AIP-FR68-A2
The AIP-FR68-A2 is a configurable industrial AI inference and training workstation from Aetina's MegaEdge PCIe Series, built around the Intel R680E platform and supporting Intel 12th, 13th and 14th Generation Core i9, i7 and i5 processors at up to 65 W TDP. With four DDR5 U-DIMM slots supporting up to 192 GB of DDR5 memory, seven PCIe expansion slots and a patented screwless chassis designed for simplified field maintenance, the AIP-FR68-A2 provides the physical architecture to run on-premises generative AI, large language model inference and computer vision workloads at the edge — without cloud infrastructure or rackmount server facilities.
Designed for GPU-intensive AI workloads in conditioned industrial environments, the AIP-FR68-A2 accommodates an NVIDIA RTX PRO 6000 Blackwell Max-Q passive workstation GPU in the primary PCIe Gen4 x16 slot and up to two Qualcomm Cloud AI 100 Ultra inference accelerators (870 TOPS INT8 each) for high-density AI inference at the edge. A built-in 10 GbE + three 2.5 GbE network stack with optional PoE PSE on two ports, out-of-band (OOB) management via InnoAgent and compatibility with Aetina's EdgeEye remote monitoring platform make it a fully managed AI workstation node deployable across distributed industrial, enterprise and smart city infrastructure.
Key Features
- Seven PCIe Expansion Slots — Maximum GPU and Accelerator Flexibility: One PCIe Gen4 x16, one PCIe Gen4 x8, two PCIe Gen3 x4 and one PCIe Gen3 x1 full-size slots accommodate NVIDIA workstation GPUs and Qualcomm Cloud AI 100 Ultra inference cards; two additional M.2 sockets (E-Key 2230 and B-Key 3052) provide wireless and cellular expansion without occupying PCIe slots
- NVIDIA RTX PRO 6000 Blackwell Max-Q and Qualcomm AI 100 Ultra Support: Primary slot validated for the NVIDIA RTX PRO 6000 Blackwell Max-Q passive GPU for AI training and computer vision; secondary slot supports up to two Qualcomm Cloud AI 100 Ultra cards at 870 TOPS INT8 each — delivering up to 1,740 TOPS of dedicated AI inference alongside the GPU without thermal interference in passive cooling mode (0–50 °C operating range)
- Up to 192 GB DDR5 Across Four U-DIMM Slots: Four 288-pin DDR5 sockets support large model residency for LLM inference, AI training checkpoints and multi-modal AI pipelines that cannot be partitioned across smaller memory pools; Intel R680E chipset provides the PCIe Gen4 bandwidth to sustain GPU memory transfer rates demanded by Blackwell-generation workstation GPUs
- 10 GbE + Triple 2.5 GbE Networking with Optional PoE PSE: Marvell AQC113-based 10G BASE-T port handles high-throughput AI data ingestion; three 2.5 GbE ports (two optionally equipped with 802.3at PSE at 30W per channel) power IP cameras, sensors and edge switches directly from the workstation — reducing external PoE switch count in compact AI inference deployments
- Patented Screwless Chassis with OOB Management: Aetina's patented screwless enclosure design enables tool-free access to GPU cards, DIMMs and storage drives for servicing without specialist equipment; built-in Innodisk InnoAgent OOB management enables remote BIOS configuration, hardware health monitoring and power control independent of the operating system state, reducing on-site maintenance visits for distributed deployments
- Industrial I/O and Quad 2.5" SATA RAID Storage: Three RS-232/422/485 COM ports (one RJ45 RS-485, two DB9), 8-bit GPIO, 8× USB 3.2 Gen2 and one USB-C cover industrial serial, discrete I/O and high-speed peripheral requirements; four 2.5" SATA III drive bays with RAID 0/1/5/10 and two PCIe Gen4 x4 NVMe M.2 slots provide flexible, redundant storage for AI model repositories and training datasets
- NVIDIA-Certified System with Multi-OS Support: AIP-FR68 carries NVIDIA Certified System (NCS) status, guaranteeing validated performance, reliability and security for NVIDIA GPU workloads; supports Windows 10 IoT LTSC, Windows 11, Ubuntu 20.04 and Ubuntu 22.04 with TPM 2.0 for hardware security and encrypted storage
Application Use Cases
On-Premises Generative AI and Large Language Model Inference
The AIP-FR68-A2 is designed for organisations that need to run large language models, retrieval-augmented generation (RAG) pipelines and generative AI applications on-premises rather than in public cloud infrastructure. The combination of up to 192 GB DDR5 for model residency, Qualcomm Cloud AI 100 Ultra inference accelerators for throughput-optimised token generation, and NVIDIA RTX PRO 6000 for mixed training and inference workloads covers the full GenAI stack from embedding generation to response serving — within a single field-deployable unit operating from a DC 24–48 V supply.
