Not even close… M5 Max vs Razer Blade 18

Product reviewed (from the video)

2026 Razer Blade 18 (the newer version) compared primarily against the Apple MacBook M5 Max.

The Razer Blade 18 is highlighted as coming with:

• RTX 5090 (upgraded from RTX 4090 last year)
• 24GB VRAM (upgraded from 16GB last year)
• Intel Core Ultra 9 (new “Core Ultra 9” platform mentioned)
• Updated design details (including “updated stickers”)

---

Key features mentioned

• High-end creator/gaming laptop configuration
• Quiet operation this year: unlike last year’s model (fans “going crazy”), this unit is described as very quiet during testing
• Power/performance modes:
  • Runs in balanced mode out of the box, and is unplugged initially
  • Has a high performance mode (selected via Windows settings/power plans), including optional command-line duplication of a “high power” / “ultimate performance” scheme
• Good keyboard + centered trackpad (noted positively even with a numpad)
• Screen described as beautiful (no specific panel specs given)
• SSD / storage considerations:
  • MacBook uses Gen 5 drives internally (described as ~2x faster sequentially)
  • Razer also has strong internals, but external Thunderbolt behavior is mixed (see cons below)
• Thunderbolt 5 / external drive testing:
  • Razer: only one Thunderbolt 5 port (and ports are not labeled; you must know which side)
  • MacBook: three Thunderbolt 5 ports

---

Performance & benchmark results (numerical highlights)

Geekbench 7 (productivity-style CPU metrics)

Razer (unplugged / balanced mode):

• Single-core: 1,840
• Multi-core: 15,118

Razer (plugged in / High performance):

• Single-core: 2,918
• Multi-core: 24,149

Razer (plugged in / Balanced): “essentially same” (slightly different figures mentioned)

• Approx. Single-core: 2,908
• Approx. Multi-core: 23,957

MacBook M5 Max (bare metal):

• Single-core: 3,756
• Multi-core: 35,623

Big takeaway stated: plugged-in vs unplugged changes Razer performance the most (not high-perf vs balanced much when plugged in).

---

“Web tooling benchmark” (browser/dev workflow style)

• Windows laptop result (Speedometer-like score):
  • Razer (claimed): 48.3
  • MacBook: 49.4
• Comparison: basically the same
  • Note: one earlier Mac number was in a VM and was corrected later.

---

JS engine / V8 related benchmark (parsings/minifiers, CPU-heavy)

Throughput (runs/sec):

• Babel: Mac 39.6 vs Razer 34.4
• CoffeeScript: Mac 37.2 vs Razer 26.06
• TypeScript (main one mentioned):
  • Mac 53.39 vs Razer 47.29

Geometric mean:

• Mac 41.9 vs Razer 35.99

Conclusion: still very good, but Mac leads in these JS/tooling tests.

---

SSD performance (internal + external)

Internal sequential reads/writes:

• Razer: ~7,000 read / ~6,300 write
• Mac: ~13,769 read / ~13,555 write
• Stated: Mac is ~2x faster sequentially

Internal random I/O (dev-relevant for compilation):

• Razer “does a little bit better than the Mac,” though specific random metrics aren’t consistently/clearly stated for internal testing.

External Thunderbolt 5 (dip on Mac vs mixed results on Razer):

• When plugged into Mac’s TB5:
  • Mac shows lower sequential numbers (~6,960 read / ~6,600 write)
• Razer external test issues:
  • Sequential: read around ~4,000 (repeated)
  • Write drops extremely low in the first test: around ~1,300 write
• Rerunning on the correct Razer TB5 port improves sequential read, but writes remain very weak (write still ~1,300)

Stated criticism: Razer Thunderbolt support “still has some work to do” (framed in the video as an Intel-box issue).

---

CPU multi-core stress (Mattelroad algorithm)

• Mac: 11.3 seconds
• Razer (Windows): 13.5 seconds

Both peg all cores; Mac finishes faster in this test. The explanation partially attributes the result to fewer active cores (Mac shown as 12 vs Razer 24).

---

.NET compilation (Devbench-style; cold/warm/incremental)

“Hello world”

• Mentioned as not going as expected on the Razer first time, then moving to larger tests.

Large repo build (100,000 classes/namespaces):

• Mac cold: 45s | warm: 8.9s
• Razer cold: 56s | warm: 15s
  • Razer slower; warm build ~2x longer

Orchard CMS project (real project):

• Razer: cold 42s, warm 9s, incremental 7.9s
• Mac: cold 21.8s, warm 6s, incremental 5.9s

Conclusion: everything faster on the Mac in the tested real projects, despite creator expectation that Razer would win.

