Welcome to the GPU Database repository for LLMs! This repository is a comprehensive collection of information about GPUs, designed to provide enthusiasts, professionals, and curious minds with a centralized hub of knowledge, specifically for those interested in Large Language Models (LLMs).
The GPU Database for LLMs is a collection of information about different GPUs, including their brand, model, RAM size, RAM speed, maximum TDP, and cost (both new and used). This information can be useful for a variety of purposes, such as:
- Comparing different GPUs to determine which one is the best fit for your LLM needs.
- Determining the cost of upgrading your GPU for LLM tasks.
- Researching the power consumption of different GPUs for LLM tasks.
- Identifying trends in GPU development and pricing for LLM tasks.
The data in the GPU Database for LLMs is organized in a table format, with each row representing a different GPU. The table includes the following columns:
- Brand: The brand of the GPU (e.g. NVIDIA, AMD, etc.).
- Model: The model of the GPU (e.g. Radeon RX 7900 XTX, etc.).
- RAM size (GB): The amount of video memory (VRAM) on the GPU, measured in gigabytes.
- RAM speed (GB/s): The speed of the VRAM, measured in gigabytes per second.
- Max TDP (W): The maximum thermal design power (TDP) of the GPU, measured in watts. This is the maximum amount of power that the GPU is designed to consume under heavy load.
- Cost new (USD): The cost of the GPU when purchased new, in US dollars.
- Cost used (USD): The estimated cost of the GPU when purchased used, in US dollars.
Brand | Model | RAM size (GB) | RAM speed (GB/s) | Max TDP (W) | Cost new (USD) | Cost used (USD) |
---|---|---|---|---|---|---|
AMD | RX 7900 XTX | 24 | 960 | 355 | 950 | 800 |
AMD | RX 7900 XT | 20 | 800 | 315 | 750 | 525 |
AMD | RX 7800 XT | 16 | 624 | 263 | 550 | 450 |
AMD | W7900 | 48 | 864 | 295 | 4000 | 3500 |
Apple | M2 Ultra | 192 | 800 | 295 | 5600 | |
Apple | M2 Max | 64 | 400 | 145 | 2400 | |
Apple | M2 Pro | 32 | 200 | 100 | 1700 | 1400 |
Apple | M1 Ultra | 128 | 819 | 215 | 4000 | |
Apple | M1 Max | 64 | 410 | 115 | 1400 | |
Apple | M1 Pro | 32 | 205 | 100 | 1300 | |
Nvidia | H100 NVL1 | 188 | 7800 | 800 | ||
Nvidia | H100 SXM | 80 | 3350 | 700 | ||
Nvidia | H100 PCIe | 80 | 2000 | 350 | ||
Nvidia | A100 SXM | 80 | 2039 | 400 | ||
Nvidia | A100 PCIe | 80 | 1935 | 300 | ||
Nvidia | L40 | 48 | 864 | 300 | ||
Nvidia | A40 | 48 | 696 | 300 | ||
Nvidia | A10 | 24 | 600 | 150 | ||
Nvidia | A16 | 4 x 16 | 4 x 200 | 250 | ||
Nvidia | RTX 6000 Ada | 48 | 960 | 300 | 6000 | |
Nvidia | RTX 5000 Ada | 32 | 576 | 250 | ||
Nvidia | RTX 4500 Ada | 24 | 432 | 210 | ||
Nvidia | RTX 4000 Ada | 20 | 360 | 130 | 1500 | |
Nvidia | RTX A6000 | 48 | 768 | 300 | 3000 | |
Nvidia | RTX A5500 | 24 | 768 | 230 | 2000 | |
Nvidia | RTX A5000 | 24 | 768 | 230 | 1000 | |
Nvidia | RTX A4500 | 20 | 640 | 200 | 700 | |
Nvidia | RTX A4000 | 16 | 448 | 140 | 700 | |
Nvidia | Quadro RTX 8000 | 48 | 672 | 300 | 2000 | |
Nvidia | Quadro RTX 6000 | 24 | 672 | 295 | 1500 | |
Nvidia | Quadro RTX 5000 | 16 | 448 | 265 | 550 | |
Nvidia | Quadro P6000 | 24 | 433 | 250 | 600 | |
Nvidia | Quadro P5000 | 16 | 288 | 180 | 350 | |
Nvidia | Tesla P100 | 16 | 732 | 250 | 150 | |
Nvidia | Tesla P40 | 24 | 694 | 250 | 200 | |
Nvidia | 2 x RTX 4090 | 2 x 24 | 2 x 1008 | 900 | 3400 | |
Nvidia | RTX 4090 | 24 | 1008 | 450 | 1700 | |
Nvidia | RTX 4080 | 16 | 717 | 320 | 1100 | |
Nvidia | 4070 | 12 | 504 | 200 | 600 | |
Nvidia | RTX 4060 Ti | 16 | 288 | 160 | 475 | |
Nvidia | RTX 3090 Ti | 24 | 1008 | 450 | 1500 | 950 |
Nvidia | 4 x RTX 3090 | 4 x 24 | 4 x 936 | 1400 | 6000 | 2800 |
Nvidia | 2 x RTX 3090 | 2 x 24 | 2 x 936 | 700 | 3000 | 1400 |
Nvidia | RTX 3090 | 24 | 936 | 350 | 1500 | 700 |
Nvidia | RTX 3060 | 12 | 360 | 170 | 275 | 225 |
I welcome contributions from the community! If you have information about a GPU that is not currently included in the database, or if you notice any errors or inaccuracies in the existing data, please open a pull request or submit an issue.
The GPU Database for LLMs is released under the MIT License.
I hope that you find the GPU Database for LLMs useful, and I look forward to your contributions! 🎉
Inital source: