Category Archives: Uncategorized

How fast is an Efficiency Core?

Last years more and more computer CPUs have two different types of Cores: Performance Cores and Efficiency Cores. As a consumer and an IT professional I am curious about the performance of such Efficiency Cores.


I will give the answer first. Under good conditions, an Efficiency Core can deliver about 20-25% of the performance of a Performance Core. This is comparable (slightly less) to what can be expected of getting an extra core with Hyper threading.

Good conditions are: 1) that the load/application is multi threaded, and 2) that there is sufficient memory cache to make use of another thread.

The difference with Hyper threading is of course that the Hyper thread is supposed to produce 30% extra performance, on top, when the CPU is already running at maximum (speed and power consumption) while an Efficiency Core is supposed to be able to run the system under low load, so no Performance Core needs to be active at all. However, efficiency Cores can under the right circumstances boost maximum performance.

Test CPUs

I have compared two mobile processors that are quite different. Both are rated at about 35W and both were released end 2021 to beginning 2022.

  • Apple M1 Pro (10 Cores – 8P + 2E)
  • Intel i5-1235U (10 Cores – 2P + 8E, and the 2P cores are HT so the OS sees 12 cores)

There are differences too. The Apple CPU is a premium chip in a computer that cost over EUR 2000 two years ago. The Intel CPU was found in a Chromebook currently costs less than EUR 700. I think it is safe to say that the Apple chip is much more expensive, and it also have much more cache. This is not an article about Apple vs Intel, I just want to explore Efficiency Cores, and for that purpose it was quite useful to have two processors with opposite (2+8 vs 8+2) configuration.

Test Algorithm

To test the CPUs I decided to use the Argon2 password hashing algorithm. The algorithm takes a number of parameters. For my purposes, two parameters were of particular interest:

  1. iteration count – double the count and the execution time doubles
  2. memory usage (kb) – double the memory usage and the execution time doubles (unless you run out of cache, then it gets much slower)

I am using “the reference implementation” that I downloaded a few years ago. It is possible that it is optimized for one architecture (x64) but not the other (ARM64), I do not know. I assume no floating point operations are involved so it is possible that workloads using the FPU would behave differently.

This algorithm is designed to require a significant amount of RAM, to prevent efficient parallelization on GPUs.

Single Core Calibration Benchmarks

The following table shows the iteration count that allows 100 hashing operations, in sequence, in 10s. I added a 4-core i7 with Hyper threading for comparison. I also added a column with Cache size but there may be more details involved not represented in the table. I also added two other benchmark scores, public Geekbench (typical scores, big variety there) and my own real day-to-day task – precommit – that I often wait for.

Cache MB
L2 / L3
M1 Pro (8P+2E)24+4 / 24210026766152300 / 1200018
i7-8809G (4C / 8HT)1 / 8290037590181500 / 4400 23
i5-1235U (2P+8E)4.5 / 124350560102222100 / 710017

I can draw two conclusions here:

  • M1 Pro seems to underperform at Argon2 compared to what would be expected from Geekbench single core performance, or my own precommit job
  • M1 Pro, with its big cache, is slowing down a bit less with bigger problem size, than the Intel chips.

For each CPU and each memory usage, I have now calibrated the iteration count so the Argon2 program always takes 0.1 seconds to run.

Adding more Cores to Benchmark

Adding more cores is now easy, as I have a command that takes 0.1 seconds to complete. If I want to activate 7 cores, I start 7 processes at the same time. As soon as any one of them dies, I start another one. Thus I always have 7 processes running (except in the end, when they die out one by one).

In an ideal CPU every extra core/process would add the same amount of work done. Lets see what happens.

On the X-axis is number of cores.
On the Y-axis is “throughput” or “performance” (measured in thousands-of-MB-iterations/second).

128k Memory

128k Memory per process is little enough to fit in the cache of all processors (not the L1 cache of course).

As we already knew, i5-1235 was ahead at 1 core and unsurprisingly the second core also performs good. After that the Efficiency cores (together with HT of the two first cores) keeps adding performance at a slower rate. The i7 displays a very linear growth over its first 4 cores, less growth when HT 5-8 kicks in, and nothing after that. M1 Pro is behind from the start and it is not until all its 8 performance cores are engaged that it has caught up with the two Intel CPUs. After 8 cores, the 2 efficiency cores of the M1 Pro give a little extra boost while i5-1235 keeps the going.

M1 Pro: 2 E-cores add about 500 each, after 8 P-cores delivered about 2500 each, so about 20% performance.
i7-8809G: 4HT add about 1500 each, after the 4 primary cores delivered about 3750 each, so about 40% performance.
i5-1235U: 8 E-cores add about as much work as the first 2 P-cores had done. It is hard to know what is HT and what is E-cores here, but about 25% performance.

