flak rss random

The Zephyrus G14 is not a MacBook Pro 14, and Sundry Observations and Insights, Keen and Mundane, Arising from an Inquiry into Power and Performance

Two years ago, all the cool tech influences got an Asus Zephyrus G14 laptop, and all I could think about was getting one, but then I didn’t. The ports were closed, the boat got stuck, whatever, but mostly the promised QHD screen model never seemed to materialize. Now finally it has, and it’s 16:10 even, so finally I got one.

One other thing that’s changed is the arrival of the Apple M1. The G14 now has a competitor in the form of the MacBook Pro 14 that didn’t exist at the time of launch.


I bought a gaming laptop because I wanted to play games, duh. But I could probably use it to do other stuff, and I’d prefer that it not become a fiery tornado. The M1 laptops, all of them, have reputations for being particularly cool and quiet machines. The MacBook Pro 14 is very similar in size and weight, making for a good comparison. I got the $1650 base model G14, while the P14 starts at $2000. So Apple is a little more, but maybe worth it for that intangible build quality, efficient operation, etc. It’s tough to make an exact comparison, but the GPU in the G14 is considerably more powerful in my estimation, requiring a step up to the M1 Max, with a starting price of $3000. So we’ve gone from slightly more to almost twice as much. Value proposition eroding quickly.

Never mind that I bought this to play games, and I’m not talking about “oh, yeah, Cyberpunk 2077 plays great on a macbook, just sign up for Stadia” gaming, but if the Pro 14 is so much faster and more efficient and generally more pleasant than my G14, I know I’ll be filled with buyer’s remorse. Is there anything that can be done? Can the G14 be a Cool’n’Quiet AMD machine after all?


The G14 likes to run hot. The good news is there’s a silent power profile which lives up to its name. It’s not completely, absolutely dead silent when the fans are spinning, but it’s pretty close to inaudible unless you are leaning in and in a very quiet room. In a public setting like a coffee shop, nobody will ever know they’re on. And frequently the fans will be off entirely. So core objective #1, quiet operation, achieved.

The silent setting doesn’t control power or performance directly. Everything runs close to full power, until it gets too hot, and then once the fans reach maximum speed, the system throttles to maintain temps and noise within limits. It will get toasty.

There’s also a performance profile, which gets a bit too loud for public display in my opinion, and a ludicrous turbo profile, which seems like benchmark bait.

hot idling

The CPU likes to jump up into its turbo frequencies, where it draws much more power and generates much more heat. It will do this when the system is near idle at the slightest disturbance, like scrolling a web page. Naturally, these are the least efficient frequencies, and the system sheds quite a bit of heat. Not usually enough to turn on the fans, but the case is hot, and as soon as you do start some real work, the fans do go on. (They are quiet, but if you keep a mouse near the side, you may feel the air.)

There is a powercfg command to disable turbo boost states. It’s per power profile, so I turned it off for silent, but left turbo enabled for the performance plan, so it’s still there in cases when I do want go extra fast. The Windows documentation for server power tuning also applies to desktop systems.

I think the theory behind race to idle is at odds with the reality of just how much power this CPU will draw, even in tiny bursts, at top frequencies, and also how much work it can get done at a mere 3.3 GHz, the top not turbo clock. Ramping up to 4.5 GHz just wastes power.

With this change, the system remains very cool, yet still extremely responsive. Battery life subjectively seems better as well. It gets hot in heavy use, to be somewhat expected, but for everyday use, I think we’ve achieved a MacBook like state of zen.

hot sleeping

Mostly. Windows is here to ruin the party. According to powercfg this laptop does not support connected standby, so I thought I’d be safe from hot flashes in the middle of the night, but apparently not. It seems to hold on to wifi signals (as seen by the AP) longer than I would have expected for a disconnected laptop.

A few times I’d close the lid at night, then find the laptop quite hot in the morning. My expectation is that a sleeping laptop should be room temperature. Pretty frying reasonable I think. Microsoft disagrees. Checking event viewer, it seems to have spent the entire night doing the guid to the guid every ten minutes.

I think it only does this on AC power, but how should I know. For now, I found it’s pretty easy to tap Fn-F12 for airplane mode when sleeping to minimize the amount of trouble it can get up to. Can’t install updates if it can’t download updates.


