Synopsys and Ansys power the future of innovation—connecting silicon to systems.
EDA
System
IP
SolvNetPlus
Training & Education
Introduction
The relentless push toward higher performance in AI accelerators, data center processors, and edge computing platforms has fundamentally transformed power delivery from a secondary, late-stage consideration into a first-order design constraint to be dealt with pre-silicon. Today's ASICs (Application-Specific Integrated Circuit), FPGAs (Field-Programmable Gate Array), and heterogeneous compute platforms must dynamically respond to fluctuating workloads while operating within increasingly stringent power and thermal envelopes. The interface that orchestrates this delicate balance directly determines system efficiency, reliability, and competitive advantage.
Adaptive Voltage Scaling (AVSBus) addresses this challenge head-on by enabling real-time voltage adjustments based on instantaneous operating conditions. AVSBus serves as the digital control plane between compute devices and their point-of-load (POL) power converters, transmitting commands and receiving telemetry to maintain optimal power states. While earlier design implementations managed single voltage rails in isolation, contemporary platforms demand coordinated control across dozens of rails and multiple power domains simultaneously — and AVSBus 2.0 represents the architectural evolution designed specifically for this complexity.
The verification challenge has also evolved in parallel. It's no longer sufficient to simply validate that voltage changes occur; engineers must confirm that voltage transitions happen at precisely the right moment, on the correct rail, targeting the appropriate device, and with deterministic acknowledgment from power-management ICs — all under the heavy, non-deterministic software workloads that trigger these power states. This level of precision enables designers to minimize voltage guardband, orchestrate smooth state transitions, and achieve measurable improvements in power efficiency without sacrificing performance headroom.
Why AVSBus Matters for Power-Sensitive Systems
Tracking workload demand with supply voltage is a highly effective way to reduce energy consumption without compromising performance, especially since dynamic power scales proportional to the square of the voltage. AVSBus provides the communication framework for that adaptive behavior: it gives the controller fine-grained control over voltage and transition rates, returns real-time telemetry such as current, temperature, and status, and closes the loop with deterministic acknowledgments and status-response fields. Manufacturer-specific commands leave room for vendor differentiation; and because every transfer is controller-initiated, behavior stays predictable and tightly aligned with the system's power-management policy.
This closed-loop adaptive mechanism proves invaluable wherever power, performance, and thermal constraints intersect:
- AI/ML accelerators and data center infrastructure: Managing power across GPU clusters, tensor processing units, and inference engines
- Hyperscale server processors: Coordinating voltage across multiple chiplets and power islands
- 5G/6G networking silicon: Balancing baseband processing loads with strict thermal limits
- Storage controllers and SSDs: Optimizing power during variable I/O workloads
- FPGA and adaptive compute platforms: Supporting dynamic partial reconfiguration with voltage scaling
- Automotive and edge AI: Meeting functional safety requirements while maximizing efficiency
In each domain, the objective transcends simple voltage adjustment — it demands controlled, observable, and verifiable voltage scaling across numerous concurrent domains.
Figure 1: Adaptive Voltage Scaling with AVSBus 2.0
AVSBus 2.0: Enabling Scalable Voltage-Control Architectures
AVSBus was earlier Part III of the PMBus 1.4.1 specification, where it connected one controller to a single target device. AVSBus 2.0 is the first release of AVSBus as a standalone specification, and it re-architects the interface for scale.
The most visible change is the multi-target addressing: a new 4-bit field gives a single controller up to 15 individually addressable targets, plus a broadcast address that reaches every target at once. This turns a point-to-point link into a shared control bus for many power converters and voltage domains — an essential capability in complex SoCs and system-level designs.
- Enhanced frame structure improves robustness as the protocol scales: The frame size expands from 32-bit sub-frames (64 bits total) to 48-bit sub-frames (96 bits total), and the 2-bit start code widens to a full 8-bit sequence (00000000b) that is far easier to lock onto. Marker bits inserted at fixed positions guarantee that no payload can accidentally reproduce that sequence, keeping frame boundaries unambiguous. The 3-bit CRC continues to protect each of the now-wider sub-frames. Together, these changes keep communication dependable, even in high-speed, electrically noisy environments.
- Deterministic recovery mechanisms replace legacy approaches: The previous resynchronization scheme is replaced by an explicit, controller-initiated frame abort. By transmitting a start code at a byte boundary, the controller can immediately terminate an in-flight transaction and return every target to a known idle state, giving the system predictable and fast error recovery.
- Coordinated update capabilities receive significant enhancement through a dedicated Commit command: Paired with Write-and-Hold, it lets the controller stage new voltages and then apply them together — synchronizing changes across multiple rails within a target, or across multiple targets via broadcast — with per-datatype or global application controlled by. The result is tightly coordinated voltage transitions, a critical capability in performance-sensitive designs where rails must move in lockstep.
