6.1. Device Overview

Tiny ML Tensorlab supports over 20 Texas Instruments microcontrollers across multiple device families.

6.1.1. Supported Device Families

Family	Core	Applications
C2000 DSP	TI C28x/C29x	Industrial control, motor drives, power systems
MSPM0	Arm Cortex-M0+	Ultra-low power, cost-sensitive IoT
MSPM33C	Arm Cortex-M33	Security-critical, high-performance edge
AM13	Arm Cortex-M33	High-performance edge with NPU acceleration
AM26x	Arm Cortex-R5	Industrial Ethernet, real-time systems
Connectivity	Arm Cortex-M33/M4	Wireless IoT, Bluetooth, Sub-GHz

6.1.2. Complete Device List

C2000 Family

Device	NPU	Description
F28P55	Yes	Recommended for complex models. 32-bit, 150 MHz, NPU-accelerated
F28P65	No	32-bit, 150 MHz
F29H85	No	64-bit C29x core
F29P58	No	64-bit C29x core
F29P32	No	64-bit C29x core
F2837	No	Dual-core, 200 MHz
F28003	No	100 MHz
F28004	No	100 MHz
F280013	No	100 MHz
F280015	No	120 MHz

MSPM0 Family

Device	NPU	Description
MSPM0G3507	No	80 MHz, ultra-low power
MSPM0G3519	No	80 MHz, ultra-low power
MSPM0G5187	Yes	80 MHz, NPU-accelerated

MSPM33C Family

Device	NPU	Description
MSPM33C32	No	160 MHz, TrustZone security
MSPM33C34	No	160 MHz, extended peripherals

AM13 Family

Device	NPU	Description
AM13E2	Yes	160 MHz Arm Cortex-M33, NPU-accelerated, TrustZone security

AM26x Family

Device	NPU	Description
AM263	No	Quad-core Cortex-R5F, 400 MHz
AM263P	No	Quad-core Cortex-R5F, 400 MHz
AM261	No	Single-core Cortex-R5F, 400 MHz

Connectivity Devices

Device	NPU	Description
CC2755	No	96 MHz Cortex-M33, wireless
CC1352	No	Cortex-M4, multi-protocol wireless

6.1.3. Target Device Configuration

Specify your target device in the config:

common:
  target_device: 'F28P55'  # or 'MSPM0G3507', 'AM263', etc.

6.1.4. NPU vs Non-NPU Devices

NPU Devices (F28P55, AM13E2, MSPM0G5187):

• Hardware-accelerated inference

• Faster execution (10-100x speedup)

• Lower power consumption

• Requires NPU-compatible models (*_NPU variants)

• Specific layer constraints (see NPU Guidelines)

Non-NPU Devices:

• Software-only inference

• More flexible model architectures

• Suitable for simpler models

• Use standard model variants

6.1.5. Choosing a Device

Consider these factors:

1. Model Complexity

   • Simple models (<1K params): Any device

   • Medium models (1K-10K params): Mid-range devices

   • Complex models (>10K params): NPU devices recommended

2. Latency Requirements

   • Real-time (<1ms): NPU devices

   • Near real-time (<10ms): Most devices

   • Relaxed timing: Any device

3. Power Budget

   • Battery-powered: MSPM0 family

   • Always-on: C2000, AM26x families

4. Existing Infrastructure

   • Motor control: C2000 family

   • Industrial Ethernet: AM26x family

   • Wireless: CC27xx, CC13xx