# A16

> NVIDIA · Ampere

Discontinued

Canonical: https://gpuninja.com/chips/a16
Data as of: 2026-10-11T10:34:56Z
Path: Home › GPU chips

> **Discontinued.** Discontinued on 1 Aug 2025. This page is a preserved snapshot of the last recorded specs, prices and benchmark results.

- Specs verified: between 11 Oct 2026, 08:46 UTC and 11 Oct 2026, 10:34 UTC

## Specifications

- Vendor: NVIDIA (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 06:57 UTC · checked 11 Oct 2026, 08:46 UTC)
- Architecture: Ampere (Verified · source en.wikipedia.org · retrieved 11 Oct 2026, 06:55 UTC · checked 11 Oct 2026, 08:46 UTC)
- Die: GA107 (Verified · source cisco.com · retrieved 11 Oct 2026, 07:02 UTC · checked 11 Oct 2026, 08:49 UTC)
- Process: Samsung 8N (Verified · source en.wikipedia.org · retrieved 11 Oct 2026, 06:55 UTC · checked 11 Oct 2026, 08:46 UTC)
- Shader cores: 5,120 (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 06:57 UTC · checked 11 Oct 2026, 08:46 UTC)
- Tensor cores: 160 (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 06:57 UTC · checked 11 Oct 2026, 08:46 UTC)
- RT cores: 40 (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 06:57 UTC · checked 11 Oct 2026, 08:46 UTC)
- Memory bus: 128-bit (Verified · source cisco.com · retrieved 11 Oct 2026, 07:02 UTC · checked 11 Oct 2026, 08:49 UTC)
- Max VRAM: 64 GB (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 06:57 UTC · checked 11 Oct 2026, 08:46 UTC)
- Reference boost clock: 1,755 MHz (Verified · source cisco.com · retrieved 11 Oct 2026, 07:02 UTC · checked 11 Oct 2026, 08:49 UTC)
- Reference TDP: 250 W (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 06:57 UTC · checked 11 Oct 2026, 08:46 UTC)
- Host interface: PCIe 4.0 x16 (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 09:53 UTC · checked 11 Oct 2026, 10:34 UTC)
- Launch MSRP: No public MSRP
- Launch date: 12 Apr 2021 (Verified · source sysrqmts.com · retrieved 11 Oct 2026, 06:55 UTC · checked 11 Oct 2026, 08:46 UTC)
- Segment: Data center (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 06:57 UTC · checked 11 Oct 2026, 08:46 UTC)
- PCI IDs: 10de:25b6 (Verified · source download.nvidia.com · retrieved 11 Oct 2026, 06:55 UTC · checked 11 Oct 2026, 08:46 UTC)
- Status: Discontinued on 1 Aug 2025 (Unverified · source docs.nvidia.com · retrieved 11 Oct 2026, 06:57 UTC · checked 11 Oct 2026, 08:46 UTC)

## Performance

- Memory bandwidth: 200 GB/s (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 09:53 UTC · checked 11 Oct 2026, 10:34 UTC)
- FP32 compute: 18 TFLOPS (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 09:53 UTC · checked 11 Oct 2026, 10:34 UTC)
- FP16 compute: 71.6 TFLOPS (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 09:53 UTC · checked 11 Oct 2026, 10:34 UTC)
- BF16 compute: 71.6 TFLOPS (Unverified · source nvidia.com · retrieved 11 Oct 2026, 09:51 UTC · checked 11 Oct 2026, 10:34 UTC)
- FP8 compute: Not supported
- FP4 compute: Not supported
- INT8 compute: 143.6 TOPS (Verified · source images.nvidia.com · retrieved 11 Oct 2026, 09:53 UTC · checked 11 Oct 2026, 10:34 UTC)

## Boards built on this chip

Board, Chip, VRAM, TDP, Length, Slots, Price, Stock

No boards recorded for this chip.

## Benchmarks

No benchmark results recorded.

## Sources

- https://images.nvidia.com/content/Solutions/data-center/vgpu-a16-datasheet.pdf — Vendor, Shader cores, Tensor cores, RT cores, Max VRAM, Reference TDP, Host interface, Segment, Memory bandwidth, FP32 compute, FP16 compute, INT8 compute; last verified 11 Oct 2026, 10:34 UTC
- https://en.wikipedia.org/wiki/Ampere_(microarchitecture) — Architecture, Process; last verified 11 Oct 2026, 08:46 UTC
- https://www.cisco.com/c/dam/en/us/products/collateral/servers-unified-computing/ucs-c-series-rack-servers/nvidia-a16-gpu.pdf — Die, Memory bus, Reference boost clock; last verified 11 Oct 2026, 08:49 UTC
- https://sysrqmts.com/gpus/nvidia/tesla/a16-pcie — Launch date; last verified 11 Oct 2026, 08:46 UTC
- https://download.nvidia.com/XFree86/Linux-x86_64/580.95.05/README/supportedchips.html — PCI IDs; last verified 11 Oct 2026, 08:46 UTC
- https://docs.nvidia.com/data-center-gpu/line-card.pdf — Status; last verified 11 Oct 2026, 08:46 UTC
- https://www.nvidia.com/content/PDF/nvidia-ampere-ga-102-gpu-architecture-whitepaper-v2.pdf — BF16 compute; last verified 11 Oct 2026, 10:34 UTC