Products — AI Computing Hardware & Edge Solutions | QS Compute

GPU & AI Accelerators

NVIDIA H100 (80GB HBM3, $22K–26K), H200 (141GB HBM3e, $28K–34K), L40S, A6000, RTX 5090. Data center GPUs for AI training, LLM inference, HPC.

H100/H200 L40S RTX 5090 In Stock

NVIDIA L40 GPU — Pricing Intelligence

48GB GDDR6 ECC, Ada Lovelace. 15-channel pricing: Auth $7,800–9,500, China ¥56K–75K, Used $5K–7K. ⭐ DigiKey/Mouser/Avnet/Arrow In Stock

48GB ECC $7.8K–9.5K 15 Channels In Stock

Jetson Edge AI Systems

Orin Nano 4GB ($149) to AGX 64GB ($1,999). 20–275 TOPS. Industrial variants with −40°C cold start, ECC, 10yr lifecycle.

20–275 TOPS $149–$1,999 In Stock

ARM Edge AI Computers

ARM Cortex + MXM NPU. 32/64 TOPS, 64-ch video decode. Fanless, −20~60°C. Rockchip RK3588, MediaTek Genio platforms.

32–64 TOPS RK3588 Fanless

Industrial Panel PCs & HMI

IPC-A310 & SDI-620C. 21.5" to 23.8", i5-14450Hx, touch, dual LAN, IP65. Intel 12/14 Gen.

Intel 12/14 Gen Touch IP65

Embedded Gateways

PE-2000U/3000U & IPC-E420. X86 fanless, DDR5, 6 COM, 9–30V DC. i3/i5/i7 UE.

i3/i5/i7 UE DDR5 6×COM

Compact Industrial PCs

T30 N95 — 1.2L ultra-compact, 7×USB, HDMI+VGA. 3yr warranty. VESA mountable.

Enterprise Storage

Samsung PM9D3a, Micron 7450, SK hynix PS1010 NVMe SSDs. DDR4/DDR5 industrial DRAM. Wide-temp, PLP, 5yr lifecycle.

NVMe DDR5 In Stock

Development Kits

Jetson Orin/Xavier/Nano DK, RK3588/3568, Atlas 200I. Ready for prototyping. Pre-flashed with JetPack SDK.

Jetson DK RK DK Atlas DK

GPU cold plates, CDU 200kW–1MW, quick-disconnect fittings, immersion cooling fluids. Full-stack thermal management for AI clusters.

H100/H200 CDU Immersion

Interconnects & Connectors

QSFP-DD 400G/800G DAC, PCIe 5.0 risers, M.2/U.2 NVMe connectors, M12 industrial Ethernet, GPU power, SMA RF, B2B mezzanine. One partner for the full connectivity stack.

QSFP-DD PCIe 5.0 M.2/U.2 M12 IP67

GPU Servers & DGX Systems

NVIDIA DGX B200, DGX H200, Supermicro 8-GPU servers. Pre-integrated AI training infrastructure with NVLink fabric. Plug in and train.

DGX B200 DGX H200 8×GPU Liquid Cooled

FPGA & AI Accelerator Boards

Xilinx Versal, Intel Agilex 7, Alveo cards. Programmable logic with integrated AI engines. Deterministic sub-ms inference for edge and datacenter.

Xilinx Versal Intel Agilex 228 TOPS

Industrial Single-Board Computers

Raspberry Pi 5 (BCM2712, 4–8GB), Arduino Portenta H7/X8/C33 (−40~85°C), BeagleBone AI-64 (8 TOPS). Industrial SBCs for edge AI, IoT, robotics.

Raspberry Pi 5 Arduino Portenta BeagleBone AI In Stock

DRAM — DDR5/DDR4 ECC Memory

DDR5/DDR4 ECC server & industrial memory. RDIMM for servers, SODIMM for edge AI. Samsung, Micron, SK hynix, Kingston.

DDR5ECC RDIMMSODIMM

Edge AI Accelerators & Modules

NVIDIA Jetson Orin/Thor, Hailo-8, Google Coral, MemryX MX3, Rockchip RK3588S. 6 to 275 TOPS edge inference.

JetsonHailo-86–275 TOPS

Edge AI Dev Kits

Turnkey dev platforms — Hailo neural processors, Rockchip NPU SBCs, Intel. Deploy vision & sensor fusion at the edge.

HailoRockchipIntel

Fanless Embedded PCs

Industrial fanless computers for edge AI, machine vision, factory automation, in-vehicle. Zero moving parts.

FanlessIP65Wide-Temp

HBM3E High-Bandwidth Memory

HBM3E 12-High 36GB stacks — Samsung & Micron. Highest-bandwidth memory for AI accelerators.

HBM3E36GB12-High

Industrial SSDs

Wide-temp M.2 NVMe, U.2, SATA SSDs. -40°C to +85°C for edge, embedded, fanless deployments.

NVMeWide-TempM.2/U.2

Industrial & server PSUs for GPU servers, AI rigs, edge, telecom. Redundant, hot-swappable, high-efficiency.

RedundantHot-Swap80Plus

Intel Xeon Granite Rapids & Sierra Forest, AMD EPYC Turin. 16–192 cores for AI servers, virtualization, HPC.

XeonEPYC16–192 Core

NVIDIA MGX compute trays for GB200 NVL72, 6.5U eight-GPU platforms, 4U liquid-cooled racks and storage chassis.

Machine Vision Cameras

Human-IPD passive stereo and industrial 3D cameras for physical AI. GMSL2/FAKRA and Gigabit Ethernet interfaces, IP67 sealing, sub-millisecond multi-camera sync.

Stereo 3D GMSL2 IP67

NVIDIA RTX PRO 6000 Blackwell 96GB

96GB GDDR7 ECC, 24,064 CUDA Cores, 125 TFLOPS FP32, 1.8 TB/s bandwidth. PCIe 5.0 workstation GPU for AI, LLM inference, professional viz.

96GB GDDR7 125 TFLOPS Blackwell PCIe 5.0

Qualcomm Dragonfly C1000

Custom Oryon cores above 5GHz, 250+ core chiplet design, over 2TB/s PCIe Gen 7 with CXL.

250+ Cores Oryon PCIe Gen 7 OCP ORv3

Qualcomm Dragonfly AI300

Third-generation rack-scale AI inference accelerator with integrated HBC Gen 2 near-memory compute.

HBC Gen 2 Rack-Scale UALink Inference

Qualcomm High Bandwidth Compute

3D-stacked near-memory architecture: 133 TB/s per card in Gen 1, 6x bandwidth per watt versus HBM.

133 TB/s Near-Memory 3D-Stacked Vs HBM

Tensordyne Napier

3nm logarithmic-math inference accelerator. 72-chip pod at 76.7 PFLOPS FP16, air-cooled rack at 306 PFLOPS.

306 PFLOPS 3nm 72-Chip Pod Air-Cooled

Turnkey rack-scale AI infrastructure: compute, PowerSwitch SN6600 networking and 10+ PB Exascale Storage.

Rack-Scale 10+ PB SN6600 6-Hour Setup

Dell PowerCool CDU C7000

4U rackmount cooling distribution unit rated for more than 220 kW and 40 degrees Celsius facility water.

220kW 4U 40°C Water Vera Rubin

LattePanda Mu Ultra

60 x 69.6mm x86 AI compute module: Intel Core Ultra 200V, 16GB LPDDR5X, up to 115 combined TOPS.

