Intel Core Ultra 7 "Lunar Lake" Performance Up By ~22% With ASUS Linux Fix

[drakonas777]

Link please. I've only seen similarly priced ones.

Just ignore sophisticles and his nonsensical posts like that one. The dude is emotional and offended by current Intel problems and tries to cope with that spreading lies and FUD about AMD. Most likely spends his evenings crying in the dark corner because of cancelled Beast Lake cores

[mateosss]

Thanks for the continuous testing on Linux! So far I've been enjoying this laptop quite a bit with Ubuntu 24.10 even though it has some micro stutters every now and then (due to the graphics driver I suppose, on Windows this doesn't happen) the battery life and responsiveness feel really good!

One comment about the benchmarks is that it might be worth, for these kind of laptops, to separate single thread and multithread benchmarks. At least for me, single thread performance was much more important as a deciding factor to get lunar lake since it gives you an idea of how "snappy" the experience might be. Plus, of course, a 12 core beast like the hx 370 will beat a 4-performance-cores laptop in multithreading workloads

I am measuring the power consumption under linux regularly and I cannot believe the constant draw being always between 2-4W when doing light tasks. An interesting thing is that moving windows around, maximizing, minimizing, etc (i.e., updating many screen pixels at once) does not increase the consumed wattage like it did in older intel cpus I had. I guess this might've been the rationale for having a beefier GPU on the chip

[ninele7]

But we were talking about the current generation.
The cheapest AMD laptop I can find costs sub 300$ (Ryzen 7000 based) (https://www.amazon.com/-/dp/B0BS4BP8FB/)
The cheapest AMD laptop which is Ryzen 8000 based is sub 600$ (https://www.amazon.com/-/dp/B0DCGMLNL4/).
So what is your point?

Never say "Ryzen 7000 based" when talking about AMD laptops, better name specific architecture.

300$ laptop that you've mentioned is a low end zen2 based model (which was available in laptops for 4 years). It's by no means current generation. In CPU name 7320U, the most important number is "2" which specifies architecture.

[Anux]

One comment about the benchmarks is that it might be worth, for these kind of laptops, to separate single thread and multithread benchmarks. At least for me, single thread performance was much more important as a deciding factor to get lunar lake since it gives you an idea of how "snappy" the experience might be. Plus, of course, a 12 core beast like the hx 370 will beat a 4-performance-cores laptop in multithreading workloads

I don't see that much benefit but I would like some indication which benchmark used how many cores (or percentage of utilization). There are rarely any pure single threaded benchmarks anymore (flac and lame come to mind) most are limited by the fastest thread but still use more than 4 threads (games) and others have a certain point where more threads don't help (video encoders).

If you just want to have a guide for a buying decision you should only look at the programs you actually gonna use and that are to slow on your current system. Flac and lame might be single threaded but also are fast enough to be used with a 10 year old cpu.

[oleid]

Never say "Ryzen 7000 based" when talking about AMD laptops, better name specific architecture.

300$ laptop that you've mentioned is a low end zen2 based model (which was available in laptops for 4 years). It's by no means current generation. In CPU name 7320U, the most important number is "2" which specifies architecture.

Well, yeah, current generation is Ryzen AI 3xx. But apparently sophisticles wasn't interested in "current generation" anymore. Hence the elder laptops.

[smitty3268]

Well, yeah, current generation is Ryzen AI 3xx. But apparently sophisticles wasn't interested in "current generation" anymore. Hence the elder laptops.

Are we sure sophisticles isn't affiliated with userbenchmark?

[retrixe]

Thanks for the continuous testing on Linux! So far I've been enjoying this laptop quite a bit with Ubuntu 24.10 even though it has some micro stutters every now and then (due to the graphics driver I suppose, on Windows this doesn't happen) the battery life and responsiveness feel really good!

One comment about the benchmarks is that it might be worth, for these kind of laptops, to separate single thread and multithread benchmarks. At least for me, single thread performance was much more important as a deciding factor to get lunar lake since it gives you an idea of how "snappy" the experience might be. Plus, of course, a 12 core beast like the hx 370 will beat a 4-performance-cores laptop in multithreading workloads

I am measuring the power consumption under linux regularly and I cannot believe the constant draw being always between 2-4W when doing light tasks. An interesting thing is that moving windows around, maximizing, minimizing, etc (i.e., updating many screen pixels at once) does not increase the consumed wattage like it did in older intel cpus I had. I guess this might've been the rationale for having a beefier GPU on the chip

Hi, I'm interested in buying this laptop and I'm curious if this 2-4W is the power draw of the total system as reported by upower or powertop or such? And if so, at what brightness do you typically operate your laptop at, and what light tasks are you typically performing

If this laptop is really able to maintain 2-4w system power when doing web browsing or watching videos or such, this is a really compelling machine I might buy, especially paired with a 75whr battery : D

[mateosss]

retrixe The 2-4W was under a very lightweight usage of 2-3 firefox tabs + mostly reading news articles and pdf with some idle times while I was doings some other stuff in my workstation, in mid brightness, balanced profile, ubuntu 24.10 kernel 6.11.0-9


Today I did a more intensive browsing session, i.e., ~15 firefox tabs of relatively high javascript execution, with no rest time since I needed to interact a lot with all websites and was jumping around quite a bit between tabs. Things like discord, gmail and other mail clients, calendar, whatsapp/telegram/matrix web, and some news sites. Average brightness, balanced power profile. I took samples for an hour and the average consumption was 5.30 watts this time. So that would be like 13hs of total battery with this workload.

I am measuring the power draw with a custom script that reads from `/sys/class/power_supply/BAT0/*`. I uploaded it here if you are interested: https://gist.github.com/mateosss/8ae...aa2f84d8432533
The script matches the remaining battery times that the ubuntu battery UI reports.

Some other nice things I've been finding about this laptop is that it is pretty good for C++ development. Granted the first build takes a lot of time due to the bad multithreading performance, but after that, recompiling a few compilation units after every change is really fast. At least a bit faster than my two-generations-old workstation. Gaming performance on windows is also another really great thing for my needs. Death stranding for example works at 40-60 fps with mid graphics settings (no XeSS) while I am on battery. For valorant I get 300-400 fps on low settings at full 2880x1440 resolution. Although of course, battery life will get reduced to ~2hs if you maximize the gpu usage like this, better to use vsync.

So far I am quite happy with this thing
Apple hardware is still better for sure, but this closes the gap more than enough for my particular preferences for x86 (mostly being able to use linux natively and play some games).