Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
namibj
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
69 ms
·
541.
▲
by
namibj
2y ago
Might be time to scale up spider silk synthesis, then? Or maybe (high strength successors to) Rayon, which as I understand should decompose like regular cotton.
542.
▲
by
namibj
2y ago
Yeah, i.e., is rayon banned? Silicones (they're very different from all the classic oil-sourced plastics).
543.
▲
by
namibj
2y ago
Can you provide some citation for the claim that x86-64 (assuming something modern like AMD Zen or Intel (post) Skylake P-core) does page table walking/TLB-filling in microcode instead of the fairly obvious state machine that can walk
544.
▲
by
namibj
2y ago
AKA x32 on AMD64.
545.
▲
by
namibj
2y ago
Well, yeah, but this actually goes beyond those that first rectify. E.g. bridge less totem-pole designs where rectifying and the active power factor correction stage are fused, would also work unless the firmware gets confused by the lack o
546.
▲
by
namibj
2y ago
You can use a current limiting buck set to pass through input voltage up until the current limit.
547.
▲
by
namibj
2y ago
What they call Dokumentenecht in Germany should be suitable; it's made so that even with technical intervention you can't feasibly bleach it without destroying the paper in the process.
548.
▲
by
namibj
2y ago
While not generally available at this voltage level, SiC VJFETs do not have an SOA limit (on drain-source current and voltage; the p-n junction from the gate to the source-drain channel has current limits for non-capacitive current (i.e., f
549.
▲
by
namibj
2y ago
Interestingly, many universal power supplies (as in, works in the US and in Europe without doing anything to the PSU) are naturally capable of digesting 200~300V unregulated DC. Beware though that DC arcs above 500mA will not easily extingu
550.
▲
by
namibj
2y ago
Inductors are needed to allow for higher output current than the input current.
551.
▲
by
namibj
2y ago
They just GaN there? I thought they where just using SiC and legacy Si for high voltage high power fixed infrastructure?
552.
▲
by
namibj
2y ago
They make 150mm wafers in quantity, up to a couple mm thick, but I don't see why they shouldn't be about to go olive thick. Other than the slow growth, that is; I think it might take a couple months to grow optical grade that thic
553.
▲
by
namibj
2y ago
AFAIK you can still only buy diamond diodes outside of secret military stuff, and even those are not at all accessible. Sadly GaN suffers from horrible p-channels (they use resistor/nmos logic as the non-differential logic family for G
554.
▲
by
namibj
2y ago
Oh they are. But heat pipes/vapor chambers have them beat. So they are only used where electrical insulation (pure diamond sheet) or mechanical strength (cheap diamond grains sorted into a densely packed single layer of diamonds embed
555.
▲
by
namibj
2y ago
No, it's substantially more expensive. At least for anything too large to just cut out of HPHT chunks, just due to CVD at optical grade being pretty slow.
556.
▲
by
namibj
2y ago
No, chromatic dispersion is due to wavelength/chroma dependant refraction. This is a separate thing from the overall/average index of refraction.
557.
▲
by
namibj
2y ago
Also, it is really easy to do the normal IO "syscall wrappers" on top of io_uring instead, even easily exposing a very simple async/await variant of them that splits out the "block on completion (after which just like no
558.
▲
by
namibj
2y ago
I'm Germany HR should be punishing managers for doing that, as a single call is basically directly an entire day of new vacation time, plus punitive damages for disturbing the employee. Of course employees who can't afford to bank
559.
▲
by
namibj
2y ago
There's another light source that isn't nearly as hard and actually delivers better quality light at far higher electrical efficiency... It's just large and most of the cost is in the large machine that's kept under vacu
560.
▲
by
namibj
2y ago
You can build a silent-enough forced convection condenser to put in the room and put the noisy part away, without water, and cheaper than the passive convection water setup. For cooling you can't use the typical passive convection radi
561.
▲
by
namibj
2y ago
It's basically two 3-way valves to flip which coil has high pressure and which coil has low pressure. If you get freezing temperatures, you may also need to do something to manage the icing that would happen outside, which may be via a
562.
▲
by
namibj
2y ago
Tbf, it's not commonly run because it increases TLB miss latency for rarely any benefit (why are you mmap-ing so much? How can you afford to keep that much data around...).
563.
▲
by
namibj
2y ago
Technically division is asymptotically as cheap as multiplication, but in practice you can't really utilize this.
564.
▲
by
namibj
2y ago
Even lasers deal with weird temperatures: they're negative.
565.
▲
by
namibj
2y ago
No, they did it the actually good/cheap way: straight PCIe.
566.
▲
by
namibj
2y ago
Also PCIe is worth mentioning for it's credit-based full reliability (in the absence of hardware failures, which are still signaled).
567.
▲
by
namibj
2y ago
You can spend over a day at each museum if you're decently thorough.
568.
▲
by
namibj
2y ago
If you include Coppicing for charcoal and building wood, along with modern knowledge, it should be possible to go straight to wind power and rush solar.
569.
▲
by
namibj
2y ago
That's a separate concern. The previously mentioned issue is that to never buffer packets in a reclocking repeater on a link, you _need_ the incoming packet rate to never be higher than the rate at which you can send them back out, or
570.
▲
by
namibj
2y ago
Sounds like they shouldn't have left port, then.
More ›