Fabric-Attached Memory System Design: Comparison Matrix

Comparison matrix

Interleave, NUMA distance, and ownership policies trade capacity gains against tail latency and RAS scope.

Use the matrix as a reasoning aid, not as a simplistic scorecard. PCIe/CXL choices are workload-sensitive: the same policy can be right for bandwidth-oriented streaming, wrong for latency-critical bursts, and risky for long-haul reliability.

+------------------+----------------+----------------+----------------+
| Approach         | Strength       | Weakness       | Best when      |
+------------------+----------------+----------------+----------------+
| Conservative     | high robustness | lower peak     | new platform   |
| Balanced         | good efficiency | needs telemetry | mixed workloads |
| Aggressive       | max throughput | tail sensitivity | bounded SKUs   |
| Hardening        | field resilience | overhead cost  | safety-critical |
+------------------+----------------+----------------+----------------+

When to choose each approach

• Choose policy from measured conflict profile, SLA targets, and reliability budget

Interview traps

• Copying scheduler recipes across unrelated traffic mixes

• Ignoring coupling between turnaround control, refresh policy, and fairness

Comparison reference

PCIe/CXL EVIDENCE MATRIX - Fabric-Attached Memory System Design

+-------------------------------+--------------------------------+--------------------------------+---------------------------+
| Evidence                      | Tells you                      | Does not prove                 | Next action               |
+-------------------------------+--------------------------------+--------------------------------+---------------------------+
| TLP type mix + credit stall counters    | protocol-layer stall cost    | link integrity and replay behavior   | inspect training margins  |
| queue age + class breakdown   | fairness and starvation risk   | command legality details       | parse command timeline    |
| LTSSM timeline + ordered set progression | timing-window pressure         | root cause by itself           | correlate with topology map|
| eye / Vref / skew snapshots   | PHY margin and drift behavior  | controller policy quality      | pair with schedule logs   |
| CE/UE + scrub telemetry       | reliability trajectory         | immediate perf bottleneck only | map to hotspot apcieesses  |
+-------------------------------+--------------------------------+--------------------------------+---------------------------+

PCIe/CXL deep dive

Memory expansion and coherency require HDM windows, ownership discipline, and NUMA-aware software policies.

Concept diagram

COHERENCY + HDM

CPU caches <-> CXL.cache <-> device memory (CXL.mem/HDM)

Metric graph

EXPANSION BOTTLENECK SHARE

remote latency      ██████
ownership retry     ████
interleave skew     ███

Reports and artifacts

• HDM decode table

• ownership transition trace

• NUMA distance profile

• RAS region policy

Mini case study

Fabric-attached memory increased capacity but p99 regressed until page placement respected NUMA distance.

Debug branches

• Map HDM windows and interleave groups

• Run ownership litmus under contention

• Correlate RAS events with region offline policy

Senior review question

Ask: which latency, bandwidth, and reliability evidence proves this PCIe/CXL topic is closed under real traffic?

Key takeaways

• Always tie controller and PHY counter shifts to application latency and throughput outcomes.

• Lock firmware timing profile, thermal condition, and DIMM state before comparing PCIe/CXL captures.

Common pitfalls

• Chasing peak bandwidth while ignoring p99 latency and fairness tails.

• Changing timing guardbands without separating SI noise from scheduling issues.

• Declaring closure without reliability gates, fault injection, and regression replay.

Principal PCIe/CXL review addendum

Fabric-Attached Memory System Design should be read as an end-to-end memory behavior, not as a single block definition. A production PCIe/CXL subsystem reflects interactions between array physics, command legality, scheduler policy, PHY margin, and reliability controls before software experiences final latency or bandwidth.

Fabric-attached memory expands capacity beyond local DIMMs with NUMA-like latency profiles. System design must balance interleave, page placement, migration policies, and error containment across the fabric. PCIe/CXL inefficiency is multiplicative: one extra ACTIVATE, one unnecessary turnaround, one weak lane margin, or one refresh collision repeated across billions of accesses can dominate product tail latency and power.

Use Effective mem bandwidth, tail latency across NUMA nodes, and RAS event rate as the opening signal, not the conclusion. A metric move only becomes actionable when paired with workload context, command traces, training telemetry, and evidence artifacts such as NUMA distance table, bandwidth/latency profile, and RAS policy doc.

Host-device coherency and HDM windows define how expanded memory behaves like first-class system memory. Senior review quality comes from proving a complete chain: request pattern -> memory-state transition -> bottleneck mechanism -> smallest owner fix -> regression-safe validation.

Review discipline should enforce a single causal chain: traffic pattern -> command-level behavior -> array/PHY effect -> measured product impact. That chain prevents tuning folklore from replacing evidence.