MED9.1 VR6 FSI

Bosch MED9.1 (Motronic MED 9.1) — VAG narrow-angle VR6 FSI direct-injection petrol (3.2/3.6, the DI VR6 sometimes grouped as EA390). Torque-structure ECU: architecturally a gasoline-direct-injection evolution of ME7, so nearly all ME7 torque-model map names/logic carry over (KFMIRL/KFMIOP/KFZW/LAMFA + two-level torque monitor) PLUS FSI-specific rail-pressure and injection-phasing maps. Same MED9.1

Reference - updated 4 August 2026

The ECU hardware

Freescale/Motorola MPC5xx PowerPC core (32-bit, BIG-ENDIAN), typically MPC562/MPC563-class, with a hardware torque/monitoring watchdog. IMPORTANT CORRECTION: MPC5xx parts have only modest INTERNAL flash (MPC563 ~512 KB, MPC565 ~1 MB), so the ~2 MB full-flash image is predominantly EXTERNAL NOR flash on the board — do not describe it as '2 MB internal flash'. Debug/recovery via the MPC5xx BDM (background-debug) port, which reaches both internal and external memory for a guaranteed dump and un-brick. Immo/component-security + adaptation live in a SEPARATE external serial EEPROM (95xxx SPI, e.g. 95040/95080/95160), sometimes with a flash shadow copy — NOT in a large MPC internal EEPROM. NOT a Tricore/MED17 (no RSA/Tprot signing on these VAG MED9.1 files).

Fuelling and load sensing

Petrol, gasoline direct injection (FSI/GDI; homogeneous stoich at mid-load with some part-load stratified/lean operation). High-pressure fuel system: cam-driven single-piston HPFP feeding a common rail (early-FSI pressure region, roughly mid-tens up to ~110-120 bar commanded), plus a low-pressure lift pump. Naturally aspirated (no factory turbo/supercharger) — this fundamentally shapes staging: NA gains are modest, and the single-piston HPFP volume is the hard fueling ceiling for any forced-induction build.

How this ECU is read and flashed

READ/WRITE paths: (a) OBD — CAN TP2.0 with KWP2000-on-CAN into the Bosch bootloader using seed-key security access; performed by aftermarket tools (KESS, KTAG, FoxFlash, MPPS, CMDflash, Galletto) and OEM VAS/ODIS SVM. OBD READ is restricted on some MED9.1 builds, and session handling matters (on MED9 certain diagnostic/programming sessions reboot to bootloader/erase and need a power cycle) — a clean full read is often taken on the bench. (b) BENCH — pull the ECU, power it via the OEM/bench connector, and do a full read/write of the flash on supported tools without opening the lid. (c) BOOT/BDM — the MPC5xx BDM (JTAG-like) debug port inside the ECU for a guaranteed full dump and, critically, recovery of a bricked/failed-checksum unit. IMMO/SECURITY: MED9.1 uses immobilizer-4-style component protection tied to the cluster/gateway. Immo-off / 'virginizing' is only needed to run a donor/bench ECU or a swap: edit immo status + component-security in the external 95xxx EEPROM (and any flash shadow), then recalc the EEPROM checksum with a MED9-aware immo/EEPROM tool (commercial: Abrites/AVDI, VVDI; or a validated community routine — there is no verified free 'MED9sum'). Always keep a verified stock full-flash + EEPROM backup before writing; a bad cal checksum or interrupted OBD write is recoverable via bench/BDM, not always via OBD.