115 TOPS Core Ultra 200V 60x69.6mm Lunar Lake

Virtium Edge AI — VMX-004 & RZ/G3E DX-M1

TAA-compliant, US-assembled M.2 AI accelerator module — 4× MemryX MX3, up to 24 TOPS at 3–10W, −40 to +85°C.

24 TOPS TAA Made in USA M.2 2280

NVIDIA Jetson Orin Nano 2

78 TOPS entry-level robotics computer: 2x Orin Nano Super inference at 40% less power, 8GB LPDDR5X, same SO-DIMM outline.

78 TOPS 8GB LPDDR5X 2x Orin Nano Super H1 2027

Unigen Cheetah SSD Series

Cheetah SCM at 60/80 DWPD and 14.7 GB/s plus Cheetah HC to 122TB. PCIe Gen5 x4 NVMe 2.0, 16 NAND channels, PLP.

122TB 60/80 DWPD PCIe Gen5 14.7 GB/s

Credo ZeroFlap AEC & Toucan Retimer

ZeroFlap active electrical cables to 1.6T with zero soft link flaps, 800G ZeroFlap optics, Toucan PCIe Gen6 retimers at 6.5ns.

1.6T AEC PCIe Gen6 6.5ns PILOT

Penguin MemoryAI KV Cache Server

Altus XE4318GT-KVC 4U CXL KV cache server: 11TB DDR5, 8x 1TB CXL AICs, dual EPYC 9005. Cuts time-to-first-token.

11TB CXL KV Cache 4U

Euclyd CRAFTWERK

Agentic AI inference SiP: 16,384 SIMD processors, 32 PFLOPS FP4, 1TB Ultra Bandwidth Memory at 8,000 TB/s. CWS 32 rack hits 1.024 exaFLOPS.

32 PFLOPS FP4 1TB UBM Exascale Rack

Axelera Europa AIPU

Second-gen Axelera AIPU: 629 TOPS INT8 from 8 D-IMC cores, 16 RISC-V vector cores, 64GB memory, PCIe Gen4 x4. Scales to 4 chips / 256GB.

629 TOPS 64GB PCIe Gen4

Moore Threads MTT S4000

48GB MUSA datacenter GPU: 768 GB/s, PCIe 5.0 x16, 200 TFLOPS FP16, 200 TOPS INT8 at 450W. Powers the 1,000-card KUAE training cluster.

48GB 200 TFLOPS PCIe 5.0

Concurrent Technologies VPX GPGPU

Rugged 3U OpenVPX GPGPU cards: Bragi with RTX PRO 4000/5000 Blackwell, 24GB GDDR7 ECC, 10,496 CUDA cores, plus Kratos Xeon 6 and Fulla 10GbE switch.

3U OpenVPX SOSA 24GB GDDR7

AMD Threadripper Halo Station

Deskside AI workstation: 96-core Threadripper PRO 9995WX, up to 2TB DDR5 and 4x Instinct MI350P for 576GB HBM3e and 16 TB/s. Trillion-parameter local inference.

96 Cores 576GB HBM3e DGX Station Rival

MSI XpertStation WS300

Deskside AI supercomputer on NVIDIA DGX Station architecture: GB300 Grace Blackwell Ultra, 748GB coherent memory, dual 400GbE ConnectX-8. Plus EdgeXpert.

GB300 748GB Dual 400GbE

MSI CG681-S6093

6U liquid-cooled AI server: dual AMD EPYC SP7 up to 256 cores and 8x RTX PRO 6000 Blackwell Server Edition at up to 1,200W. MGX, DC-MHS and Open Compute.

6U 8x 1200W GPU MGX

EIZO Condor Thor 3U VPX

Rugged SOSA-aligned 3U OpenVPX single board computer on Jetson Thor T5000: 2,070 FP4 TFLOPS, 2,560 CUDA cores, 128GB LPDDR5X, VITA 46/65 and CMOSS.

2,070 FP4 TFLOPS 3U OpenVPX VITA 46/65

ASRock NUC Ultra 300 BOX

Panther Lake compact edge computers built on Intel Core Ultra Series 3 — Core Ultra X7 358H or Ultra 5 325 with in-band ECC, up to 128GB DDR5-7200, PCIe Gen5 NVMe, four independent 8K displays, dual 2.5GbE and Wi-Fi 7.

Panther Lake-H 128GB DDR5 Quad 8K

ARBOR FPC-9309W-G5

Rugged edge AI platform that accepts an NVIDIA RTX 50 series GPU up to 575W: Intel Core Ultra Series 2, 192GB DDR5, dual SATA plus dual NVMe with RAID 0/1/5, 4× GbE/2.5GbE and 24–48V DC input.

RTX 50 ≤575W 192GB DDR5 RAID 0/1/5

ARBOR ARES-2100

Ultra-slim 1U fanless edge AI system on Intel Core Series 3 (Wildcat Lake) with Xe3 graphics and NPU 5.0 — up to 40 TOPS, triple 2.5GbE, 9–36V DC input and MIL-STD-810H durability.

40 TOPS 1U Fanless MIL-STD-810H

AAEON UP Xtreme PTL Edge

Rugged Panther Lake edge AI system: Intel Core Ultra Series 3 with Arc Xe3 graphics and NPU 5.0 for up to 180 TOPS, 128GB DDR5-7200, USB4, dual 2.5GbE, dual COM and a 40-pin GPIO header.

180 TOPS Core Ultra S3 USB4

Edge AI system-on-module on the NXP i.MX 95: hexa-core Cortex-A55 with integrated GPU/VPU/ISP and an 8 eTOPS NPU, 8GB LPDDR5, 32GB eMMC, CAN-FD and PCIe Gen3 in a -40°C to +85°C industrial module.

8 eTOPS NPU i.MX 95 -40~85°C

Alif Balletto B1 / E1C StartKits

~$50 tinyML starter kits from Alif Semiconductor: 160 MHz Arm Cortex-M55 with Helium plus an Ethos-U55 NPU, 2MB SRAM, Arducam camera header, mikroBUS Click socket, dual PDM mics and an onboard Segger J-Link.

Ethos-U55 Cortex-M55 BLE 5.3

NVIDIA IGX Orin 700

Industrial-grade Orin platform with 2048-core Ampere iGPU, 12-core Arm Cortex-A78AE, 64 GB LPDDR5 ECC, ConnectX-7 dual 100GbE, BMC and a functional safety MCU. Up to 1705 sparse TOPS with an optional RTX 6000 Ada GPU.

1705 TOPS ConnectX-7 ISO 26262

Aetina AIE-MIX640

NVIDIA-Certified IGX Orin inference system for industry and healthcare: 248 TOPS base, scalable to 1705 TOPS with an RTX 6000 Ada Generation GPU and 200 Gb/s ConnectX-7 networking.

IGX Orin 248 TOPS Certified

Advantech MIC-735-IT

Rugged AI inference system on NVIDIA IGX Thor T5000: up to 2070 FP4 TFLOPS, onboard functional safety MCU, Holoscan Sensor Bridge, 4x10G SFP plus 5G Ethernet, -30 to +60C, 10-year lifecycle.

2070 TFLOPS IGX Thor -30~60C

Dedicated Computing M1300

Embeddable NVIDIA IGX Orin 700 edge AI server for medical imaging and machine vision: ConnectX-7 200GbE with GPUDirect RDMA, 600+ TOPS with an RTX A6000, Holoscan software, medical-grade design.

IGX Orin 700 600+ TOPS Holoscan

Syslogic RPC RS A4NA

IP67 rugged computer on NVIDIA Jetson Orin Nano with a purpose-built shock and vibration rated carrier board, screw-on M12 connectors and a sealed fanless housing for mining and construction vehicles.

