Developer setup for AI platform desktop application
Developer setup for AI platform desktop application
Hi Linus, during the COVID-19 pandemic I’ve made it a point to revisit my interests in technology. I recently started working with IBM’s Watson AI platform to enhance my skills and also explored NVIDIA’s AI solutions. While considering what to build—a custom AI desktop—I wondered if seeking expert advice would help. My focus is on CPU options for deep learning AI platforms. I’m leaning toward AMD Ryzen because of its cost-effectiveness and strong gaming performance, which seems more aligned with AI applications compared to other types of processors. Please share any insights or old videos that could guide my project better. Looking forward! -Skywizard
Depends I'm not actually familiar with this work load. So I don't know if ryzen or intel will perform better. If it's more a multi threaded task them ryzen will most likely pull ahead. What I can recommend though is Waiting. If you absolutely don't need a rig now then wait for zen 3 and check what it can offer.
The goal here is unclear. For home machine learning, GPUs and ASICs are essential; CPUs are less important. You can run models on any PC, but for complex tasks it's often better to use cloud services from providers like Amazon, Microsoft, or Google. You pay based on usage and can access high-end Tesla GPUs that cost more than a car. For basic needs such as OpenCV or deepfakes, standard Nvidia GPUs suffice. The stronger the card, the quicker results. Nvidia also offers Tegra development kits for embedded applications.
Model training isn't heavily dependent on CPU speed. A basic Celeron with 32GB RAM and any NVIDIA GPU can get you going fast. If you're uncertain about investing in that level of hardware, exploring platforms like Google Colab or Kaggle notebooks is a good option—they're free and offer decent GPU power (P100 or T4).