Industrial AI Training at the Edge
For manufacturing and process automation programmes requiring localised model training on production data — without transferring proprietary datasets to cloud infrastructure — the AIP-FR68-A2 provides the GPU memory, PCIe bandwidth and storage RAID architecture of a small server in a wall- or desk-mountable workstation chassis. The four 2.5" SATA drive bays with RAID 5 support training dataset resilience, and the PCIe Gen4 x16 primary slot sustains the GPU-to-CPU bandwidth required by full fine-tuning and transfer learning workloads.
AI-Enabled Computer Vision and Video Analytics
Eight USB 3.2 Gen2 ports connect USB3 Vision cameras and frame grabbers for multi-stream video analytics, whilst the optional PoE PSE on two 2.5 GbE ports powers IP cameras directly from the workstation. The NVIDIA RTX PRO 6000 Blackwell GPU runs NVIDIA DeepStream video analytics pipelines at scale, and the 10 GbE uplink handles the bandwidth demands of high-frame-rate multi-camera ingestion without network saturation in factory and surveillance installations.
Smart Factory and Industrial Automation AI Server
Industrial serial ports (RS-232/422/485), 8-bit GPIO and DC 24–48 V power input bridge the gap between enterprise AI compute and the fieldbus connectivity requirements of production environments. The EdgeEye platform provides centralised monitoring of deployed AIP-FR68-A2 units across factory sites, alerting operations teams to hardware events, GPU utilisation anomalies and thermal threshold breaches without manual inspection rounds. RAID storage redundancy protects AI model files and inference logs from single-drive failure during continuous 24/7 operation.
Enterprise Edge AI for Healthcare, Retail and Security
NVIDIA Certified System status and compatibility with Windows 10 IoT LTSC and Windows 11 integrate the AIP-FR68-A2 into enterprise IT procurement and compliance frameworks for healthcare diagnostics AI, retail analytics and physical security inference. TPM 2.0 hardware security satisfies data-at-rest encryption requirements for regulated environments, and the OOB management capability allows IT teams to apply firmware updates and security patches across distributed deployments without dispatching on-site engineers.
Why Choose the AIP-FR68-A2
Server-Class AI Compute in a Field-Deployable Workstation
The AIP-FR68-A2 addresses the gap between compact AI box PCs — which lack the PCIe expansion and memory capacity for serious training and LLM inference workloads — and rackmount servers, which require controlled data-centre environments and AC power infrastructure. At 340 × 215 × 279 mm with DC 24–48 V input, wall or desk mounting and industrial serial I/O, it installs in factory control rooms, retail back-offices and edge data hubs where a 2U server cannot go. The combination of up to 1,740 TOPS from dual Qualcomm AI 100 Ultra cards, NVIDIA RTX PRO 6000 Blackwell GPU support and 192 GB DDR5 makes it the most compute-dense platform in this form-factor range.
The Screwless Chassis — A Maintenance Advantage at Scale
Aetina's patented screwless enclosure design is not a cosmetic feature. In deployments where dozens or hundreds of AIP-FR68-A2 units are fielded across distributed sites, the ability to replace a GPU, add a DIMM or swap a drive without tools — without specialist field technicians and without a full service call — translates directly into lower maintenance cost and shorter downtime per incident. Combined with OOB remote management via InnoAgent, most software-level interventions can be completed remotely, reserving physical access for hardware changes only.
NVIDIA-Certified, Multi-Framework and Multi-OS from Day One
NVIDIA Certified System status means the AIP-FR68-A2 has been validated by NVIDIA for performance, stability and security with NVIDIA GPU workloads — not self-certified by the manufacturer. For enterprise procurement teams and AI programme managers, this certification eliminates a validation step from the project timeline and provides a clear escalation path for GPU-related technical issues. Windows 11, Ubuntu 22.04, TensorRT, DeepStream and the broad Qualcomm AI SDK ecosystem are all supported from the first boot, without custom driver work or OS porting effort.
