GPU Database for Large Language Models (LLMs) 💻🖥️

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).

Table of Contents

• About
• Data Format
• GPU Data
• Contributing
• License

About

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.

Data Format

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.

GPU Data

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

Contributing

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.

License

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:

• reddit