GeekRadar / The comparison desk

COMPARE.

GPU

Nvidia

A40

Graphics processor

Video memory: 48 GB^*
Memory bus: 384 bit^*
Rated memory bandwidth: 696 GB/s^*

Benchmark results

Geekbench GPU (reported compilation) · version unreported · CUDA205,120^*Platform unreported Compare results
Geekbench GPU (reported compilation) · version unreported · CUDA205,120^*Platform unreported Compare results
Geekbench GPU (reported compilation) · version unreported · OpenCL185,972^*Platform unreported Compare results
Geekbench GPU (reported compilation) · version unreported · OpenCL149,033^*Platform unreported Compare results
Geekbench GPU (reported compilation) · version unreported · OpenCL185,972^*Platform unreported Compare results

Processing & platform

Architecture: Ampere^*

Memory & storage

Video memory: 48 GB^*
Video memory type: GDDR6^*
Memory bus: 384 bit^*
Rated memory bandwidth: 696 GB/s^*

System configuration

Detailed specifications

Specifications 1 · 20 fields

GPU architecture — NVIDIA A40 — 48 GB GDDR6

Ampere^*

Theoretical peak: BF16 tensor dense per GPU — NVIDIA A40 — 48 GB GDDR6

149.7^* TFLOPS

Scope and conditions

Accumulation: FP32^*

Theoretical peak: FP16 tensor dense per GPU — NVIDIA A40 — 48 GB GDDR6

149.7^* TFLOPS

Theoretical peak: FP32 non-tensor dense per GPU — NVIDIA A40 — 48 GB GDDR6

37.4^* TFLOPS

Native hardware support: FP4 tensor — NVIDIA A40 — 48 GB GDDR6

No^*

Native hardware support: FP64 tensor — NVIDIA A40 — 48 GB GDDR6

No^*

Native hardware support: FP6 tensor — NVIDIA A40 — 48 GB GDDR6

No^*

Native hardware support: FP8 tensor — NVIDIA A40 — 48 GB GDDR6

No^*

Theoretical peak: INT8 tensor dense per GPU — NVIDIA A40 — 48 GB GDDR6

299.3^* TOPS

Theoretical peak: TF32 tensor dense per GPU — NVIDIA A40 — 48 GB GDDR6

74.8^* TFLOPS

Physical configuration — NVIDIA A40 — 48 GB GDDR6

PCIe, dual-slot FHFL^*

GPU interconnect bandwidth (NVLink 3) bidirectional — NVIDIA A40 — 48 GB GDDR6

112.5^* GB/s

GPU interconnect bandwidth per direction — NVIDIA A40 — 48 GB GDDR6

56.25^* GB/s

Scope and conditions

Calculation: 112.5 / 2^*

PCIe interface declaration — NVIDIA A40 — 48 GB GDDR6

PCIe Gen4 x16^*

Maximum board power — NVIDIA A40 — 48 GB GDDR6

300^* W

Peak memory bandwidth per GPU — NVIDIA A40 — 48 GB GDDR6

696^* GB/s

Memory interface width per GPU — NVIDIA A40 — 48 GB GDDR6

384^* bit

Memory capacity per GPU — NVIDIA A40 — 48 GB GDDR6

48^* GB

Memory technology — NVIDIA A40 — 48 GB GDDR6

GDDR6^*

GPU model — NVIDIA A40 — 48 GB GDDR6

NVIDIA A40^*

Specifications 2 · 6 fields

Device

A40^*

Scope and conditions

benchmark_context

suite: Geekbench GPU (reported compilation)^*
backend_columns: CUDA^*
Metal^*
OpenCL^*
Vulkan^*
test_configuration_verified: No^*
comparable: No^*

Manufacturer

Nvidia^*

Scope and conditions

benchmark_context

suite: Geekbench GPU (reported compilation)^*
backend_columns: CUDA^*
Metal^*
OpenCL^*
Vulkan^*
test_configuration_verified: No^*
comparable: No^*

Device

A40^*

Scope and conditions

benchmark_context

suite: Geekbench GPU (reported compilation)^*
backend_columns: CUDA^*
Metal^*
OpenCL^*
Vulkan^*
test_configuration_verified: No^*
comparable: No^*

OpenCL

149,033^*

Scope and conditions

benchmark_context

suite: Geekbench GPU (reported compilation)^*
backend_columns: CUDA^*
Metal^*
OpenCL^*
Vulkan^*
test_configuration_verified: No^*
comparable: No^*

CUDA

205,120^*

Scope and conditions

benchmark_context

suite: Geekbench GPU (reported compilation)^*
backend_columns: CUDA^*
Metal^*
OpenCL^*
Vulkan^*
test_configuration_verified: No^*
comparable: No^*

OpenCL

185,972^*

Scope and conditions

benchmark_context

suite: Geekbench GPU (reported compilation)^*
backend_columns: CUDA^*
Metal^*
OpenCL^*
Vulkan^*
test_configuration_verified: No^*
comparable: No^*

Specifications 3 · 4 fields

Manufacturer

Nvidia^*

Scope and conditions

benchmark_context

suite: Geekbench GPU (reported compilation)^*
backend_columns: CUDA^*
Metal^*
OpenCL^*
Vulkan^*
test_configuration_verified: No^*
comparable: No^*

Device

A40^*

Scope and conditions

benchmark_context

suite: Geekbench GPU (reported compilation)^*
backend_columns: CUDA^*
Metal^*
OpenCL^*
Vulkan^*
test_configuration_verified: No^*
comparable: No^*

CUDA

205,120^*

Scope and conditions

benchmark_context

suite: Geekbench GPU (reported compilation)^*
backend_columns: CUDA^*
Metal^*
OpenCL^*
Vulkan^*
test_configuration_verified: No^*
comparable: No^*

OpenCL

185,972^*

Scope and conditions

benchmark_context

suite: Geekbench GPU (reported compilation)^*
backend_columns: CUDA^*
Metal^*
OpenCL^*
Vulkan^*
test_configuration_verified: No^*
comparable: No^*

* Provisional data