IP67 M12 Orin Nano

Kontron VX30101

SOSA-aligned 3U VPX single-slot computing board with Intel Core Ultra Series 3 up to 16 cores, on-die NPU and Xe graphics, 32-64 GB ECC LPDDR5x, 100GbE fabrics, VITA 46.11 Tier 3 IPMI, 10-year availability.

3U VPX 100GbE -40~85C

SECO SOM-SMARC-Dragonwing-IQ8

Industrial SMARC Rel. 2.1.1 module on Qualcomm Dragonwing IQ-8275 with up to 40 TOPS, 32 GB LPDDR5/X, 1 TB UFS 3.1, GbE plus 2.5GbE, PCIe Gen4, MIPI-CSI and CAN-FD.

40 TOPS SMARC 2.1.1 IQ-8275

Enclustra Andromeda XRU50

AMD Zynq UltraScale+ RFSoC module with 8 RF-ADCs to 5 GSPS and 8 RF-DACs to 9.85 GSPS for direct RF synthesis, 8 GB ECC DDR4, 16 PL GTY to 28.21 Gbps, PCIe Gen4 x8, 80x64 mm industrial.

RFSoC 9.85 GSPS PCIe Gen4

Axiomtek mBOX603

Medical-grade edge system with 13th/12th Gen Intel Core i9-i3 up to 65W, 96 GB DDR5-5600 ECC, dual PCIe x16 for GPU and capture cards, 500W/900W medical PSU, EN/IEC 60601-1 certified.

EN 60601-1 96 GB ECC Dual PCIe x16

Corvalent CoreAI Systems

US-built industrial AI platforms with a field-upgradable CoreAI module to 104 TOPS, Hailo-8 and Hailo-10H 40 TOPS M.2 accelerators or NVIDIA GPU cards, 7-slot rackmount and CorGrid Edge form factors.

104 TOPS Hailo-10H US-built

Wincomm WMP Medical AI Panel PCs

19 to 23.8 inch medical AI panel PCs on Intel Bartlett Lake-S Core Series 2 up to 24 cores with 96 GB DDR5, optional NVIDIA MXM RTX PRO graphics, EN 60601-1 certified and IP65 front anti-bacteria housing.

EN 60601-1 IP65 front MXM RTX

Lanner Falcon-H8

PCIe AI acceleration card carrying 4, 5 or 6 Hailo-8 processors behind an onboard PCIe switch for up to 156 TOPS of multi-stream inference, in commercial and industrial grade SKUs.

156 TOPS 6x Hailo-8 PCIe

Microsoft Maia 200

Azure's 216GB HBM3e inference accelerator at 7TB/s with native FP4/FP8/FP6 tensor cores, 272MB of on-die SRAM and a 750W envelope for RL post-training and token generation.

216GB HBM3e 10.15 PFLOPS FP4 Azure XPU

Untether AI speedAI 240 Slim

75W low-profile PCIe inference card using at-memory compute with 1,456 RISC-V cores and up to 30 TFLOPS/W, for cloud and edge inference on x86 or Arm hosts.

75W At-Memory Compute Arm Host

SiMa.ai Modalix MLSoC

50 TOPS multimodal edge MLSoC at sub-10W. SoM from $449 pin-compatible with NVIDIA Jetson Orin NX/Nano carriers, plus a PCIe Gen5 HHHL card and a $1,499 devkit.

50 TOPS Sub-10W Orin Pinout

AAEON MIX-PTLWV1

Panther Lake industrial mini-ITX board at 15–25W with up to 128GB DDR5, dual 10GbE, quad DisplayPort 1.4, PCIe Gen5 x8 and 12–24V wide DC input.

Panther Lake Dual 10GbE 4x DisplayPort

StarFive VisionFive 2 Lite

Cost-reduced RISC-V SBC on the quad-core JH7110S at 1.25GHz with 2/4/8GB LPDDR4, Imagination BXE-4-32 GPU, M.2 M-key, GbE and 40-pin GPIO from $19.90.

RISC-V JH7110S M.2 M-key

ECS LIVA Mini PCs

Elitegroup's LIVA line-up: Z11 PLUS on Core Ultra 7 255H with up to 96GB DDR5 and dual LAN, Z15 PLUS on Wildcat Lake with NPU, LGA1851 One H810, fanless Z4F and USB-C Q4.

Core Ultra 255H Wildcat Lake NPU LGA1851

MikroTik CRS804-4DDQ-hRM

Compact 1U switch with four 400G QSFP56-DD cages on a Marvell 98DX7335, dual 10G management ports, hot-swap PSUs and fans, running RouterOS v7 for AI and storage fabrics.

4x 400G RouterOS v7 1U

FS AI Optics Portfolio

AI fabric optics in two tiers: 1.6T OSFP-DR8/2FR4 modules verified on NVIDIA Quantum-X800, 800G ZR/ZR+ coherent to 500km, the D7070 muxponder and BOX 5 Ultra toolkit.

1.6T OSFP 800G Coherent X800 Verified

Orbbec Physis 257g

Human-IPD passive RGB stereo camera with a 63mm baseline, dual 1920×1200 global-shutter streams at 60fps, 0.1–10m depth range and IP67 sealing for humanoid robots.

63mm Baseline IP67 GMSL2

Orbbec Gemini 338Le / 338Lg

Industrial stereo 3D cameras with enhanced ambient-light rejection — Gemini 338Le over Gigabit Ethernet for fixed cells, 338Lg over GMSL2/FAKRA for robot payloads.

Stereo 3D Gigabit Ethernet GMSL2 FAKRA

OpenAI Jalapeño inference ASIC

OpenAI's first in-house LLM inference ASIC, built with Broadcom: 13.4 PFLOP/s MXFP4 matrix compute against 216 GiB of HBM4 at 15.4 TB/s, scaling to 27 EFLOP/s across a 2,048-chip system.

216GiB HBM4 13.4 PFLOPS MXFP4 Broadcom ASIC

AWS Graviton5 192-core Arm server CPU

AWS's fifth-generation Arm server processor places 192 Neoverse V3 cores in a single socket on 3nm with ~180MB of shared L3, 12-channel DDR5-8800 and PCIe Gen 6 for cloud AI workloads.

192 Cores Neoverse V3 DDR5-8800

AMD Helios rack-scale AI system

AMD's first complete rack-scale AI system: 72x Instinct MI455X accelerators with ~31TB of HBM4, ~1.4PB/s bandwidth and ~2.9 FP4 exaFLOPS per rack over open UALink and Ultra Ethernet fabrics.

72x MI455X 31TB HBM4 2.9 EF FP4

Google Ironwood TPU v7 inference accelerator

Google's seventh-generation TPU is the first built inference-first: 192GB of HBM3E at 7.37TB/s and 4,614 FP8 TFLOPS per chip, with 9,216-chip superpods reaching 42.5 FP8 ExaFLOPS.

192GB HBM3E 4614 FP8 TFLOPS 9216-chip pod

D-Robotics RDK S600 robot development kit

A 560 TOPS embodied intelligence development kit with an 18-core Arm Cortex-A78AE CPU, six MIPI camera inputs, six USB 3.0 ports and four PCIe 3.0 interfaces for advanced robotics.

560 TOPS 18x A78AE Robot Kit

Broadcom Tomahawk 6 Davisson switch

The first 102.4 Tbps Ethernet switch with third-generation co-packaged optics — 224G SerDes, 200 Gbps per optical channel and Ultra Ethernet Consortium compliance for AI-scale fabrics.