1024k Memory

Now each Argon2-process uses 1MB of RAM. This should keep most of the work inside the L3 cache of all CPUs.

Again, and with much higher RAM load, the M1 Pro displays impressive linear growth over 8 cores, and very similar characteristics as for 128k. Same goes for the i7-8809G, however HT is less helpful now when the RAM pressure is higher. The i5-1235U again gets a ahead with two blazingly fast cores, after that HT (or E-cores) help a bit for a few more processes, but it is very obvious that all the 8 E-cores do not contribute to maximum performance. Up to 3 processes i5-1235U wins and M1 Pro is last. From 8 processes and up, it is the other way around.

4096k Memory

4MB per process is going to challenge the caches, especially on the Intel CPUS.

Now this looks very different. For 1 process, i5-1235U is still fastest, although the gap has tightened to both the competitors. Apple M1 Pro again manages to sustain linear growth through 8 P-cores, with some extra performance from the 2 E-cores on top. The Intel CPUs, because of smaller cache sizes, peak at 4 processes, and there is no reason to think that the E-cores contribute at all for the i5-1235U.

16384kb Memory

16Mb per process should destroy the cache of all the processors.

Again i5-1235U wins at single core performance, and Apples curve starts looking a bit more shaky. Interestingly for the i5-1235U, now when all cores suffers from too little cache, you can actually see that the E-cores contribute a bit from 4 to 10 processes.


First I think it is important to say that the low cost Chromebook with i5-1235U performs exceptionally well, both in absolute terms, for my specific daily (precommit) task, and for its low price.

The M1 Pro delivers impressively solid scalability up to 8 cores. If you can parallelize your problem in 8 threads/processes, you can count on the M1 Pro to deliver close to 8x performance. I suppose for heavy applications (like video processing) this is a very valuable property both to the developer and the end user. With 2 E-cores and 8 P-cores, the E-cores have little impact on maximum performance, but rather run the computer (laptop) using little power under little load. An E-core seems to have about 20-25% performance compared to a P-core.

The i7-8809G also scales very nicely over its 4 cores. In many cases HT delivers a significant extra performance (up to 40%). However, as an application developer, you may not want to create more than 4 threads if each thread is using much memory. I now understand that an HT-core typically delivers more performance than an E-core, but they serve different purposes.

The i5-1235U displays more… unpredictable… multi-process performance. The 2 main cores are awesome! But as an application developer I would hesitate how much more to parallelize. As a user of a modern operating system, having two super fast cores that get rid of heavy tasks in short time, while having a bunch of E-cores that can handle lesser tasks in the background, is appealing. I can imagine having multiple applications, tabs and services running (all on E-cores) and still getting full performance from my demanding application in focus. That is, however, quite much up to the operating system to enable.

What puzzles me with this i5-1235U CPU is – why SO MANY E-cores? The standard M1 chip has 4P+4E cores. We have noticed above that a E-core delivers about 25% of a P-core. So the M1 chip can run 1-2-3-4 E-cores, and then activate a P-core instead. Why have 8 E-cores? Perhaps there are good reasons, I would probably have made it 2P+4E and a some more cache instead.

Elemental Dice – Cerium Problem

After a few months in a box in a close, it is obvious that Cerium has a problem.

I have received a replacement Cerium die, with cerium in resin.

Trying tmux

It seems screen is old and tmux is what I should use. Here are some findings and notes.

Cheat Sheet

I found a decent Cheat Sheet.

macOS backspace issue

There seems to be a problem with backspace in tmux on macOS. I installed tmux via pkgin, so if you use brew or something, perhaps the situation is different.

The simple fix I found here was to create a ~/.tmux.conf and add one line:

set -g default-terminal "screen"


set -g default-terminal "screen-256color"

Other solutions fixing tmux-256color with infocmp and tmp failed for me. I probably just didn’t use the right versions of the commands in the right way.

macOS resizing panes

As I understand it, panes are resized with CTRL+B+ArrowKey. But CTRL+ArrowKey does something else on macOS. I have not decided if I need to solve this yet.


Scrolling was always a hassle in screen. In tmux, this is a nobrainer for me (again ~/.tmux.conf):

set -g mouse on

Smögen 5 private casks

I got samples of 4 private casks from Smögen and a Smögen 8YO. I will try them all and see what comes out of it. I will have a standard Longrow available as reference, and that is a quite high bar for a Swedish whisky.

From most pale to most dark: Longrow, 8YO, P79, P135, P119, P103… at a quick glance.


P79: Quite fruity, no peat really, some vanilla. More bourbon after a while.

P103: Slightly more mellow and sweet than P79.

