The Inno Score, explained
Every score on InnoReviews is computed, not opined. We take each device's published specifications and typical benchmark results, normalize them against fixed reference ranges, weight them, and blend them into sub-scores and one overall score out of 100. The exact ranges and weights are published below — if you disagree with a weight, you can see precisely how it moves the result.
1. Normalize. Each metric v maps to(v − min) / (max − min)clamped to 0…1 against its reference range. Wide-range benchmarks (like AnTuTu) use a logarithmic curve so mid-range devices aren't flattened. “Lower is better” metrics (weight, thickness, process node, TDP) are inverted.
2. Weight & blend. Normalized metrics are combined into sub-scores (Performance, Display, Battery, …) using the weights below. Sub-scores blend into the overall score using per-category weights. If a device is missing a metric, that metric is skipped and the remaining weights are renormalized — devices are never punished for missing data, only for weak data.
Minimum evidence. Renormalizing has one limit. Sub-scores markedrequired below (single- and multi-core for laptop and desktop processors, CPU and GPU for mobile chipsets, rasterization for graphics cards) need at least one published benchmark result before a device receives an overall score. For a graphics card that means a 3DMark Time Spy Graphics result or measured frame rates. For a mobile chipset that means a Geekbench 6 result for the CPU, and for the GPU either a 3DMark Wild Life Extreme result or an AnTuTu total, because every AnTuTu run includes a GPU test. A chip with AnTuTu but no 3DMark result is scored without a GPU sub-score, and its AnTuTu total still counts only once, under Benchmarks. Until a device has that evidence it is listed as “awaiting benchmarks” and left out of the rankings, so no chip is ever ranked on its spec sheet alone. Sub-scores built purely from spec-sheet figures — TDP, process node, cache, clocks, a vendor's AI TOPS claim — are labelled spec-based on comparison pages and are never cited as the reason one device beats another.
3. Scale. The blended 0…1 result is mapped onto a 20–100 editorial scale (so a hypothetical “worst possible device” sits at 20, the reference-range ceiling at 100), rounded to one decimal.
Data sources. Manufacturer spec sheets plus typical published benchmark results (AnTuTu, Geekbench 6, Cinebench, 3DMark, PassMark). Benchmarks vary by unit and firmware; we use representative figures and update them as better data lands. A laptop processor's iGPU sub-score is the 3DMark Time Spy Graphics result of its integrated GPU (Notebookcheck's average for that GPU model); Apple GPUs cannot run Time Spy, so Macs carry no iGPU sub-score.
AnTuTu versions. Every AnTuTu total shown with a CPU / GPU / Memory / UX breakdown is one row, breakdown included, of AnTuTu's own AnTuTu 11 (V11) device ranking at antutu.com/web/ranking, the July 2026 edition (stamped “Updated: 2026.7”). The Wayback Machine archived that edition on 24 August 2026, and every breakdown on this site matches it digit for digit. We take one mainstream representative device per chip, never a fan-cooled gaming variant, and those pages say “AnTuTu 11”. AnTuTu refreshes the table monthly, so its live figures can drift a few percent from the edition we cite.
Totals shown without a breakdown are typical published results whose AnTuTu version we have not verified (most appear to come from earlier AnTuTu versions), labelled “AnTuTu (typical)”. The two are not on one scale: where we had both for the same device, the V11 row was typically 13–16% higher, and up to 43% higher on mid-range chips. So we never set one against the other. The bar and “Top N%” beside an AnTuTu total rank it only among AnTuTu 11 results; a typical figure carries no rank; and a comparison page shows both numbers without naming a winner when the versions differ. We re-source typical figures to V11 as devices appear in AnTuTu's ranking.
AnTuTu in the Inno Score. The score still reads both kinds of total, as a deliberate trade-off. When a device's older figure was replaced by its V11 row, its overall score rose by about 0.6 points on average (1.9 at most). Dropping the typical figures from scoring instead would strip AnTuTu from about 250 devices and move their scores by about 2 points on average (up to 4.6), roughly three times the distortion it would remove. The mix shrinks as totals move to V11.
Reusing our data. Inno Scores, sub-scores and ranking positions are published under CC BY 4.0: quote, chart or republish them with a credit and a link to the page you took them from. Every ranking page offers a dated CSV download. Third-party benchmark results (AnTuTu, Geekbench, 3DMark and others) belong to their publishers, are not covered by this licence, and are not included in the downloads.