Frequently Asked Questions
What processors does the AIP-FR68-A2 support?
The AIP-FR68-A2 supports Intel 12th, 13th and 14th Generation Core i9, i7 and i5 processors at up to 65 W TDP on the Intel R680E chipset. The system is supplied as a bare platform; the CPU, memory, storage, GPU and power supply are configured to application requirements at the time of order.
Which GPU and AI accelerator cards are compatible with the AIP-FR68-A2?
The primary PCIe Gen4 x16 slot supports the NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition GPU in passive cooling mode. The PCIe Gen4 x8 slot supports Qualcomm Cloud AI 100 Ultra inference accelerators (870 TOPS INT8 each); up to two AI 100 Ultra cards can be installed simultaneously, delivering up to 1,740 TOPS of dedicated inference throughput alongside the primary GPU.
What is the maximum memory configuration of the AIP-FR68-A2?
The AIP-FR68-A2 supports up to 192 GB of DDR5 memory across four 288-pin U-DIMM slots (maximum 64 GB per slot). DDR5 provides the memory bandwidth required to sustain large language model inference and AI training workloads at the GPU performance levels delivered by Blackwell-generation workstation graphics.
Does the AIP-FR68-A2 support remote management?
Yes. The AIP-FR68-A2 includes built-in out-of-band (OOB) management via Innodisk InnoAgent, enabling remote BIOS access, hardware health monitoring and power control independent of operating system state. It is also compatible with Aetina's EdgeEye platform for centralised fleet monitoring, alerting and management of multiple deployed units.
What operating systems and AI frameworks does the AIP-FR68-A2 support?
The AIP-FR68-A2 supports Windows 10 IoT LTSC, Windows 11, Ubuntu 20.04 and Ubuntu 22.04. It is an NVIDIA Certified System (NCS) and is compatible with NVIDIA TensorRT, DeepStream, the CUDA toolkit and the Qualcomm AI SDK. TPM 2.0 is built in for hardware security and encrypted storage across all supported operating systems.
The AIP-FR68-A2 is a configurable industrial AI inference and training workstation from Aetina's MegaEdge PCIe Series, built around the Intel R680E platform and supporting Intel 12th, 13th and 14th Generation Core i9, i7 and i5 processors at up to 65 W TDP. With four DDR5 U-DIMM slots supporting up to 192 GB of DDR5 memory, seven PCIe expansion slots and a patented screwless chassis designed for simplified field maintenance, the AIP-FR68-A2 provides the physical architecture to run on-premises generative AI, large language model inference and computer vision workloads at the edge — without cloud infrastructure or rackmount server facilities.
Designed for GPU-intensive AI workloads in conditioned industrial environments, the AIP-FR68-A2 accommodates an NVIDIA RTX PRO 6000 Blackwell Max-Q passive workstation GPU in the primary PCIe Gen4 x16 slot and up to two Qualcomm Cloud AI 100 Ultra inference accelerators (870 TOPS INT8 each) for high-density AI inference at the edge. A built-in 10 GbE + three 2.5 GbE network stack with optional PoE PSE on two ports, out-of-band (OOB) management via InnoAgent and compatibility with Aetina's EdgeEye remote monitoring platform make it a fully managed AI workstation node deployable across distributed industrial, enterprise and smart city infrastructure.