P119: More bourbon and caramel here.

P135: More subtle, unusual aroma, fruit or chemical, a bit sour. Feet.

8YO: Slightly peated, dry, wood.

Based on my nose alone, winner would be P119 followed by, P79, P103, 8YO, P135.


P79: Clearly peated flavour, quite rich and soft in flavour.

P103: Both peat and I would say sherry. A bit more raw than P79.

P119: Lightly peated, more bourbon than sherry.

P135: Mostly dominated by peat, but not so much, sour.

8YO: This is a peated whisky. Really nice.


From best to worst: P119, Longrow, 8YO, P79, P103, P135

Framing it

I compared P135 with Floki, and found P135 better.
I compared P119 with Bushmills 16, and found Bushmills better.


Some private casks are much better than others. The buyer of P119 picked american white oak and got a very good result, the buyer of P135 settled for “bourbon” and got a really bad result.

More tests

P135 vs Deanston 15YO Organic: Smögen is darker and with a clearly peated aroma. Deanston has a more clean, alcohol dominated aroma. Tasting Deanston it has some nice saltiness and maltiness but it is quite thin. P135 has a quite fine classic peat flavour, and it beats Deanston.

P135 vs Svensk whisky för Ukraina: Smögen slightly darker. Ukraina has a more raw wood aroma, Smögen a bit more peated. Ukraina, slightly raw flavour, but quite complex and balanced with a nutty sweetness. Smögen has a more solid peated flavour, and wins.

P135 vs Bergslagen Gast: Bergslagen has a more raw wood aroma, P135 more peat. Bergslagen has a quite dry clean smoky flavour. Smögen a bit sour in a not so pleasant way. Bergslagen wins.

P103 vs Bergslagen Gast: Smögen is darker and more red. Bergslagen has a kind of raw wood aroma. Smögen a bit sweeter (perhaps not softer), probably some sherry. I taste Smögen after a splash of water, it tastes young and raw. Bergslagen has a pretty dry, somehow classic quite balanced peated flavour. I like Bergslagen better.

P103 vs Balcones: Balcones is darker, sweeter and softer. Smögen is more raw wood and some peat. Balcones has quite much bourbon flavour, with some unusual kick. It tastes nicer than Smögen.

P103 vs Glenmorangie 10: Glenmorangie is much paler, with a light fruity aroma. Smögen is heavier and sweeter on the nose. I taste Glenmorangie and this is a whisky I want to like, but it kind of disappoints me, tasting alcohol like a blend, not having the malty caramel flavours I expect. Smögen wins.

P103 vs Andalusia: Andalusia a bit darker, with a soft sweet aroma, some vanilla. Smögen is more raw and peated on the nose. Andalusia also tastes quite soft. Smögen is more for the experienced and curious drinker. Quite close, I prefer Andalusia.

P79 vs Bergslagen Gast: Smögen is darker, with a softer and sweeter aroma. It also tastes pretty sweet and soft. Bergslagen is more raw, sour. I like Smögen better.

P79 vs Bergslagen Two Hearts: Very similar color. Bergslagen has a more bourbon like sweet classic aroma. Smögen more raw young wood. I had added water to both but after tasting I add more. Smögen has a more raw wood flavour. Bergslagen is a more soft mature whisky, not peated. I prefer it to Smögen.

P79 vs Highland Park Valfather: Smögen is a bit darker. HP is more lightly peated and a softer aroma, and the same is true when tasting. Smögen is younger, more experimental and more peated and you can obviously prefer that. I think HP is the better whisky.

P119 vs Johnny Walker Blue Label: Smögen a bit darker and more red. JW smells soft like velvet, Smögen is a bit sweeter, more raw and more peated. I taste Johnny Walker, it is flawless but a bit boring. Smögen more flavour, first peat and a bit sour with a nice bourbon finish. JW may be more smooth and easy to drink, but it just does not taste that good, so I recommend Smögen.

P119 vs Springbank 15 Rum: Springbank is much paler. Smögen has a more oily raw wood aroma. Springbank a bit lighter, not really fruity but honey or something. More peat in Smögen. I taste Springbank and find it salted, balanced, hint of sulphur (so I add more water) and pretty nice. Smögen is more burnt, raw and much more young tasting, but with a surprisingly nice bourbon finish. Springbank tastes even better with extra water, like a desert wine of whiskies. I prefer Springbank.

P119 vs Tobermory 12: Smögen a little darker. On the nose Tobermorys main impression is sweet caramel, nice and balanced. Smögen is much more raw on the nose with some peat. Tasting Tobermory it has a rich balanced flavour, hints of peat perhaps. Smögen is more both sweet and sour, less balanced, but more flavour indeed. Tobermory is much more of an ordinary whisky, but I prefer it.

