Artemis II is basically a test mission for Orion. And while flippantly Orion isn't doing anything that Apollo didn't do first, it definitely does it with a lot more margin, more living space, more safety and redundancy, and an actual toilet instead of gross poop bags you had to manipulate your waste into.
While the speed increases weren't as dramatic, do note that even in single core speed, unlike the clocks would suggest the Ryzen 7 is much, much more than 1.23X faster than the P4. The P4 was a particularly fragile architecture, and achieved IPC on real code was typically well below 1, often closer to 0.5. The x3d variants of Ryzen have been measured at running above 3 average IPC on real, complex loads. So the single-core uplift from that P4 to a modern AMD core is about the same as from that 300MHz Pentium to the 3.8 P4, it just took 20 years, not 8. Of course, now we also have 8 times the cores.
Your explanation is also an oversimplification that leaves out a lot of key details.
TDF is ran by a board. The board is supposed to contain 10 people, it currently has 7. This board is expected to be elected by members on a regular schedule. The elections are late, because the rump board has twice delayed the elections. Instead of holding elections to fill out the board, the rump board chose to change the bylaws, through a legally questionable process (properly, they would have to hold a vote of trustees, but chose not to), to allow them to exclude people from voting in the elections. Then they use the new bylaws to exclude many of their political opponents, on very flimsy grounds⁰.
You don't need to even consider which side of this conflict is technical or non-technical to see that there is something rotten here.
0: And yes, the grounds are very flimsy indeed. Excluding people in case of active litigation sounds sensible, until you consider that the litigation was started by the TDF board, and is frivolous. Collabra is using the trademarks under valid license.
LibreOffice exists because the devs of OpenOffice forked it. If the project leadership now ejects the devs, I think that the new fork will be the living one.
Well, then something's wrong. I click on different pages in the documentation and the whole page gets rerendered. Seems like it's not delivering what's promised.
LTPO has problems with uniformity of brightness, that get worse the larger the panels are. On a phone screen, this is usually not perceivable, but if you made a 27" screen out of it, most such screens would be visibly brighter in some corner or other.
Centralized inference is more economically efficient⁰, and should be cheaper for most users once competition squeezes the air out of token prices. It remains very valid for anyone who wants to maintain their privacy, ofc.
0: Because the only way to get cache locality out of a LLM is to batch invocations. A centralized system where the server handles thousands of invocations at the same time only needs a tiny fraction of the total memory throughput as having all of those invocations run locally on different machines would.
It wasn't just that. Mendocino wasn't a low bin, it was an entirely separate die from Pentium II. At the time, Pentium II used a 512kB off-die L2 cache, running at half cpu clock. To save on costs, Celeron 300A moved to a 128kB L2, which was integrated on the CPU die, and which ran at full cpu clock.
And it turns out that for a lot of software, a smaller but faster L2 was actually better than the bigger one. And because there were no fast products that used the Mendocino die, even the fastest of them were sold as Celerons. 300A was particularly nice because very nearly all of them could run at 450, and 100MHz FSB motherboards were widely available to pair with the fixed 4.5 multiplier of the CPU.
reply