102.4 Tbps Co-Packaged Optics 224G SerDes

MSI XpertStation WS300T60L — Grace Blackwell Ultra AI Workstation

Deskside DGX Station architecture: 72-core Grace CPU, Blackwell Ultra GPU, up to 748 GB coherent memory and dual 400GbE.

GB300 748 GB 400GbE

Radxa rCore-Q8550 — 48-TOPS Qualcomm Dragonwing Module

36 × 32 mm QCS8550 SoM with up to 48 dense INT8 TOPS, 16 GB LPDDR5X, PCIe Gen 4 and six MIPI-CSI camera lanes.

QCS8550 48 TOPS BGA SoM

IEI NANO-X100 — AMD Ryzen AI Embedded X100 EPIC SBC

Zen 5, RDNA 3.5 and XDNA 2 NPU on a 115 × 165 mm EPIC board with 64 GB LPDDR5X-8533 and dual 2.5GbE.

Zen 5 XDNA 2 Dual 2.5GbE

ASRock Industrial Jetson AGX Orin Developer Kit

Industrial AGX Orin carrier for the Industrial, 64 GB and 32 GB SOMs up to 275 TOPS, with four GMSL II camera inputs and 12 PoE ports.

275 TOPS 12x PoE GMSL II

LattePanda Mu MXM GPU Carrier Kit — Arc A380 / RTX 3060

DFRobot MXM carrier for the LattePanda Mu x86 module with Arc A380 6 GB or RTX 3060 12 GB graphics and six 4K displays.

MXM 6x 4K RTX 3060

ASUS Ascent GX10 — NVIDIA GB10 Desktop AI Supercomputer

GB10 Grace Blackwell Superchip, 1 PFLOP AI performance, 128 GB unified LPDDR5X for 200B-parameter models, up to 4 TB NVMe and ConnectX-7 in a 150 x 150 x 51 mm chassis.

1 PFLOP 128 GB GB10

Qualcomm Dragonwing IQ10 — 350 TOPS Edge AI Robotics

18-core Qualcomm Oryon CPU with NPU and GPU, up to 350 TOPS dense, Safety Island error correction, 12x GMSL2 cameras and TSN/EtherCAT/CAN-FD deterministic I/O.

350 TOPS Robotics Edge AI

Premio RCO-1000-ASL — Intel Atom x7000 Rugged Edge PC

Ultra-compact fanless industrial computer with 8-core Intel Atom x7835RE, DDR5 up to 32 GB, dual 2.5GbE, -40 to 70 C and 9-36 VDC input.

Fanless 2.5GbE Wide-Temp

AMD EPYC 8005 Sorano — 8-84 Core Single-Socket CPU

Zen 5 edge and telco CPUs from 8 to 84 cores in 70-225 W, 6-channel DDR5-6400 ECC, 96 PCIe Gen5 lanes, NEBS-ready SP5 single socket.

Zen 5 SP5 84-Core

RealSense D585 Pro — AI-Native Stereo Depth Camera

Gen 5 SoC depth camera for AMRs and humanoids: 120 x 100 degree FOV, 12 cm to 10 m+ range, IP65 standard, GMSL2 and USB-C with sync, dual RGB and on-camera person detection.

IP65 GMSL2 Stereo Depth

DCX FDU V2AT2 — 8.15 MW Facility CDU for 45 C Warm Water

Facility distribution unit delivering up to 8.15 MW at a 2 C approach temperature, with 45 C warm-water operation for NVIDIA NVL72 GB200/GB300 and Vera Rubin and chillerless heat rejection.

8.15 MW 45 C Water CDU

KEYENCE VS Series — AI Smart Camera with Optical Zoom

IP67 smart camera with a 19-lens ZoomTrax optical system, 1.6 to 15 MP modes, 85 fps, PoE IEEE 802.3at and AI segmentation, detection, classification and OCR alongside rule-based tools.

IP67 AI Vision PoE

Zivid 3 Series — 3D Colour Cameras for Bin-Picking

8 MP structured-light 3D colour cameras delivering point clouds, 2D images, normals and depth in 50 ms. Zivid 3 XL250 covers 2.5 m2 out to 4.5 m with IP65 and 15G shock ratings.

3D Vision 8 MP IP65

Kingston Industrial SSD — i-Temp and Gen 4 NVMe

i-Temp SATA drives from 128 GB to 1 TB at -40 to 85 C plus M.2 2280 TLC PCIe Gen 4 NVMe up to 2 TB at 7,000/6,500 MB/s and 1,280 TBW, all under locked-BOM control.

Wide-Temp 1,280 TBW Locked BOM

Qualcomm Dragonwing IQ-X — 45 TOPS Industrial SoC

Industrial processors with 12 Oryon cores at 3.4 GHz, a Hexagon NPU delivering up to 45 TOPS, Adreno GPU and up to 64 GB LPDDR5x, rated -40 to 105 C in a 58 mm FCBGA package.

45 TOPS -40 to 105 C Industrial SoC

Darveen MIC-9000 — Modular Industrial PC with 300 W GPU

14th to 12th Gen Intel Core on the Q670 chipset with up to 64 GB DDR5, a PCIe x16 slot for a full-length 300 W GPU, RAID 0/1 removable storage and 24 V DC input for mobile installs.

300 W GPU DDR5 Machine Vision

Premio RCO-6000-RPL — Rugged AI Edge Inference PC

Fanless and cable-less LGA1700 platform with up to 96 GB DDR5 including ECC, modular hot-swappable EDGEBoost Nodes for NVMe or low-profile GPU acceleration, UL Listing and EN 50155 rail approval.

96 GB ECC EDGEBoost EN 50155

Moxa RKP-C220 — AI-Ready 2U Rackmount IPC

Moxa's first AI-ready rackmount IPCs: 13th Gen Intel Core up to 24 cores on the Q670E chipset, up to 64 GB DDR4, a PCIe x16 Gen 4 slot for full-length discrete GPU cards, up to 8 LAN and 10 serial ports, isolated CAN, TPM 2.0 and fanless -30 to 60 C operation.

PCIe x16 GPU 24 Cores Fanless 2U

LiquidStack CDU-1MW — 1,350 kW Direct-to-Chip CDU

Row-form-factor coolant distribution unit delivering up to 1,350 kW of direct-to-chip cooling through a 316SS plate heat exchanger with dual hot-swappable VSD pumps, integrated ATS with supercapacitor, dual 25-micron filtration, PN16 rating and MODBUS/BACnet/SNMP controls.

1,350 kW Dual VSD Pumps BACnet

Accelsius NeuCool — Two-Phase Direct-to-Chip

Two-phase direct-to-chip cooling across the IR80 in-rack system at 80 kW+, the IR150 at 150 kW and the MR250 at 250 kW per rack, cooling 4,500 W per socket and hotspot fluxes above 500 W/cm2 with a dielectric fluid and leak-protection warranty.

250 kW/Rack 4,500 W Socket Dielectric Fluid

Abaco SBC3901/SBC3902 — Rugged 3U VPX AI SBC

SOSA-aligned 3U VPX single board computers built on NVIDIA Jetson AGX Orin with 2048 CUDA cores, up to 275 TOPS, 64 GB DRAM, 64 GB eMMC, optional 2 TB NVMe, a 100GbE plus 25GbE data plane and MIL-STD-810 ruggedization.

275 TOPS 3U VPX MIL-STD-810

Mercury RES-XR6 — Rugged Rackmount Server

Rugged EnterpriseSeries rackmount servers with up to four Intel Xeon Scalable processors, 6 TB DDR4 ECC, 240 TB of hot-swap storage, seven PCIe slots, NVIDIA GPU and BlueField DPU options, dual redundant power and MIL-STD-810G/901D/167 qualification.

