GeekRadar / The comparison desk
COMPARE.
GPUNVIDIA A16
Graphics processor
Benchmark results
- Geekbench GPU (reported compilation) · version unreported · OpenCL33,748*Platform unreported Compare results
Processing & platform
- Architecture
- Ampere*
- Video memory per GPU
- 16 GB*
Memory & storage
- Video memory type
- GDDR6*
System configuration
Detailed specifications
Specifications 1 · 18 fields
- GPU architecture — NVIDIA A16 — 16 GB GDDR6
- Ampere*
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Theoretical peak: FP16 tensor dense per GPU — NVIDIA A16 — 16 GB GDDR6
- 17.9* TFLOPS
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Theoretical peak: FP32 non-tensor dense per GPU — NVIDIA A16 — 16 GB GDDR6
- 4.5* TFLOPS
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Native hardware support: FP4 tensor — NVIDIA A16 — 16 GB GDDR6
- No*
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Native hardware support: FP64 tensor — NVIDIA A16 — 16 GB GDDR6
- No*
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Native hardware support: FP6 tensor — NVIDIA A16 — 16 GB GDDR6
- No*
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Native hardware support: FP8 tensor — NVIDIA A16 — 16 GB GDDR6
- No*
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Theoretical peak: INT8 tensor dense per GPU — NVIDIA A16 — 16 GB GDDR6
- 35.9* TOPS
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Theoretical peak: TF32 tensor dense per GPU — NVIDIA A16 — 16 GB GDDR6
- 9* TFLOPS
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Physical configuration — NVIDIA A16 — 16 GB GDDR6
- PCIe, dual-slot FHFL, four GPUs per board*
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Native hardware support for GPU interconnect — NVIDIA A16 — 16 GB GDDR6
- No*
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- PCIe interface declaration — NVIDIA A16 — 16 GB GDDR6
- PCIe Gen4 x16*
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Maximum whole-board power for four GPUs — NVIDIA A16 — 16 GB GDDR6
- 250* W
- Peak memory bandwidth per GPU — NVIDIA A16 — 16 GB GDDR6
- 200* GB/s
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Memory interface width per GPU — NVIDIA A16 — 16 GB GDDR6
- 128* bit
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Memory capacity per GPU — NVIDIA A16 — 16 GB GDDR6
- 16* GB
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- Memory technology — NVIDIA A16 — 16 GB GDDR6
- GDDR6*
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
- GPU model — NVIDIA A16 — 16 GB GDDR6
- NVIDIA A16*
Scope and conditions
- Configuration
- Per-GPU value; the board contains four GPUs.*
Specifications 2 · 2 fields
- Device
- NVIDIA A16*
Scope and conditions
- benchmark_context
- suite
- Geekbench GPU (reported compilation)*
- backend_columns
- CUDA*
- Metal*
- OpenCL*
- Vulkan*
- test_configuration_verified
- No*
- comparable
- No*
- OpenCL
- 33,748*
Scope and conditions
- benchmark_context
- suite
- Geekbench GPU (reported compilation)*
- backend_columns
- CUDA*
- Metal*
- OpenCL*
- Vulkan*
- test_configuration_verified
- No*
- comparable
- No*
0 selected Compare 0
* Provisional data