SpaceTiD Calculating

EcoBox Computing Power Forecast

Enter the baseline forecast inputs

Project Name
Application
Industry
Concurrent Users users
Model Size
Request Length
Project A
Finance / Dialog Assistant
This estimate uses a 14B model with medium-length requests.
0 cards
Recommended GPU Count
By compute: 0 cards; by memory: 0 cards
0
Peak Compute Demand
Estimated from token throughput and active model parameters.
0
Total VRAM Demand
Includes parameter residency, activation memory, and KV cache.

Token Generation Rate token/s
Model Precision
GPU Utilization
Redundancy Factor
GPU TFLOPS
Text Ratio
Image Ratio
Image FLOPs Multiplier

Trend Analysis

Concurrent User Growth Trend

Required GPUs and VRAM for Different Model Sizes

Request Length vs Compute Demand

Saved Forecasts

Logo
Spacetid Digital Technology Co., Ltd
Contact Us
Tel: 021-80411593
Mail: official@spacetid.com
Copyright ©2023 SpaceTiD
Logo
Scan the QR Code
Follow Us on WeChat