I am quite possibly imagining this, but it seems some of the mid-night madness heat is a bug where the dGPU turns on and then doesn’t turn off. I’ve noticed the system cools off when switching to iGPU only. It doesn’t immediately get hot after turning the dGPU back on, however. It’s not too many clicks, but it’s annoying. I found some powershell to accomplish the same result.

Get-PnpDevice | where {$_.friendlyname -like "AMD Radeon RX 6700S"} | Disable-PnpDevice

Get-PnpDevice | where {$_.friendlyname -like "AMD Radeon RX 6700S"} | Enable-PnpDevice

I don’t believe this is particularly wise.


For me, depending on task, battery life can be irrelevant, sufficient, or extraneous. Heavy load on the G14 reduces battery life to around two hours, which is irrelevant. Just bring a charger. Going the other way, after 8+ hours I find additional battery life to be of limited use. I’m never using it that long. G14 thankfully falls into this realm. For browsing and light programming (read documentation, write code, stare out window) I think it’s at least 10 hours. Six hours of use only dropped battery to 50 percent.

The included power brick is an absolute monster. It seems to weigh as much as the laptop itself. That’s what it takes to power this thing in the absurd turbo mode. For portability, a USB-C charger is recommended.

One important caveat to note is that the system will happily use more power than USB-C provides, drawing down the battery. A 65W charger is definitely not enough. Even a 100W charger will fail to keep up at times, even on silent profile. It’s a noise profile, not strictly a power profile, if the cooling can keep up. I estimate between 110W and 120W is the limit for silent. (Total system draw, including screen and peripherals.)

I found one of those supercool cables with the builtin LED power meter really useful. Makes it possible to monitor power usage in real time, maybe turn down shadows or ambient occlusion, to keep things under 90W, without having to switch programs. (Or to observe power while the lid is closed. 20W with the lid closed is a rather restless sleep.)

The Asus software has a manual mode as well, which does allow direct control over CPU and GPU power limits, but I’ve decided not to fuss with that too much.

Concretely, I found I could play Frostpunk at max settings, but the battery would slowly drain as a consequence of the relentless pursuit of 60 fps. Turning down the snow and particle effects, which I can’t tell the difference, keeps system power draw down around 85W. I find this preferable to imposing a GPU power limit because I don’t mind short bursts of activity requiring more power. My criteria are never loud, controlled by the silent profile, and average power less than 100W, controlled by game settings.


I am not the first to notice that increasing power to a CPU results in increasing performance, but the return is sublinear.

There’s one slide at the 3:00 minute mark in 6900HS vs M1 that compares reduced power. Here’s the table form.

Single coreMulti core
G14 70W159513736
G14 45W154312395
M1 Pro153212381

Another video, another 3:00 slide compares with Intel, but also includes more data points for 6900HS power and performance in Blender. I’ve added some relative numbers. The ratio only gets better as power decreases.


Anandtech has several slides analyzing the same thing on page 4 of their 6900HS review.

Conveniently, all these tech reviewers used the same G14 laptop that I have for their testing.


The 2021 G14 is on sale for quite a bit less than the 2022 model. For me, I wouldn’t consider it because I want at least QHD resolution, and 16:10 is a very nice bonus. The 2022 model uses an AMD GPU, and it’s maybe a little more powerful than the Nvidia RTX it replaces, but its ray tracing is much weaker and it lacks DLSS. For exclusive 1080p@120Hz gaming, the older model may be better fit.

The 6900HS processor is nearly identical in terms of performance to its predecessor. Zen 3+ is identical to Zen 3, except for some additional power saving tech, especially while idle. So this is obviously of interest to me as well.


The Zephyrus G14 is not a MacBook Pro 14.

Yet, it can come fairly close. It takes a few tweaks, and the mimicry is incomplete, but there’s a cool, quiet, efficient number crunching pixel pumping computer in here somewhere.

There’s a tendency for everyone to look at the redline max power numbers when comparing systems, but one need not operate the system at that level. Apple made the right decision here, making an efficient processor, and then keeping it efficient, delivering reliable and consistent performance in a variety of models. Not everyone will spend the time to fuss with settings, and instead complain about how hot and loud their laptop is. Alas, laptops that show up slightly lower on benchmark graphs probably don’t sell as well. It’s nice to have options, though. I think AMD is not so far behind as people think.

Posted 08 Apr 2022 13:47 by tedu Updated: 05 May 2022 18:44
Tagged: computers