Frequently Asked Questions
TensorRigs is an independent guide to hardware for AI and deep learning. It publishes GPU and component buying guides, sample workstation builds, setup documentation, and a cloud GPU comparison. It is a content site, not a software product or a hardware shop.
TensorRigs covers NVIDIA RTX 50 series cards, used RTX 30 and 40 series cards, professional RTX workstation cards, and data center GPUs like the H100, along with CPUs, memory, storage, power supplies, and motherboards. The sample builds range from single-GPU desktops to Threadripper PRO multi-GPU workstations.
TensorRigs does not run its own test lab. Performance figures are compiled from published benchmarks and manufacturer specifications, and are labeled as estimates where no measurement exists. Prices are US street prices checked against retailers and price trackers, and guides state the month they were last checked.
The documentation section covers DALI pipelines, data loading, GPU memory, batch size, learning rate, and multi-GPU training, with scripts for monitoring and backups. Start with the data loading and GPU memory pages if training is slower than expected.
Scores in the guides are relative, usually with the RTX 5090 set to 100, so they show how cards compare with each other and not an absolute speed. Real results change with drivers, framework versions, and batch sizes, so test your own workload on a rented GPU before a large purchase.
Yes. The setup documentation covers NVIDIA drivers, CUDA, PyTorch, and TensorFlow, and the blog compares PyTorch, TensorFlow, and JAX. The hardware recommendations apply to all three frameworks.
Yes. There are guides on multi-GPU training, multi-GPU Ollama setups, and motherboard and PCIe lane requirements, plus documentation for working on HPC clusters with Slurm.
Start with the GPU buying guide to decide how much VRAM you need, then look at the workstation build guide or the systems page. If you are not ready to buy, the cloud GPU page compares rental providers. Once you have hardware, follow the setup documentation for drivers and frameworks.