FAQ & Support

It depends on the model you want to run. Memory comes first: the model's weights must fit in GPU memory, with room to spare for its context and the software running it. At 16-bit precision a 7B model needs about 14 GB for its weights and a 13B model about 26 GB; 4-bit quantization cuts that to about a quarter. Among cards with enough memory, compare memory bandwidth next.

It depends on the card. Data-center cards are built for servers: some are passively cooled and rely on the server's fans, and many draw more power than a typical desktop supplies. Before putting one in a workstation, check its power connectors, its cooling and whether it fits the case.

Our team offers custom configurations, from single GPUs to large clusters, and technical consultation on your use case, infrastructure design and technical requirements. Send us a technical inquiry describing what you need.

Every spec on the site shows its source and the date it was checked, and a value we have not sourced is marked Unknown. Benchmark results are measured results only, each with its test setup, source and date. We never show an estimate as a result.

Most LLM tools are built first for NVIDIA's CUDA, with growing support for AMD's ROCm. Older cards can drop out of support in newer drivers and libraries, so check that the tools you plan to use still support the card.

Yes. Send a sales inquiry through the contact page for pricing, availability and custom quotes for GPUs, servers and complete solutions.