Show Idle (>14 d.) Chans

← 2019-09-19 | 2019-09-21 →
asciilifeform: !q uptime
snsabot: asciilifeform: time since my last reconnect : 35d 1h 18m
asciilifeform: ohai dllud
asciilifeform: dllud: answr'd comment.
dllud: asciilifeform, damn quick!
asciilifeform: dllud: btw do you actually have a 'talos' ? lotsa folx seem to like to bring it up, but ~no one has
dllud: ahhah I fit in that category, I don't have one
asciilifeform: i've yet to run into anyone who'd admit to having bought it.
asciilifeform: ( or for that matter who could explain the appeal. ibm is ~at least~ as much a tentacle of the octopus as intel..)
dllud: I think I know just one person that bought it, by the nick of Jeremy_Rand_Talo, hangs around on some IRC channels.
dllud: Indeed. Although for Intel we are sure the ME is there and we need to feed it with blobs.
asciilifeform: who's to say what blobs are baked straight into the iron in the ibm box.
dllud: Yes. But I can argue the same about the old Opterons, can't I?
asciilifeform: but unlike e.g. the opteron -- given the cost, nobody's likely to ever pull ibm's apart.
dllud: Did someone do it for the Opterons?
asciilifeform: here for instance is some dissection of the newer, plainly-boobytrapped opteron.
snsabot: Logged on 2016-10-04 14:41:09 kmalkki: note that RtmPubSigned.key[0x14..0x23] == AmdPubKey.bin[0x04..0x13]
asciilifeform: the older one, someone recently made progress in reversing the microcode (which loads w/out any cryptotrap), but i haven't the link handy
asciilifeform: the classic opterons included 0 oddball on-die periphs (nics, remote managers, etc) unlike the newer, and have published init code. so hard to picture of what would consist an effective mine in'em
asciilifeform: could, i suppose, contain a 'magic instruction' that goes from ring3 to 0, but all cpu are 'guilty until proven innocent' of containing this, and besides, process isolation on linux is rubbish anyway
dllud: Thanks for the links! Have you seen similar research for the POWER chips? Any idea of the peripherals baked into them?
asciilifeform: dllud: that's the thing, nobody seems to have any idea , given as they cost weight in gold
asciilifeform: vivisection required supply of reasonably-inexpensive burnable units, unless yer truly loaded
asciilifeform: ( and if yer loaded, for coupla 10 k $ you can have own cpu baked )
dllud: As for the passive cooling. Your requirements are just too high (8 SSDs, 2nd CPU, etc.), ahahha.
dllud: Damn, have you ever thought about doing it the other way around?
dllud: I.e. grabbing an old power transformer from the scrapyard, strip it's guts out, and machine some copper blocks to connect the CPUs to the transformer's outer layers?
asciilifeform: dllud: experimented with similar. problem with oil submersion is that no pc mobo is designed for it -- track impedance changes, and capacitors dissolve
asciilifeform: not to mention that yer fucked if you ever want to swap a part
asciilifeform: ... whereas if you don't submerge, but run heat pipes, box will die under load, all kinds of parts (e.g. voltage regulators) presume ~some~ airflow
dllud: ohhh, that's what I was thinking about, solid copper blocks as heat pipe connecting to the ex-transformer shell.
dllud: Didn't think about all those tiny components besides the CPU.
asciilifeform: dllud: consider this photo .
dllud: If it was just the CPU I bet you could get a custom heat conductor for a decent price.
asciilifeform: notice, north/south bridge, and the voltage regs (right/center of pic) are quite warm.
dllud: CNC machines are pretty common nowadays. I've seen lots of SMEs with them. Shouldn't cost that much to have your custom copper thing machined.
asciilifeform: dllud: i have cnc mill right here.
dllud: ahahhaah ohh well, x-ray, CNC. Do you also happen to have a laser cutter? xD
asciilifeform: problem is that one can't easily cool the vregs / inductors / all the little pieces of shit, that way
asciilifeform: dllud: laser cutter (tho dun see what good it does for this problem)
dllud: Taking that into account, I wonder how those guys from Compulab get away with it.
dllud: Hmm.. it has "some" airflow.
asciilifeform: i've yet to see 1 of those w/out any air holes, aha
dllud: It doesn't seem to be sealed.
dllud: Yup, this one isn't.
asciilifeform: and, again, much easier when everything is soldered down, single cpu, et
asciilifeform: sorta academic from my pov, tho, i won't be buying intel cpu (and esp. not intel nics)
dllud: asciilifeform, but do those tiny holes pose any problem to you? I bet it takes years before they accumulate too much dust.
asciilifeform: well, say i want radio-shielded box.
