Mx Series FracTLcore Server Systems
Cornami's Mx Series FracTLcore Server Systems are a next-generation computing platform designed for real-time, scalable performance across AI, data analytics, and secure computing workloads.
Publisher review
Cornami's Mx Series FracTLcore Server Systems are a next-generation computing platform designed for real-time, scalable performance across AI, data analytics, and secure computing workloads. They target enterprises and developers needing high-throughput, low-latency processing for encrypted AI, zero-knowledge proofs (ZKPs), fully homomorphic encryption (FHE), and blockchain applications. The systems are purpose-built for edge-to-cloud deployments, handling data-intensive tasks like private LLM inference, generative AI, and smart contract execution on encrypted inputs. Cornami positions the Mx Series as a solution for organizations that require both massive parallelism and cryptographic security without compromising on power efficiency or scalability.
The core technology is the FracTLcore Compute Fabric, a dynamically reconfigurable, instruction-set programmable architecture that scales from thousands to millions of cores. Unlike traditional designs, it eliminates shared memory and bus bottlenecks by avoiding cache subsystems, enabling linear scalability with no performance penalty as cores increase. The fabric supports all 55 forms of parallelism and can dynamically reconfigure cores based on workload needs—streaming execution, native security, or cryptographic acceleration. In a strategic collaboration with Ingonyama, integration with the ICICLE cryptographic library achieves 10× to 100× improvements in throughput and energy efficiency for ZKP generation and verification compared to GPU-based accelerators. The system also accelerates FHE workloads, ensuring quantum-secure privacy-preserving computation on encrypted data without exposing plaintext.
In the high-performance computing and cryptographic acceleration market, the Mx Series competes with GPU-based solutions from NVIDIA (e.g., A100, H100) and ASIC-based accelerators from companies like Intel (e.g., SGX) and specialized FPGA providers. Cornami's architecture claims superior linear scalability and energy efficiency over GPUs, which suffer from SIMD limitations and memory bottlenecks for irregular workloads like ZKPs. However, the Mx Series is a relatively new entrant with limited ecosystem maturity compared to NVIDIA's CUDA and established cryptographic libraries. Its reconfigurable nature offers flexibility over fixed-function ASICs, but may not match the raw throughput of dedicated hardware for specific algorithms.
Key trade-offs include a higher upfront investment for custom hardware integration, as the Mx Series requires specialized software optimization to leverage its reconfigurable fabric fully. The platform's reliance on Cornami's proprietary architecture may lock users into a single vendor's toolchain, limiting portability. While it excels at cryptographic and parallel workloads, general-purpose computing tasks may see less benefit compared to traditional CPUs or GPUs. Additionally, the lack of publicly available pricing and limited third-party benchmarks make it difficult to validate performance claims independently.
How it works
-
Dynamically reconfigurable cores
Cores can be reconfigured on-the-fly to match workload demands, optimizing for streaming execution, security, or parallelism.
-
Linear scalability to millions of cores
Architecture scales from thousands to millions of cores with no performance penalty, avoiding shared memory bottlenecks.
-
Fully Homomorphic Encryption support
Enables quantum-secure computations on encrypted data without exposing plaintext, critical for regulated industries.
-
10x-100x ZKP acceleration
Integration with Ingonyama's ICICLE library achieves 10× to 100× throughput and energy improvements over GPUs for zero-knowledge proofs.
-
No cache subsystems
Eliminates cache coherence overhead, reducing latency and power consumption for massively parallel pipelining.
-
Supports all 55 forms of parallelism
Handles diverse parallel patterns including data, task, and instruction-level parallelism, adapting to complex algorithms.
-
Edge-to-cloud scalability
Designed for deployment from edge devices to cloud data centers, maintaining real-time performance across environments.
Strengths and trade-offs
Strengths
- Achieves 10× to 100× improvements in throughput and energy efficiency for ZKP workloads compared to GPU-based accelerators.
- Scales linearly from thousands to millions of cores without performance degradation, unlike traditional shared-memory architectures.
- Supports Fully Homomorphic Encryption (FHE) for quantum-secure privacy-preserving computation on encrypted data sets.
- Dynamically reconfigurable cores optimize power and performance for specific workloads, including streaming execution and native security.
Trade-offs
- Requires specialized software optimization to fully utilize the reconfigurable fabric, increasing development time and cost.
- Proprietary architecture may create vendor lock-in, limiting portability of applications to other hardware platforms.
- Limited third-party benchmarks and ecosystem maturity compared to established GPU and CPU solutions from NVIDIA and Intel.