P119 vs Hazelburn 2007-2021: Smögen is darker, with a heavier, more raw-wood, peated aroma. Hazelburn is dry, malty, flawless, but when it comes to power it can not match Smögen. I taste Hazelburn and find what I found with my nose, also hints of peat and tropical fruits, and with a kick. Smögen has a more peated flavour, more mature than the aroma indicated, less complex than Hazelburn. Smögen is good, but I prefer Hazelburn in every way.

Replacing MacBook Air 2014 in 2022

I have a MacBook Air 11-inch 2014. It runs Big Sur and it is stuck there (not Monterey or Ventura) so I can guess I have security updates for another year (to autumn 2023). This means I am beginning to look for a new MacBook, but I am not desperate.

My 11-inch was bough as an outgoing model, discounted, sometime in 2015 so it has given me 7 years of good use so far. When I bought it I paid not so little extra money get 8GB of RAM (instead of std 4GB) but I left it at standard 128GB of SSD. Later I have bought a 256GB minimal USB-key that is constantly plugged in for a total of 384GB of storage.

Looking at Apples baseline today (autumn 2022) that is 8/256 GB, twice as much as it was 8 years ago (where is Moores Law?). Using the same logic as in 2015 I would be buying the M1 Air with 16/256.

Apple Support

Apple typically supports a computer for 7 years, then releasing security updates for another 2 years. But this is no rule. Some models are luckier than others. We may assume the M1 Air will be abandoned before the M2 Air. But it could be different. Maybe Apple will support Apple silicon indefinitely because of environmental reasons. Maybe the M4 will be very different, and M1-M3 will all be dropped with the same future release of macOS. We do not know.

My 2014-Air is approching end of support. But I have to admit that is not the only reason I want to replace it.

Apple Upgrades

Apple computers can not be upgraded when it comes to RAM, SSD or CPU. And Apple charges quite much for upgrades when you order. So it may make more sense to buy a computer that will last you for a few years (and a little longer if you are lucky) than to pay much extra for a computer that may last a few more years (if it is still alive and supported).

Being Cheap Strategies

Assuming you want value-for-money when you buy a new Apple laptop there are a few strategies:

Buy the cheapest (outgoing) model that can support your needs today. Since Apple computers are relatively expensive this gives you all advantages of a Mac and a new computer at the lowest price. This is also a wise choice in the sense that the computer may fail, be lost or be accidentally broken.

Buy a discounted (outgoing) more premium model that can be expected to be useful for longer. The drawback is primarily that the computer will be supported by Apple for a shorter time after you bought it.

Buy a new model just when it shows up. Apple keeps models for 18 months or longer, and rarely lowers the prices much. So it makes sense to wait until they release new models and buy them immediately. This maximizes Apple support time.

Regardless, the best time to buy a Mac is when new models just came out.

Right now, Autumn 2022, is not the best time based on the above. But we have a situation with inflation (and I would guess the upcoming M2 Pro and M3 Air will be more expensive than current lineup) and recession (Black Friday offers may be unusually good 2022).

As the M2 Air was introduced this year I looked at it and decided to keep my 11-inch 2014 until the M3 Air is introduced in the future, but a good enough Black Friday deal could perhaps change my mind.

Apple compters can not be upgraded (RAM/SSD). The problem with future-proofing by buying a very expensive computer (say M1 Ultra, 32/2048) is that it may take several years until you have any need for it, and by that time it may be the wrong thing anyway (or it has broken). Paying for a 2TB SSD today that you will not make use of until 3-4 years, or never, is simply not so smart.

My 2014-Air and Requirements

The status of my 11-inch 2014 is:

  • 8GB RAM is enough for now
  • 128GB SSD is too little, mini USB-key helps
  • Screen 1366×768 is beginning to feel underwhelming (as is webcam)
  • Approaching end of support
  • Can not use my iPad as second screen

So when I bought the 2014-Air I knew these things and I knew it would perhaps last so long. In that perspective, 7 years of service is good.

What are my new requirements?

  • RAM
    • 8GB would currently be ok, but I doubt it is a smart buy
    • 16GB is ok
    • 32GB would be for a need or usecase I am not aware of today (I have a Windows computer for playing games with 32GB of RAM)
  • SSD
    • 256GB – I could squeeze myself down from 384GB, but it feels wrong
    • 512GB is ok for needs as I know them today
    • 1024GB would allow me to not think so much about storage organisation as I do today
    • I have a NAS and other computers, so I can work with less SSD, and I do not really need 1024GB of high-performance storage.
  • CPU
    • I currently do not do any CPU-intensive tasks where M1, M1 Pro or M2 would make a significant difference

The problem with CPU is… that it is currently the most limiting factor of my 2014-Air (i5@1.4Ghz), but no other laptops in 2015 would really have made a difference today anyway. When you are out of SSD, or when you are out of RAM, it really limits you and you need to replace the machine or change how you work with it. When your CPU gets old it is just a slightly degraded experience. What my computer is not capable of is to use my iPad as a second display, and that is just because it is a too old generation.