dllud: ahhh hahha, better start small then. I bet that finding a way to get rid of the vacuum cleaner you have nowadays (i.e. fans) would be a good first step. Then onto the sealed thing.
asciilifeform: or, say, to run it in a packed rack, with buncha hot boxen.
asciilifeform: i noticed that all sorts of 'consumer' rubbish in fact overheats if placed in a reasonably warm confined space, rather than ideal room temperature desk.
dllud: yup, it does. Tried that while fitting stuff into a small video/audio rack.
asciilifeform: at any rate, currently 'workstation' to me means at least 2 pre-2013 opterons. and swappable nics , as cards. and expandable ecc ram.
asciilifeform: so, not much interest in the various bricks. tho i've used, in various applications, small single-board arms and the like.
asciilifeform: they simply aint workstation replacements.
dllud: Yup. And now I can see why you are into a dead end about the cooling. But couldn't you ditch the "radio shielded" requirement, and build a box with some holes that give you enough airflow for the voltage regulator, plus heatpipes for CPU, GPU (and RAM?)
asciilifeform: hypothetically, a pc mobo designed for classic opterons, w/ a heavy aluminum ground layer , designed to be sat down on top of a 1x1metre hedgehog, would do the job. but of course dun exist presently.
asciilifeform: dllud: the 'heatpipes for x,y,z, some airflow' box is actually pictured in the article we're discussing.
asciilifeform: i was speaking of actual solution.
asciilifeform: the 'here's how to solve your problem, why dontcha give up and settle for half' is uninteresting , imho
dllud: hahah, indeed, got me
dllud: BTW, since I am wasting your time, I'll take the chance to ask about this:
dllud: Did you meanwhile found any better candidate for a laptop?
asciilifeform: what about it ?
asciilifeform: dllud: nope.
asciilifeform: not found.
dllud: Bahh :(, sad. But that's the answer I was expecting.
asciilifeform: someone earlier linked to a self-assemble-kit arm lappy, but turned out to be blob-booting chinese thing.
asciilifeform: (and with rubbish lcd.)
asciilifeform: state of the lappy market, even in 'price no object' category, is pretty sad these days -- keyboards that wouldn't look out of place in a toy store, etc (not even to mention intelism)
asciilifeform: i'd happily buy a lappy that consists of ips lcd + ice40 fpga + some dram sockets. but, of course, no one makes.
asciilifeform: would even buy a 'rockchip' lappy, if could be had w/out any googlisms in it.
dllud: asciilifeform, Well, there are some attempts. Like this:
dllud: Basically the same SoC and in the Purism Librem phones.
asciilifeform: ugh brick
dllud: But still a promise. Not something you can buy.
asciilifeform: 1990s want their chassis back
asciilifeform: imho '3d printed' plastics feel like shit to the touch, btw
asciilifeform: proper lappy oughta be milled, or at least stamped, aluminums
dllud: yup, good for prototyping and nothing else. Then make a mould out of it and inject with proper plastic.
asciilifeform: or even that
asciilifeform: the products of starvation-budget 'kickstarters' etc tend to be horrifying from simple mechanical pov
dllud: Yup, you will be in for a sad ride in the next few years. The laptops that may come out without Fritz chips will be those kinds of community projects with bulky shitty chassis.
asciilifeform: if they simply made e.g. replacement mobo for c101pa chassis/screen -- would be potentially interesting.
asciilifeform: could cost even less than the clunker in the linked www.
asciilifeform: there's nothing wrong with the c101pa's chassis/screen, why not use it.
dllud: There is stuff like that for Thinkpad chassis:
dllud: albeit with Intel Inside.
asciilifeform: imho x86 in lappy is lunacy to begin with. why the fuck put 100 watt clothes iron in lappy when arm etc. exist.
asciilifeform: e.g. rk3328 draws <3watt (entire machine) .
dllud: Because people like me hop too frequently from place to place during the work day to ever consider buying a desktop. A laptop with okish performance is a blessing. You can compile lots of stuff in situ, and only leave the really heavy duty jobs for your server (whenever you can sit in a place a good enough/cheap enough internet connection).
dllud: With an ARM-based laptop I would need an always on, low latency, large bandwidth connection. That's something that does not exist yet. (Perhaps 5G or Starlink?)
asciilifeform: dllud: the rk3328 actually performs very (by my standards) decently. (e.g. gcc compiles at roughly 1/2 the speed of my desktop)
asciilifeform: the lack of expandable ram is annoying tho
asciilifeform: if it were available in a usable laptop form factor, i'd buy it (already have a proper linux port for it, etc)
asciilifeform: ( 'usable' here means, something ~other~ than 3kg of 'glue gun' printed plastic w/ camcorder batteries rattling around inside )
← 2019-09-19 | 2019-09-21 →