jmb99

"The kilogram, symbol kg, is the SI unit of mass. It is defined by taking the fixed numerical value of the Planck constant h to be 6.62607015×10−34 when expressed in the unit J s, which is equal to kg m2 s−1, where the metre and the second are defined in terms of c and ∆νCs.[1]"

The base SI unit for power is the watt. The base SI unit for mass is the kilogram. Yes, this is dumb, but it's the way it is.

[1] https://www.bipm.org/utils/common/pdf/si-brochure/SI-Brochur...

Unfortunately, in the real world, # of bugs solved per unit time does not scale linearly with # of developers - and, you eventually reach enough people that you can't effectively coordinate changes without wasting more time on processes than you're gaining by adding another person.

I've never worked at Apple and I don't know anyone on the Photos team, but I imagine a company at that scale probably has a good idea as to the optimal number of developers working on one application. Optimal to Apple probably involves optimal money spent:money earned ratio, not most bugs fixed per unit time, but I would wager those numbers are pretty similar anyways.

Iirc the FFT step uses AVX, and on Zen 5 that’ll be AVX-512. It should keep 100% of the required data in L1 caches, so you’re keeping the AVX units busy literally 100% of the time if things are working right. The rest of the core will be cold/inactive, so if you’re dumping an entire core’s worth of power into a teeny tiny ALU, which is gonna result in high temps. Most (all?) processors downclock under heavy AVX load, sometimes by as much a 1GHz (compared to max boost), because a) the crazy high temperatures results in more instability at higher frequencies, and b) if the clocks were kept high, temperatures would get even higher.

But in general yes, PCIe vs RAM bandwidth is like comparing apples to watermelons. One’s bigger than the other and they’re both fruits, but they’re not the same thing.

Generally people don’t talk about random-access PCIe latency because it generally doesn’t matter. You’re looking at a best-case 3x latency penalty for PCIe vs RAM, usually more like an order of magnitude or more. PCIe is really designed for maximum throughput, not minimum latency. If you make the same tradeoffs with RAM you can start tipping the scale the other way - but people really care about random access latency in RAM (almost like it’s in the name) so that generally doesn’t happen outside of specific scenarios. 500ns 16000MT/s RAM won’t sell (and would be a massive pain - you’d probably need to 1.5x bus width to achieve that, which means more pins on the CPU, which means larger packages, which means more motherboard real estate taken and more trace length/signal integrity concerns, and you’d need to somehow convince everyone to use your new larger DIMM...).

You can also add more memory channels to effective double/quadruple/sextuple memory bandwidth, but again, package constraints + signal integrity increases costs substantially. My threadripper pro system does ~340GB/s and ~65ns latency (real world) with 8 memory channels - but the die is huge, CPUs are expensive as hell, and motherboards are also expensive as hell. And for the first ~9 months after release the motherboards all struggled heavily with various RAM configurations.

Yeah... And that’s a pretty damn big difference. A connector is always going to result in worse signal integrity than a high-quality solder joint in the real world.

I know of at least two cars with 800k km with original engines. Both GM small blocks (Gen 2, multiport fuel injection so computer-controlled). Neither engine has been opened since they rolled off the floor in the 90s. They’re not particularly efficient (only about 270HP out of 5.7L) but if taken care of, they probably will go forever.

Definitely an exception, though. Very little else on those cars is still original. But it can be done.

It depends on the cost.

Translating the OS to Slovene really just costs whatever a translation service costs, plus review & a bit of ongoing maintenance. Even on the high end, I doubt it would push 8 figures up front, and probably 6 in yearly maintenance.

If compliance with the Japanese law means every iOS build in every region must support alternate browser engines, what's the cost to Apple? They clearly must believe it's high (otherwise they would have done so already). Brand risk? Maybe, and hard to price that, but at Apple's valuations and revenues that could be 9+ figures depending on severity. Loss of revenue from Google (if they still pay to be the default search engine in Safari)? That could be a large number too.

It all comes down to the numbers. How much will Apple lose by pulling out of Japan (possibly betting on such a move being temporary, hoping Japan changes their mind, and also hoping that their reputation worldwide doesn't suffer) vs allowing alternate browser engines worldwide?

> I don't see how running internet over wires to rural housing could ever be seen as a good expenditure of money.

In pure ROI, it's the same as running electrical service to rural areas. But still very important in general

Do you have evidence to back this up?

In Canada, many bank branches don't carry cash except at ATMs, which means 100% of the cash transactions at those branches go through ATMs. Bills are inserted without an envelope, and are counted one by one, which means they're already being optically scanned (read: photographed) to determine the denomination. It's not a stretch that the serial number could be captured at this phase. When bills are withdrawn, they're withdrawn off the top (or bottom) of a stack of bills, so it is known which elements are removed from the stack. Again, it would not be infeasible to track all of the serial numbers in the stack, in order, and associate those numbers with withdrawals.

I do not have evidence that this occurs, but I've always assumed it was at least possible. It's technically trivial. But if you're claiming that it's either impossible or it doesn't happen, I'd need some convincing evidence that that's the case.

Yes! They’re called solar panels, and our best ones are about 4x more efficient than the most efficient photosynthesis processes in nature, afaik.

>Turns out nobody was really using it because discoverability sucked..

Sure, but then redesign the UI after removing 3D Touch to not be equally undiscoverable but less precise. Even on the latest iOS beta with its full redesign, there's still many, many actions that require a long press that are completely undiscoverable. (For example, if you don't have the Shazam app installed, go find the list of songs Siri has recognized when asked "What's this song?" Don't look up the answer.)

Edit: And also, to get even half as much memory, you need to spend $10k. If you want to host actually-large LLMs (not quantized/distilled versions), you'll need to spend close to that much. Maybe you can get away with 256GB for now, but that won't even host full Deepseek now (and I don't know if 512GB either, with OS/etc overhead, and a large context window).

Maybe if I was on a well I would need to filter my own water, but then I definitely wouldn’t trust that job to a fridge.

Be careful what you wish for.

When buying my current fridge, I specifically tried to go out of my way to avoid complexity, but the opening for my fridge is an odd size so my choices were limited. The only fridge I could buy that didn’t have a bunch of crap that’s was guaranteed to break (ice maker, water dispenser [seriously? aren’t most fridges right next to a faucet?], LCD-covered glass panel, etc) that also had a thermostat had a digitally-controlled thermostat. No knob or physical buttons, just a capacitive surface for temperature adjustment and some LCD screens showing the fridge and freezer set temps. (Not the actual measured temps, that would be too useful, just the set temps.)

In hindsight, I probably should’ve just gotten one with a regular dial, but I was a bit fixated on the “real” thermostat. So now I’ve got that to look forward to breaking in 4-5 years and figuring out where the hell to source a discontinued fridge LCD panel from.

(If you need even more memory you could equip one of those servers with 6TB of DDR3 but you’ll lose a bit of bandwidth if you go over 2TB. DDR4 is also a slightly faster option but you’re spending 4x as much for the same capacity.)

So should everyone under the age of 25 (roughly when your brain becomes fully developed) be prohibited from talking about any kind of medical treatment?

Also, how do you expect people to develop critical thinking skills if they’re never presented with challenging concepts or, you know, required to think critically about things?