Terms of Service & Disclaimers

1. Acceptance of Terms

By accessing or using GPU Garden ("we," "our," or "the Site"), you agree to be bound by these Terms of Service ("Terms") and our Privacy Policy. If you do not agree to these Terms, do not use the Site.

2. "As-Is" Mathematical Estimations & No Warranty

THE SITE, CALCULATORS, BENCHMARK ESTIMATES, AND ALL CONTENT ARE PROVIDED ON AN "AS IS" AND "AS AVAILABLE" BASIS WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, ACCURACY, TITLE, OR NON-INFRINGEMENT.

Specifically:

• Memory Footprints: VRAM calculations (weights + KV-cache + activation buffers) represent theoretical minimums. Real-world inference runners (vLLM, Ollama, llama.cpp, TensorRT-LLM, ComfyUI) may experience Out-of-Memory (OOM) faults due to memory fragmentation, CUDA context overhead, or display server allocation.
• Inference Speeds: Quoted tokens-per-second (“tok/s”) are synthetic approximations based on peak memory bandwidth and standard architectural assumptions. Actual performance varies with thermal throttling, PCIe bus bandwidth, prompt length, and batch size.
• Hardware Pricing: Hardware pricing figures and cloud hourly rates change continuously. We do not guarantee price accuracy or product availability.

3. Electrical, Power, and Hardware Safety Disclaimer

Nothing on GPU Garden constitutes electrical, thermal, structural, or professional computer engineering advice. High-power GPUs (including but not limited to RTX 3090, RTX 4090, and multi-GPU configurations) introduce extreme thermal and power demands:

• Power Supply Requirements: High-end GPUs exhibit severe microsecond transient power spikes that may trip low-spec power supplies (PSUs). Always adhere strictly to the PSU wattage and 12V rail specifications published by the graphics card manufacturer.
• 12VHPWR / 12V-2x6 Connectors: High-draw 16-pin power connectors require full seating, adequate bend radius, and zero terminal strain. We accept zero liability for damaged, melted, or burnt hardware components, cables, or power supplies.
• Consult Qualified Technicians: When in doubt, always consult official manufacturer technical documentation and certified system builders before installing or modifying computer hardware.

4. Third-Party Services & Cloud Compute Instances

GPU Garden provides direct links to third-party marketplaces and compute providers, including Amazon, RunPod, and Vast.ai:

• We are an independent informational publication. We are not a seller, reseller, cloud hosting provider, or agent of Amazon, RunPod, Vast.ai, or any hardware manufacturer.
• Any purchase, rental, contract, or transaction you enter into is strictly between you and that third party, subject to their respective terms and conditions.
• Cloud Billing Responsibility: YOU ARE SOLELY RESPONSIBLE FOR MONITORING, LIMITING, PAUSING, AND TERMINATING ANY CLOUD INSTANCES, PODS, OR VIRTUAL MACHINES YOU PROVISION ON THIRD-PARTY PLATFORMS. Cloud compute runs on hourly or second-based billing models. GPU Garden bears no responsibility or liability for third-party billing charges, unstopped instances, or lost data.

5. FTC & Affiliate Disclosure

Mandatory Amazon Associates Disclosure: As an Amazon Associate I earn from qualifying purchases.

GPU Garden participates in affiliate and referral programs with Amazon Services LLC Associates Program, RunPod, and Vast.ai. When you click an affiliate referral link on this site and complete a hardware purchase or fund a cloud compute account, we may receive a small commission or account credit at zero additional cost to you. All referral links are clearly annotated with rel="noopener sponsored" in compliance with Federal Trade Commission (FTC) 16 CFR Part 255 guidelines.

6. Nominative Trademark Fair-Use Notice

All product names, brand names, logos, and trademarks displayed on GPU Garden are the property of their respective owners:

• NVIDIA, GeForce, RTX, and CUDA are registered trademarks of NVIDIA Corporation.
• AMD, Radeon, Ryzen, and ROCm are trademarks of Advanced Micro Devices, Inc.
• Apple, Mac, M-Series, Apple Silicon, and Metal are registered trademarks of Apple Inc.
• Meta and LLaMA are trademarks of Meta Platforms, Inc.
• DeepSeek is a trademark of Hangzhou DeepSeek Artificial Intelligence Co., Ltd.
• Amazon is a trademark of Amazon.com, Inc.
• RunPod and Vast.ai are trademarks of their respective owners.

The use of these trademarks on GPU Garden is strictly nominative and descriptive, used only to the extent necessary to identify specific hardware components, architectures, and open-weight models for accurate comparative analysis. GPU Garden is an independent technical resource and is not affiliated with, endorsed by, sponsored by, or partnered with NVIDIA, AMD, Apple, Meta, or any other manufacturer.

7. Limitation of Liability

TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW, IN NO EVENT SHALL GPU GARDEN, ITS OPERATORS, EMPLOYEES, CONTRACTORS, OR AGENTS BE LIABLE FOR ANY INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, OR PUNITIVE DAMAGES, INCLUDING BUT NOT LIMITED TO LOSS OF DATA, LOSS OF PROFITS, HARDWARE DAMAGE, ELECTRICAL MALFUNCTION, CLOUD OVERAGE CHARGES, OR INCOMPATIBLE HARDWARE PURCHASES ARISING FROM OR RELATED TO YOUR USE OF OR INABILITY TO USE THE SITE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.

IN ALL EVENTS, GPU GARDEN'S TOTAL AGGREGATE LIABILITY ARISING FROM OR RELATING TO THE SITE SHALL NOT EXCEED ONE HUNDRED U.S. DOLLARS (US $100.00).

8. Governing Law & Severability

These Terms shall be governed by and construed in accordance with the laws of the United States and the State of Delaware, without regard to conflict of law principles. If any provision of these Terms is found to be unlawful, void, or unenforceable, that provision shall be deemed severable and shall not affect the validity of the remaining provisions.

9. Contact Us

For inquiries regarding these Terms of Service, legal notices, or technical corrections, please contact our team: