4 key features I miss after switching from Nvidia to AMD GPUs

AMD and Nvidia make some excellent graphics cards for gaming, AI development, and more. Whether you plan to have a conversation with an LLM to aid with coding or enjoy some immersive virtual experiences with the latest PC games, you'll need one of the best GPUs from either brand to leverage all the latest and greatest technology advancements.

I recently made the switch from Nvidia to AMD on my primary PC, and although things have been nothing short of amazing (especially on Linux), there are a few things I do miss from Nvidia's GPUs.

Higher resale value

Slower depreciation

When upgrading the GPU (and some other parts of your PC), an option is to fund the purchase of the replacement component by offloading the old one. For graphics cards, this can provide much of the deposit for the upgrade and is certainly worth considering. The only drawback for AMD GPUs is the slightly higher depreciation of the cards themselves.

Whether it's brand recognition, well-established resale networks, or some of the points raised right here, Nvidia GPUs tend to hold their value better than AMD Counterparts, offering better resale value for owners. That's not to say you won't recoup some of the cost of an AMD GPU, but you may find it easier to sell your Nvidia GPU at a higher price, and it's something I've noticed when browsing classified listings.

Nvidia ecosystem

Compelling creative software

Nvidia has developed a powerful suite of tools for creators and gamers alike. There's Nvidia Broadcast, Reflex, and ShadowPlay that can elevate your experiences to new levels. AMD has bolstered its Adrenalin software suite with many features found in Nvidia apps, but there are still some ways to go to fully realize the potential for targeted audiences on AMD-powered systems.

Nvidia makes use of artificial intelligence (AI) to power many of these features and has years of experience to date to almost perfect its execution. Then there are some cool functions, such as Nvidia's NVLink for deploying multi-GPU configurations, which is incredibly important for working with AI and LLMs, something Nvidia is keen to continue capitalizing on, thanks to its CUDA technology.

Although driver support on the AMD Front has improved immensely, Nvidia is known for reliable drivers, although its support for Linux leaves much to be desired.

Ray tracing (and DLSS)

RT Core magic

Nvidia was first to bring ray tracing to mainstream gaming with its RTX 30 series, hence the name switch from GTX. The company's RT Cores are now on their 4th generation and have been refined to the point that they offer some incredible performance when used in conjunction with DLSS and frame generation. Compared to earlier RT versions, Nvidia's GPUs allow for some utterly stunning gaming experiences.

Switching to AMD would require disabling most of the advanced graphical features to ensure ample performance is available, as AMD's FidelityFX Super Resolution (FSR) and its ray tracing cores weren't as refined and didn't leverage the power of machine learning. FSR 4 is about as good as Nvidia's DLSS and can even be considered better in some cases, so the company has definitely made some ground, but again, Nvidia has a more mature platform.

CUDA

The almighty Nvidia cores

Compute Unified Device Architecture (CUDA) is Nvidia's proprietary parallel computing platform. Unlike normal cores that handle all the graphical work inside games, these cores are better suited for artificial intelligence work, machine learning (hello, DLSS!), and scientific computing. Whenever you've seen Nvidia mentioned in connection with something, CUDA is often involved, which is why Nvidia is the brand of choice for AI development.

Radeon Open Compute (ROCm) is available on AMD's hardware, but it's not quite as widely supported or as mature as Nvidia's solution. It is getting there, however, as my esteemed colleague Adam Conway realized through getting his RX 9700 XTX running through Proxmox for an Open Web UI LXC and AI inference. So, you do technically lose something with CUDA, but ROCm can be a viable replacement if supported

AMD has come a long way

Although AMD graphics cards used to be viewed as the cheap alternatives to Nvidia counterparts, often with weaker specs and software support, AMD has poured resources into getting its GPUs up to the same level. AMD's Adrenalin software is vastly more refined than it used to be. The actual hardware is also excellent for gaming, even at higher resolutions with the newer RX 9070 series. There's plenty of appreciation with Radeon, but there are still some things I miss with Nvidia.