4x Xeon Scalable 240 TB MIL-STD-810G

Arduino Nicla — Edge AI Boards at 22.86 mm

Three 22.86 x 22.86 mm edge AI boards: Nicla Voice with the Syntiant NDP120 neural decision processor for always-on speech, Nicla Vision with a dual-core STM32H747 and 2 MP camera, and Nicla Sense ME with the Bosch BHI260AP self-learning AI sensor.

NDP120 TinyML 22.86 mm

Luxshare 224G OSFP — 1.6 Tbps per Port

224 Gbps PAM4 OSFP connector, 8 channels per port and 36 ports per 1U slot, backward compatible with OSFP112G for an 800G to 1.6T upgrade on the existing fabric. KOOLIO CPC/NPC and Intrepid cable cartridge alongside.

224G PAM4 1.6 Tbps OSFP112G

Samtec Si-Fly HD 224G Backplane

224 Gbps PAM4 cable backplane carrying up to 2,048 differential pairs per open rack unit at 1.75 ps/m max twinax skew, with NovaRay 128G PAM4, ExaMAX 112G and XCede HD in the same family.

224G PAM4 2,048 pairs/OU 7.2 Tbps

Eaton 800 VDC AI Factory Power

800 VDC reference architecture built to NVIDIA's specification around a 2 MW 15 kV class MVSST running above 97% efficiency, with ORV3 busbar distribution, supercapacitor backup and dedicated DC connectors.

800 VDC 2 MW MVSST Grid-to-Chip

Chicony CRPS / M-CRPS Server PSU

1U common redundant power supplies from 550 W to 2400 W in a 73.5 x 40 x 185 mm envelope: 94% Platinum, fully digital PMBus/I2C control, hot-plug N+1, and Titanium designs ready for ErP Lot 9.0.

550-2400 W PMBus N+1 hot-plug

Broadcom MegaRAID 9600 Series

Third-generation tri-mode NVMe/SAS/SATA RAID adapters driving 240 SAS/SATA or 32 NVMe devices per controller at 6.4M IOPS, with 60x faster rebuilds and CacheVault CVPM05 flash cache protection.

MegaRAID 24G SAS 32 NVMe

Vertiv Geist UPDU

Universal rack PDU that accepts 16 A to 63 A at 120 V to 415 V worldwide — the input cord is detachable, so one SKU covers multiple regional power standards in 11 kW and 22 kW models.

22 kW 415 V Detachable Cord

Server Technology PRO4X PDU

Zero-U intelligent rack PDU with up to 54 HDOT Cx outlets — more than 20% smaller than standard outlets, so two PDUs replace three or four — plus billing-grade +/-0.5% metering to IEC 61557-12.

54 Outlets HDOT Cx 100 A

Eaton CDU Series

Row-level coolant distribution units to 2 MW at 3 °C approach built to OCP Project Deschutes 5th-generation specification, plus 22-110 kW rack-level liquid-to-liquid and liquid-to-air CDUs.

2 MW N+1 Pumps L2L / L2A

Stäubli UQD & UQDB Quick Disconnects

OCP-compliant universal quick disconnects for direct-to-chip cooling: blind-mate UQDB self-aligns to ±2.5-5 mm so GPU trays connect on insertion. 16 bar, -20 to +150 °C, NVIDIA-approved.

UQD02-08 Blind-Mate 16 bar

Siemens SIMATIC IPC227G Nanobox PC

Sealed fanless Nanobox industrial PC in about one litre: Intel Atom X6413E, 8/16 GB DDR4, 240/480 GB SSD, 3x GbE, TPM 2.0, rated for 24/7 operation at up to 55 °C.

Fanless TPM 2.0 55 °C

Ingrasys NVIDIA GB300 NVL72 Rack

72 Blackwell Ultra GPUs plus 36 Grace CPUs across 18 compute trays and 9 NVLink switch trays, with liquid-to-liquid cooling rated to 2,500 kW.

7 2 G P U s , 3 6 G r a c e C P U s , 2 . 5 M W C o o l i n g

H3C UniServer R5350 G6

4U dual AMD EPYC 9004 AI server supporting up to 10 double-width GPUs, 24 DDR5 slots to 6 TB, CXL 1.1 and twelve PCIe 5.0 slots.

1 0 G P U s , 1 2 8 - C o r e E P Y C , 6 T B D D R 5

IEIT Systems NF5468G7 (Inspur)

4U dual-platform AI server: up to 10 double-width GPU cards with either 192-core AMD EPYC or 112-core Intel Xeon, DDR5, CXL 1.1 and 400G NDR InfiniBand.

1 0 G P U s , 1 9 2 C o r e s , C X L 1 . 1

Kioxia XL1 CXL Memory Expansion

512 GB CXL 2.0 Type 3 memory expansion in E3.S built on XL-FLASH Generation 2 NAND — 755 ns latency and close to 2M 512B random-read IOPS.

5 1 2 G B , C X L 2 . 0 , 7 5 5 n s

Quadric Chimera GPNPU

Licensable GPNPU IP with 512 MAC to 8K MAC cores, 1 to 6,912 TOPS, configurable 1-16 MB L2 and a single ONNX-plus-C++ toolchain.

1 - 6 , 9 1 2 T O P S , I N T 8 + F P 1 6 , C + + P r o g r a m m a b l e

AVerMedia AVerAI Embedded Systems

AVerAI carrier boards for NVIDIA Jetson AGX Orin and AGX Thor with 120-pin GMSL2 camera input, 10G Ethernet, 10GigE Vision and -40 to 85 °C operation.

G M S L 2 x 8 , 1 0 G + 1 0 G i g E , - 4 0 t o 8 5 C

Chenbro RM252 MIO Series

2U 450 mm short-depth edge AI chassis for two full-height GPGPUs on a PCIe Gen5 riser, with front- or rear-I/O variants.

2U / 450 mm 2x Full-Height GPGPU PCIe Gen5 Riser

Compact Vision 5

SECO Compact Vision 5 Dragonwing IQ-2390

5-inch IP65 panel PC on Qualcomm Dragonwing IQ-2390 with a 1.1 TOPS NPU, dual GbE with TSN and RS232/RS485/CAN.

1.1 TOPS NPU IP65 Front Panel Dual GbE TSN

AAEON GENESYSM-MTH6

0.9 kg subcompact edge AI system on Intel Core Ultra with Arc graphics, 96 GB DDR5 and three M.2 2280 M-Key slots.

Intel Core Ultra 3x M.2 2280 9-36V DC

20-inch short-depth tower server chassis with eight full-height slots and optional 4-bay 3.5-inch hot-swap storage.

20" Short Depth ATX / mATX 1+1 CRPS

SECO SBC-Dragonwing-IQ-2390

47 x 114 mm board-level industrial computer on Dragonwing IQ-2390 with 1.1 TOPS NPU, dual GbE TSN and camera expansion.

47 x 114 mm 1.1 TOPS NPU Dual GbE TSN

Suitcase-sized portable AI supercomputer with six swappable sled slots, NVIDIA L40S accelerators and 246 TB NVMe per storage sled.

91.6 TFLOPS FP32 6 Sled Slots <55 lbs

Silicom Marbella AI-N / AI-F140

Short-depth 1U edge AI analytics appliance with Intel Xeon D-1700, an NVIDIA L4 Tensor Core GPU at 485 INT8 TOPS or Intel Flex 140, and up to 4x 25G SFP28 uplinks.

