ARK gives the PCIe configuration as up to PCI Express 4.0 x16 while Intel's own datasheet for the same card says "Gen 4.0 x16 (x8 Electrical)". The datasheet is the more specific statement and is what pcie_lanes records. The two Intel throughput figures are again quoted at different unstated clocks: 5.02 TFLOPS FP32 implies roughly 2,451 MHz, ARK's 69 dense INT8 TOPS implies roughly 2,100 MHz, and ARK's stated graphics clock is 1,550 MHz.

Intel Arc Pro A40 6GB

Alchemist PCIe 2022 6nm Spec confidence: Official
FP32
5
TFLOPS
VRAM
6 GB
GDDR6
Bandwidth
192 GB/s
memory
TDP
50 W
100.4 TFLOPS/kW
Download datasheet

Overview

The Arc Pro A40 is the smallest Alchemist workstation card, a half-height half-length single-slot board that needs no power connector at all inside its 50 W envelope. It runs 8 Xe-cores and 128 XMX matrix engines on a 96-bit bus with 6 GB of GDDR6, and Intel rates it at 5.02 TFLOPS FP32 and 69 dense INT8 TOPS. It is a professional display and light acceleration card rather than an inference card, and it is listed here for completeness of the Arc Pro family. Its useful trait is density: 168 mm long, one slot, no auxiliary power, and Intel's certified workstation driver, which is why OEMs put it in small form factor business desktops.

Performance

Peak theoretical throughput by precision type

PrecisionPeak
FP64
No verified data available
FP32
32-bit floating point
5TFLOPS
TF32
No verified data available
BF16
No verified data available
FP16
No verified data available
FP8
No verified data available
FP6
No verified data available
FP4
No verified data available
INT8
8-bit integer
69TOPS

Every figure here is dense. The vendor documents no structured (2:4) sparsity mode for this part, so there is no second number to quote.

Specifications

Architecture

Alchemist

Form Factor

PCIe

Launch Year

2022

Process Node

6nm

Memory

6 GB GDDR6

Bandwidth

192 GB/s

TDP

50 W

Max Power

50 W

Xe Cores

8

Spec Confidence

Official

Full Specifications

Compute Engine
Xe Cores 8
Tensor Cores 128 (Xe-HPG XMX)
Boost Clock 1.55 GHz
Memory
VRAM 6 GB
Memory Type GDDR6
Bandwidth 192 GB/s
Interface Width 96-bit
Memory Clock 16 GT/s
Interconnect & I/O
PCIe 4.0 x8
Power & Thermal
TDP 50 W
Max Board Power 50 W
Power Connector None
Cooling Active
Enterprise Features
Sparsity No
Compute APIs oneAPI, OpenCL, Level Zero, OpenVINO
Physical & Media
Card Length 168 mm
Width 1-slot
General
Form Factor PCIe
Architecture Alchemist
Process Node 6nm
Launch Year 2022

Datasheet & Resources

Data Provenance

Every figure traced to a source

Primary Source

Publisher
Intel
Published
No verified data available

Data Quality

Spec confidence
Official
Clock basis
Boost
Core precisions with figures
2 of 9
Normalization
All values in TFLOPS

Compare cloud providers offering on-demand GPU instances for AI training, inference, and HPC workloads.

Browse GPU Cloud Providers

Similar GPUs

Intel Arc A770 16GB

PCIe · 2022
FP32: 20 TFLOPS
Compare vs Arc A770

Intel Arc Pro A60 12GB

PCIe · 2023
FP32: 10 TFLOPS
Compare vs Arc Pro A60

Intel Arc A770 8GB

PCIe · 2022
FP32: No verified data available
Compare vs Arc A770

Intel Arc A750 8GB

PCIe · 2022
FP32: No verified data available
Compare vs Arc A750

Intel Arc A580 8GB

PCIe · 2023
FP32: No verified data available
Compare vs Arc A580

Frequently Asked Questions

How many TFLOPS does the Intel Arc Pro A40 have?

The Intel Arc Pro A40 delivers 5 TFLOPS FP32 at peak. Flopper does not currently have verified FP16 and FP8 throughput figures for it.

What is the power consumption of the Intel Arc Pro A40?

The Intel Arc Pro A40 has a TDP (Thermal Design Power) rating of 50 watts.

How much memory does the Intel Arc Pro A40 have?

The Intel Arc Pro A40 is equipped with 6 GB of VRAM with 192 GB/s of memory bandwidth.

What architecture is the Intel Arc Pro A40 based on?

The Intel Arc Pro A40 is based on the Alchemist architecture, launched in 2022.

Stay Updated on GPU Releases

Get notified when new GPUs are added or specifications are updated.

Loading verification...

No spam, unsubscribe anytime.

Back to GPUs