- General-purpose computing tasks may not see significant performance gains, as the design is optimized for parallel and cryptographic workloads.
Pricing context
Not specified in sources.
Getting started with Mx Series FracTLcore Server Systems
-
Request access to Mx Series
Contact Cornami's sales team through their website to request access to the Mx Series FracTLcore Server Systems. Provide details about your workload requirements, such as AI, ZKP, or FHE tasks, to receive evaluation units or cloud-based trial access.
-
Set up development environment
Install the Cornami software development kit (SDK) and toolchain on your development machine. Follow the provided documentation to configure the environment for your operating system and connect to the Mx Series hardware or simulator.
-
Configure FracTLcore fabric for workload
Use the SDK to define the core configuration for your specific workload, such as streaming execution or cryptographic acceleration. Set parallelism parameters and enable features like FHE or ZKP support to match your application's needs.
-
Run a sample ZKP acceleration task
Load a sample zero-knowledge proof generation or verification task from the SDK examples. Execute it on the Mx Series system, monitoring throughput and energy efficiency to verify the 10x-100x improvement over GPU baselines.
-
Deploy workload to production
Integrate your optimized application into your production pipeline, deploying the Mx Series system at the edge or in the cloud. Schedule regular performance checks and update core configurations as workload demands evolve.
Frequently Asked Questions
What are Cornami Mx Series FracTLcore Server Systems?
They are next-generation computing platforms for real-time AI, data analytics, and secure workloads. They use a dynamically reconfigurable FracTLcore Compute Fabric that scales from thousands to millions of cores, targeting encrypted AI, zero-knowledge proofs, and blockchain applications.
How does the FracTLcore architecture achieve linear scalability?
It eliminates shared memory and bus bottlenecks by avoiding cache subsystems. This design allows scaling from thousands to millions of cores with no performance penalty, supporting all 55 forms of parallelism and enabling linear scalability for massively parallel workloads.
What performance gains does the Mx Series offer for zero-knowledge proofs?
Through integration with Ingonyama's ICICLE library, it achieves 10× to 100× improvements in throughput and energy efficiency for ZKP generation and verification compared to GPU-based accelerators, making it highly efficient for cryptographic tasks.
Can the Mx Series handle fully homomorphic encryption workloads?
Yes, it supports FHE for quantum-secure privacy-preserving computation on encrypted data without exposing plaintext. This capability is critical for regulated industries needing secure data processing while maintaining confidentiality.
How does the Mx Series compare to NVIDIA GPUs for cryptographic tasks?
It claims superior linear scalability and energy efficiency over GPUs, which suffer from SIMD limitations and memory bottlenecks for irregular workloads like ZKPs. However, it has a less mature ecosystem compared to NVIDIA's CUDA and established libraries.
What are the main drawbacks of the Mx Series server systems?
It requires specialized software optimization to fully use the reconfigurable fabric, increasing development time. The proprietary architecture may cause vendor lock-in, and limited third-party benchmarks make independent validation of performance claims difficult.
Alternatives
How Mx Series FracTLcore Server Systems compares
Direct head-to-head against 2 competitors. Picked by 7wData.
Mx Series FracTLcore Server Systems
- Pricing
- Not specified in sources.
- Target
- Cornami's Mx Series FracTLcore Server Systems are a next-generation computing platform designed for real-time, scalable performance across AI, data analytics, and secure computing workloads.
- Strength
- Achieves 10× to 100× improvements in throughput and energy efficiency for ZKP workloads compared to GPU-based accelerators.
- Watch for
- Requires specialized software optimization to fully utilize the reconfigurable fabric, increasing development time and cost.
NVIDIA
- Pricing
- Custom/Contact sales; GPU cloud rentals from ~$2.50/GPU/hour (H100).
- Target
- AI inference, training, and HPC workloads in data centers and cloud.
- Deployment
- On-prem, cloud, edge.
- Strength
- Dominant GPU ecosystem with CUDA and massive developer community.
- Watch for
- Vendor lock-in via CUDA; pricing escalation on high-demand GPUs.
Dell
- Pricing
- Custom/Contact sales; PowerEdge servers start ~$10,000.
- Target
- Enterprise AI server deployments with integrated support.
- Deployment
- On-prem, data center.
- Strength
- End-to-end enterprise support and global supply chain.
- Watch for
- Long lead times for custom configurations.
User reviews
No user reviews yet. Be the first to write one.
Sources
Reporting on this tool draws on these publicly available sources.