L4 24GB 485 INT8 TOPS 4x 25G SFP28

Silicom Toledo IoT Edge Appliance

Intel Atom C3000 IoT edge appliance with 4-16 cores, up to 64GB ECC LPDDR4, 4x 10G SFP+ or 10GBase-T, dual M.2 NVMe, IPMI 2.0 BMC and industrial temperature options to -30 C.

4x 10G SFP+ IPMI 2.0 TPM 2.0

Rambus DDR5 Server DIMM Chipset

Complete DDR5 server module silicon: registering clock drivers to 9600 MT/s, MRDIMM MRCD and MDB to 12800 MT/s, three PMIC grades, SPD hub and temperature sensor.

DR5RCD6Gxx 9600 D5MRCD2Gxx PMIC5030

Rambus SOCAMM2 Server Chipset

Voltage regulators and SPD hub for SOCAMM2: P2815Gxx 12A and P2813Gxx 3A with the SPD1605Gxx hub, bringing LPDDR5X efficiency to serviceable AI server modules to 9600 MT/s.

P2815Gxx 12A P2813Gxx 3A SPD1605Gxx

Blackmagic DeckLink Capture Cards

Broadcast-grade PCIe capture cards for AI video ingest: DeckLink 8K Pro G2 with 4x bi-directional 12G-SDI and HDMI 2.1, plus Quad 2, Duo 2 and Quad HDMI Recorder models.

12G-SDI x4 HDMI 2.1 8K capture

AJA KONA 5 PCIe Video I/O

8-lane PCIe 3.0 video I/O card with bidirectional 12G-SDI at SMPTE-2082, HDMI v2.0 at 6 Gbps per component and HDR10/HLG support, alongside KONA X, KONA IP25 and KONA HDMI.

12G-SDI SMPTE-2082 HDMI 2.0 HDR10 / HLG

Gateworks Catalina GW9200 — i.MX 95 Industrial SBC

Industrial i.MX 95 single board computer with a 2 TOPS eIQ Neutron NPU, 10GbE and dual field-swappable FSA sockets.

i.MX 95 2 TOPS NPU 10GbE

Zhihe A210 — 12 TOPS RISC-V Edge AI SoC

Octa-core RISC-V RV64GCV edge AI processor with a 12 TOPS INT8 NPU, 4Kp120 AV1 decode and 8 tokens/s DeepSeek-7B inference.

RISC-V 12 TOPS Edge AI

Amlogic A123X and C305X2 — Cortex-A320 AIoT SoCs

Armv9 Cortex-A320 AIoT SoCs pairing a 4 TOPS vision NPU with an 8 TOPS Transformer NPU and AI-ISP to 4Kp30.

Cortex-A320 8 TOPS AI-ISP

MediaTek Genio 360 / 360P — Cortex-A76/A55 AIoT SoCs

Cost-sensitive AIoT processors with a MediaTek 8th-gen NPU up to 7.4 TOPS GenAI, dual MIPI DSI or DP outputs and GbE with TSN.

7.4 TOPS GenAI Cortex-A76 TSN

Toradex OSM and Lino — 30x30 mm i.MX 93 SoMs

30x30 mm i.MX 93 or i.MX 91 modules with a 0.5 TOPS Ethos-U65 NPU, inline ECC memory and -40 to +85C rating.

i.MX 93 30x30 mm OSM Size-S

Cornelis Networks CN6000 — 800G Multi-Protocol SuperNIC

First 800G multi-protocol SuperNIC running Omni-Path, RoCEv2 and Ultra Ethernet concurrently on PCIe 6.0 at 1.6 billion messages/sec.

800G PCIe 6.0 Omni-Path

Kalray MPPA DPU Cards — Coolidge v2 / K300 / TC4

Production MPPA DPU cards for AI inference and storage offload, 40–160 TOPS INT8 with NVMe-oF, encryption and erasure coding and an open SDK.

160 TOPS KVX cores NVMe-oF

Quectel SE200ZC-AP — Rockchip RV1126B Vision Smart Module

Rockchip RV1126B smart module with 12M ISP HDR, 8M AI-ISP and a 3 TOPS NPU at INT4/8/16 and FP16 with 4K H.265 multi-stream encoding.

3 TOPS NPU AI-ISP 4K H.265

NVIDIA MMS1X00-N5400 400Gbps QSFP112 Optic

NVIDIA 400Gbps QSFP112 optic at 1310 nm with 4× 112G PAM4 and MPO-12, validated on ConnectX-8 and Spectrum-X for GB300 AI Station links.

400G 112G PAM4 Spectrum-X

Innodisk EXEC-Q911 COM-HPC Mini Kit

200 TOPS Qualcomm IQ-9075 edge AI kit with 36 GB LPDDR5X, dual 2.5GbE and -40 to +85 C operation.

200 TOPS IQ-9075 9-36V

Thundercomm TurboX C7790 Dev Kit

Qualcomm Dragonwing Q-7790 edge AI platform with 24 TOPS NPU, 12 GB LPDDR5X and triple 21MP ISP.

24 TOPS Q-7790 LPDDR5X

Jetway ATX-ARS1-W880 Industrial ATX

Arrow Lake-S ATX board on Intel W880 with 36 TOPS, 256 GB ECC CU-DIMM DDR5 and three 2.5GbE ports.

36 TOPS W880 CU-DIMM

Jetway SMC-ARK1 SMARC 2.1 Module

Rockchip RK3588 SMARC module with 6 TOPS NPU, 16 GB LPDDR4x, dual GbE and a 314-pin MXM edge connector.

RK3588 6 TOPS SMARC 2.1

Foremay Rugged SSD Series

Military and space grade SSDs with radiation-hardened options from 0-70 C to -55 to +125 C and 3U VPX formats.

SC199/TC166 -55 to +125C 3U VPX

AMD Versal Premium Gen 2 MoP

2VP3422 / 2VP3522 / 2VP3622 adaptive SoCs with 32 GB LPDDR5X in-package, PCIe Gen6 and CXL 3.1.

32GB LPDDR5X PCIe Gen6 CXL 3.1

8devices 8Sight T100

UAS-optimised 320×240 LWIR thermal camera with an STM32N6 Neural-ART NPU for fully offline edge AI inference. NDAA/TAA compliant, 38 g.

320×240 LWIR 60 FPS 600 GOPS NPU

Delta GoCool CDU Series

Delta GoCool coolant distribution units from 80 kW liquid-to-air to 3,000 kW liquid-to-liquid, plus CoolDoor RDHx and RowCool row coolers.

80 kW–3 MW L2A / L2L RDHx

Delta ORV3 Server Power

Open Rack V3 48 VDC power for AI server racks: 3 kW and 5.5 kW hot-swap PSUs plus 18 kW and 33 kW power shelves with N+1/N+N redundancy.

ORV3 48 VDC Up to 33 kW

TechNexion TELOS-AI4000

Enclosed 40 eTOPS edge-AI accelerator module on the NXP Ara240 DNPU with 16 GB LPDDR4 and a 12 W envelope.

40 TOPS Ara240 12 W

ASRock NUC 300 Series

Four Intel Core Series 3 Wildcat Lake platforms — NUC BOX-320/304 mini PCs and NUC-320/304 boards with up to 40 TOPS.

Wildcat Lake 40 TOPS dual 2.5GbE

AMD Ryzen AI Max PRO 400 Series

Zen 5 + RDNA 3.5 halo platform with up to 192 GB unified memory and 160 GB VRAM — runs 300B-parameter LLMs locally.

