Disclosure: Most of the following report was generated using an LLM trained on my previous power meter and smart trainer tests (both successes and failures). I’ve verified the observations and am confident in its ability to identify the issues I look for when reviewing and comparing power sources. Outdoor tests are yet to be performed. 

The InPeak Twin2 was tested against two established power references:

• Garmin Tacx NEO 3M: Control unit and primary reference
• Favero Assioma PRO RS-2: Dual-sided pedal-based reference
• InPeak Twin2: Dual-sided crank-based power meter under test

The test covered steady-state ERG intervals, high-power efforts, short sprints, variable riding, cadence measurement, left/right balance and zero-power behaviour.

Overall Results

All three power traces aligned closely after correcting the recording timestamps.

Compared with the NEO 3M during active riding:

• The Assioma PRO RS averaged approximately 0.76% higher.
• The InPeak Twin2 averaged approximately 0.23% lower.
• Both meters followed ERG steps, cadence changes and sprint efforts well.
• Neither showed obvious thermal drift, residual power after sprints or cadence-related power failures.
• No sustained power oscillation was observed during the steady-state ERG sections.

The InPeak’s overall accuracy was excellent. However, its position fractionally below the NEO is mechanically unusual: a crank-based power meter would normally read slightly higher than a trainer due to drivetrain losses.

The difference remained comfortably inside the InPeak’s stated ±2% accuracy specification.

Recording Alignment

The Assioma trace led the NEO by approximately one second during rapid power changes. This is expected because the pedals measure power before it passes through the drivetrain, while the NEO measures it at the trainer and introduces its own reporting latency.

The RMSE values are heavily influenced by one-second timing differences during sprints and power transitions. They should not be interpreted as steady-state accuracy figures.

Steady-State ERG Accuracy

200 W ERG Interval

Eight-minute comparison:

The Assioma remained consistently 2–3 W above the NEO. This is directionally correct for a pedal-based meter, although the difference is smaller than the drivetrain loss normally expected.

The InPeak was essentially level with the NEO, sitting only 0.55 W lower on average.

250 W ERG Interval

The relationship tightened at 250 W. The Assioma remained slightly above the NEO, while the InPeak remained fractionally below it.

There was no evidence of either meter progressively scaling away from the NEO as power increased from 200 W to 250 W.

225 W ERG Interval

This was the InPeak’s largest steady-state difference from the NEO. Even here, the discrepancy was less than 2 W and remained well inside its stated accuracy tolerance.

Power Stability and Oscillation

Ten-second-smoothed variation was assessed during the main ERG sections:

The pedal and crank meters naturally showed slightly more variation than the NEO. The trainer controls the resistance and consequently produces a smoother power stream.

There was no evidence of a repeating oscillation unique to either the Assioma or InPeak. Most small fluctuations appeared simultaneously in both meters, indicating genuine pedalling variation rather than measurement instability.

The InPeak did not display obvious power hunting or unstable reporting.

Consistency Over Time

The relationship between the meters remained stable throughout the test:

Both meters moved slightly downward relative to the NEO, but the change was approximately half a percentage point. That is too small to classify as meaningful thermal drift from a single test.

Sprint Performance

First Sprint

The Assioma recorded the highest initial peak. This is expected from a pedal meter because it measures power before the drivetrain and responds immediately to changes in force.

The InPeak also recorded a higher one-second peak than the NEO, although its five-second result was lower.

Across the complete sprint:

• Assioma energy was 0.73% above the NEO.
• InPeak energy was 0.56% below the NEO.

The close agreement across the complete effort indicates that the shorter-duration differences were primarily caused by response timing and sample placement.

Sustained High-Power Effort

This effort was exceptionally well aligned.

Across the complete effort:

• Assioma: +0.28%
• InPeak: −0.29%

Final Sharp Sprint

The InPeak reported more power through the three-to-ten-second window than both the NEO and Assioma.

Across the broader sprint event:

• Assioma: +0.13%
• InPeak: +1.80%

This appears more consistent with a difference in transient response or power-decay timing than residual torque. The InPeak returned to agreement immediately after the effort and did not remain artificially elevated.

It is worth monitoring across repeated short, sharp sprints, but this single event is not enough to identify a definite sprint-accuracy problem.

Variable Riding

Across the final 18-minute variable-power section:

Both meters followed repeated accelerations, low-power sections and cadence changes very well.

There was no indication that the InPeak’s cadence measurement produced corresponding power errors during variable riding.

Cadence Accuracy

Both meters reached approximately 135 rpm, compared with 136 rpm from the NEO.

The Assioma produced the cleaner cadence trace and followed the trainer more closely. The InPeak displayed slightly more one-second variation, but no meaningful cadence spikes, dropouts or sprint-related failures were observed.

Left/Right Balance

The Assioma and InPeak produced almost identical average left/right balance results:

During steady-state riding, the balance readings differed by approximately 0.6 percentage points per sample on average.

During variable riding, the instantaneous difference increased to approximately 1.4 percentage points. This is unsurprising because the meters measure at different locations and may use different reporting or averaging methods.

The broader left/right trend agreed extremely well. Neither meter reported a sustained imbalance that was absent from the other.

Coasting and Residual Power

There were 169 aligned samples where the NEO reported zero power:

• The Assioma reported zero for all but two samples.
• The InPeak reported zero for all but three samples.
• The highest non-zero value from either meter was only 4 W.

There was no evidence of sticky watts, residual torque or prolonged non-zero power after the sprint efforts.

Accuracy Claims in Context

The Assioma PRO RS has a claimed accuracy of ±1%, while the InPeak Twin2 is rated at ±2%.

Based on this test:

• The Assioma remained inside its claimed range for much of the ride.
• Its largest steady-state difference was approximately +1.3% at 200 W.
• Combined device tolerances and the difference in measurement location can readily account for this.
• The InPeak remained comfortably inside its ±2% specification throughout.
• The InPeak’s slightly negative relationship to the NEO is mechanically unusual, but the difference was too small to establish an accuracy fault.
• The Assioma’s positive offset was also smaller than the drivetrain loss normally expected between pedals and a trainer.

Accuracy specifications apply to each device independently. They should not be interpreted as a guarantee that two meters will always agree within the accuracy claim of either individual unit.

Verdict

The Favero Assioma PRO RS performed as expected. It was stable and responsive, produced excellent cadence data, captured plausible sprint peaks and returned cleanly to zero after each effort. Its average power was slightly above the NEO, although generally by less than the drivetrain-loss difference normally expected.

The InPeak Twin2 also performed very well. Its steady-state power was exceptionally close to the NEO, its left/right balance closely matched the Assioma, and it showed no obvious drift, oscillation, cadence failure or residual torque.

Two observations warrant further testing:

• The InPeak consistently sat fractionally below the NEO during steady-state riding.
• It reported more power than the other meters during the decay of the final sharp sprint.

Neither observation was large enough to classify the InPeak as inaccurate. Based on this dataset, the InPeak Twin2 operated comfortably inside its stated ±2% specification and delivered a strong overall result.