Engine codes in this family

• AXZ — 3.2 FSI VR6, ~250 PS, VW Passat B6 3.2 V6 FSI 4motion (the less common of the FSI VR6 codes), MED9.1 (true FSI DI)
• BLV — 3.6 FSI VR6, ~280 PS, Passat B6 (North America) / CC, MED9.1
• BWS — 3.6 FSI VR6, ~300 PS, Passat R36 (transverse, 4motion, DSG), MED9.1
• CDVA / CDVB — 3.6 FSI VR6, ~260-280 PS, later Passat CC / Touareg / Skoda Superb 4x4 (longitudinal apps often paired with a torque-converter auto, not DSG), MED9.1
• MIS-ID CAUTION: 'BUB' is NOT MED9.1 FSI. BUB = Mk5 Golf R32 / TT / A3 3.2 VR6 with PORT (multi-point) injection on Bosch ME7.1.1 — different ECU, non-DI fueling, no rail-pressure maps. Do NOT apply MED9.1 FSI defs/tunes to a BUB (or BDB/BMJ/BHE/CBRA) port-injection 3.2 VR6.

Staging - what each stage really gives

• stage1 - Software-only on stock hardware. Optimize KFZW toward MBT, tidy LAMFA high-load enrichment, raise/align KFMIRL + KFMIOP and the torque-monitor limiters together (so extra output isn't clawed back by intervention), fine-tune dual-VVT cam targets, sharpen KFPED, raise the rev limit slightly, de-restrict VMAX. Requires higher octane (98 RON / 93 AKI) to support the added timing. Realistic NA gains are MODEST — on the 3.6 (R36/BLV/BWS) roughly +10-15 hp and ~+15-25 lb-ft, biggest in the mid-range; the 3.2 (AXZ) scales similar but smaller. (Vendor-published NA figures such as APR's 3.6 Stage-1 are single-to-low-double-digit hp; treat any specific number as approximate/verify — stock R36 is ~296 h
• stage2 - Stage-1 recalibration plus bolt-on breathing: high-flow/panel intake, less-restrictive exhaust (cat-back, high-flow cats or headers/downpipes — the stock manifolds/cats are the main NA restriction), sometimes a lightweight pulley/lower-temp thermostat. Re-tune timing/fuel/VVT for the new airflow and keep post-cat/O2 monitors satisfied (or re-define if hardware legitimately changed). Being NA, cumulative gains stay moderate (order of +15-30 hp over stock depending on parts); diminishing returns are inherent.
• stage3 - Forced induction or major internal work — where real numbers appear. Supercharger or turbo kit (HPA / VF / custom) taking the VR6 FSI to ~400-550+ hp, OR big cams/head/displacement for high-strung NA. Requires full recalibration for boost: new load/boost model, fueling (the single-piston HPFP is the ceiling — typically add auxiliary PORT injection or upgrade fueling), timing pulled for boost, and the torque model + limiters rebuilt to match. Mandatory supporting mods: clutch/flywheel or DSG upgrade (torque now exceeds transmission protection), intercooling/charge cooling, and often heavily reworked MED9 (or standalone) strategy. Watch low-speed pre-ignition (LSPI) at low-rpm/high-load. Clutc

