A couple of things. I have updated the story with a little more information. The reason I didn't include AMD laptops in the first draft is because we haven't tested any that lasted more than 9 hours on a charge within the last 12 months. I went back and looked at our colleagues at Laptop Mag, which uses our same test and test conditions, and found two AMD laptops they reviewed in the last year that were more than 9 hours. Only one exceeded 10 hours.
I have also included a few paragraphs on historical testing Laptop Mag did that shows x86 (Intel in this case) laptops getting more than 17 hours on a charge, due to high battery capacity. Yes, that's a trade-off and Snapdragon is pretty efficient per Whr. However, it is not always more efficient than Intel.
Obviously, we can all disagree upon what constitutes "all day" battery life or two-day battery life. I'm definitely guestimating when I say 14 or 15 hours on our test would be one solid workday plus maybe a little more. What's a workday to you? Are you putting the laptop to sleep for some of it? Are you pumping up the brightness or doing more than web surfing? Like anything else, our battery test is a snapshot of a particular, repeatable workload. I would impressed if we crossed the 20-hour rubicon. The 15-hour rubicon has been crossed by Intel PCs in the recent past.
For me to jump on the Snapdragon PC laptop bandwagon, would basically require a repeat of what happend on the smartphone side
Until and including the Snapdragon 820, everything from the first Galaxy S had me exposed to "battery anxiety": Just using the phone during the day (no antisocial networks, nor audio or video, mostly just browsing and e-mail and actually a bit of phoning) had it drop visibly during an average working day, to a point where I thought I costantly had to manage around the remaining capacity.
That changed with the first 835 based OnePlus 5, which just never got below 70% charge even on a long day no matter what I did with it (except hours of navigation). And since then it's never come back, even with a bit more daily browsing time and today's full sized browsers waging epic battles against ad-servers.
It's always on (although not the screen), instantly usable 24x7 and would last for days of regular use on a single charge.
And it's a very different story on laptops so far. There I still manage the battery charge, which typically involves bringing and even using a charger in a conference. And when I should have forgotten it (it happened more often that I care to admit), it involves hunting around for someone with an extra USB-C, while increasingly all these fruity cult guys don't even bring their incompatible ones to multi-day events.
So I try to gauge and measure my usage, have it go to standby on a lid close and then use the lid to have it switch. On lunch breaks I might even have it suspend to disk and then worry, if that suspend/resume cycle won't actually eat more energy than having it stay in suspend-to-RAM for 30 minutes.
And then, when I wanted to watch a relaxing movie or do some Kindle reading after that long day in the hotel, the power might just run out, unless I bought a new charger at the corner store.
For me the ARM promise isn't top Cinebench figures, even if I'd rather have the laptop beat my OnePlus 5+ (~€230 with VAT for the 32GB version) in terms of performance at 5-10x the price. It's about sipping
phone amounts of Wattage at every opportunity to darken most of its silicon. And switching between next to zero to full attention in the time it takes me to operate the lid, while with the lid closed, it should quite simply last the week my mobile phone will last with nothing going on. In short: Power buttons on a phone are serving many purposes these days, but rarely ever to actually turn phones on or off. I want the same for my other personal computer, if it can't just be an extra screen for various corporate and private PC enclaves on my phone (which I'd still prefer).
I've forgotton to shut down my backup phones, which I sync and charge before bringing them along on multi-day trips. Only to discover that they were basically still full while switched on for the week, before I stored them away. And those phones have 16GB of RAM, too, SoC stacked mobile variants and slightly more expensive than an SO-DIMM, but if I need to pay for soldered, I at least want similar power savings.
I don't tend to buy 10 Watt notebooks, e.g. machines with only 2P cores. I go for machines that will support a
maximum of 28-45 Watts, because I want the ability to also use them as workstations with plugged-in power. That's why I also tend to be very generous with RAM and SSD storage, so I don't need a separate machine for that on the road. They obviously can't replace the really big iron or even a gamer rig, but for that I got plenty of remote connect if needed.
But on a modern laptop I expect even a 8-12 core machine capable near workstation workloads at 50 Watts to use phone wattage, when nothing is going on. It shouldn't matter how many transistors your SoC has, if you can turn off all you do not need. That's where x86 might have never quite achived what mobile chips managed, but that's exactly where I expect a
Snapdragon to deliver.
My Ryzen 5800U allows 28Watt peaks but drops to 15Watts on sustained 8 core loads. It tends last much longer than its Alder Lake competition with an i5-12500H, while both are mostly twiddling thumbs. And when it comes to peak loads, Alder Lake takes a much deeper swipe from the bottle for results that aren't much better.
As with the desktops to me that's indicated a much better general energy efficiency from AMD than Intel, but that says nothing about minimal energy consumption on a desktop idle system that remains ready to be used instantly. And so much of that isn't even dictated by the CPU cores, which should remain stopped for most of the time.
Both have failed me on a full conference day, and I find that very disappointing given the lightness of the actual load I exposed them to, with mostly a bit of note taking and reference-surfing.
Current testing seems to focus on maximum loads or medium constant loads, because those are traditional and 'easiest' to measure. I put the quotes, because I fully understand that measuring in a repeatable and comparable manner as such is becoming an ever greater nightmare.
But with this architecture it's the ability to suck very near zero power for hours if not days or indeed a week, yet 'instantly' deliver near desktop computing power for a couple of seconds and perhaps ordinary notebook power for minutes or a few hours after that, only to go back to catatonic consumption at every opportunity, which has people spend the extra buck.
And at least in my case, it won't be Copilot, nor very likely anything NPU related for the life-time of the device. Because for that stuff, I got big iron and one of the other things that would keep me from buying "Elite" notebooks, is the inability to turn the NPU off in the BIOS and all those "intruvasions" that M$ keeps throwing at what should be treated as sovereign owners of personal computers, not victims of home intrusion and coersion.
Facebook
Twitter
Instagram