Video summary
Why 3 different core types in this laptop computer
Main summary
Key takeaways
Laptop Reviewed (Sponsored by Geekom)
- Model: Geekom Geekbook M16
- CPU: Intel Core Ultra 9 185H (with “three core types” behavior)
- Chassis & basics:
- Aluminum chassis
- 16-inch display at 1080×1200
- 16GB RAM
- 512GB SSD
- Upgrades available (RAM/SSD/display) for an additional cost
Key Technological Concept: Intel “Three Core Types” Behavior
The Intel Ultra 9 185H uses:
- Performance cores
- Efficiency cores
- Low power cores
Windows core scheduling quirks (important finding)
- The reviewer couldn’t reliably force tasks onto logical core 0; Windows treated it more generically.
- Low power cores (two of them) were basically not used unless explicitly targeted (e.g., by specifying core numbers such as 20 or 21).
- This suggests Windows is not fully adapted to this heterogeneous core setup.
Benchmark Methodology & Results (Core-by-Core Analysis)
Integer benchmark (forcing workload onto specific cores/logical cores)
- Windows showed a progression of observed speeds:
- First “fast” cores
- Slower efficiency cores
- Two very slow low power cores
- Hyperthreading effects were noted as part of how cores were grouped in the test.
CRC benchmark (computing CRCs on multiple buffers concurrently)
- Single buffer:
- Performance difference between performance vs efficiency cores was only ~10%.
- Multiple buffers in parallel (more “pipeline friendly”):
- Performance cores pulled ahead more than efficiency cores.
- Efficiency cores remained ~half as fast as performance cores at best.
- The reviewer notes this gap is smaller than in an older Intel i9 Ultra comparison (previous generation had a larger efficiency gap).
- Hyperthreading was disabled in these comparisons (one task per physical core).
- Two tasks sharing hyperthreads together could reduce performance slightly.
“All-core” Performance Comparisons
- Compared against another Geekom model mentioned: IT13 with M16
- ~13% faster using all cores
- Low-power cores were excluded due to scheduling limitations
- Single-core performance still favors the M16, but direct comparison is limited because the other CPU context may differ (power limits/thermals).
Potential upside from low-power cores
- If Windows could use the two low-power cores effectively, the reviewer estimates up to ~5% net improvement.
- However, it doesn’t appear to happen automatically.
Why Low-Power Cores Weren’t Used (Interpretation)
Based on the reviewer’s research, the conclusion is:
- Low-power cores may be on a separate chiplet/die
- When idle, the system can run on the low-power die
- When background tasks occur (example given: Program Manager updates), the scheduler tends to move work to the main cores
- Possible cache behavior:
- Low-power cores may share Layer 3 cache with the main CPU
- When enabled, that could involve slower cache operation
- The reviewer didn’t see clear cache effects in simple benchmarks, likely because those tests didn’t stress L3 heavily
Process Node / Efficiency Discussion
The reviewer compares Intel lithography nodes across devices:
- One system uses 10nm lithography
- Another uses Intel 4, described as ~7nm
- A competitor CPU was mentioned with 4nm, but the reviewer notes it isn’t focused on efficiency
Overall takeaway:
- Smaller nodes are presumed to improve efficiency
- The reviewer finds the Ultra 9 remains competitive efficiency-wise versus some older/larger-node comparisons
Practical Performance & Workflow Tests
Daily use + video editing/rendering
- Workload: Kdenlive
- Comparison target: reviewer’s fastest desktop (A9 Max)
- Result:
- Laptop rendering took about 15% longer
- Described as “not bad”
- Noise perception:
- Fans were less noisy than on the desktop during that workload
Connectivity & Expansion
Ports
- Two USB-C (one used for charging)
- HDMI
- Two USB-A
Missing
- No built-in SD card reader (external required)
Included accessory
- A dock-like multi-port adapter enabling:
- charger + device connections
- fewer cables
Power Use, Battery, and Battery-Life Estimates
Power measurement method
- Measured using a TP-Link smart plug on the power adapter
- Adapter efficiency estimated around ~90%
Measured approximate power draw:
- Idle / “sitting around”: ~7.8W
- Kdenlive workload: ~20W
Battery claim vs observed behavior
- Laptop battery: 100Wh (flight-legal threshold)
- Reviewer expectation:
- ~10 hours for normal use
- ~5 hours for video editing (claims may be optimistic)
More realistic estimate (measured at the laptop level)
External monitor/peripherals disconnected:
- Reading: 3.5–4W
- Typing: ~7W
- Watching video (VCL player): ~7W
- YouTube: ~12W
- Video editing: ~20W
Battery life conclusion
- A very low-power “reading” scenario could last for many hours
- Reviewer cites a theoretical ~24 hours reading-only possibility
- Realistic work scenarios draw more power, reducing total runtime
User Experience Notes
- Keyboard preference:
- Reviewer dislikes very thin keyboards
- Prefers deeper key travel (doesn’t want a “flat keyboard” feel for daily work)
- Touchpad:
- Improved
- Supports multi-touch gestures like pinch-to-zoom
- Fingerprint sensor:
- Present
- Reviewer needed to configure it for their preferred finger
Hardware Serviceability / Internals
Back panel removal
- Held by 19 tiny star screws plus snap tabs
Material feel
- Back cover feels light
- Likely stamped aluminum sheet rather than machined
Upgrade access
- No obvious RAM slot → RAM likely soldered
- PCIe SSD underneath tape
- Extra M.2 slot available for an additional SSD
- Possible larger battery mod mentioned, but:
- flight legality suggests staying under 100Wh
Overall Review Stance (Evaluation)
- The reviewer’s impression:
- Exciting less due to novelty, more due to practical performance/power behavior
- Value argument:
- Offers more value than a similar Dell model
- Cited differences include:
- CPU performance
- battery size
- Dell using Windows 11 Home vs this laptop using Windows Pro
Main Speakers / Sources
- Primary speaker/source: The video reviewer (first-person), reviewing the Geekom Geekbook M16
- Company source mentioned: Geekom (sponsorship and product origin)
- OS/benchmark context: Windows (scheduler behavior discussed)