What options do I have today?

M2 Air 16/1024GB: this would be a convenient choice. However it is so expensive (twice as expensive as the computer it replaces).

M1 Air 16/1024: this would be a reasonable choice at a good discount. But this is a computer that already used 2 years of expected support. Buying it at full price (which is cheaper than the M2 Air) would have made sense when it was just released but not today.

M1 Air 16/256: At a good discount, this is the equivalent decision of buying my current 2014-Air in 2015, which turned out ok.

M1 Pro 14-inch 16/512: This is a machine that I usually consider too expensive (because I can lose it, pour Coca Cola in it, or it may just break after warranty). However at a good discount it definitely seems to give value for money.

M1/M2 Pro 13 Inch: This machine would be a “pro” option if it had SD-slot and Magsafe.

Black Friday Discounts

I live in Sweden so the prices are in SEK. Black Friday season has started and we see lower prices. But given that these are models that are 1-2 years old, to me it looks more like an adjustment for an older product than a real bargain.

The M1 Pro is down to 5/6 of the price. But what if that gives you just 5 years of supported usage instead of 6? Then I basically pay the same per day of usage, but I missed the first year when the model was the best. On the other hand, if the Pro is the right way to go (14-inch, better loudspeakers, SD-slot), this may be the last opportunity ever to pick up a Pro at this price point.

Models and prices

I have to say that Magsafe is something I really want. I am so happy Apple brought it back (otherwise I could considera Dell XP13 at a great Black Friday price). So M1 Air is a bit off the table.

I also like the idea of an SD-slot, not only to import photos now and then, but also to use for extra semi-permanent storage in the future.

Given what this looks like the only thing that tempts me is an M1 Pro.

M1 AirM1 ProM2 Air
16/1024 (10 Core)29990
SD-Slot (and HDMI)Yes
Age / Support-2 years-1 years
Better Display & AudioYes
Faster CPU/GPUYesYes

Decision and Conclusion

I caved in an replaced my working, supported, fit-for-purpose MacBook Air 2014 with a MacBook Pro 14-inch, 10-core M1, 16/1024GB. I have used it for a month now, and I have not looked back.

I like the display, the audio, the keyboard, how it sits in my lap, that it is fast and has seemingly limitless battery life. I have not taken it out of my home or pressed it very hard yet though.

The most surprising thing is that I barely use the Mag Safe. I have an iPad Pro with a USB-C-charger. So when I put away my MacBook Pro I just put it on top of the iPad and plug in the same USB-C-charger. Since battery life is so good I don’t really feel the need to keep it plugged in when I work with it, thus the need for MagSafe is limited.

I have moved in all files from my NAS using syncthing that made some sense to copy to the laptop. That is less than 300GB used, out of 1TB. I think I will make use of more than 512GB eventually, but the extra cores compared to the cheapest M1 Pro or even the M1 will probably never make a difference.

I have used the SD-card once to import pictures from a camera.

All in all, I now feel I would have been ok with any of the 16/512 configurations above, and given my MagSafe revelation, the M1 Air would have been a sensible choice. Now I paid 30-35% more for the Pro with more storage. If it does not die on me I am good with that.

Switching from Ubuntu to Fedora

I have been using Ubuntu since 6.06 I think, and before that I mostly used Debian. Before that I used Slackware, and I have occasionally tried Red Hat Linux (version 6 or 7, not RHEL) long ago as well.

I have mostly used Xubuntu because it is light, clean, traditional desktop-minded, and somewhat similar to macOS, and my current Linux PC is a Hades Canyon that I installed Xubuntu 19.10 to. That Xubuntu has been updated a few times, lastly from 22.04 to 22.10, and while I like a Debian based system there are somethings that have bugged me:

  • HDMI Audio got noisy after upgrading to 22.10 (this used to be a problem earlier)
  • Audio IN (and Bluetooth Audio) has never worked well with Xubuntu.
  • Some I/O errors/warnings occasionally when booting, no real problems though
  • Background image in Xubuntu sometimes is replaced with a generic background, and a few days later it came back (not feeling 100% solid)
  • Firefox is SNAP, updated separately and manually (so I get a notification to close the browser, but that does not update it, so I have to do a manual update and then restart my browser)

Some of these problems are perhaps caused by a system having been updated for several versions over a few years so it was anyway time for a clean install.

