Encrypted Computing at Scale

Solving today’s critical data privacy and security problems requires advanced solutions. Fully Homomorphic Encryption (FHE) and Artificial Intelligence (AI) are being deployed to support ever-increasing data sets. Keeping data encrypted “in use” presents new opportunities to extract valuable insights while protecting data privacy in highly regulated markets.

Cornami’s breakthrough FracTLcore Computing Fabric seamlessly scales from thousands of cores on a single chip to millions of cores across systems, transparent to the software. This delivers a “reboot” to the computing industry to scale performance without penalties. 

Cornami’s unique technology can scale to deliver real-time computing with low power consumption, latency, and cost; delivering the highest silicon efficiency.

Scale to deliver real-time Fully Homomorphic Encryption (FHE) for quantum-secure privacy-preserving computing on encrypted data sets vital for cloud security. 

Process privacy-preserving machine learning (PPML) models using data from multiple owners without ever exposing the underlying plaintext data.

• Data owners maintain control of their decryption key.
• Data privacy, confidentiality, and control are all preserved.

Dial the size of the FracTLcore Computing Fabric up or down to meet

• performance vs. cost requirements. 
• Add additional systems for penalty-free linear scaling up to 64 million cores.

Unique FracTLcore Computing Fabric Technology 

• Share-nothing Architecture – avoids bottlenecks inherent in legacy computing architectures.
• Massively Parallel – supports all forms of parallelism without restriction (SIMD, MIMD, Data Flow, Pipelining, Systolic, MapReduce, etc.)
• Streaming processing – Streaming data architecture fully exploits all the parallelism and pipelining available within the application domain.

Leading Industry Frameworks Supported with Seamless Software Integration

• Industry-standard frameworks supported to accelerate software without code changes. 
• TruStream SDK is available to support the programming of new applications.

Encrypted Computing at Plaintext Speed 

Linear regression is a supervised machine-learning algorithm that learns from the labeled datasets and maps the data points to the most optimized linear functions. These data points can be used for prediction on new datasets.

Cornami enables commercially viable FHE for data collaboration without disclosing the underlying confidential plaintext data. This includes protecting high-value AI and ML models. The data remains encrypted during use. The availability of practical FHE 

introduces a paradigm shift to the cloud computing marketplace and the entire information processing industry. This enables a wide range of services to become accessible while maintaining full confidentiality of the data, the application, and service results. Cornami’s technology can linearly scale from thousands of cores on a chip to 64 million cores across systems via its FracTLcore 

Computing Fabric. This is transparent to the software and allows end users to dial up or down the FracTLcores to balance performance and cost requirements.

Specifications

FracTLcore Computing Fabric Mx⁸⁰ System

• Scale system up to 10 Boards for full capacity
• Total Computing Cores: Up to 80,000 per system

Performance

• Mx⁸⁰ delivers Real-Time FHE PPML 
• 123 Dense FP64 TFLOPs 
• 492 Dense FP32 TFLOPs  
• 983 Dense FP16 TFLOPs
• 1,966 ML int8 TFLOPs      

Dimensions 

• Height: 7” (178mm) 
• Width: 17.2” (437mm)
• Depth: 29” (737mm)

System Power Usage

• 3.2 kW 

Drive Bays / Storage (Hot-Swap)

• 24x 2.5” NVMe/SATA/SAS drive bays
• (8×2.5” NVMe dedicated)
• M.2 2 M.2 NVMe OR 2 M.2 SATA3

Software: Linux OS

• SCIFR API*
• Supports Standard Industry Frameworks with no code changes
• TruStream SDK for programming of new applications

Operating Environment

• Operating Temperature: 10°C: 35°C (50°F: 95°F)

Specifications are preliminary and subject to change without notice. * SCIFR (Secure Computing Interface FRamework)