The norm thirty years ago for a hardware store battery was zinc-carbon, with premium batteries being alkaline. The norm today is alkaline, with fancy batteries having a lithium chemistry. So it's absolutely true that the "regular AA" battery you put in your flashlight back then had something like an 800 mah capacity; there is nothing on the market today that is that weak.

In any case that's primary cells, which have zero relevance to this topic. We're mainly interested in secondary cells, and there the improvements in the common rechargeable battery has been dramatic and continual. Thirty years ago the standard hardware store rechargeable was Ni-Cad; a AA probably had about 700 mah capacity. A modern alkaline AA has a capacity of 2000 mah or so roughly 3x as much. This understates the case because modern rechargeable alkalines can typically be recharged easily twice as much as a 1990s NiCad. And *rechargeable* alkalines are getting significantly better almost year to year.

Of course the hardware store battery only has minimal relevance to what we're talking about. What we really care about is Li-ion, and capacity, lifespan and cost for *those* are improving faster than any other battery technology ever has.

You're not going to Gish Gallop your way out of this one. You're the one who brought up your personal experience with the price of batteries at the *hardware store* as proof that batteries have not gotten cheaper. I'm actually being charitable in assuming you're talking about shopping for primary cells; if we're talking *rechargeable* cells the argument is even stronger because they are recharged over and over again which means the steady increase of capacity and lifespan in secondary cells over the decades dramatically lowers your lifetime costs.

As documented in my links above, the cost per energy stored of secondary storage has gone done dramatically in the past twenty years, over 90% since 2000. As for why the Tesla Powerwall isn't dirt cheap yet, customers report waiting months from order to delivery; Tesla already has more customers for this product than it can handle at the current price, why would they drop it? This is Tesla milking the early adopter market segment for a product that they can't produce in high enough volume to sell to the pragmatist market segment.

In any case we're not talking about home storage, we're talking *utility* scale grid storatge with is three orders of magnitude larger. There have been economically successful grid storage projects for years now. Hornsdale in Australia earned back its construction costs in just two years [source]. That's probably close to an ideal econmic situation for grid storage, but as costs continue to drop more and more projects that wouldn't quite clear the normal profit bar will become economically feasible.

Not to mention silliness like requiring the credit card used to pay for the flight to match the exact name of the person flying. (Why do they care if someone else pays for your flight? What business is it of theirs? Why do the credit card companies even allow them to put in additional restrictions like that?) My mom goes by her middle name, and all her cards are in that name. This causes all manner of h**l every time she tries to book a flight from American (and only American).

And any time you end up calling them, you're guaranteed hours of hold time, because they don't have a tenth as many agents as they need to handle their average call volume. And if you're really lucky, they'll fix the problem. If you're not, you'll find out that the flight you thought was booked and paid for didn't actually go through, and you have no ticket and no recourse.

I swear, the best thing that could happen for the industry would be for American Airlines to go under and be replaced by about five smaller airlines that are actually run by people with at least a tiny bit of technical competence. Because you're right. It's obvious even as an outside observer that their computer systems are a steaming pile of excrement. I can't even imagine how much horror must lurk beneath the surface that we can't see. I can fully imagine it being the sort of thing where senior engineers take one look at it and defenestrate themselves rather than touch it.

I'm pretty sure you'll slice your fingers off when you try to pick it up. :-D

Latency of RAM (when measured in nanoseconds) is also proportional to the length of a clock cycle, and shorter buses are more likely to be stable at faster clock speeds. So at least in theory, if all else is equal, you should be able to get lower latency with shorter bus lengths. I think.

It's true you're paying about the same for a AA battery in the hardware store than you were 30 years ago, if you account for inflation. However a 1990s AA battery would have a capacity of around 800 mAH whereas a modern AA battery offers 2000 mAH or more for the same (adjusted for inflastion) price. So while it *looks* like you're paying more for batteries, you're not if you account for inflation. If you actually look at the number of batteries you to buy over the course of time to power some device, you're actually paying less than 1/3 the price *for the stored energy you get*.

In any case we're not talking about the primary (non-rechargeable batteries) you are buying in the hardware store. We're talking secondary (rechargeable) batteries. In secondary cells the price/per capacity deflation is dramatic. The cost of kWH of lithium ion battery went down by 92% since 2000 [source]. Projections are grid storage costs will continue to drop at dramatic, albeit at somewhat lower rates, so we'll see a cost reduction of about one half in the next seven years [source].

Note this is a conservative projection of of lithium ion technology's evolution. There are multiple promising technologies in the pipeline that could significantly beat this projection. Some of these technologies (e.g. molten metal batteries) promise to be an order of magnitude cheaper if the bugs can get ironed out.

