Innodisk Product Group

Edge AI Systems (APEX)

The APEX series brings AI computing to where data is created. Eight systems cover the whole range, from a 36 TOPS Intel box PC to a 2U short-depth server with two NVIDIA H200 NVL accelerators, all built with Innodisk's own SSDs, DRAM and I/O modules and supported by the Innodisk AI SDK.

We integrate APEX systems into POLYRACK enclosures and cabinets, pair them with the right cameras and storage, and plan the cooling together with our thermal management partners.

SectorsEdge AI Applications Industrial Automation Medical Industrial E-Mobility Railway & Transportation Defence Industry Data Centre Telecommunications Multimedia & Broadcasting Measurement & Control Green Energy

All Edge AI Systems (innodisk.com) Selection Guide (PDF, innodisk.com)

NVIDIA · 2U short-depth server

APEX-S100

2U short-depth edge AI server built on the NVIDIA MGX modular architecture for training, fine-tuning and LLM/SLM inference close to the data.

Up to 2 × NVIDIA H200 NVL, RTX PRO 6000 Blackwell Server Edition or L40S
Hot-swappable E1.S SSDs and redundant 2 × 2,000 W Titanium power supplies
NVIDIA BlueField-3 DPU and NVLink Bridge support
Optimised for deep learning, RAG pipelines and LLM/SLM inference

CPU	Intel Xeon 6 (6731P / 6730P / 6736P), up to 36 cores
GPU memory	48 GB GDDR6 · 96 GB GDDR7 · 141 GB HBM3e
Memory	Up to 1 TB DDR5 6400 RDIMM (8 slots), MRDIMM 8000
Storage	4 × E1.S PCIe Gen5 (RAID 0/1/5/10) + 2 × M.2
Networking	2 × 10/25 GbE; 2 × 200 GbE with BlueField-3
Power	2 × 2,000 W CRPS, AC 220–240 V
Operating temperature	0 ~ 30 °C
Dimensions	438 × 420 × 88 mm, rackmount

All Models in This Category

NVIDIA · workstation-class AI system

APEX-X100

Intel-based AI system with an NVIDIA RTX PRO 6000 Blackwell Max-Q or RTX 6000 Ada accelerator for on-premise LLM and SLM applications.

Up to 24,064 CUDA, 752 Tensor and 188 RT cores with 96 GB GDDR7
Intel 14th Gen Core i9, up to 192 GB DDR5 and a 2 TB NVMe SSD pre-installed
10 GbE plus 3 × 2.5 GbE, two ports with 30 W PoE for cameras

CPU	Intel Core i9-14900 / i9-13900E / i7-13700E
AI accelerator	RTX PRO 6000 Blackwell Max-Q (96 GB) or RTX 6000 Ada (48 GB)
Memory	Up to 192 GB DDR5 4400 UDIMM
Expansion	PCIe Gen4 x16 / x8, Gen3 x4 / x1; M.2 M, E and B key
I/O	8 × USB 10 Gbps, USB-C 20 Gbps, 3 × COM, 8-bit DIO
Power	DC 24–48 V
Operating temperature	0 ~ 40 °C
Dimensions	340 × 279 × 215 mm, wall or desk mount

All Models in This Category

Qualcomm · Cloud AI 100 Ultra

APEX-X100-Q

Intel-based system with the Qualcomm Cloud AI 100 Ultra accelerator: 870 TOPS and 128 GB of on-card memory for on-premise LLMs and generative AI.

870 TOPS (INT8) / 288 TFLOPS (FP16) with 128 GB LPDDR4x (ECC)
Intel Core i7-13700E with 192 GB DDR5 and a 2 TB NVMe SSD pre-installed
Single-slot, full-height PCIe Gen4 x16 accelerator

CPU	Intel Core i7-13700E, 16 cores
AI performance	870 TOPS INT8 · 288 TFLOPS FP16
Memory	192 GB DDR5 4400 UDIMM
Networking	10 GbE + 3 × 2.5 GbE (2 × PoE)
Power	DC 24–48 V
Operating temperature	0 ~ 40 °C
Dimensions	340 × 279 × 215 mm

All Models in This Category

NVIDIA · RTX 5000 Ada MXM

APEX-P100

Compact Intel system with an NVIDIA RTX 5000 Ada MXM module for factory AOI and AI medical imaging recognition.

9,728 CUDA, 304 Tensor and 76 RT cores with 16 GB GDDR6 (ECC)
5 × 2.5 GbE and 5 × DP++ for multi-camera, multi-display inspection
4 × RS-232/422/485 and 8-bit GPIO for machine integration

CPU	Intel Core i7-13700E, 16 cores
AI accelerator	NVIDIA RTX 5000 Ada, MXM 3.1 Type B
Memory	Up to 64 GB DDR5 5200 SODIMM (32 GB installed)
Storage	2 × M.2 PCIe Gen4 / SATA (512 GB installed)
Power	DC 24 V
Operating temperature	0 ~ 50 °C
Dimensions	280 × 270 × 148.2 mm, 5.7 kg, wall mount

All Models in This Category

NVIDIA · compact RTX 2000 Ada

APEX-P200

The most compact APEX with a discrete NVIDIA GPU: RTX 2000 Ada in a 182 × 158.6 × 90 mm chassis for AI medical imaging and object detection, rated -20 ~ 60 °C.

