Throughput Vs latency. Hardware RAID tends to be fast for processing large volumes of data, but the overhead increases the latency.

It's a bit different with DRAM though because the controller is complex and built into the CPU. AMD, for example, already transparently encrypts data in RAM with no performance loss, all in hardware.

That's the point of the CXL memory expansion parts. Something like a Marvell Structera X 2404 is a CXL/PCIe 5 device with DDR4 controller so that you can add DDR4 to a system that would otherwise only support DDR5. There's a latency penalty and the part itself isn't free; but it allows you to make some use of DDR4 that you already have or can get for less than DDR5 without needing some custom CPU that combines current-gen features with a last gen memory controller.

The mention of a NUMA node has me unclear on whether this is a software or hardware implementation. There are several vendors either offering or showing around proofs of concept to try to generate interest, for CXL memory devices(which all essentially look like a NUMA node that is nothing but memory controller); mostly aimed at hyperscaler use cases where you want the improved performance and platform features of current gen servers; but your last gen servers were beefy enough that just disposing of them would be getting rid of a relatively titanic amount of DDR4 that is still faster than most NVMe. Some of those attempt to offer software-transparent compression, some just put CXL on one side and memory controller on the other and allow you to do (mostly) agnostic mixing of memory generations on anything new enough to have some PCIe lanes support CXL extensions.

I suspect that it's mostly my ignorance that is to blame; but what I don't understand about implementing memory compression(especially the schemes that try to do it transparently with high speed general purpose compression/decompression in hardware) is how you compensate for the fact that the amount of memory you need is no longer predictable without significantly more effort(potentially untenably more if you are using RAM because latency is critical).

With normal uncompressed RAM it just demands RAM in direct proportion to its size. Potentially hideously wasteful if there's some memory-backed XML-spew log that would zip down to 10% of its nominal size; but predictable unless you've got memory safety bugs. If you are compressing your RAM suddenly you've got some uses of RAM that are functionally uncompressable and require 100% of the space their nominal size suggests they will; other things might compress exceptionally well and save you 90%; and some might change unpredictably from moment to moment depending on what needs to be stored there.

When you are doing FS compression and dedupe for backups and stuff that is normally manageable; on average enough compression is possible enough of the time that you can safely assume that you'll do better than you would if you just didn't try; and you can just keep an eye on how quickly the big SAN appears to be filling up as versions accumulate and (if tapes or other fixed-sized media are involved) just change the backup media more or less quickly; do you do that for RAM? Just use the big slack area for FS caching or something you can drop on short notice in case a burst of incompressible data comes in? Do you essentially have to rebuild how your entire workload handles RAM so that it considers, at least approximately, how well a given use of memory will compress ahead of time?

Given the sorts of files and data structures you commonly encounter I can easily enough believe that compression is worth the trouble a lot of the time; but unless you are leaving a lot of slack in your available RAM it seems like a situation where, under certain circumstances, you could go from being fine because most of what is in-memory compresses well to suddenly needing as much physical RAM as you do logical RAM because you have a bunch of already-compressed or incompressible data would markedly increase the complexity or potential for things to go wrong.

Thanks, I'll have a watch of that later.

I was interesting by the article's comment that it could lead to smaller clocks, but that doesn't seem likely from what you are saying. At the moment a "chip scale" atomic clock is a metal can maybe the size of a consumer CPU, and pulling quite a bit of power.

Maybe the real end of the world, if Trump reacts badly enough.

Population growth is levelling off and we will end up somewhere between 10 and 12 billion by the end of the century. It's sustainable as long as we are careful... Which we tend not to be. Most of the growth is in Africa, which has a lot of scope for it.

In fact the main reason population is still growing is people are living longer. The global fertility rate has dipped below 2.1, the steady state level due to childhood deaths.

Not only did she develop software without the aid of modern debugging tools, on a new architecture that was limited in ways that most modern embedded developers can't even imagine, she pioneered a lot of the software safety techniques we still use.

No, not at all. The gender pay gap is in the training data thousands of times, from online comments, studies, news articles, etc. Since AI is a dumb machine replicating what it has in its training data, it replicates that.

I'm pretty sure you can find religious believe, superstitions of all kinds, human fallacies and whatever in AI if you provide it with the right setup.

That argument only works if there either isn't a statistically measurable and consistent bias, or if you assume that bias is because men just have better ideas.

I'd like to see some comparisons between power consumption for JXL and JPEG. Constant image quality, file sizes will vary, total energy consumed by the download and decoding process.

We have known about this for decades. It's called internalized sexism. Back when women where trying to get the vote, some women opposed them. They made the same arguments against it that men did.

It's understood to be largely a product of systemic sexism in our culture. It's been getting better, but it's still there. An example is how ideals of feminity were portrayed in the 1950s. The housewife, the home maker, the mother. Obedient to her husband. That's why there was so much focus on representation, so that girls didn't grow up only seeing that as the natural and proper order of things.

In the 90s there was Girl Power, and for the last 25 years we have had many, many badass action women in movies. They aren't really what we need now though. Just more competent women doing "normal" jobs, or as normal as it gets in TV shows and movies.

That's not how LLMs work. While they are trained on existing data, they are also capable of reasoning. They can produce novel solutions to problems.

Just because they don't work the same way as a the human brain doesn't mean that they are not capable of this sort of thing.

Leap hours are just a legal thing to satisfy countries that require UT to be synchronized with the orbit of the Earth around the sun. There won't be one for centuries, i.e. it's so far out that it can be ignored.

Seinfeld went to Israel and visited the IDF camps where they train people to participate in genocide. That's not a joke, that's becoming directly involved in a crime against humanity.

That's the fig leaf that is always used. "It's just a joke! I'm not hurting anyone!" Then they use the platform and money to hurt people. JK Rowling is another example. "It's just a children's book!" (never mind that she also writes shit adults novels under a male name). She funds hate groups and their lawsuits, which has recently resulted in an extremely damaging change in the law in the UK.

Speech does not happen in a vacuum.