Mechanism to understand

Mechanism for Refresh & Bandwidth Efficiency focuses on effective bandwidth, row-hit rate, refresh stall percentage. The goal is to connect the observable symptom to protocol mechanism, ownership, and regression risk.

refresh, bank conflicts, turnaround, and command scheduling reduce useful bandwidth below headline bus width. Think of it as a contract enforced at boundaries: the sender promises stability and legality, the receiver promises forward progress, and the fabric in between promises not to silently change identity or ordering.

• Identify the transaction boundary: request, data, response, completion, or retry.

• Identify the flow-control boundary: valid/ready, grant, credit, FIFO depth, or lane state.

• Identify what the receiver is allowed to assume and what the sender must hold stable.

Layered view

PROTOCOL STACK VIEW — Refresh & Bandwidth Efficiency

software / firmware intent
        |
        v
transaction semantics: address, ID, length, attributes, ordering
        |
        v
link / channel behavior: handshake, credits, backpressure, retries
        |
        v
physical or timing layer: clocking, reset, pins, lanes, PHY
        |
        v
observability: waveform, VIP transaction, counter, analyzer trace

Debug rule: never jump layers without carrying the transaction identity with you.

Where DDR bandwidth is lost

EFFECTIVE BANDWIDTH BREAKDOWN

peak bus            ████████████████████████  100%
- refresh stalls    ██████████████████████     ~92%
- read/write turn   ███████████████████        ~78%
- row miss penalty  ██████████████             ~58%
= effective         ██████████████             ~58%

Fix targets: better interleave, batch same-direction traffic, page policy.

Layer responsibilities

LAYER RESPONSIBILITY — Refresh & Bandwidth Efficiency

layer          owns                         common failure
-----------    --------------------------   -----------------------
software       intent, ordering needs       wrong assumption
transaction    id/addr/len/attributes       ordering / outstanding
link/channel   handshake, credits, retry    backpressure / deadlock
physical       clock/reset/lanes/PHY        timing / training / SI
observability  waveform/log/counter         missing evidence

Protocol deep dive

DDR bandwidth is scheduler + PHY: rows, banks, refresh, and turnarounds eat headline data rate.

Concept diagram

MEMORY PATH

masters -> controller scheduler -> PHY -> DRAM banks
              |                      |
         refresh/QoS            training/margin

Scheduler sees transactions; PHY sees picoseconds.

Metric graph

BANDWIDTH LOSS WATERFALL

peak              ████████████████████████
refresh           █████████████████████
turnaround        ██████████████████
row miss          ██████████████
effective         ██████████████

Quote the bottom bar in reviews.

Metrics and artifacts to collect

• effective BW

• row hit rate

• refresh stall %

• training margin

• ECC error log

Mini case study

Video workload lost half effective bandwidth after firmware enabled aggressive low-power refresh. Scheduler and firmware QoS had to be co-designed.

Debug branches

• If ECC errors, check training margin and address interleave first.

• If BW low with high row hit, suspect port arbitration not DRAM.

• If boot fail, stop at training step in transcript.

Senior review question

Ask: what is the first transaction that deviates, and which spec rule does it test?

Key takeaways

• Connect every protocol claim to a transaction identity and measurable metric.

• Store the artifact (waveform, log, counter) next to every signoff decision.

Common pitfalls

• Debugging timeouts without finding the first bad transaction.

• Quoting peak bus width without payload efficiency and retry overhead.

• Treating VIP compliance as a substitute for system integration replay.

Mechanism deep dive

refresh, bank conflicts, turnaround, and command scheduling reduce useful bandwidth below headline bus width.

Walk the transaction forward: request accepted → data moves → response completes → software visible effect.