His underlying point holds up well. Use headline specs to build a shortlist, then let independent testing decide. "Ignore" is too strong for several of the five, though. Some of his specific claims also go further than the evidence supports. Here is each one, with context.
Monitor peak brightness: ask "how bright, where, and for how long?"
Singh says a monitor's advertised peak brightness usually refers to a tiny 1% or 2% window of the screen. He says full-screen brightness is often only 20–30% of that figure. The window-size point is sound. Treat the 20–30% ratio as his estimate, not a rule, because it varies by panel type and test conditions.
VESA's DisplayHDR criteria show why one number can't carry the whole story. The current criteria measure an 8% center patch for small-area peaks. The same standard also tests full-screen output in two ways. One is a quick flash test, representing brief moments like an explosion in a game or movie. The other is a long-duration test of stable full-screen luminance. Peak and sustained brightness are different measurements. Both are legitimate, and they answer different questions.
A small-window peak does matter for bright HDR highlights, such as sparks, sun glints and neon signs. It says much less about a bright snowy scene or a white spreadsheet.
What to do:
- Look for reviews that measure brightness at several window sizes, not just the smallest one.
- Check sustained output, because some displays dim themselves during large bright scenes.
- Match the numbers to your room. A dim cave needs less brightness than a sunlit desk.
Motherboard Wi-Fi 7: a feature, not a scam
Singh calls paying extra for Wi-Fi 7 on a motherboard "absurd right now". That is a defensible opinion for people on Ethernet or older routers. Calling the technology useless is a stretch.
Intel describes Wi-Fi 7 as 802.11be. Its headline features are 320 MHz channels and 4K QAM, which Intel says give up to 2.4 times the peak throughput of Wi-Fi 6/6E. There is also Multi-Link Operation (MLO), which lets a device use multiple bands at once. Intel says MLO improves throughput, latency and reliability.
Intel's own footnote is a useful reality check. The 5.76 Gbps maximum for a 2x2 client is a theoretical figure. That figure assumes 320 MHz channels, 4096-QAM and MLO. Your garage-bound desktop will not see that number.
The practical question is whether you have, or plan to buy, a Wi-Fi 7 access point. If you don't, the feature does nothing for you today. If your PC can't run an Ethernet cable, or you're buying a router anyway, it is cheap future-proofing.
What to do:
- If you game over Ethernet, don't pay extra for Wi-Fi 7.
- If you rely on Wi-Fi and will upgrade your router, it is worth considering.
- If you're on Wi-Fi 6 or 6E and happy, you're fine.
One of Singh's claims, that Wi-Fi 8 specifications are already being worked on, is something I couldn't confirm in the sources I consulted. I've left it out.
VRM phase count: the number vendors love to inflate
Singh is on solid ground here. A headline phase count can mislead, and the details matter more.
One detail is phase doublers. Phase doublers let a controller with fewer native outputs advertise a bigger number. One KitGuru analysis found a board built on a seven-phase controller with four doublers. Another source on a high-end MSI board says the CPU still only sees 13 interleaved phases, not the advertised 26. Doubled designs aren't automatically bad, since they can spread the load across more power stages. The catch is that the marketing number isn't comparable to a board with true native phases.
Amperage matters as much as count. Tom's Hardware says you need both phase count and MOSFET output. It gives the example of a 12-phase VRM with 80A MOSFETs outputting up to 960A, against 2,200A for a 20-phase design with 110A parts. That is theoretical capacity, not a verdict on any one board.
Real-world results can be underwhelming. In one Tom's Hardware reader comment (not the article's own testing), a user swapped a 12-phase 80A board for a 16-phase 90A board with a mildly overclocked Core Ultra 7 265K and Core Ultra 9 285K. They report that cooling the CPU silicon was the thermal limit in both cases, not the board's power delivery. It is anecdotal, but it fits the broader point.
Tom's Hardware also notes the thermal stakes. Entry-level boards that list flagship CPU support can throttle under heavy workloads without adequate cooling.
