HackerTrans
TopNewTrendsCommentsPastAskShowJobs

shdudns

no profile record

comments

shdudns
·3 tháng trước·discuss
Man I fell for it until I saw your post. In fact, I was just about to post what the man's name means in Italian.

I was even going to point out how ironic it is that the mans first name and last name fit together so well.

In my defense I only got as far as idraulico and missed the "Mamma mia, they're super!"
shdudns
·3 tháng trước·discuss
I guess it'd only work in places where the locals don't use AC. Otherwise it'd be a killer to run in the summer, and so you'd have capital goods rotting in a persons basement half the year.

If the user has AC for the summer, then running the AC in reverse for the winter will provide the homeowner with far more heating per joule than running a computer.

So the issue boils down to, there's not enough people living in places cold enough to make it worthwhile.

Of course, Id argue AI data centers aren't worth it full stop.
shdudns
·4 tháng trước·discuss
"now run that unshielded wire 50 meters past racks of GPUs and enjoy your EMI"

Multipole expansion scales faster than r^2.

Also, im not in the field (clearly) but GPUs cant handle 2.4 kHz? The quarter wavelength is 30km.

"nothing in that catalog is rated for 100kW–1MW rack loads at 800Vrms"

Current wise, the catalog covers this track just fine. As to the voltages, well that's the whole point of AC! The voltage you need is but a few loops of wire away.

"you still need an inverter-based UPS upstream, which is the exact conversion stage DC eliminates"

So keep it? To clarify, this is the "we're too good for plebeian power, so we'll transform it AC->DC->AC", right?

"SiC solid-state DC breakers are shipping today from every major vendor"

Of course they do. They're also pricey, have limited current capability (both capital costs and therefore irrelevant when the industry is awash with GCC money) and lower conduction, and therefore higher heat.

They're really nice though.

"wide-bandgap converters are at 95%+ with no moving parts"

transformers have no moving parts. Loaded they can do 97%+ efficiency, or 2MW of heat eliminated on a 100MW center.
shdudns
·4 tháng trước·discuss
At 100 000 A for a 100 MW data center at 1000 V, speaker wire is a joke.

You obviously need at least a dozen stands in parallel!!

Clearly skin effect scales with frequency but, 400 Hz is still low, only 2.5x lines frequency (the scale is by the root); so the skin depth is 3mm. 3mm on each side makes for a pretty hefty rectangular cross-section.
shdudns
·4 tháng trước·discuss
How is DC better than a three phase delta 800Vrms, at 400Hz?

- Three conductors vs two, but they can be the next gauge up since the current flows on three conductors

- no significant skin effect at 400Hz -> use speaker wire, lol.

- large voltage/current DC brakers are.. gnarly, and expensive. DC does not like to stop flowing

- The 400Hz distribution industry is massive; the entire aerospace industry runs on it. No need for niche or custom parts.

- 3 phase @ 400Hz is x6 = 2.4kHz. Six diodes will rectify it with almost no relevant amount of ripple (Vmin is 87% of Vmax) and very small caps will smooth it.

As an aside, with three (or more) phase you can use multi-tap transformers and get an arbitrary number of poles. 7 phases at 400Hz -> 5.6kHz. Your PSU is now 14 diodes and a ceramic cap.

- you still get to use step up/down transformers, but at 400Hz they're very small.

- merging power sources is a lot easier (but for the phase angle)

- DC-DC converters are great, but you're not going to beat a transformer in efficiency or reliability
shdudns
·4 tháng trước·discuss
So what Iran did in the Gulf

Cheap drones overwhelming defenses until the billion dollar radars and airfields got hit.

Then methodically hit everything according to a plan that forces allied forces to retreat to reliable water sources.

Whatever one thinks of Iran, the way they're waging this war is a masterclass in strategy.
shdudns
·4 tháng trước·discuss
But they can steer, videos show that.

This really shouldn't surprise anyone. Iran graduates as many engineers as the US (70% women), but few of them are working on front-end A/B optimization of some boutique dating site.

And, having taken grad classes with folks graduated from Iranian universities, their training is excellent. The Persian kids were always at the top of their class.

EDIT: for the record the class I merely audited was graduate level (rational) mechanics - the class par excellence if you're going to build a hypersonic.

Some observations:

Half the class was Chinese, the academically better half was Persian.

I was the only Westerner (albeit also foreigner)

The girls were wearing veils.

According to the professor, the best mecanist (?) of the 20th century, Clifford Truesdelle, was an American

The Professor was Iranian.
shdudns
·4 tháng trước·discuss
[flagged]
shdudns
·4 tháng trước·discuss
To be fair, a golf ball's trajectory is hardly ballistic given its relatively large surface to volume ratio. Never-mind the dimples are there to cause a turbulent boundary layer to lower drag.
shdudns
·4 tháng trước·discuss
No?

Today's your lucky day, you get to learn about the Mpemba effect.

(Although the why of the effect is disputed, the trivial counter to your point is that boiling water loses mass quickly so there's less mass to cool)
shdudns
·4 tháng trước·discuss
A full model of a cup of water cooling is, in fact, incredibly difficult.

Impossible, since it is chaotic.

But a T(t) model should not be too hard for an LLM with a basic heat transfer book in its training set.
shdudns
·4 tháng trước·discuss
The problem is both highly complex, but fairly easy to model. Engineers have been doing this for over a century.

Of all the cooling modes identified by the author, one will dominate. And it is almost certainly going to have an exponential relationship with time.

Once this mode decays below the next fastest will this new fastest mode will dominate.

All the LLM has to do, then, is give a reasonable estimate for the Q for:

$T = To exp(-Qt)$

This is not too hard to fit if your training set has the internet within itself.

I would have been more interested to see the equations than the plots, but I would have been most interested to see the plots in log space. There, each cooling mode is a straight line.

The data collected, btw, appears to have at least two exponential modes within it.

[The author did not list the temperature dependance of heat capacity, which for pure water is fairly constant]
shdudns
·4 tháng trước·discuss
Racist? That's new. Because he's conservative?

In the 90s he sued one of his franchisees for defamation of character because they were racist.
shdudns
·4 tháng trước·discuss
[flagged]
shdudns
·4 tháng trước·discuss
Thorn would solve so many problems.

For example, how do I write an h sound after t?
shdudns
·4 tháng trước·discuss
I worked for a startup once for a brief time.

Engineering would make tools for the sales guys to make websites. We hated those bastards because they would litter the pates with adds.

"We need revenue" they'd protest. Engineering would respond that past two ads, the revenue was too small to be worth destroying the brand.

What I got out of that was that business folks, often, don't give a shit about the product reputation on a timescale longer then their ownership of it.
shdudns
·4 tháng trước·discuss
Which is why I stressed that successful WFH is one of many enablers of off-shoring.
shdudns
·4 tháng trước·discuss
I dunno, I'm just an amateur photographer.

The calculation isn't too hard though. The width of a pixel divided by the velocity of the subject on the sensor is the maximum delta(T) between real and ghost frames.

But, again, you dont have to shoot faster. You just have to drop the 180-180 degree phase between a real and ghost frame to be 10-350 degrees. Then your 24 fps is capturing the background as if it were 870 fps
shdudns
·4 tháng trước·discuss
Sorry, I'm a bit of a noob on llm. What is "prefill"? As opposed to what?
shdudns
·4 tháng trước·discuss
Two problems:

- It'll bleed on fast motion. Hair in the wind would just not work.

- Incandescent lights are out.

You could solve both by having two ghost frames shot very close to the real frame (no need to evenly space the frames, after-all) and using strobing a high powered laser.

You'd need very fast sensor or another one optically on the same position.