According to Tom's Hardware, a homelab builder recently achieved 100.8 tokens per second running Qwen 3.6 35B on three used Nvidia V100 GPUs in a custom setup costing $1,042. The decommissioned GPU configuration outperformed Apple's M5 Max Mac Studio (priced at $5,399 with 40-core GPU and 128GB memory) at longer context windows, delivering 72.7 tokens per second at 128K context compared to the Mac's 61.3 tokens per second. The builder tested the setup using llama.cpp across multiple context sizes (4K to 128K) and quantization levels, with the V100 cluster consuming approximately half the energy of equivalent P100-based systems.
Disclaimer: This content is for informational purposes only and does not constitute investment advice or encourage the buying or selling of crypto assets. It is intended to provide information about industry developments and is not directly related to the services offered by Gate TR. Although information obtained from third-party sources has been carefully reviewed, Gate TR assumes no responsibility for any errors or omissions. Crypto assets involve high risks and may experience significant price fluctuations. Before making any investment decisions, you should assess your financial situation and risk tolerance, conduct your own research, and make decisions independently.