Reference ranges & weights by category
Phones
- AnTuTu300,000–4,200,000 · w3 · log
- Geekbench 6 (single)1,200–4,000 · w2
- Geekbench 6 (multi)3,000–11,500 · w2
- RAM (max)6–24 GB · w1
- Pixel density300–560 ppi · w1
- Refresh rate60–165 Hz · w2
- Peak brightness800–5,500 nits · w2
- Screen-to-body80–95 % · w1
- Main camera12–200 MP · w1.5 · log
- Ultra-wide8–64 MP · w1
- Telephoto8–200 MP · w1 · log
- Optical zoom1–10 x · w1.5
- Front camera8–50 MP · w0.75
- OIS0–1 · w1
- Capacity3,000–7,000 mAh · w2
- Wired charging18–120 W · w1.5 · log
- Wireless charging0–80 W · w0.75
- Active use endurance5–15 h · w2.5
- OS update pledge2–8 years · w2
- 5G0–1 · w1
- NFC0–1 · w0.5
- Stereo speakers0–1 · w1
- eSIM0–1 · w0.5
Laptops
- Geekbench 6 (single)1,800–4,200 · w1.5
- Geekbench 6 (multi)8,000–28,000 · w2
- RAM8–128 GB · w1 · log
- Storage128–4,096 GB · w0.5 · log
- Refresh rate60–300 Hz · w1.5
- Peak brightness200–1,600 nits · w2
- Capacity40–100 Wh · w1
- Web browsing endurance4–19 h · w2.5
- Weight↓ 0.9–2.8 kg · w2.5
- Thickness↓ 10–27 mm · w1
Tablets
- AnTuTu300,000–3,500,000 · w2.5 · log
- Geekbench 6 (multi)2,000–15,000 · w1.5
- RAM4–16 GB · w1
- Pixel density160–360 ppi · w1
- Refresh rate60–144 Hz · w1.5
- Peak brightness400–1,600 nits · w1.5
- Capacity7,000–12,000 mAh · w1.5
- Endurance8–16 h · w2
- Charging10–130 W · w1
- Weight↓ 440–750 g · w1.5
- Thickness↓ 5–8 mm · w1
- Keyboard support0–1 · w1
- Front camera4–20 MP · w1
- Rear camera4–24 MP · w0.5
Mobile Chipsets
- Geekbench 6 (single)500–4,500 · w2
- Geekbench 6 (multi)1,500–15,000 · w2
- Max clock2–5 GHz · w1
- 3DMark Wild Life Extreme200–10,000 · w3
- GPU clock500–1,300 MHz · w0.5
Requirement also met by AnTuTu (scored under Benchmarks)
- AnTuTu250,000–6,000,000 · w3 · log
- AI performance5–100 TOPS · w2 · log
- Process↓ 2–7 nm · w2
- TDP (typical)↓ 4–15 W · w1
Laptop Processors
- Geekbench 6 (single)900–4,700 · w2
- Cinebench R23 (single)600–2,500 · w1.5
- Geekbench 6 (multi)2,000–30,000 · w2
- Cinebench R23 (multi)1,000–42,000 · w1.5
- PassMark3,000–62,000 · w1
- Process↓ 2–14 nm · w1
- Base TDP↓ 9–55 W · w2
- 3DMark Time Spy Graphics (iGPU)300–11,000 · w1 · log
- Max memory8–512 GB · w1 · log
- L3 cache4–128 MB · w1 · log
- Threads4–40 · w1
Desktop Processors
- Geekbench 6 (single)1,000–3,550 · w2
- Cinebench R23 (single)950–2,470 · w1.5
- Cinebench R23 (multi)1,500–41,000 · w2
- PassMark3,400–67,000 · w1.5
- Geekbench 6 (multi)2,800–24,000 · w1
- L3 cache8–96 MB · w3 · log
- Boost clock3.3–6.5 GHz · w1
Graphics Cards
- 3DMark Time Spy2,000–55,000 · w2
- 4K Ultra10–160 fps · w2
- 1440p Ultra15–250 fps · w1
- 3DMark Speed Way1,000–16,000 · w2.5
- VRAM2–36 GB · w1.5
- Bandwidth100–2,500 GB/s · w1.5 · log
- TDP↓ 115–600 W · w1.5
- Process↓ 2–20 nm · w0.5
- PassMark G3D5,000–45,000 · w1.5
- Shading units1,024–32,768 · w0.5 · log
↓ marks lower-is-better metrics · “w” is the metric's weight inside its sub-score · “log” marks logarithmic normalization. Ranges are periodically recalibrated as hardware generations move; recalibration recomputes every score on the site automatically.