The HDMI noise was something I could not tolerate so I tried a Live Ubuntu 22.10 USB, and that worked fine. But I have recently read some things that made me curious about Fedora so I tried a Fedora 37 Live USB as well, and that also gave me good audio out. Both come with a new version of Gnome (Fedora more standard than Ubuntu).


I have never really been a Gnome fan. Checking out the Live USBs I realise that Windows 11, Gnome and ChromeOS are surprisingly similar. And in a way they are all pretty similar to macOS.

Deep inside I would like to use the Awesome Window manager. But I am not really willing to pay the price of learning and of not having the convenience (of a modern desktop). I have not tried Awesome so I do not really know.

So I am willing to give Gnome a chance instead of lingering on Xfce and Xubuntu. Those people who design Gnome must hate high information density – I have a problem with that.

Advantage Fedora

Given I want to try Gnome I see some advantages with Fedora:

  • No Firefox-SNAP-situation
  • Standard Gnome
  • Wayland (?) – at least it makes me curious
  • Trying something new

A few days with Fedora

Installing Fedora 37 (BETA) was easy. I few thoughts after a few day:

  • My immediate impression is that Wayland (or Gnome) is faster than X11/Xfce. There is a snappier more immediate feeling to the UI.
  • Pressing the start-button on the keyboard, and the clicking a common application or searching for anything is not so bad (like Chrome OS and Windows 11).
  • Even though the information density is low there is little cognitive noise. When running the browser, the tabs are on top of the window and Gnome itself occupies minimal space, I like that.
  • Settings windows is dead simple, very much like ChromeOS but even more minimalistic. Appearance: I can change background image and light/dark theme. That is it. Change the font or font size of window headers? Nope.
  • Audio Input on the Hades Canyon NUC is still not working properly/easily. Perhaps a USB headset would work, I have none available to try.
  • On shutting down the computer it asks me if I want to install updates before it powers off – this feels like the right thing to do.
  • The new package manager seems nice, I used it to install node.js, and it seems solid.

On my Mac there is an application called Performance Monitor. In Gnome I have System Monitor. This is for a technical audience but System Monitor in Gnome feels underwhelming. This is where I would like higher information density, options to dock it the the menu (or something), and simply a better application. Honestly, Task Manager in Windows NT4 felt more professional.

On my Mac there is the “About this Mac” menu option. I like it much. I can immediately see details about my computer, both hardware and OS. It would make so much (more so than in macOS) sense to have this in Fedora: RAM chips, CPU model, Motherboard model, Installed hard drives, GPU model and VRAM, and so on. And what is my Gnome version? It may sound like a joke, but then you want to install and configure stable-diffusion and you need to know you GPU spec.

In the “Software” application I can see that there is a kernel update from to But it does not tell me I am on Fedora 37. And lsb_release is not available, as it is on Ubuntu.

I suppose it is possible to install more things and configure Gnome to make it a bit more tech-savy.

Simple Vegetable Oil Lamp

WARNING: The lamp prototypes suggested below may not be safe for general use: especially not around children, left unattended, or close to anything flamable.

Oil Lamps

I got a beautiful Oil lamp that I use much.

Oil Lamp

This lamp uses Lamp Oil (kerosene, paraffin oil). When I bought that I was a little chocked with two thing:

  1. The price (compared to vegetable oil)
  2. How seriously poisonous it is (to the point I dont like to handle it, and I wonder if I want it at home at all)

However this “real” Oil lamp does not run well on vegetable oil (I have tried canola oil). It runs for a while but I think the problem is that the viscosity is too high so the oil does not flow properly upwards through the wick as required.

Vegetable Oil

I can buy canola oil for 25% of the price of lamp oil. And it is obviously not dangerous (since it is for cooking). However it is thicker and has a higher flash point. It is also supposed to not burn cleanly (leaving smoke and smell). So I was curious if I could design a simple practical and not too ugly oil lamp for simple (unused) cooking oil.

Skipping the failed designs here are the ones that kind of work.

A can lamp

What you see in the picture are five components:

  1. a metal can
  2. canola oil
  3. a few candlewicks
  4. a metal washer (the flat metal ring with a small hole in it)
  5. a metal “bridge”

placed inside a fireplace. This burns well: no smoke, no smell, burns for hours. I have read that vegetable oil consumes the wick faster than lamp oil. Perhaps that is true, but nevertheless the wick lasts much longer than it would have in a normal candle.

A little bottle lamp

How about moving the metal washer with the wick to a small bottle?

This is a very simple design and as you can see in the (somewhat unsharp) picture it burns nicely. But it only burns nicely for about 60min, and then it burns barely for another 60 minutes and then it dies.