Key Features
- Seven PCIe Expansion Slots — Maximum GPU and Accelerator Flexibility: One PCIe Gen4 x16, one PCIe Gen4 x8, two PCIe Gen3 x4 and one PCIe Gen3 x1 full-size slots accommodate NVIDIA workstation GPUs and Qualcomm Cloud AI 100 Ultra inference cards; two additional M.2 sockets (E-Key 2230 and B-Key 3052) provide wireless and cellular expansion without occupying PCIe slots
- NVIDIA RTX PRO 6000 Blackwell Max-Q and Qualcomm AI 100 Ultra Support: Primary slot validated for the NVIDIA RTX PRO 6000 Blackwell Max-Q passive GPU for AI training and computer vision; secondary slot supports up to two Qualcomm Cloud AI 100 Ultra cards at 870 TOPS INT8 each — delivering up to 1,740 TOPS of dedicated AI inference alongside the GPU without thermal interference in passive cooling mode (0–50 °C operating range)
- Up to 192 GB DDR5 Across Four U-DIMM Slots: Four 288-pin DDR5 sockets support large model residency for LLM inference, AI training checkpoints and multi-modal AI pipelines that cannot be partitioned across smaller memory pools; Intel R680E chipset provides the PCIe Gen4 bandwidth to sustain GPU memory transfer rates demanded by Blackwell-generation workstation GPUs
- 10 GbE + Triple 2.5 GbE Networking with Optional PoE PSE: Marvell AQC113-based 10G BASE-T port handles high-throughput AI data ingestion; three 2.5 GbE ports (two optionally equipped with 802.3at PSE at 30W per channel) power IP cameras, sensors and edge switches directly from the workstation — reducing external PoE switch count in compact AI inference deployments
- Patented Screwless Chassis with OOB Management: Aetina's patented screwless enclosure design enables tool-free access to GPU cards, DIMMs and storage drives for servicing without specialist equipment; built-in Innodisk InnoAgent OOB management enables remote BIOS configuration, hardware health monitoring and power control independent of the operating system state, reducing on-site maintenance visits for distributed deployments
- Industrial I/O and Quad 2.5" SATA RAID Storage: Three RS-232/422/485 COM ports (one RJ45 RS-485, two DB9), 8-bit GPIO, 8× USB 3.2 Gen2 and one USB-C cover industrial serial, discrete I/O and high-speed peripheral requirements; four 2.5" SATA III drive bays with RAID 0/1/5/10 and two PCIe Gen4 x4 NVMe M.2 slots provide flexible, redundant storage for AI model repositories and training datasets
- NVIDIA-Certified System with Multi-OS Support: AIP-FR68 carries NVIDIA Certified System (NCS) status, guaranteeing validated performance, reliability and security for NVIDIA GPU workloads; supports Windows 10 IoT LTSC, Windows 11, Ubuntu 20.04 and Ubuntu 22.04 with TPM 2.0 for hardware security and encrypted storage
Application Use Cases
On-Premises Generative AI and Large Language Model Inference
The AIP-FR68-A2 is designed for organisations that need to run large language models, retrieval-augmented generation (RAG) pipelines and generative AI applications on-premises rather than in public cloud infrastructure. The combination of up to 192 GB DDR5 for model residency, Qualcomm Cloud AI 100 Ultra inference accelerators for throughput-optimised token generation, and NVIDIA RTX PRO 6000 for mixed training and inference workloads covers the full GenAI stack from embedding generation to response serving — within a single field-deployable unit operating from a DC 24–48 V supply.
Industrial AI Training at the Edge
For manufacturing and process automation programmes requiring localised model training on production data — without transferring proprietary datasets to cloud infrastructure — the AIP-FR68-A2 provides the GPU memory, PCIe bandwidth and storage RAID architecture of a small server in a wall- or desk-mountable workstation chassis. The four 2.5" SATA drive bays with RAID 5 support training dataset resilience, and the PCIe Gen4 x16 primary slot sustains the GPU-to-CPU bandwidth required by full fine-tuning and transfer learning workloads.
AI-Enabled Computer Vision and Video Analytics
Eight USB 3.2 Gen2 ports connect USB3 Vision cameras and frame grabbers for multi-stream video analytics, whilst the optional PoE PSE on two 2.5 GbE ports powers IP cameras directly from the workstation. The NVIDIA RTX PRO 6000 Blackwell GPU runs NVIDIA DeepStream video analytics pipelines at scale, and the 10 GbE uplink handles the bandwidth demands of high-frame-rate multi-camera ingestion without network saturation in factory and surveillance installations.
Smart Factory and Industrial Automation AI Server
Industrial serial ports (RS-232/422/485), 8-bit GPIO and DC 24–48 V power input bridge the gap between enterprise AI compute and the fieldbus connectivity requirements of production environments. The EdgeEye platform provides centralised monitoring of deployed AIP-FR68-A2 units across factory sites, alerting operations teams to hardware events, GPU utilisation anomalies and thermal threshold breaches without manual inspection rounds. RAID storage redundancy protects AI model files and inference logs from single-drive failure during continuous 24/7 operation.
