The challenges of selling GPU power for Rendering or AI workloads

GPU mining rigs can be repurposed for AI and machine learning workloads, but there are some challenges and considerations to be aware of. The suitability of a GPU mining rig for AI tasks will depend on the specific hardware components, their performance capabilities, and the requirements of the AI workloads. Here are some challenges that a GPU miner might face when adapting their hardware for AI:

  1. Hardware compatibility: While most mining rigs use GPUs, not all GPUs are suitable for machine learning tasks. For AI workloads, you’ll want GPUs that support CUDA (for NVIDIA) or ROCm (for AMD), and have a sufficient amount of VRAM, typically the higher the better, but 16GB of VRAM is the minimum. Some popular GPUs for AI workloads include NVIDIA’s Tesla, Titan, and RTX series or AMD’s Radeon Instinct series.
  2. Hardware performance: The performance of GPUs in a mining rig might not be enough to handle the computational demands of some AI tasks. Deep learning, for example, often requires high-performance GPUs with large memory bandwidths and fast tensor processing capabilities. You’ll need to assess if your GPUs meet these requirements or if you need to upgrade them.
  3. Cooling and power management: AI workloads can generate a significant amount of heat, similar to cryptocurrency mining. Adequate cooling solutions should be in place to prevent hardware damage or thermal throttling. Additionally, AI workloads can consume a lot of power, so you’ll need to ensure your power supply can handle the load. Furthermore, AI requires greater power consistency and guaranteed uptime, similar to a traditional data center.
  4. Software and libraries: AI and machine learning typically rely on specific software and libraries like TensorFlow, PyTorch, and CUDA. You’ll need to familiarize yourself with these tools and ensure that your hardware is compatible with the required software stack.
  5. Scalability: AI workloads can scale across multiple GPUs or even across multiple machines. If you plan on running large-scale AI tasks, a typical mining rig will need to be reconfigured to support multi-GPU setups or invest in additional hardware. Additionally, the CPU will need to be upgraded to a unit with a high core count, along with the power delivery and ideally, the motherboard and amount of RAM for most typical GPU mining rig builds.
  6. Maintenance: Running AI workloads can be taxing on your hardware over time. Be prepared to invest in regular maintenance, updates, and potentially hardware replacements to ensure your system remains functional and efficient.
  7. Skillset and knowledge: To pivot into an AI machine learning business, you’ll need to develop or hire the appropriate skillset and knowledge in AI, machine learning, and deep learning. This includes understanding different algorithms, neural networks, and optimization techniques, as well as keeping up with the latest research and developments in the field.

In summary, this is a use case that has potential – but it comes with a few hurdles. There are viable markets for renting out your rigs for AI such as Vast.ai, among other startups, with a reasonable potential for growth on the horizon. A typical GPU mining rig will require some hardware modifications. If you have powerful Nvidia GPUs with 12GB+ VRAM, that’s a good start. But AI training requires a lot more than raw GPU power. You need a fast CPU with lots of cores, tons of ECC memory, high bandwidth data links, lots of fast SSD storage, and ideally a fiber optic internet connection.

We are beginning to see signs of the potential in entry-level service provision tapping out. As of this writing, there are now waitlists on AI rig rental sites, and even once your rig is approved, as AI rig rental is not permissionless like mining, there are additional requirements that make this avenue more resource & time-consuming than mining. The demand side of the market is increasingly demanding the most powerful GPUs available, with waning interest in using lower-end cards with less VRAM for training. While there’s always the possibility this changes with advances in training and distribution of the workload, it appears to increasingly be a use case that you build specifically for, rather than repurpose existing hardware. If you have a farm full of 24GB RTX 3090s or A5000s you have a much more viable starting point because VRAM is king for training, but this market comes with a higher bar to clear that has turned off most miners we’ve worked with. 

Rendering or AI inference is less demanding on hardware but poses additional challenges with client management.

Pivoting a GPU mining rig to a rendering business can be a viable option, as both mining and rendering rely on GPUs for high-performance parallel computing. However, there are still some challenges and considerations to keep in mind when repurposing your mining rig for rendering tasks:

  1. Hardware compatibility and performance: Ensure that the GPUs in your mining rig are suitable for rendering tasks. Popular GPUs for rendering include NVIDIA’s GeForce RTX and Quadro series or AMD’s Radeon RX and Radeon Pro series. You may need to upgrade your GPUs if they don’t meet the requirements of the rendering software you plan to use.
  2. Cooling and power management: Rendering workloads can generate significant heat and consume a lot of power, similar to mining and AI workloads. Make sure you have adequate cooling solutions in place and that your power supply can handle the load.
  3. Software compatibility: Familiarize yourself with the rendering software and ensure that your hardware is compatible with the required software stack. Popular rendering software includes Blender, Octane Render, Redshift, V-Ray, and Arnold. Some of these applications have specific GPU requirements or work better with certain GPU architectures.
  4. Scalability: Depending on the scale and complexity of the rendering projects you plan to undertake, you may need to expand your setup to include multiple GPUs or machines to handle the workload efficiently. Be prepared to invest in additional hardware if needed.
  5. Networking and storage: Rendering tasks often involve working with large amounts of data, such as high-resolution textures and complex 3D scenes. You’ll need to ensure your rig has sufficient storage and that your network can handle the data transfer requirements.
  6. Skillset and knowledge: To pivot into a rendering business, you’ll need to develop or hire the appropriate skillset and knowledge in computer graphics, 3D modeling, animation, and rendering techniques. This includes understanding different rendering engines, lighting techniques, and optimization methods.
  7. Client management and marketing: As a rendering business, you’ll need to manage client relationships, negotiate contracts, and market your services to attract new clients. This may involve developing a portfolio, setting up a website, and promoting your business through social media or other marketing channels.

By addressing these challenges and considerations, you can successfully pivot a GPU mining rig into a rendering business and make the most of your existing hardware investment.

In our experience working with hundreds of miners, we have found that most GPU miners do not have a competitive edge in addressing the rendering, AI & machine learning markets. The few we know who have successfully pivoted from GPU mining to establishing an AI data center have addressed the above challenges, typically years ahead of the Merge – scaling their operation with the capital and personnel to address the additional demands of client acquisition and management, securing performance hardware & investing in the infrastructure upgrade.

If you’re holding onto your hardware for the next bull market, it might be worth reading our latest update on the GPU market. Because GPUs still have resale value, selling your hardware is a viable option to get liquidity for your next venture. Contact us for a no-obligation offer on your hardware.

Get Offer