Shattering the Memory Wall
Shattering the Memory Wall Stefano Giaconi
The First Fully Clockless Interconnect Ecosystem
About Us
Chronos Tech is US-based semiconductor IP design company developing game-changing IPs and fabric interconnect solutions to enable next-generation intelligent SoCs.
We target a wide range of applications-from cloud computing and AI to mobile and automotive-leveraging best-in-class latency and throughput performance for modern heterogeneous systems across multiple dies.
Technology Advantage
Best-in-class-performance and testability
Latency
- 0.37 ns/mm
- Lowest latency on the market
- No clock penalties
- PPA Calculator available for quick & accurate analysis
Throughput
- 5GHz
- Link speed not limited by distance
- Support for wide data buses
Power
- 0.1pJ/bit/mm
- Low Power
- Self clockâgating technology
- Clock tree removal
- Substantial saving with AVFS when combined with proprietary "O n Die Telemetry"
Hybrid-Bonding
- Correct by construction
- No MD-STA needed
- Pre & Post Bonding integrated die level testing
- No clock travelling across dies
- Low Power
Routing Reduction
- 50% (Based on a large mobile SoC)
- Bus width reduction
- Patented compression technology
- Crosstalk mitigation
Area Reduction
- 10% (Based on large SoC example)
- Reduced top level routing -7%
- Clock tree removal
- Additional savings inside IPs
Testability
- On Die Telemetry: on chip per-link performance measurement
- Simplified margining and binning
- Easier DPM analysis
- Next Generation AVFS
Ease of Deployment
- Deployable with standard Digital Flow and Tools
- Soft IP
- Automated generation of netlist and constraints
- Automotive friendly
- SoC protocol compatible (ARM, RISC-V, etc.)
Safe & Secure
- Low EMI profile
- Protect against Differential Power Analysis (DPA) attacks
- Hardware Trojan detection
- Soft Failure behavior
Design Space
Our solution covers a wide range of applications:
Cloud AI: Training / Inference
- Best-in-class latency for accelerators
- Unlimited scalability (Very large dies)
- Reduces Latency in CXL Expanders
- Easy integration of next generation Neuromorphic Processing Units (NPU) such as Spiking Neural Network (SNN)
Data Center / HPC
- High Performance Interconnect
- High Throughput Low Latency
- Simplified pipelining and design closure
- Ideal for CXL Expanders or High-Speed Switches
- Unlimited Scalability (no clock limitations)
3D Interconnect ( Wormhole )
- High Performance
- PVT Resilient Interface (ideal for die to die)
- Correct by construction (No MD-STA needed)
- Low Power
- Excellent test capability (pre- and post-Hybrid-Bonding)
Automotive / Mobile / Edge-Computing
- Ideal for complex heterogeneous systems
- Supports High Throughput
- Low Latency, Safe and Secure
- Built-in silicon self test and margining
- Soft Failure behavior
Arm & RISC-V Based Solutions
- SoC protocol compatible (AXI, ACE, CHI, Tilelink, OPB, etc.)
- Tunable PPA per channel enables optimization & flexibility
- Xplorer (GUI) provides early architectural investigation
- Strong partnership within the Arm ecosystem