3,072 CUDA, 96 Tensor and 24 RT cores with 8 GB GDDR6 (ECC)
Intel Core i7-13800HE with 32 GB DDR5 and a 512 GB NVMe SSD pre-installed
3 × 2.5 GbE, two with 30 W PoE for IP cameras
RTX PRO 2000 Blackwell version planned

CPU	Intel Core i7-13800HE, 14 cores
AI accelerator	NVIDIA RTX 2000 Ada, MXM 3.1 Type A
Memory	Up to 64 GB DDR5 5200 SODIMM
Power	DC 19 V
Operating temperature	-20 ~ 60 °C
Dimensions	182 × 158.6 × 90 mm, 3 kg

All Models in This Category

NVIDIA · GeForce RTX 5080

APEX-X200

Intel Core Ultra system with an NVIDIA GeForce RTX 5080 for cost-efficient generative AI and LLM/SLM workloads.

10,752 CUDA, 336 Tensor and 84 RT cores with 16 GB GDDR7
Intel Core Ultra 9 285 / Ultra 7 265, 64 GB DDR5 and a 1 TB PCIe Gen5 SSD pre-installed
HDMI 2.1 and DP 2.1 for 8K displays

CPU	Intel Core Ultra 9 285 / Ultra 7 265
AI accelerator	NVIDIA GeForce RTX 5080, PCIe Gen5 x16
Memory	Up to 96 GB DDR5 6400 CSODIMM
Storage	M.2 PCIe Gen5 (1 TB installed) + M.2 PCIe Gen4
Networking	2 × 2.5 GbE
Power	AC 100–240 V
Operating temperature	0 ~ 45 °C
Dimensions	381 × 185 × 235 mm

All Models in This Category

Qualcomm · Dragonwing IQ-9075

APEX-A100

Rugged Qualcomm Dragonwing system with 100 TOPS for AGVs, AMRs and in-vehicle AI: -40 ~ 70 °C, 9–36 V DC input and chipset support through 2038.

100 TOPS dense / 200 TOPS sparse via 2 × Hexagon Tensor Processor
36 GB LPDDR5X and 128 GB UFS 3.1 onboard
4 × 4-lane MIPI CSI-2 camera inputs and CAN FD
Qualcomm chipset supported through 2038

SoC	Qualcomm Dragonwing IQ-9075, 8 × Kryo up to 2.36 GHz
AI performance	100 TOPS (dense) / 200 TOPS (sparse)
Memory / storage	36 GB LPDDR5X · 128 GB UFS 3.1
Networking	2 × 2.5 GbE
Power	DC 9–36 V
Operating temperature	-40 ~ 70 °C
Operating system	Yocto Linux, Ubuntu 24.04
Dimensions	180 × 108.7 × 69 mm, 1.7 kg

All Models in This Category

Intel · Core Ultra with NPU

APEX-E100

Compact Intel AI box PC with an integrated NPU: up to 36 TOPS for AMR/AGV, AI surveillance and visual inspection, with direct support for Innodisk MIPI over Type-C cameras.

Intel AI Boost: up to 36 TOPS across CPU, GPU and NPU
16 GB DDR5 and a 512 GB NVMe SSD pre-installed
2 × MIPI over Type-C camera ports and 2 × HDMI 2.1

CPU	Intel Core Ultra 7 165HL / Ultra 5 135HL / Ultra 3 105UL
AI performance	Up to 36 TOPS
Memory	Up to 96 GB DDR5 5600 SODIMM
Storage	M.2 PCIe Gen4 / SATA (512 GB installed)
Networking	2 × 2.5 GbE
Power	DC 12 V
Operating temperature	-20 ~ 60 °C
Dimensions	188 × 140 × 56 mm, 1.6 kg

All Models in This Category

Which APEX for Which Task?

Model	Platform	AI compute	Typical workloads
APEX-S100	NVIDIA	Up to 2 × H200 NVL, RTX PRO 6000 or L40S	Model training, fine-tuning, RAG; LLM/SLM/VLM server
APEX-X100	NVIDIA	RTX PRO 6000 Blackwell Max-Q or RTX 6000 Ada	Advanced AI inference and on-premise LLM/SLM
APEX-X100-Q	Qualcomm	Cloud AI 100 Ultra, 870 TOPS	Large-scale inference, LLMs and generative AI
APEX-P100	NVIDIA	RTX 5000 Ada (MXM)	High-precision analysis, factory AOI, medical imaging
APEX-X200	NVIDIA	GeForce RTX 5080	Generative AI and LLM/SLM at lower cost
APEX-P200	NVIDIA	RTX 2000 Ada (MXM)	ADAS, defect detection, medical imaging assistance
APEX-A100	Qualcomm	Dragonwing IQ-9075, 100 TOPS	Image and speech analysis, AGV/AMR, in-vehicle AI
APEX-E100	Intel	Core Ultra with NPU, 36 TOPS	Low-power inference: face, object and plate recognition

Why Björn Tech?

One Integrated System

We combine Innodisk storage, memory and I/O with POLYRACK enclosures and our thermal management partners in a single, tested system.

The Right Grade for the Application

Together with you we choose the temperature grade, form factor, endurance and protection features the application really needs.

Long-Term Supply

Innodisk's long-term supply and fixed-BOM options support products that stay in the field for years.

Talk to our engineering team about your project.