Enterprise Edge AI for Healthcare, Retail and Security
NVIDIA Certified System status and compatibility with Windows 10 IoT LTSC and Windows 11 integrate the AIP-FR68-A2 into enterprise IT procurement and compliance frameworks for healthcare diagnostics AI, retail analytics and physical security inference. TPM 2.0 hardware security satisfies data-at-rest encryption requirements for regulated environments, and the OOB management capability allows IT teams to apply firmware updates and security patches across distributed deployments without dispatching on-site engineers.
Why Choose the AIP-FR68-A2
Server-Class AI Compute in a Field-Deployable Workstation
The AIP-FR68-A2 addresses the gap between compact AI box PCs — which lack the PCIe expansion and memory capacity for serious training and LLM inference workloads — and rackmount servers, which require controlled data-centre environments and AC power infrastructure. At 340 × 215 × 279 mm with DC 24–48 V input, wall or desk mounting and industrial serial I/O, it installs in factory control rooms, retail back-offices and edge data hubs where a 2U server cannot go. The combination of up to 1,740 TOPS from dual Qualcomm AI 100 Ultra cards, NVIDIA RTX PRO 6000 Blackwell GPU support and 192 GB DDR5 makes it the most compute-dense platform in this form-factor range.
The Screwless Chassis — A Maintenance Advantage at Scale
Aetina's patented screwless enclosure design is not a cosmetic feature. In deployments where dozens or hundreds of AIP-FR68-A2 units are fielded across distributed sites, the ability to replace a GPU, add a DIMM or swap a drive without tools — without specialist field technicians and without a full service call — translates directly into lower maintenance cost and shorter downtime per incident. Combined with OOB remote management via InnoAgent, most software-level interventions can be completed remotely, reserving physical access for hardware changes only.
NVIDIA-Certified, Multi-Framework and Multi-OS from Day One
NVIDIA Certified System status means the AIP-FR68-A2 has been validated by NVIDIA for performance, stability and security with NVIDIA GPU workloads — not self-certified by the manufacturer. For enterprise procurement teams and AI programme managers, this certification eliminates a validation step from the project timeline and provides a clear escalation path for GPU-related technical issues. Windows 11, Ubuntu 22.04, TensorRT, DeepStream and the broad Qualcomm AI SDK ecosystem are all supported from the first boot, without custom driver work or OS porting effort.
Frequently Asked Questions
What processors does the AIP-FR68-A2 support?
The AIP-FR68-A2 supports Intel 12th, 13th and 14th Generation Core i9, i7 and i5 processors at up to 65 W TDP on the Intel R680E chipset. The system is supplied as a bare platform; the CPU, memory, storage, GPU and power supply are configured to application requirements at the time of order.
Which GPU and AI accelerator cards are compatible with the AIP-FR68-A2?
The primary PCIe Gen4 x16 slot supports the NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition GPU in passive cooling mode. The PCIe Gen4 x8 slot supports Qualcomm Cloud AI 100 Ultra inference accelerators (870 TOPS INT8 each); up to two AI 100 Ultra cards can be installed simultaneously, delivering up to 1,740 TOPS of dedicated inference throughput alongside the primary GPU.
What is the maximum memory configuration of the AIP-FR68-A2?
The AIP-FR68-A2 supports up to 192 GB of DDR5 memory across four 288-pin U-DIMM slots (maximum 64 GB per slot). DDR5 provides the memory bandwidth required to sustain large language model inference and AI training workloads at the GPU performance levels delivered by Blackwell-generation workstation graphics.
Does the AIP-FR68-A2 support remote management?
Yes. The AIP-FR68-A2 includes built-in out-of-band (OOB) management via Innodisk InnoAgent, enabling remote BIOS access, hardware health monitoring and power control independent of operating system state. It is also compatible with Aetina's EdgeEye platform for centralised fleet monitoring, alerting and management of multiple deployed units.
What operating systems and AI frameworks does the AIP-FR68-A2 support?
The AIP-FR68-A2 supports Windows 10 IoT LTSC, Windows 11, Ubuntu 20.04 and Ubuntu 22.04. It is an NVIDIA Certified System (NCS) and is compatible with NVIDIA TensorRT, DeepStream, the CUDA toolkit and the Qualcomm AI SDK. TPM 2.0 is built in for hardware security and encrypted storage across all supported operating systems.
Get a Quote