Protocol Aspect | PMBus 1.4.1 | AVSBus 2.0 | Verification Impact |
Topology | Point-to-point | Multi-target (up to 15) | Address-aware stimulus and checking |
Frame Size | 64-bit (32-bit sub-frames) | 96-bit (48-bit sub-frames) | Larger frame parsing, marker validation |
Start Code | 2-bit (01b) | 8-bit (00000000b) | Robust detection + abort support |
Synchronization | Resynchronization via idle pattern | Explicit frame abort | Deterministic recovery validation |
Modes | 2-wire and 3-wire | 3-wire only | Mandatory return-path verification |
Coordination | Write-and-Commit, Write-and-Hold | Adds dedicated Commit (per-type or global) | Complex multi-step sequencing |
Table 1: AVSBus Evolution: From PMBus 1.4.1 to AVSBus 2.0
Verification Challenges in AVSBus 2.0
Each capability enhancement proportionally expands the verification challenge surface. With multiple targets sharing a single bus, testbenches must model correct target address interpretation, proper broadcast command handling, and independent acknowledgment and status responses from each target. Additionally, verification must exercise corner cases emerging from partial address matches, resource unavailability, or invalid commands — elevating address decoding, partial responses, and multi-target synchronization to first-class verification concerns.
The enriched frame format raises integrity checking requirements. Beyond basic command-response validation, verification must confirm:
- Marker bits appear at correct positions with proper values
- Start-code detection distinguishes valid frames from malformed patterns
- CRC-3 catches corrupted transfers
- Framing errors and abort conditions trigger deterministic recovery
These bit-level requirements apply to every single frame. Coordinated updates introduce additional sequencing complexity. Write-and-Hold and Commit operations create scenarios including:
- Held values overwritten before taking effect
- Commits applying per data type or globally
- Broadcast commits spanning multiple targets
- Commit interactions with unavailable or invalid resources
Validating behavior across all combinations distinguishes plausible implementations from provably correct ones.
AVSBus 2.0 Verification IP
Synopsys AVSBus 2.0 Verification IP (VIP) is a native SystemVerilog UVM solution built to address these challenges across IP, subsystem, and system-level verification environments. A single configurable system environment models a complete AVSBus topology — one controller agent and up to 15 target agents — enabling engineers to verify controller DUTs, target (POL/PMIC) DUTs, or complete multi-target subsystems with the same VIP.
Figure 2: Synopsys AVSBus 2.0 Verification IP
- The environment is configuration-driven rather than hard-coded for each project: Teams can set the number of targets and their addresses, per-rail Vout minimum, maximum, and reset values, clock and timing parameters, broadcast-response policy, and many other parameters — and the same environment adapts. Underneath, an address-aware, per-rail model tracks voltage, transition rate, and power mode — including the distinction between held and committed values — so the VIP always holds a golden reference for what each rail should be doing.
- Protocol coverage is comprehensive: The VIP handles multi-target and broadcast addressing, read, write-and-commit, write-and-hold, and dedicated commit flows, status-response frames, handling, and target-initiated interrupts, along with the 96-bit frame parsing, frame abort, pipelined (back-to-back) transfers, clock suspend, and bus-timeout behavior. In other words, the corner cases described above are part of the VIP's normal operating range, not features a team manually builds for itself.
- Built-in protocol checks power the VIP's verification services: With full command, frame, and response awareness, the VIP applies checks, coverage, and error injection across the complete protocol. More than 100 checks cover start-code and marker validation, CRC-3, reserved bits, encoding, status-response structure, setup/hold and clock timing, framing errors, X/Z detection, and more.
Figure 3: Synopsys AVSBus 2.0 Verification IP Features
- Address-aware functional coverage: The coverage model tracks command and data types, target address, rail select, acknowledgment codes, status and prefix fields, commit modes, pipelined transfers, bus timeout, frame abort, and many others — so teams can measure real progress against the specification.
- Configurable error injection: An extensive library injects CRC corruption, marker and reserved-bit errors, invalid , status and prefix corruption, clock stall, mid-byte framing errors, X/Z on data and clock, and others — exercising a design's error-detection and recovery paths on demand.
- Layered sequence library: Random and directed sequences for read, write, write-and-hold, commit, broadcast, and manufacturer-specific flows make tests fast to author and easy to constrain.
- Native integration with the Synopsys Verdi® Protocol Analyzer provides transaction-level, protocol-aware waveform view as well as debug. Layered transaction logs, XML tracing, transaction reports, and a waveform debug port make it straightforward to trace an issue back to its source, whether that is target-address decoding, rail selection, an acknowledgment or status mismatch, a held-data or commit interaction, or a bit-level framing error.
Figure 4: VIP configuration to coverage closure
Conclusion
AVSBus 2.0 is a significant step forward for voltage-control interfaces. By adding multi-target addressing, coordinated update mechanisms, and more robust communication, it enables the scalable, deterministic power management that modern compute systems need. As power delivery becomes more dynamic and more system-wide, verifying AVSBus behavior is no longer optional — it is essential to first-pass success.
Synopsys AVSBus 2.0 Verification IP helps engineering teams meet that bar, giving them confidence that voltage-control communication across targets, rails, and operating conditions is correct, reliable, and ready for real-world deployment.
Synopsys is partnering with early customers and collaborators to enhance the standard architecture for their next-generation designs, incorporating new features now available with the latest specifications.
Synopsys AVSBus 2.0 VIP is natively integrated with the Synopsys Verdi® Protocol Analyzer debug solution for protocol-aware, transaction-level debug. For system-level power-management scenarios that demand a faster, hardware-based pre-silicon platform, Synopsys transactors and hybrid and virtual solutions extend these use cases to the industry's fastest verification hardware, Synopsys ZeBu® emulation and Synopsys HAPS® prototyping systems.
Continue Reading
Reduce Risk and Time-to-Market for DDR6 Designs with Synopsys DDR6 Verification IP