---

Visual Studio / Windows dev workflow

• “Create new project” / startup:
  • Razer “wins” by a small margin in one attempt
  • UI behavior: Visual Studio repeatedly prompts “Project configured to use SSL… don’t ask again” (described as happening multiple times)
• Subsequent warm build:
  • “About the same,” with Razer “just a tad bit faster”
• Framing: mixed, but sometimes favoring Razer due to Windows tool familiarity/workflow.

---

AI / local model inference results (high-level)

• LM Studio + Qwen 34B (multimodal):
  • Mac MLX attempt at 262,000 context length:
    • extremely slow (~25 tokens/s)
    • memory spilling occurs
  • After reducing context length to 68,000:
    • Razer is described as much faster
    • Mac reaches roughly 129–181 tokens/s (varies by run)

Key stated AI conclusion:

• Razer wins for:
  • Prompt processing (initial inference) on some larger models
  • Token generation (Razer “wins here, and it’s not even close”)

Token generation example numbers (from video):

• Qwen 3.6 9? (model name partially garbled in captions):
  • Razer: ~169.1 tokens/s
  • Mac: ~95 tokens/s

Unified memory advantage (Mac trick):

• Mac can run a much larger model via Llama CLI:
  • Qwen 3.5 122B mentioned
  • Prompt processing: ~139 tokens/s
  • Generation: ~47 tokens/s
  • Uses ~94GB of memory (explains GPU usage/growth)
• Claim: typical laptops can’t run that large a model—Mac’s 128GB unified memory enables it.

Overall AI verdict in the video:

• Razer is very strong where its GPU excels—especially generation speed
• Mac has a unique advantage for extremely large models, driven by memory capacity/unified memory

---

Pros (unique points mentioned)

• Big GPU upgrade this year (RTX 5090, 24GB VRAM) aimed at both gaming and AI
• Much quieter than last year’s model; impressive cooling / no fan chaos
• Strong high-performance behavior when plugged in
• Keyboard/trackpad quality (centered trackpad even with a numpad)
• Excellent screen (described as beautiful)
• Windows/dev workflow can benefit (Visual Studio steps sometimes faster on Razer)
• AI performance can be excellent, particularly token generation speed (Razer “not even close” in the tested case)
• Overall framing: “wins some, loses some”, but called a good buy for gaming + AI

---

Cons (unique points mentioned)

• Value/price concern: very expensive; last year’s complaint was what you get for the money
• Unplugged performance is significantly weaker vs when plugged in (Geekbench unplugged much lower)
• CPU/tooling and compilation benchmarks favor the Mac:
  • Better Geekbench multi-core
  • Better web tooling / JS engine throughput
  • Faster .NET compilation on tested real projects
• Thunderbolt/external storage performance issues on the Razer:
  • Writes drop very low (~1300 in repeated external TB5 tests)
  • Ports not labeled; correct port matters
  • “Still have some work to do” on Thunderbolt support
• Visual Studio SSL prompt behavior annoyance (multiple “don’t ask again” prompts)
• Potential buyer trap: ensure you buy the correct model year (see below)

---

Comparisons made

• Main comparison throughout: Razer Blade 18 vs MacBook M5 Max
  • Mac wins: Geekbench, JS tooling/V8 tests (geometric mean, TS), multi-core stress (Mattelroad), and .NET compilation benchmarks
  • Razer wins / trades off: Visual Studio Windows workflow sometimes faster; AI token generation often faster on Razer
• Context: prior year’s Razer vs Mac was disappointing; the video says this year improves, citing quiet operation + RTX 5090 + VRAM

---

Model-year / buying advice (important “watch out”)

• Buy guidance:
  • If shopping, buy the 2026 model with Intel Core Ultra 9 290HX, not the 2025 edition, which the speaker says will disappoint
• Mentions a related video comparing against the 2025 version

---

Overall verdict (concise recommendation)

Best fit: The 2026 Razer Blade 18 is recommended if you need a single laptop for serious gaming and local AI work, and you’ll typically keep it plugged in for peak performance.

Why not everyone should buy it: For pure productivity CPU-heavy work (browser tooling, JavaScript build steps, and especially .NET compilation), the MacBook M5 Max generally outperforms it in these tests. External storage/TB5 performance also looks problematic on the Razer (especially writes).

Final framing from the video: “This year it absolutely does” justify the price for the tested AI/gaming workload, and “you win some, you lose some” depending on priorities.

---

Unique points by speaker

• Single primary speaker (“Alex” referenced): all test results, comparisons, hardware observations, AI conclusions, and the buying warning about 2026 vs 2025 model.
• Brief second-person interjections: only rhetorical audience comments addressed by Alex (e.g., unplugged vs high power, and “I don’t do web development why should I care”).