as a budget-friendly Blender workstation, it is important to compare its performance against that of modern GPUs. Using a variety of common Blender benchmark scenes, we conducted a comparison against GPUs such as the RTX 2060, AMD 6700 XT, and a 12-core 3900X CPU with slight overclocking. The results were both surprising and impressive.

Benchmarking with common Blender scenes

In the BMW scene, which is relatively quick to render on most GPUs, the Tesla K80 completed the rendering in 1 minute and 13 seconds, on par with an RTX 2070. In terms of CUDA performance, the RTX 2060 outshined the K80 with a render time of only 39 seconds. However, when considering the Tesla K80's significantly lower price in comparison to the RTX 2070, its performance becomes incredibly compelling.

Moving on to more complex benchmark scenes like the classroom and barbershop, the Tesla K80 continued to showcase its power. In the classroom scene, it completed the render in 4 minutes and 20 seconds, comfortably positioned between the RTX 2070 and the 2070 Super. Surprisingly, the relatively wide gap between the 2070 and the 2070 Super became evident, with the latter offering almost half the render times in many scenarios.

Price-to-performance comparison

In the Current market plagued by GPU shortages and inflated prices, the NVIDIA Tesla K80 stands out as a remarkable option. It significantly outperforms the RTX 2070, which currently sells for around 6 to 700 euros on eBay, while the Tesla K80 can be found at approximately 250 euros. Although the RTX 3060 would be the better investment at its manufacturer's suggested retail price (MSRP), current market conditions make it considerably more expensive than the Tesla K80. For those looking to build a budget-friendly Blender workstation without the wait, the Tesla K80 offers unparalleled price-to-performance value.

Considerations for using the Tesla K80

Before rushing to purchase the Tesla K80, there are a few important considerations to keep in mind. First and foremost is cooling. As the card was designed to be used in data centers with constant airflow, ensuring adequate cooling in a personal workstation setup is essential. Custom cooling solutions using radial fans, such as 92mm low-profile Noctua fans, can help maintain optimal temperatures during long rendering Sessions.

Additionally, due to the absence of video outputs on the Tesla K80, a secondary graphics card is required for monitor output. This can be an integrated GPU on the CPU or another separate graphics card, preferably from NVIDIA's GTX or RTX series. However, compatibility between the Tesla K80 and a secondary graphics card can be challenging, as Microsoft's game-ready drivers may conflict with the Tesla card's presence. Workarounds with modified drivers exist but require further testing.

Compatibility with the motherboard is another crucial factor. The Tesla K80 requires a motherboard that supports above 4G decoding in the BIOS. Furthermore, a PCIe 3.0 slot with an 8x speed is the minimum requirement. Some motherboards offer PCIe 4.0 or 3.0 16x speeds for the primary GPU connected to the CPU, while the secondary slot often operates at 4x speeds connected to the chipset. For optimal performance, using a PCIe 4.0 8x slot is recommended.

Lastly, the power supply must be capable of handling the Tesla K80's 300-watt TDP (thermal design power). Additionally, an 8-Pin CPU connector is required for power delivery. It is crucial to ensure the power supply can handle the load and has the necessary connectors.

Conclusion

In conclusion, the NVIDIA Tesla K80 proves to be an excellent investment for Blender rendering, particularly in the current market landscape. With its unmatched price-to-performance ratio and impressive specifications, the Tesla K80 offers an affordable option for those looking to build a budget Blender workstation. However, it is essential to consider cooling requirements, secondary graphics card compatibility, motherboard support, and power supply capacity before investing in this enterprise-level GPU. Despite its age, the Tesla K80 continues to excel in the face of modern GPUs, delivering exceptional performance in demanding computational workloads.

Future testing possibilities in different workloads

While this article primarily focused on the Tesla K80's performance in Blender rendering, its potential in other workloads remains intriguing. Machine learning, data science, and photogrammetry are among the areas that could greatly benefit from the Tesla K80's computational power. Future testing in these domains could provide valuable insights into the card's performance and expand its range of applications. If You have any suggestions for benchmarking or thoughts on this article, please share them in the comments below.

Highlights:

• The NVIDIA Tesla K80 is an enterprise-level graphics card capable of delivering exceptional computational performance.
• Despite its use of the older Kepler architecture, the Tesla K80's specifications make it a compelling choice for budget-friendly Blender workstations.
• In comparison to modern GPUs, the Tesla K80 offers unparalleled price-to-performance value, especially in the current GPU shortage market.
• Cooling requirements, secondary graphics card compatibility, motherboard support, and power supply capacity are key considerations when using the Tesla K80.
• The Tesla K80's potential in other workloads, such as machine learning and data science, warrants further exploration and benchmarking.

FAQ:

Q: Can the Tesla K80 be used for gaming? A: No, the Tesla K80 lacks video outputs and is primarily designed for computing tasks in data centers. It is not suitable for gaming purposes.

Q: How does the Tesla K80 compare to modern NVIDIA GPUs in terms of performance? A: While the Tesla K80 may be an older card, it still offers impressive performance. It can compete with mid-range modern GPUs and outperforms some in terms of price-to-performance ratio.

Q: Is the Tesla K80 compatible with all motherboards? A: The Tesla K80 requires a motherboard that supports above 4G decoding in the BIOS and a PCIe 3.0 slot with an 8x speed. Some motherboards may have limitations, so compatibility should be checked before purchasing.

Q: Does the Tesla K80 require additional cooling solutions? A: Yes, the Tesla K80 requires adequate cooling due to its high thermal design power (TDP). Custom cooling solutions using radial fans or other methods are recommended to ensure optimal temperatures during prolonged use.

Q: Can the Tesla K80 be used alongside another graphics card for multi-GPU rendering? A: Yes, the Tesla K80 can be used alongside another graphics card for multi-GPU rendering. However, compatibility and driver issues may arise, so thorough testing and research are recommended before implementing such a setup.