Singh offers specific thresholds: 12 true phases with 60–70A stages for most gamers, and 16 phases above 70A for enthusiasts. I couldn't verify these as universal rules. Treat them as one builder's rule of thumb. The right answer depends on the CPU, since a 65W part has very different needs from a flagship chip under all-core load.
What to do:
- Pick your CPU first, then shop for boards.
- Find a review of that exact board that reports VRM temperatures under sustained load.
- Check for a real heatsink, not a decorative shroud.
- Be skeptical of any phase count you can't match to a power-stage rating.
GPU AI TOPS: a spec for a different kind of buyer
TOPS means trillions of operations per second, a theoretical measure of AI throughput. Nvidia leaned on it heavily when it unveiled the GeForce RTX 50 series. Its newsroom announcement said the RTX 5090 provides over 3,352 AI TOPS. It also said DLSS 4 Multi Frame Generation can generate up to three frames per rendered frame. The announcement also listed prices next to TOPS figures: the RTX 5090 with 3,352 AI TOPS at $1,999, and the RTX 5080 with 1,801 AI TOPS at $999. That pairing is what Singh found jarring.
These are manufacturer claims, not game benchmarks. TOPS may matter if you run local AI models or AI-driven creative tools. For a pure gaming buy, the useful questions are different:
- What are the frame rates at your resolution and settings?
- How do the games you play handle upscaling and frame generation?
- How is the image quality and latency with those features on?
- How much VRAM does the card have, and what does it cost?
Singh's remark that the RTX 50 launch was the "official transition" of Nvidia into an AI chipmaker is his opinion. His claim that "no one cared" about TOPS is a generalization. I wouldn't treat either as established fact. The marketing emphasis is real, though.
CPU boost clock: a ceiling, not a promise
Singh says boost clocks are effectively fiction for gaming. The accurate version is that they are conditional.
Intel says as much. Intel's Turbo Boost page states that a given CPU may not always reach its Max Turbo Frequency, because the speed depends on workload and thermal headroom. AMD's Ryzen Max Boost, per its own definition, is the maximum frequency of a single core running a bursty, single-threaded workload. Neither vendor promises all-core operation at that number.
Singh is right that all-core speeds sit lower, and that power and thermal limits cap the clock. He goes too far in saying boost clocks "don't exist" in real use. Single-thread boost behavior can matter in lightly threaded games. A maximum boost number just can't predict frame rates by itself. Cache size, memory, the game engine and GPU limits all matter, often more.
What to do:
- Compare benchmarks in the games you play, at your resolution.
- Look at power draw and temperatures, not just clocks.
- Don't compare boost numbers across different architectures and expect them to line up.
The shared principle
Across all five, the pattern is the same: a spec is a clue, not a result.
| Spec | What it really tells you | What to check instead |
|---|---|---|
| Monitor peak brightness | Small-window highlight capability | Multi-window and sustained brightness tests |
| Motherboard Wi-Fi 7 | Capability if you have matching gear | Your router, your use case, Ethernet availability |
| VRM phase count | Marketing count, possibly doubled | Power-stage amps, heatsink, VRM thermal tests |
| GPU AI TOPS | Theoretical AI throughput | Game benchmarks, VRAM, price, feature support |
| CPU boost clock | Single-core ceiling under ideal conditions | Gaming benchmarks, power and thermal behavior |
A fair critique of the "ignore" framing
There's a risk in reading this too literally. Singh's piece is an opinion column from a veteran builder, and it leans on his own experience and a few rules of thumb. Some of his numbers (the 20–30% full-screen figure, the VRM thresholds) aren't backed by a cited test. His conclusion is stronger than his evidence in places.
His method is the part to keep. Shortlist with specs, then verify with testing relevant to your use. That advice holds up whether you're building a Windows 11 gaming rig or just picking a monitor for it. Reputable reviewers measure brightness at multiple window sizes, log VRM temperatures, and run actual games. Do your final comparison there, and let the marketing slide stay on the slide.