But why do people want a thinner, lighter laptop? The pushback from high-end users against the Retina MacBook Pro was so bad that even Apple has realized that making laptops thinner isn't automatically a good idea.

Lighter, maybe, but you're not going to make enough weight difference by removing SO-DIMM slots to be noticeable.

And battery life is almost entirely a red herring. A RAM chip draws the same amount of power to keep it running and refreshed whether it is soldered onto the motherboard or onto a separate SO-DIMM. You do save a tiny bit of power on the memory bus itself by not sending the signals as far, but that tiny savings would be entirely lost in the noise compared with, for example, the power consumed by the CPU itself.

No, there are really only three reasons:

• Reliability. Soldered-on RAM won't ever have problems caused by the RAM getting unseated. Realistically, this is a minor concern, though, because that almost never happens in the real world.
• Memory bus latency. If and only if the soldered-on RAM is closer to the CPU, reduced memory latency could improve performance in some workloads.
• Disposability. Soldered-on RAM can't ever be upgraded, so when you realize that you need more, you have to buy a new computer instead of being able to upgrade it. Similarly, soldered-on RAM can't be replaced, so if you start getting errors from your RAM, you have to buy a new computer (or at least replace the entire motherboard for a sizable percentage of the cost of a new computer) to fix it.

Manufacturers love that last one. Why would they want to make it easier for consumers to keep using their products longer? That means they make less money. If consumers are still willing to buy computers with non-upgradeable, non-replaceable RAM, why wouldn't they take maximum advantage of that to earn more revenue?

The only time I could see it being a real benefit to users is when the RAM is integrated into the CPU, because that could provide significantly lower memory bus latency and maybe higher memory bus speeds. But even that would only really be beneficial for users who frequently do some operation that is impacted significantly by memory latency, and only if the CPU's pipeline is efficient enough to take advantage of the lower latency.

Otherwise, soldered-on RAM seems like a rather large net negative from my perspective.

Yup. The Ampere Altra Max can run unmodified Windows 11 Pro out of the box. So presumably the unmodified Ubuntu UEFI/SBBR builds will run on it, too. Unfortunately, that support tends to mostly be limited to ARM server hardware.

That's probably because you've only dealt with consumer-grade ARM hardware built using silicon from CPU manufacturers that either don't support UEFI/SBBR or don't support it properly (e.g. Raspberry Pi and every clone thereof).

It even references the same apnews article, although this link has a heap of automatically generated garbage tacked onto the end for identification purposes.

I wouldn't say it soared. I would say that it recovered a bit, because the market had previously overcorrected. It is still below its highest March 2024 close, and barely half of its peak price.

As I understand it, most ARM devices don't have anything like BIOS/EFI/UEFI/Open Firmware to provide information about what hardware is present, so you configure the OS with a custom device tree file that provides that info instead. Some server hardware actually does have UEFI (SBBR), so presumably could support a truly generic boot image, I think, but someone more familiar with it may correct me on that point.

Sure they do. Ampere Altra Max has 128 cores of ARM goodness. The benchmarks show it mostly running about half the speed of recent AMD and Intel offerings, and actually beating the Xeon in some tests, but using significantly less power to do it (resulting in better performance per watt).

And with more and more server workloads depending on outboard GPUs and TPUs for most of the interesting workload, raw server CPU performance is likely to take a back seat to power consumption anyway at some point.

Server CPUs are far more expensive than their general market equivalent. I wonder how much of the resistance to change here is down to companies trying to keep their IT hardware homogenous, in other words, once their loss of market share there hits a certain point it will be very difficult to repair the damage.
Pat Gelsinger saw the problems and is trying to fix them, but he's dealing with a supertanker and it takes a very long time to turn things around.

Your argument is literally beside the point. The images are nonconsensual because the subjects of those images did not agree to the images being created in the first place, nude or non-nude makes no difference. The fact that they involve nudity just makes the nonconsent 99.99999999% likely.

They *are* making them. The dealers are not *selling* them. And if the dealers aren't selling them, they're not going to make larger quantities of them. And as I said, the dealers have every reason to *not* want to sell them. They don't make nearly as much money off of EVs on an ongoing basis, because they require far less service (fewer major malfunctions, no oil changes, fewer brake jobs, etc.).

Also, unionized car companies are under heavy pressure from the unions to drag their heels on EVs because there are fewer parts to put together, and the cars require less human labor to build, which means fewer workers. (Pedantically, they don't require significantly less human labor, but a big chunk of the labor shifts to the battery manufacturers.)

No conspiracy is needed. The car manufacturers have unions pushing them on one side to not make EVs and their dealer network pushing them on the other side to not make EVs. What possible incentive, other than being compelled to do so by law, could possibly get them to build more EVs that cost more (and by extension, will sell fewer units even in an ideal world) under those circumstances?