Only the canola oil in the bottleneck is consumed. After that it appears the height difference between the oil level and the washer/fire prevents the oil from ascending the wick (fast enough).

A used candle jar

I tried filling an old candle jar with about 1cm of canola oil, and used a wick and a metal thing for this result.

This burns nicely! The sides of the candle jar does not get very hot, and the bottom of and the oil remains quite cool. The metal thing from a hardware store is obviously designed for another use.

The good thing with this design is that it is simple (jar+metal thing+wick) and that not so much oil goes into the lamp. You can easily reuse pretty candle jars that are already designed for the purpose.

Spirit Burner

I would not guess that most spirit burners (or oil lamps) work well. But SPIRI-1 from Böhm Stirling-Technik works perfectly with canola oil. The good thing is that it is (roughly) the size of a tealight so you could replace your disposable tealights. The bad thing is that it is quite expensive.


First I think vegetable (canola) oil is underestimated for decorative light at home. However I can see that tealights can be sold and managed in a safe way and are easier to use.

It often requires two matches to light the canola, because the flash point is very high. However I think the high flashpoint is also good for safety.

Cheap candles and tealights are made of petroleum and they don’t necessarily burn cleanly without leaving unhealthy particles in the air. I can not guarantee that the canola oil also does not leave any particles in the air, but the oil itself is not toxic at all.

Getting Ripped with Jordan Peterson

I have never been the exercising type. I walk quite much. But I don’t run, I don’t lift weights, I don’t like to get exhausted and I don’t like when it hurts.

Having passed the age of 40 I realized I am not getting younger, healthier or stronger. Although a little bit heavier – not overweight at all, but skinny-fat.

In the autumn of 2018 I listened to Jordan Peterson talking about his book 12 Rules for Life. This particular lecture was about Rule #4: Compare yourself to who you were yesterday, not who someone else is today. Peterson mentioned in his lecture that if you make small consistent improvements over a long period of time, you will eventually make significant improvements in the long run.

Inspired by this I set up a scheme of exercising that works for me:

  • I have a number of simple exercises I can do at home, in a hotel room, or elsewhere: pushups, squats, planks and things like that. Most are body-weight only exercises.
  • Every week I do at least as many of each exercise as I did last week.
  • Every week, in total, I do more than last week.

That is it!

I started out with very low ambition. The first weeks I did ridiculously little exercise. But after 8-9 weeks I saw some improvements, I had got the habit, and I did not want to fail and give up. I have now done this for a little more than a year, and not a single week I have failed to improve.

A few more things:

  • A week starts on Sunday (so it is easy to get a good start) and ends on Saturday (so I have time to catch up after a bad week)
  • Some exercises I do better/heavier after a while, even though the amount is the same. For example, there are different pushup techniques and I allow myself to change between them, generally this makes my exercise harder than it was a year ago. I get more quality out of the same time or repetitions.
  • Sometime I add a new exercise to the list rather than doing more of the old ones.

This is my weekly (every 4 week printed) progress:

Week  49    1    5    9   13   17   21   25   29   33   37   41   45   49
a)   195  320  380  400  420  435  440  450  460  460  460  460  460  460
b)                        30   30   35   50   60   65   70   75   80   80
c)    40   75  105  120  120  120  120  120  125  130  130  140  140  140
d)    29   80  100  120  155  160  165  165  171  175  175  180  180  180
e)    60   85  105  120  120  120  120  120  120  125  125  125  125  125
f)    30   45   60   75   80   90  100  100  120  120  135  150  155  170
g)                   20   30   35   40   45   55   60   65   70   70   80
h)                                  30   45   50   50   50   50   50   50
i)                                                 15   20   25   25   30
j)                                                 15   30   35   35   40
k)                                                                60   60

Total 354  605  750  855  955  990 1050 1095 1161 1215 1260 1310 1380 1415

The different exercises are here named a-k (the important thing is that you find exercises that you like) and the number can be seconds or repetitions (sometimes this is for two sides so I do twice as many). So I think it is a good guess that I exercise almost one hour per week, but that is effective time. I would not be able to do this in a gym in one hour, that would be too heavy.

As you can see it levels out a bit. It is hard, I sometimes hate it, but I improve and I do not give up!

In the beginning I set up a few goals: new personal records when it comes to pushups and planks for example. I reached those and now my goal is to be able to walk on my hands. So I do see results! Also my body looks and feels different.

Discipline, systems and motivation

Motivation will not take you too far. There comes a day when motivation simply fails you and you can lose a good habit. However, if you are disciplined about systematic improvement, you do not fail even when your motivation is low.


If you already exercise regularly and you are happy about it, you are probably already better than I will ever be. But if you really do not exercise and you are aware that you and your body would benefit from it, I think this is a method for you.

