85 karma · joined November 17, 2017
Not exactly a programming language feature, but it would be nice to have a scheduler implemented somehow if that's possible. You can sort of train yourself to do this, at the cost of context switching inefficiency. Context switching is really expensive for the brain in time but can be feasible if done at boundaries of "task units" and only every few seconds.
Where I think the MacBook Pro could do better is if they adopted more of a vapor chamber strategy for the heatsink, but as you mentioned the radiators are probably going to be the limiting factor, seeing as there's really not a huge temperature delta going on here.
I suppose one strategy could be to perhaps up the fan speed yet again, maybe at the cost of a little bit more noise, and use heat sinks with fins as finely pitched as the MacBook Air. Of course this will make the heat sink more heavy and expensive, but given the robustness of old Thinkpads that did not go easy on the copper, it could work.
The only issue is that these things are typically designed to be engineered along with the systems they control, and are difficult for the average user to implement or use.
It's interesting to see desktop machines move back to the BTX model of thermal design, with the CPU fan attached to the case at the front and positively pressurizing the case.
In my opinion this is how more or less ATX should have been designed, with a wind-tunnel concept, and standardized CPU cooler position.
Samsung devices (this is a Note 5 btw) limit charge current when the screen is on, probably to reduce the heat generated by the device (though I would much have preferred using actual SoC and battery thermistors to dynamically throttle this), but under load the device seems to be able to pull power from the charger as needed. When the screen is off, the battery charges at full current but beyond 80% usually slows down below the rated (Entering CV phase from CC) However, the battery chargers on all devices might be set to lower than 4A or whatever the rating is, and without the system load it won't hit the full input current.
Also note that typically devices (esp. Apple) expect 4.8V or so to keep increasing current draw. Most cables are quite thin and unable to meet this requirement, dropping under 4.7 at 1.5-2A. The charge controller will likely then throttle the current to stay above this threshold.
I have had Anker PowerLine (2?) USB A to Micro B cables that were able to handle 4A (rated 2.4) before falling to 4.85V (tested with 5 port usb charger modified to bypass load side switches so theoretically 12A output). Neither connectors, cable, or other components became hot or warm.
Most other cables can barely deliver 4.76V at 1.7 ish amperes.
This probably correlates with parent comment's experiences.
With this said, had this been a lower end product, the SoC would probably have been an Allwinner or MTK with a soldered Realtek SDIO WiFi solution.
Generally I use manual mode because I know exactly what kind of shutter speed I need given my motion and lighting (avoid blur), while attempting to stay under ISO 500, as the images become poor for pixel peepers over that.
I think if I were to be forced to use an Auto mode I would probably want to be able to specify "optimization priorities" to the camera. Maybe "Keep the ISO low at the expense of aperture" or "Get a fast shutter speed at any cost". But then it's hard and unnatural to set up, than to flick a few dials and see what the screen shows.
It also helps that I shoot with post in mind, so sometimes I under or overexpose on purpose when I know I can recover those highlights and shadows in post. It helps to take pictures at diff exposure levels and play with detail recovery to get a sense of how the sensor behaves with noise when boosting shadows or recovering highlights, and correlate that to what you see in live view.
1) Touch screen UI that doesn't seem that bad 2) Power over USB-C and charge at the same time 3) Record to external SSD. I don't like trusting SD cards so this is a huge one.
With this said, if these features aren't important to you and you're just starting out, a Sony NEX can usually be had for like 150 CAD $ or so, and will produce quite good shots and can shoot RAW.
What I've noticed in consumer electronics over the past few years is increased part integration - pretty much every device is just a logic board with a case and display, along with some support components. Pretty much every logic board is just a collection of switching power supplies, soldered ASICs, mem/disk, and connectors.
Even things like power supply ICs are becoming proprietary,with customer-specific part numbers that are hard or impossible to source. Furthermore anything with a microcontroller and firmware on it is perma-bricked should something go wrong on that. This is something Louis Rossmann (the famous MacBook repair guy) talks about a lot.
Now don't get me wrong, I totally support repair documentation, but being able to get replacement parts and materials is probably more of an issue to those "reasonably skilled in the art" than manuals teaching how to mechanically open something (usually possible to figure out) or even what's connected to what (worst comes to worst, you "beep it out" with diode mode and compare with a working exemplar [REWA technology on YouTube does this]).
The issue with parts and materials is that since we're moving towards proprietary ICs, the motive of manufacturers to not provide documentation/parts is even greater as they don't necessarily want to talk about the technology they are using, whether or not it was really novel. It provides a decent excuse for the manufacturers to fight Right to Repair. Furthermore, security issues and proprietary interests come into play, since you don't want hacked fingerprint sensors and Apple probably doesn't want to sell A-series processors.
Futhermore complicating matters is that component and board level repair remains more of a science project than a process. It is logistically difficult for many companies to bother dealing with selling parts (should they sell just subassemblies at 70% of the replacement cost of the product? Resistors for ten cents?) especially to people outside their network of people. And it's hard to deliver consistent experiences when you actually need to try to fix stuff; a lot of people I know are happy with Apple store service because they DON'T try to fix stuff and instead give you a NEW one, without all the warts and scratches, and it's hard to compete with this if you have to swap parts over and over again with subpar or defectively-designed "refurb" parts like I sometimes see happen with the PC manufacturers.
So I'm sort of half and half on Right to Repair every time it gets brought up. People seem to be warming up to first party repair. I think the biggest realistic benefit to RtR at this point is going to be pressure on first party repair to lower prices and improve service. And that's a good thing.
I think the other issue facing RtR discussions is, not that it's anyone's fault or anything but often it is very difficult to have the perspective of actual design and business challenges faced by the engineers and people delivering products. This makes it easier for companies to refute the claims being made. The people actually working on this stuff from the companies' side usually and understandably never talk about it - so we're only representing largely the end user side, or the reverse-engineering side.
Currently there are two companies, Shaw (cable) and Telus (fibre to the home), and some smaller ISPs that use either Shaw cables or Telus DSL.
Internet pricing here tends to look like $20-35 CAD per month on the low end up to about $110 for a high end fibre symmetrical connection, with most plans coming in around $50-75 if you play the "bounce from ISP to ISP every few months" game. Data cap costs appear to look reasonable on most plans and you can get them without caps.