Guides: short answers about renting GPUs

• How does per-minute GPU billing work on CloudGPU?How per-minute GPU billing works on CloudGPU: rounded up to the minute, 1-minute minimum, one hour pre-authorised, unused minutes back to your balance.
• How long does a rented GPU take to start?A CloudGPU instance takes about 2 minutes from Deploy to reachable. What adds time, such as downloading a large model, and how per-minute billing applies.
• How much VRAM do I need for FLUX and Stable Diffusion XL?VRAM for FLUX.1 and SDXL image generation: weight sizes at 16-bit and FP8 from the publishers, what adds memory in a workflow, and where to read more.
• How much VRAM do I need to run an LLM?VRAM needed to run an open LLM: weight sizes at FP16 and Q4_K_M for 8B to 120B models, how context length adds memory, and the formula to work it out.
• How do I choose a rental GPU for ComfyUI?How to choose a rental GPU for ComfyUI: what to check for SDXL and FLUX workflows, VRAM need by precision, startup time, and saving outputs before you finish.
• How do I choose a rental GPU for Ollama?How to choose a rental GPU for Ollama: match VRAM to the quantised model weights plus context, check startup time, and pull models after the instance starts.
• How can I pay for GPU rental without a US card?CloudGPU takes card, Alipay, Google Pay, Apple Pay and USDT on TRC-20 for one prepaid balance. Sign up with email or Google; no phone number or ID needed.