Start out with very low ambition. You need to negotiate with yourself (as Peterson says). Perhaps you can do 10 pushups and 10 situps the first week? And 11 the next. Do it, you have nothing to lose. And after a few months giving up on your good development is harder than doing those damn pushups.

I think for me a weekly goal has been good. Some days are just not good days but I can make my weekly goal anyway.

Install Catalina 10.15 on unsupported MacBook Pro

I have previously run Mojave 10.14 on an unsupported MacBook Pro thanks to Mojave Patcher. Now Catalina (10.15) is out, and so is Catalina Patcher (1.1.7).

The only story to tell is a short story of success:

  1. I downloaded Catalina Patcher
  2. I created a bootable SD-card
    • you can obviously use a USB drive instead
    • you can use a previously downloaded Install Catalina-app, or Catalina Patcher will help you to dowload
  3. I restarted the computer, held down Alt, and started the Catalina Installer
  4. There was no upgrade option, so I just picked “Install”.
  5. It took about an hour, a few restarts but I needed to do nothing
  6. All well! My user, all data, all configuration, all programs perfectly in place.

No need to patch manually, to choose hardware, or anything. All just smooth.

2020-04-05: Upgrade to 10.15.4

I was running 10.15(.0) for a while, but it never upgraded minor versions. With MojavePatcher, macOS upgraded itself, often resulting in an unbootable system (until patched again with a more current MojavePatcher). With CatalinaPatcher it seems macOS simply does not upgrade. I will not tell you what did not work, but what I finally did.

  1. Download latest Catalina Patcher 1.4
  2. Run Catalina Patcher
    1. let it download full Catalina from Apple
      (despite I already had a 10.15 install App since before)
    2. install on a USB drive
  3. Restart the computer using the Catalina USB Drive
  4. Reinstall macOS on top of itself.
    (got a question about some time-machine-thing that I accepted)

After this, my MacBook Pro 6,2 came up with 10.15.4 and all seems fine. Even TimeMachine.

So upgrade minor version seems to be about reinstalling full versions.

2020-05-30: Upgrade to 10.15.5

This was an easy update, I did just as for 10.15.4.

  1. Download latest Catalina Patcher 1.4.4
  2. Run Catalina Patcher
    1. let it download Catalina from Apple
      (not using previously downloaded Catalina)
    2. install on a USB drive
  3. Restart the computer using Catalina USB Drive
  4. Reinstall mac os on top of itself

Everything seems to be working just fine.

2020-08-29: Upgrade to 10.15.6

Same as for 10.15.5. It is actually the same version of Catalina Patcher (1.4.4) that I ran again to download latest Catalina from Apple.

Wonderdraft – first impressions

As a D&D Dungeon Master I occationally need to create maps, and I discovered Wonderdraft. I have tried it a bit and I will make notes in this post (and update as I learn) hoping it could be useful for other people thinking about getting Wonderdraft.

Initial productivity

I spent a few hours the first evening I got Wonderdraft and I produced two real maps for my D&D campaign. One is a town, and I think I need to learn more to make good town maps, but it is ok. The other one is a more black and white map with towns, paths, mountains, rivers and a few named places.

One feature that surprised me was that I can import a picture (a PNG scan of a map) and it makes a map of it. It does not get good, but at least you get the basics. If you have a map with coasts and forests and towns and you just want the proportions right this is useful.

I learnt a little later that you can rotate a symbol using they keys: . and , before placing it.


The Wonderdraft web page is quite clear that a powerful computer with a decent GPU is needed or recommended. I have tried a few different computers and reasonably small maps (1920×1280).

  • Painfully slow
    • Mac Book Pro Mid 2010
    • NUC 54250WYKH (4th generation i5), 2013 (Ubuntu)
  • Working fine
    • MacBook Air 2014
    • MacBook Air 2015
  • Working perfectly
    • NUC Hades Canyon NUC8i7HVK (8th gen i7), 2018 (Windows)

What I am saying here is that for my initial, not so large or complex maps, I am satisfied with the performance of 2014 computers and newer, running macOS and Windows: Wonderdraft was snappy and immediate. With the painfully slow machines there is a 1s latency on everything. But it works.

Map size matters much though! I tried some assets on the MacBook from 2010 and used 800×600 maps. Then performance was acceptable (although I had occational crashes).

Extra Assets

There is a community for extra free assets with Wonderdraft. I did things backwards, but I recommend you make it easy for yourself.

  • Click “community links” in the menu
  • Click “Mythkeeper”
  • Install Mythkeeper (macOS and Windows only)
  • Run Mythkeeper

I started working with manual downloads, manual unzip, manual placement of folders in my asset-folder, and Mythkeeper simplifies it all very much.