In forced-induction Subaru EJ engines, the intake manifold can quietly limit output once targets exceed roughly 400 whp. Restrictions in runner distribution, plenum volume, or entry geometry lead to uneven cylinder filling, higher exhaust gas temperatures, and a ceiling on power despite strong supporting mods.
These tests used a SuperFlow 1020 flow bench at a standardized 28 inches of water depression, with corrected CFM measured individually through each runner at full throttle body opening. The set covers OEM parts, common aftermarket options, and heavily reworked examples (some reversed for better flow distribution).
Scope disclaimer: This is strictly a flow-bench dataset—not the definitive ranking of manifolds. It lacks dyno-correlated power numbers and omits several popular pieces (e.g., the AMS intake). Treat it as a useful but incomplete reference point in the ongoing EJ airflow discussion.
Runner-Specific Flow Data
Individual runner flows, averages, and runner-to-runner variation (absolute CFM and percentage of average). All values corrected CFM at 28" H₂O.
Runner data highlights:
OEM pieces (Stock STI, 09 WRX) show the lowest variation (<4%), ideal for balanced, street-friendly tuning.
Higher-capacity manifolds trade some uniformity for volume—variation reaches 8–9%, still tunable via individual-cylinder corrections on modern ECUs.
The light-ported 2.5i NA stands out for preserving velocity while approaching premium aftermarket numbers through relatively simple work.
Runner Variation Summary
Absolute max differences span 10 CFM (OEM) to 30+ CFM (high-flow customs). Percentage-wise, best cases sit ~3%, while top performers reach ~8–9%. Sub-6% is optimal for tight AFR control; 8–9% is generally manageable with targeted tuning.
Manifold Influence on Head Flow (Cylinder 3)
Tested bolted to a ported Spec V25 EJ257 head, focusing on cylinder 3 (often the most restricted):
Head-only flow represents the unrestricted ceiling.
2.5i light-ported adds the least restriction—stays closest to head-alone across the lift range.
Stock STI, Cosworth, and Forsa impose more loss (typically 10–20% at higher lifts), with stock showing the largest penalty.
Key takeaway: A high-flow manifold can still hurt performance if it introduces turbulence or poor port matching. Restrained, velocity-focused porting (as on the 2.5i example) frequently outperforms aggressive volume increases.
A Note on Boost Behavior After Manifold Upgrades:
One counter-intuitive but important reality: installing a significantly better-flowing manifold can result in lower observed boost pressure at the same throttle position and RPM compared to the previous setup.
Boost is not a direct measure of power—it's primarily an indication of pressure created when airflow exceeds what the engine can consume. A more efficient manifold (higher volumetric efficiency) allows the engine to ingest more air for a given manifold pressure. The turbo therefore doesn't need to produce as much pressure to deliver the same (or more) mass flow.
In practice, many owners see 2–5 psi less peak boost after a good manifold swap, yet the car makes noticeably more power, spools faster, and runs cooler EGTs. Dyno tuning is essential to recalibrate targets and take full advantage.
Practical Implications for EJ Builds
Sub-350 whp targets — Realistically, unless you're in a classed race car...don't bother. If you are racing: Light-ported 2.5i NA: strong flow-to-cost ratio, excellent balance, and good part-throttle response.
400–550 whp range — Cossie, Process West or Forsa (with 75 mm TB where applicable): meaningful step up in capacity without overcomplication.
600+ whp territory — Both the ASF Monstrosity (reversed, highest measured flow) and Process West (solid 360 CFM with large plenum) belong here. The choice depends on priorities: absolute peak flow vs. plenum volume and packaging.
Ultimately, raw CFM matters, but distribution, velocity preservation, and head compatibility often dictate real-world outcomes. These numbers help shift the conversation from hype to data.
If your build involves one of these (or something untested), drop the details—happy to brainstorm how the flow numbers might translate. More bench time incoming.