A100
NVIDIA's previous-generation GPU accelerator; still widely used for LLM training and inference due to lower cost than H100.
Think of it like
The budget option; older but still good.
Example
A100 (40GB or 80GB) is cheaper than H100; used for inference, smaller training runs.
How it actually works
A100 has 40GB or 80GB HBM2 memory; slightly lower bandwidth than H100. Deprecating in favor of H100 but still plenty in production.
For product teams
Second-hand A100 market is active; used for cost-sensitive workloads.
For engineers
A100 is solid for inference; training on A100s is slower but still viable.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome