Question 1

What AI models can run on P100?

Accepted Answer

The P100 (16 GB VRAM) can run models including paligemma-3b-ft-vizwizvqa-448-jax, Llama-3.2-3B-Instruct-QLORA_INT4_EO8, Llama-2-7b, Llama-3.2-1B-Instruct-SpinQuant_INT4_EO8, Llama-3.2-1B-Instruct-QLORA_INT4_EO8 and more — 10 compatible models indexed so far. Compatible models are determined by whether the model's memory footprint fits within the GPU's 16 GB VRAM, potentially using quantization (Q8, Q6, or Q4) to reduce memory usage.

Question 2

What is the cheapest cloud option for P100?

Accepted Answer

The cheapest on-demand option for the P100 is Google Cloud at $1.46/hr (). Spot/preemptible pricing is available at $0.66/hr. We track pricing across 7 providers to help you find the best deal.

Question 3

Is P100 good for fine-tuning?

Accepted Answer

The P100 can handle fine-tuning with some constraints. With 16 GB of VRAM, it's suitable for LoRA/QLoRA fine-tuning of models up to ~7B parameters. Full fine-tuning is limited to smaller models (~1-3B parameters). Techniques like gradient checkpointing and mixed precision help maximize what fits.

Question 4

How much VRAM does P100 have?

Accepted Answer

The P100 has 16 GB of VRAM (HBM2). It also offers 732 GB/s memory bandwidth, 21.2 TFLOPS of FP16 compute.

Question 5

What quantization formats work on P100?

Accepted Answer

The P100 supports the following quantization formats: FP16 (full precision), Q8 (8-bit), Q6 (6-bit), Q4 (4-bit). FP16 works for models up to ~3B parameters. Q4 quantization can fit models requiring up to ~64 GB at FP16 into 16 GB.

P100

Specifications

Cloud pricing

Compatible models

Compare this GPU

Frequently asked questions

VRAM	16 GB
Memory type	HBM2
Bandwidth	732 GB/s
FP16 TFLOPS	21.2
FP8 support	No