Flopper.io Flopper.io

NVIDIA P100 PCIe 16GB vs NVIDIA Tesla P40 24GB

The Tesla P40 delivers 29% more FP32 throughput than the P100 (12 vs 9.3 TFLOPS dense). The Tesla P40 also carries 8 GB more memory (24 GB vs 16 GB).

P100: Pascal, 2016 Tesla P40: Pascal, 2016
FP32 dense lead
+29%
Tesla P40: 12 vs 9.3 TFLOPS
Memory
16 vs 24 GB
8 GB more for the Tesla P40
Bandwidth
732 GB/s vs 346 GB/s
P100 moves data faster
TDP
250 W vs 250 W
Same power budget
See rental pricing from $0.08/hr Download comparison sheet

Share this page

NVIDIA P100 PCIe 16GB vs NVIDIA Tesla P40 24GB

A post with this link shows the card on the Image tab as its preview.

Post on X Share on LinkedIn
 Printable version →

NVIDIA P100 PCIe 16GB

Full specs →

What is the NVIDIA P100?

The P100 is a GPU from NVIDIA on the Pascal architecture, an accelerator for AI training, inference and high-performance computing. It is a PCIe add-in card, so it fits standard servers and workstations. P100 full specifications →

Can I rent the P100 in the cloud?

Yes. Cloud providers tracked by Flopper rent the P100 by the hour, and the pricing page compares their current rates and the price history side by side. P100 rental prices →

How does the P100 compare with other accelerators?

This page sets the P100 against the Tesla P40. Both parts launched in the same year. Both are NVIDIA parts, and both ship as a PCIe card, so they fit the same class of system. The compare tool builds the same comparison for any two parts in the catalogue. All GPU comparisons →

Where do the P100 figures come from?

Every P100 figure on this page traces to NVIDIA's own datasheet or specification table. How Flopper verifies specifications →

NVIDIA Tesla P40 24GB

Leads on FP32 Full specs →

What is the NVIDIA Tesla P40?

The Tesla P40 is a GPU from NVIDIA on the Pascal architecture, an accelerator for AI training, inference and high-performance computing. It is a PCIe add-in card, so it fits standard servers and workstations. Tesla P40 full specifications →

Can I rent the Tesla P40 in the cloud?

Yes. Cloud providers tracked by Flopper rent the Tesla P40 by the hour, and the pricing page compares their current rates and the price history side by side. Tesla P40 rental prices →

How does the Tesla P40 compare with other accelerators?

This page sets the Tesla P40 against the P100. Both parts launched in the same year. Both are NVIDIA parts, and both ship as a PCIe card, so they fit the same class of system. The compare tool builds the same comparison for any two parts in the catalogue. All GPU comparisons →

Where do the Tesla P40 figures come from?

Every Tesla P40 figure on this page traces to NVIDIA's own datasheet or specification table. How Flopper verifies specifications →

Specifications

P100
Tesla P40
Architecture
PascalPascal
Launch Year
20162016
Form Factor
PCIePCIe
Memory
16 GB24 GB
Memory Bandwidth
732 GB/s346 GB/s
TDP
250 W250 W
Process Node
—16nm

Performance (TFLOPS)

P100
Tesla P40
4.7 TFLOPS No verified data available
9.3 TFLOPS 12 TFLOPS
No verified data available No verified data available
No verified data available No verified data available
19 TFLOPS No verified data available
No verified data available No verified data available
FP6
No verified data available No verified data available
No verified data available No verified data available
No verified data available47 TOPS

Figures are dense. Where a vendor publishes a figure it sits under the dense one and never decides a winner. Hover any precision for its definition, or read the GPU glossary.

FLOPS by Precision

What actually differs

The P40 and P100 split local inference down the middle: the P40 holds more model, the P100 reads it faster. The P40’s 24 GB of GDDR5 fits a quantised 30B-class model on one card, which the P100’s 16 GB cannot. On anything that fits both, the P100’s HBM2 moves 732 GB/s against 346 GB/s, so it generates tokens roughly twice as fast.

The trap is precision. Pascal has no tensor cores, and the P40 runs FP16 at a small fraction of its FP32 rate, so tools that compute in half precision crawl on it. Runtimes that dequantise into FP32 suit the P40; FP16 paths favour the P100 at 18.7 TFLOPS FP16.

Both are passive 250 W server cards with no display outputs, so a desktop needs a fan shroud on each. To test before buying, the P40 rents from $0.11 per GPU-hour on the pricing page.

New to the terms? The glossary explains FP64, FP32, FP16, INT8, Sparsity and TFLOPS. For how the generations relate, see NVIDIA architectures explained.

Frequently asked questions

Is the P100 faster than the P40?

For LLM decode, yes, on any model that fits both. The P100's HBM2 moves 732 GB/s against 346 GB/s on the P40, roughly twice the bandwidth, and token generation is bound by memory bandwidth. The P40 leads only on raw FP32 and INT8 rates, which matter less for decode.

Should I buy a P40 or a P100 for local LLMs?

Buy the P40 if your target model needs more than 16 GB, since its 24 GB holds a quantised 30B-class model on one card. Buy the P100 if your models fit in 16 GB and you want faster generation or FP16 software support. Neither card has tensor cores.

Does the P40 support FP16?

It runs FP16, but at a small fraction of its FP32 rate, so software that computes in half precision runs slowly on it. Runtimes that dequantise weights and compute in FP32 suit the P40. The P100 runs FP16 at full speed, rated at 18.7 TFLOPS FP16.

How much does it cost to rent a P40 or a P100?

Flopper tracks one P40 listing, Vast.ai community capacity at $0.11 per GPU-hour. The P100 PCIe 16GB has no live rental listings, so second-hand purchase is the main route to one. Both are 2016 Pascal parts whose NVIDIA product pages are now retired.

Where to Rent

NVIDIA P100 PCIe 16GB

Rent from

$0.09 / GPU-hr

Lowest of 2 live listings across 2 providers

See all P100 rental prices →

Price history and every provider on the pricing page. Some links are affiliate links.

NVIDIA Tesla P40 24GB

Rent from

$0.08 / GPU-hr

Lowest of 1 live listing across 1 provider

See all Tesla P40 rental prices →

Price history and every provider on the pricing page. Some links are affiliate links.

Get Comparison Updates

New GPUs added weekly. Be the first to see how they compare.

Open in Full Compare Tool → P100 full specs Tesla P40 full specs Estimate cluster cost in the calculator →

© 2026 Flopper.io - Compare the GPUs Powering AI