192 GB 160 GB VRAM 300B LLM

Efinix Sapphire RV64

64-bit RISC-V soft SoC for Titanium, Topaz and Trion FPGAs with an optional Sv39 MMU for embedded Linux.

RISC-V 64-bit LPDDR4x

Avaota F2 RISC-V SBC

First Allwinner V861 SBC — dual XuanTie C907 RISC-V cores, RVV 1.0, a 1 TOPS NPU and 4K H.265 encode for AI cameras.

Allwinner V861 1 TOPS 4K H.265

DCX ECDU Series CDUs

Rack-based enterprise coolant distribution units from 600 kW to 2.6 MW for direct-to-chip AI data-center cooling.

600 kW–2.6 MW in-rack direct-to-chip

Allied Vision FXO 100GigE

100GigE QSFP28 global shutter cameras: FXO925 24.5 MP at 442.7 fps, FXO926 12.3 MP at 660.8 fps, fiber runs to 10 km.

24.5 MP 660 fps QSFP28 RDMA

XIMEA MX1048 105 MP Camera

105.5 MP Sony IMX927 global shutter at 10,272 x 10,272 px and 56 fps over PCIe Gen4 x4, no frame grabber.

105.5 MP 56 fps PCIe Gen4

LUCID Triton10 8K Line Scan

8,192 px line scan at 55 kHz in a 44 x 44 x 82 mm IP67 body, with 10GigE RDMA streaming and PoE.

8K line scan 55 kHz IP67 PoE

LUCID Atlas10 SWIR Cameras

10GigE SWIR cameras on Sony SenSWIR IMX992/IMX993 InGaAs sensors, 400-1700 nm, with integrated TEC cooling.

5.2 MP 400-1700 nm TEC cooled

Allied Vision Alecs VSWIR

IP67 open smart camera pairing a Sony SenSWIR imager with an NVIDIA Jetson Orin NX 16GB (100 TOPS) for on-camera AI.

Jetson Orin 100 TOPS IP67

Dell PowerEdge XE9785

10U AI server with 8x NVIDIA HGX B300 NVL8 (270 GB) or 8x AMD MI355X, dual EPYC 9005 and up to 6 TB DDR5.

8x B300 2.1 TB HBM 10U

Frore Systems AirJet / LiquidJet

Solid-state active cooling chips and direct-to-chip coldplates: AirJet Mini G2 removes 7.5 W per chip; LiquidJet Nexus cuts thermal-stack weight 65%.

Solid-State 7.5 W/chip 53C Inlet

Accton 800G AI/ML Fabrics

Tomahawk 5 Ethernet and Jericho3-AI / Ramon3 VoQ fabrics scale to a 32K-GPU 800G/400G cluster with a software migration path.

51.2 Tbps 64x800G DDC

AEWIN Network Appliances

Intel Xeon 6 and AMD EPYC network appliances, rack-scale AI servers and two-phase immersion cooling up to 20kW.

Xeon 6 8x Gen5 2PIC 20kW

Allied Vision HR 100GigE

100GigE Vision area-scan cameras up to 104.9 MP and 225 fps with RDMA streaming over QSFP28.

104.9 MP 225 fps RDMA

e-con NetraVis 3D Depth Camera

Active-stereo depth camera with 100 mm baseline, 1% accuracy at 2 m and an integrated IMU for AMRs and humanoids.

Active Stereo 0.19-6 m USB 3.0

Texas Instruments Edge AI SoCs

TDA4VM (8 TOPS, ASIL-D), AM68A and AM62A vision processors with TIDL scale from one to eight cameras.

8 TOPS ASIL-D TIDL

Advantest V93000 EXA Scale ATE

Advantest V93000 EXA Scale SoC test system and T5801 memory platform for AI accelerators, HPC, GDDR7/LPDDR6/DDR6 DRAM and PMICs.

ATE GDDR7/LPDDR6 SmarTest 8

Rohde & Schwarz MXO + RTO6 Scopes

Rohde & Schwarz MXO 3/4/5/5C and RTO6 oscilloscopes to 2 GHz with 1 Gpts memory and a new three-phase power-analysis option.

PCIe 6.0 1 Gpts K333

NI PXI Test Platform

New NI PXIe-1081 18-slot chassis, PXIe-8842/8862 controllers, PXIe-5108 oscilloscope and PXIe-5842 VST for automated test.

PXIe 18-slot VST

Delta 800 VDC Power Rack

Delta 800 VDC In-Row 660 kW power rack for NVIDIA MGX with six 110 kW shelves, 480 kW embedded BBU and 98%-efficient PSUs.

800 VDC 660 kW MGX

Opto Engineering ITALA G

Opto Engineering ITALA G GigE Vision PoE cameras from 0.4 to 31.5 MP with VIS–SWIR 400–1700 nm, polarized, IP67 and liquid-lens variants.

Hioki PW8001 Power Analyzer

DC-5 MHz power analyzer, 8 channels, 15 MS/s 18-bit sampling, ±0.03% accuracy.

8 ch ±0.03% 15 MS/s

SIGLENT SDS7000A Oscilloscope

3-8 GHz 12-bit oscilloscopes with 20 GSa/s, 2 Gpts/ch, eye/jitter and compliance test.

8 GHz 12-bit 20 GSa/s

RIGOL DHO900 12-bit MSO

12-bit economical MSO, 4+16 channels, 50 Mpts, 1,000,000 wfms/s capture.

12-bit 50 Mpts Bode plot

Fluke 1770 Power Quality

Three-phase power quality analyzer, IEC 61000-4-30 Class A Ed.3.

Class A 20 MHz 8 kV

OWON XDS3000 14-bit Scope

Benchtop DSO, switchable 8/12/14-bit ADC, 60-300 MHz, 40 Mpts.

14-bit 2.5 GSa/s 40 Mpts

Luxima LUX High-Speed Sensors

Global shutter CMOS sensors at 1,000+ fps, 1.3-9.5 MP.

1,000 fps Global shutter 12-bit

SALPHOENIX/SALEAGLE FPGAs: 60-400K LUTs, 12.5G SerDes, MIPI, ISO 26262 certified.

60-400K LUTs 12.5G SerDes ISO 26262

17nm Chinese-made DDR5, ~90% yield, demonstrated above 9,000 MT/s at CL30.

DDR5 17nm 9000 MT/s

Lumentum AI Optics

1.6T DR4 prototype, SHP CPO laser and 16-channel DWDM ELSFP light sources.

1.6T DR4 SHP laser ELSFP

8-core UltraRISC UR-DP1000 RISC-V Mini-ITX with 64 GB ECC DDR4 and PCIe Gen4 x16.

8-core RISC-V 64 GB ECC PCIe 4.0 x16

64-core SOPHGO SG2042 RISC-V mATX with 128 GB ECC and dual PCIe Gen4 x16.

64-core RISC-V 128 GB ECC mATX

OrangePi 6 Plus

12-core CIX CD8180 ARMv9 SBC with 28.8 TOPS NPU and 45 TOPS combined AI.

45 TOPS 12-core dual 5GbE

OrangePi AI Station

Ascend 310 board with 176 TOPS, 16 CPU cores and 48/96 GB LPDDR4X for dense inference.

176 TOPS Ascend 310 96 GB

PINE64 StarPro64

RISC-V SBC on ESWIN EIC7700X with 19.95 TOPS NPU and up to 32 GB LPDDR5.

19.95 TOPS RISC-V 32 GB LPDDR5

IBASE IB966 3.5" SBC

3.5-inch Panther Lake SBC with up to 180 TOPS, 128 GB DDR5-7200 and dual 2.5GbE.

