Dear community,
I’m currently daily driving an 11th Gen FW13 (Intel i7-1165G7) with no upgrades undertaken so far. I’m considering purchasing a new i5/Ryzen 5 mainboard but would like to know, which one will provide the best battery life? Is the Ultra better than the Ryzen AI 5/previous Ryzen Gen? I’m also considering switching from Win 10 to Linux, if that matters.
I’ll for sure upgrade the battery and the HDMI adapter, but will continue with the 60 Hz display.
Thanks in advance!
Intel Core Ultra or AMD 7040U, Ultra is better at lowest idle power, Ryzen 7040U is better when loaded. They are both more efficient than 11th gen, loaded or idling.
Thanks for the info! Is it your own experience or what’s it based on? Because the sources I’ve found are a bit contradictive… Also it seems that AMD has issues with sleep on Linux, isn’t it?
I tend towards the Ultra, as I’m also not regularly running demanding tasks.
I use 7840U FL13 and the energy efficiency is pretty good. Intel Core Ultra version has S3 sleep [deep] while AMD Ryzen 5/7 version only has s0ix [s2idle]. On actual use they should have similar consumption on sleep if you use Linux, on Windows, modern standby is going to drain your battery quicker and occasionally heat up your laptop. The latest Ryzen AI 300 has higher consumption on idle, compare the “idle average” here, this may be solved later on OS and driver optimization but I think it’s preferable to wait for a while
Thank you for the info! What’s your real world battery life roughly?
Those numbers for idle power consumption on AMD look like something was wrong when they ran that test. They’re normally like half that for me.
I have a similar question to the OP, except I’d like to specify I’d be willing to tinker (i.e. undervolt, minimal linux install (probably nixOS but maybe debian)) if possible to achieve the best possible battery life. I was initially leaning towards intel, but after reading this post I’m less sure. I suppose I could buy a 7040u and upgrade the battery and camera, but I’m not sure. Just to be clear, I’m split between the Ultra 5 125H and the 7640u. Or if the ryzen ai 340 is better than the other I could spring for that instead.
Are you using AI 9 HX 370? I got mine and can’t go lower than 6W at idle on empty F42 SB, most of the time I get 8-10W at idle on my configured system. Mission Center shows 4-5W being consumed by iGPU (maybe this stat is combined with whole SoC idk). Fedora Silverblue 42, GNOME 48.2 Wayland, VRR enabled 120Hz, 6.14.9-300.fc42.x86_64.
The 340 is mostly worse than the 7640u and definitely worse than the 7840u unless you really need the npu for something.
Power efficiency wise it may get better than the 7x40u series in time but currently it doesn’t seem to be.
The notebookcheck min idle number for the 7840u is way higher than mine but I assume those measurements were taken at launch. The ai ones are ridiculously high so there is likely a software issue there that’ll get resolved at some point.
The battery life, i.e. full charge to empty, is determined by the amount of watts the CPU, RAM, Keyboard backlight, display, ports etc. draw.
So, to reduce Watts, you can use less RAM, and turn the keyboard backlight off and put the CPU power slider in Power saver mode.
There are power draw improvements being made to the Linux kernel for AMD CPUs, but it is not bug free yet. I.e. Panel Self Refresh and CPU Idle.
In Linux, one can shut down individual CPU cores, but I have not seem much power savings using that.
That made very little difference in my testing.
As it looks even windows has some serious issues left if notebookcheck somehow got min idle numbers that high.
Probably cause it is very good at doing that itself and when you actually need to do stuff having more cores running slower tends to be more efficient than trying to do the same with less cores.
They got intel 1x gen, AMD Ryzen 7, Intel Core Ultra and AMD Ryzen AI all high. You can still estimate the idle power by comparing the relative power draw
I think it might depend a bit on which manufacturer of RAM you use.
Some people, on another thread, were seeing some differences when they used smaller RAM chips. It was having quite a significant difference to battery life while on suspend-to-idle.
Notebookcheck uses pre-installed (non-DIY) version, the newer RAMs are probably all ADATA AD5S560032G-S (change the 32 to different value according to the RAM size)
That is a whole different story, I was talking about on numbers. While on the bulk of the memory power draw seems to be from the memory controller and not the ram (4800 vs 5600 makes no difference, 16gb vs 32gb sticks makes no difference but running only one dimm does). The suspend numbers are pretty terrible no matter what you do.
Something relatively non obvious is that depending upon what software you use to report power you can end up with higher power. For example I prefer to use warp terminal, but I find running anything that monitors power by sampling (like powertop or btop) through warp terminal is much higher than using Konsole.
This is because warp uses graphics to render but Konsole doesn’t.
I don’t disagree that Phoenix is running higher than Strix, but the Notebook check numbers seem too high for what I expected.
I would suspect that this improves over time with BIOS updates though.
I tend to do the measurements over ssh so I can use the screen for other test scenarios. Networking seems to make a very small difference.
PCI Wireless network adapter or USB network? On AMD platforms active USB will block the deepest idle states.
The ax210 is pcie afaik. In my testing wifi/bt connected/off/entirely removed made no measurable difference on min idle. May have an impact on sleep power draw but I am not testing or using that.
Doing heavy upload or download does produce measurable differences compared to not but that’s not idle.
Yeah no concerns with that. I was just checking whether you were using a USB Ethernet adapter.