I am running some comparisons of various PCs (two desktops, two laptops). Specifically a Jupyter Notebook in VSCode with 474 code cells: 14 simulations of MTK models (rigid models, PDEs), with plotting, one animation. So (for me) a realistic mix of code. All computers are set up with Julia Num Threads in VSCode; I don’t know if it still uses single core, though.
Kind of surprising: both desktops are slower if I re-run without restarting the notebook. The desktops have 64 GB RAM and 32 GB RAM (intel vs. AMD). The laptops have 32 GB RAM and 64 GB RAM (lunar lake vs. M5).
Apart from this, the performance of the MacBook is impressive. I will do some testing of the battery life of the MacBook vs. the Lenovo laptop with Lunar Lake.
Are the above results as one should expect?
Any ideas of why desktops appear to be faster when I restart the notebook? (Not restarting should save time by avoiding compilation; not restarting may leave “large” data structures?)
Apart from this, the laptops use default power settings. For the MacBook, I used “caffeinated” setting.
Your guess is as good as anyone else’s with little information. Larger persisting allocations could explain the same allocations taking more time, but it could also be the operating system prioritizing other processes the longer the Julia process lasts. I wouldn’t worry too much about it unless you’re observing RAM usage growing when it shouldn’t.
So, my impression is that the MacBook is by far the fastest and most snappy of the machines in this test, not only from the numbers. As one other example: Inkscape opens much faster on the MacBook than on any of the Windows machines.
Of course, another important thing is what I feel about the software system (OS, but also software I use). So far, I have very mixed feelings about this.
VSCode: I run Jupyter notebooks a lot. Jupyter notebook is notoriously poor in 2 areas… (i) loosing contact with the Julia server and defaulting to Python, where it often gets stuck (both Windows and Mac). (ii) Search and replace is a nightmare in the sense that it does not move linearly through the occurrences in the file (both Windows and Mac). A third area that appears tricky on MacOS… (iii) Sometimes, running Julia code results in zero response from the Mac. Possibly because of sandboxing, or something.
LyX: Installation went ok, but I had to fix a lot of things due to different paths for OneDrive/cloud syncing on Windows and Mac. However, I spent a few hours trying to find out why Mac Preview didn’t work properly. In the end, I found that Preview lacked some vector fonts in Font Book. I have never had such problems on Windows. Perhaps MikTeX is better than MacTeX?
Mac keyboard. I use a non-US keyboard. Several important keys are missing from the keyboard… piping, tilde, bracket ([) and brace ({), etc., etc. So I have to reshape my typing style, and also try to figure out what keyboard shortcuts I should use. I miss the Windows Alt key to highlight menu letters that activate a menu.
One thing that almost qualifies for a prison sentence for Mac developers… “Cmd + Ctrl + Shift + 4” to make a screen capture to the clipboard??? 4 left hand keys??? (I installed Raycast and made a shortcut).
Etc.
I haven’t had time to compare battery life. “Everyone” says I should expect a full workday battery life, which to me means 7-9 hours. I expect to be disappointed.
So far, my thinking is that, say, an Asus Expertbook Ultra, or similar Windows machines, would probably offer at least half the computational power, and weigh in at around 1 kg instead of 1.5-1.6 kg. And be sufficient for my kind of work.
Anyways, I will test the MacBook some more time. Some of my frustrations are, of course, due to a new OS.
Why would ‘everyone’ expect a full days battery life?
In my experience most people sit at desks with power sockets. Or use docking stations.
Sure- if you are travelling then long battery life is highly useful. But surely you enable best power saving features.
In more of an answer to your post. I spent my career setting up servers for benchmarking and customer acceptance of HPC systems. BIOS settings are critical for performance.
In the old days the wisdom was to disable hyper threading.
On the latest AMD system I worked on we left hyper threading enabled.
Well, most Mac users I ask claim “all day” battery life, or even claim they haven’t charged in 2 days. This may reflect on their use of the laptop. I don’t sit at power sockets. I don’t use docking stations. I have a free position, and often work in cafes. Many cafes have limited access to power outlets.
If I were sitting at a desk with a power socket, I probably wouldn’t ask my employer for a laptop. So to me, a long battery life is important. What I get at Lunar Lake laptops, is some 3-3.5 hours. At best.
Of course, there is nothing magic about battery life. Processors have a maximum rated power draw, and given the battery capacity, it is fairly easy to find the maximal battery life – assuming full power usage. The MacBook I am testing can draw up to some 80 W (cpu, gpu, npu, display, etc.), which means worst case less than 1 hour. With my use of mixed word processing and running Julia code, I expect to get considerably longer battery life. So I find it interesting what real battery life I can expect.
People have different life/work situations. So I guess we have different needs/requirements. Some people advocate for using a NUC or Mac Mini or something, and a keyboard/monitor. I have my backpack with my laptop, and may go for a 25 km walk and stop in a cafe and do some work :-). I have met people working from a Starbucks, in between going to meet with clients. OK, in Starbucks, often a power outlet is available. But other cafes restrict the power availability (at least where I go), possibly because they don’t like clients that just buy a single espresso and sit for hours .
so we do not have a falling out.
.