180 TOPS Panther Lake DDR5-7200

Banana Pi BPI-M7 RK3588 SBC

Rockchip RK3588 SBC with 6 TOPS NPU, dual 2.5GbE and M.2 NVMe.

6 TOPS RK3588 2.5GbE

Banana Pi BPI-3A6000 Loongson Mini PC

Loongson 3A6000 4-core 8-thread mini PC, up to 32GB DDR4, dual HDMI and dual Gigabit LAN.

LoongArch 32GB DDR4 Dual LAN

Luckfox PicoKVM IP-KVM

Palm-sized IP-KVM with 1080P60 H.264/H.265 capture, HID emulation and Wake-on-LAN.

IP-KVM 1080P60 H.265

Banana Pi BPI-R4 WiFi 7 Router

MediaTek MT7988A WiFi 7 router board with dual 10G SFP, 4x GbE and dual M.2 slots.

WiFi 7 Dual 10G SFP M.2

Banana Pi BPI-CM6 RISC-V Module

SpacemiT K1 8-core RISC-V CM4-compatible module with 2 TOPS AI and 16GB LPDDR4.

RISC-V 2 TOPS CM4

Luckfox Pico AI Camera Boards

RV1106/RV1103 Linux AI camera boards with 0.5-1 TOPS NPU and up-to-5MP ISP.

1 TOPS RV1106 MIPI CSI

POET Teralight 1.6T Optical Engine

1.6T DR8/2xFR4 optical engines plus Blazar ECL light source for CPO; 4-laser transmit design.

Hisense Broadband 1.6T OSFP

1.6T DR8 RTLR under 21W, 800G 2xFR4 LPO, silicon photonics, dual MPO-12/APC, 500m.

1.6T DR8 800G LPO SiPh

Keysight AI DC 1.6T Test

INPT-1600GE 1.6T Ethernet test platform plus ITS software validating 100GE-1600GE AI fabrics.

Vecow ECX-3000 PEG

Intel Core 200S AI workstation, up to 4 PCIe slots, 4x 2.5G PoE+, 4 front SSDs, -20-45C.

4x PCIe 4x PoE+ 65W

Synopsys HBM4/4E IP

HBM4/4E PHY+Controller IP, 12 Gbps/pin, over 3 TBps; first HBM4 test chip at 9.2 Gbps.

HBM4 12 Gbps 3 TBps

GOWIN Arora V FPGA (GW5A / GW5AT)

SRAM-based FPGA with up to 138K LUTs, 12.5 Gbps CDR SerDes, PCIe 2.0 hardcore and MIPI D-PHY hard PHY.

12.5 Gbps SerDes PCIe 2.0 138K LUTs

Matrox LUMA Pro Video Wall Cards

Intel Arc powered PCIe 4.0 cards driving 4-8 DisplayPort 2.0 outputs, up to 4 cards framelocked per system.

8x 5K60 12GB GDDR6 DP 2.0

Matrox Mura IPX & Mura C4K Capture

4-channel 12G-SDI capture and IP encode/decode plus 4x HDMI 2.1 4K60 capture with H.264 level 5.2 and SRT streaming.

H.264 L5.2 4x HDMI 2.1 Fanless

Pleora iPORT External Frame Grabbers

Convert Camera Link, analog and block cameras to GigE Vision, 10 GigE or USB3 Vision with no PCIe slot needed.

10 GigE USB3 Vision eBUS SDK

Anritsu MT1040A Network Master Pro

Portable B5-size tester spanning 10M to 400G Ethernet, OTN/SDH, eCPRI and 1G-16G Fibre Channel.

400G ETH OTN/SDH eCPRI

Rosewill Rackmount Server Chassis

2U-4U chassis with 12 Gbps hot-swap SATA/SAS backplanes, E-ATX/SSI-EEB support and high-airflow PWM cooling.

RSV-H424 12 Gbps SAS 2U-4U

ASUS Ascent QN10 — 80 TOPS Snapdragon X2 Elite AI Mini PC

ASUS Ascent QN10 mini PC on Qualcomm Snapdragon X2 Elite: 18-core Oryon CPU, Adreno X2-90 graphics and an 80 TOPS Hexagon NPU with up to 32 GB LPDDR5X-9600 and dual M.2 NVMe in a 130 x 130 x 40 mm chassis.

80 TOPS NPU Snapdragon X2 Elite Wi-Fi 7

Aetina AIB-AT78/68 — Jetson Thor T5000 / T4000 Edge AI Platform

Aetina autonomous edge AI platform on NVIDIA Jetson Thor: up to 2,070 FP4 TFLOPS, 128 GB LPDDR5X, a 14-core Arm Neoverse CPU, up to 20 camera inputs and 8 lanes of PCIe Gen5 for physical AI.

2070 FP4 TFLOPS 128 GB LPDDR5X Jetson Thor

Cincoze MD-3000 — Scalable DIN-Rail Machine Vision Computer

Cincoze MD-3000 DIN-rail computer: 14th Gen Intel Core at 35-65 W with up to 96 GB DDR5 ECC at 5600 MHz, dual NVMe slots and 2/4/6-slot SED expansion decks in a 150 mm PLC-height chassis.

150 mm DIN-Rail 96 GB ECC 6-Slot SED

Phison Pascari D206V — 245.76 TB PCIe Gen5 Data Center SSD

Phison Pascari D206V read-intensive NVMe SSD packing up to 245.76 TB into a single U.2 drive at 14 GB/s sequential read, with NVMe 2.0 dual-port support, full power-loss protection and TCG Opal 2.0.

245.76 TB PCIe Gen5 14 GB/s

ASUS PE6000G — Rugged Edge AI GPU Computer (450 W GPU)

ASUS IoT PE6000G rugged edge AI computer supporting a full-height 450 W GPU on the Intel R680E platform with 64 GB DDR5, four 2.5-inch SSDs, -20 to 60 C operation and 8-48 VDC ignition power.

450 W GPU -20-60 C MIL-STD-810H

Aetina AIP-FR68S — Edge AI Agent Workstation, 3x Cloud AI 100 Ultra

Aetina AIP-FR68S edge AI agent workstation with a PCIe Gen5 switch driving up to three Qualcomm Cloud AI 100 Ultra cards at 870 TOPS each, or NVIDIA RTX PRO GPUs, on a built-in 2000 W supply.

870 TOPS/Card PCIe Gen5 Switch 300B LLM

Chenbro NVIDIA MGX AI Server Chassis

1U/2U liquid-cooled GB200 NVL72 and NVL36 compute trays plus 4U air-cooled chassis for up to eight double-width GPGPUs.

GB200 NVL72 8-GPU NVIDIA MGX

InWin IW-RG658 PRO 6.5U 8-GPU Chassis

Eight NVIDIA RTX 6000 Pro GPUs on dual 5th Gen Xeon Scalable, 32 DDR5 slots, 16 PCIe Gen5 slots and 9.6 kW Titanium CRPS.

RTX 6000 Pro 9600W E1.S NVMe

SilverStone RM44 4U Rackmount Chassis

SSI-EEB steel 4U with 424 mm GPU clearance, 360 mm front radiator support, eight PCI slots and PS2 or redundant power.

SSI-EEB 424mm GPU 360mm Rad

ATP Industrial Enterprise Gen4 NVMe SSD

N651Sie in M.2, U.2 and E1.S: 240 GB to 7.68 TB, PCIe Gen4 x4 up to 6,450 MB/s, -40C to 85C and 1 DWIP with Level 4 PLP.

Gen4 NVMe -40C to 85C 7.68TB