Safety limits

• KNOCK: zero sustained knock. Watch per-cylinder knock retard (KLR/zwout path); if retard climbs under load, pull KFZW timing. Aggressive timing REQUIRES 98 RON / 93 AKI or better — never run a high-timing map on low octane.
• LAMBDA/AFR: enrich high-load cells to ~λ0.85-0.88 (~12.5-13.0:1 AFR) NA for EGT + knock protection, richer (~λ0.80 / ~11.5:1) under Stage-3 boost; NEVER lean under high load. Mid-load DI may sit at stoich (λ1.0) for economy/cats — but never at WOT.
• EGT: keep exhaust-gas temperature below ~950 °C to protect the turbine-less manifolds, valves and catalysts (the NA VR6 FSI cats are the weak point) — leaning out or over-advancing timing spikes EGT.
• RAIL-PRESSURE / FUELING CEILING: do not command beyond the HPFP mechanical/flow limit (~110-120+ bar region stock). The single-piston HPFP's fuel VOLUME is the hard ceiling — any Stage-3 FI build will go lean at high rpm/boost unless fueling is upgraded (auxiliary port injection). Lean-at-load = detonation = engine damage.
• HPFP CAM-FOLLOWER / MECHANICAL WEAR (FSI-specific): raising commanded rail pressure increases peak force on the cam-driven single-piston HPFP and its follower — excessive/aggressive rail-pressure demand accelerates follower and cam-lobe wear (a known FSI/TFSI failure mode). Keep rail-pressure increases modest on the stock pump.
• REV LIMIT: keep ≤ ~7000-7200 rpm without upgraded valvetrain — valve float, not the ECU, is the real limit; set the NMAXF diagnostic threshold above the cut.
• TORQUE-MODEL CONSISTENCY: keep KFMIOP/KFMIRL and the torque monitor (miszul/mimax path) internally consistent. Raising limiters alone just moves intervention from Level-1 (timing cut) to Level-2 (fuel/throttle) and triggers limp — the model must actually match the new output.
• TRANSMISSION (DSG/CLUTCH): on the transverse R36/Passat DSG (DQ250) stock torque protection is ~350 Nm (the 3.6 already sits near it); a DSG tune safely lifts it toward ~410 Nm, beyond which clutch/DMF and oil-pressure work are required. Longitudinal apps (Touareg/others) use torque-converter autos with their own limits; manual clutches are near stock rating on a torquey Stage-1. A Stage-3 engine tune MUST be matched by a clutch/DSG upgrade — exceeding these = slip / DSG limp / driveline damage.
• LSPI (Stage-3 forced induction): boosted DI at low-rpm/high-load risks low-speed pre-ignition — enrich mixture and pull timing in the low-rpm/high-load cells. NA is largely immune, but FI is not.
• BRICK-SAFETY / BACKUPS: always keep a verified stock full-flash AND stock EEPROM backup before writing; never flash a file whose cal checksums weren't recalculated (checksum DTC/limp/reject); never interrupt an OBD write (stable battery/charger required — a half-written flash can leave the ECU only recoverable via bench or MPC5xx BDM, not always OBD).
• EMISSIONS/CATS (street): don't run so rich it overheats/destroys the cats; keep secondary-air and cat/O2 monitors functional unless hardware is legitimately removed and re-defined. Note DI intake-valve carbon buildup is inherent to these engines (mechanical, not a flash setting).

Every car using this platform (24)

Stock power and typical remap figures per car. Stage results assume a healthy engine on good fuel.

How confident is this?

SOLID: ECU family and processor class (Freescale/Motorola MPC5xx PowerPC, BIG-ENDIAN, BDM-capable, external NOR flash with a ~2 MB full-flash image); torque-structure architecture with ME7-shared map names and lockstep (KFMIRL/KFMIOP/KFZW/LAMFA + two-level torque monitor) — well documented (s4wiki) and directly applicable; NON-cryptographic correctable checksums (no RSA/Tprot on VAG MED9.1) fixable in WinOLS/FoxFlash/KESS/KTAG; OBD+bench+BDM access paths and the 'performance flash needs no immo-off' boundary; NA staging logic and the DQ250 ~350→410 Nm torque-protection reality; and the engine-code correction (BUB = Mk5 R32 3.2 VR6 PORT injection on ME7.1.1, NOT MED9.1 FSI; true MED9.1 FSI codes are AXZ (3.2) and BLV/BWS/CDVA/CDVB (3.6)). CORRECTED FROM DRAFT: the ~2 MB image is external NOR flash, not MPC 'internal' flash; there is no verified tool named 'MED9sum' (ME7sum is ME7-focused; use WinOLS/flasher built-in for MED9 flash checksums and a commercial immo tool for the EEPROM); 'OLS288 plugin' and 'TracqiTechnology MxT' were not verifiable and were replaced with real references; LAMFA is load×rpm and commonly BE u16 (a known dtype trap). THINNER / VERIFY-PER-FILE: exact flash