« first day (3961 days earlier)      last day (1055 days later) » 
00:00 - 12:0012:00 - 00:00

12:00 AM
@forestdistrustsStackExchange there's poor miners?
There's people who buy entire warehouses of GPUs, and we can't compete anyway
 
By "poor" I mean "not those three Chinese billionaires and their pools".
> The environmental impact of Bitcoin blockchain has always been in discussion but the topic got fire mid-May when electric automaker Tesla stopped receiving Bitcoin (BTC) as payment for its vehicle purchase citing the environmental impact of Bitcoin. But these studies findings suggest that Tesla should stop accepting bank issued currencies to protect climate.
 
Musk is a bit nutty.
Mars will have a cryptocurrency tho
 
Processing non-cryptocurrency transactions costs more in terms of CO2 than processing cryptocurrency transactions. If someone really wanted to go green, cryptocurrency is better than traditional cash/credit/etc.
@JourneymanGeek So the solution is to use a hash that works poorly on GPUs, like Argon2d.
It's quite easy, even trivial, to ensure that a GPU is less cost efficient than a consumer CPU.
 
@forestdistrustsStackExchange which would drive up CPU costs. Or in the case of chia... storage costs
 
It wouldn't drive up CPU costs as much because you'd need more CPUs, making buying them in bulk less and less economical (compared to buying GPUs for other cryptocurrencies).
 
12:05 AM
@forestdistrustsStackExchange queue chinese motherboards with 10 CPUs ... (I kid I kid...)
 
If it can't fit into the L2 cache, having multiple cores isn't necessarily beneficial if your memory bus is already saturated. Then the limiting factor is that bus, not your CPUs.
 
@forestdistrustsStackExchange I mean literal CPUs :D
 
Oh, like using QPI/UPI for inter-chipset communication?
 
'inter chipset communication?' ....
 
IIRC Intel can use up to what is it, 4 individual chipsets?
 
12:07 AM
"Lets slap on as many CPUs, EMMC and maybe ethernet onto a board'
 
Yeah. When you have a board with multiple chipsets ("physical" CPUs, not CPU cores), they have to have a protocol to communicate if they want to be able to run at the same time.
 
heck if it runs linux, maybe even an SD card boot
Or just run a shedload of seperate instances at once
 
But there comes a point when it's cheaper to buy 10 COTS computers than it is to design a custom one with 10 chips and not enough RAM.
@JourneymanGeek Think about the cost calculations if you need 1 GiB to mine at a reasonable speed. Good luck accelerating that on GPUs or (affordable) ASICs.
Now compare that to Bitcoin or Ethereum where you can use GPUs and ASICs, meaning you're buying up lots of silicon but not putting much money back into the industry for other components.
 
@forestdistrustsStackExchange so 1-2 sticks of ram, a CPU + cooler and storage in as small a space as possible. Throw in a switch...
 
Whereas if you have to buy multiples of entire computers rather than multiples of specific silicon-heavy components like GPUs, you're giving more back to the industry for the same amount of silicon you're using, so the effect on the shortage is reduced.
 
12:10 AM
@forestdistrustsStackExchange basically the enthusiasm for cryptocurrency is so high any component used will be bought in bulk, and kept out of the retail chain
 
@JourneymanGeek But if each CPU you buy also requires you to buy a motherboard, RAM, etc., then it'll be a lot harder to buy it all in bulk. I doubt miners have enough resources to buy out the entire computer industry. GPUs alone? Sure. But if you use a memory-hard hash with well-chosen parameters, they'll be forced to buy the entire computer to make it worthwhile, which means instead of having 100,000,000 GPUs and 1,000 CPUs, they'll have 100,000 full computers.
 
GPU makers have to meet contractually obligated quotas to make for OEMs (laptops and prebuilt desktops). The remaining is sold as retail stock. Lower GPU production capacity means nothing for contractual obligations for OEMs, so retail stocks are the first to go.
 
Interesting. Didn't know that.
Although it makes sense.
So it's really only impacting the "srs gamerzzz" who need a bunch of the newest GPUs, and not people who are buying computers that come with their own discrete GPUs?
 
Rob
Costs vary, but an out of date estimate is: Mining just one bitcoin in South Korea costs $26170, but only $531 in Argentina; so simply moving earns over $25K per coin. It's a case of electricity, cooling, and labor costs (if you need a lot of people). --- For some it's extremely worthwhile, for others you make next to nothing.
 
@Rob Those don't sound like correct numbers unless your mining system is really efficient.
 
Rob
12:19 AM
Simply copy the sentence and paste it into Google.
 
Looks like that's only for electricity prices.
It's not what the average person could expect if they plugged in a computer and started mining.
 
Same applies for practically all other parts (CPUs, storage devices, RAM, etc.)
 
@SonictheAnonymousHedgehog Not sure about GPUs, but for those others, they also sell the ones which failed various QA and stability tests under different names and at a lower price.
 
True, typically those ones go to OEMs so that they can fill their obligations.
 
So they have a RAM stick and if it remains stable even when overclocked, they sell it as gaming RAM and modify the SPD to report that speed. If it needs to be underclocked to be stable, they sell it as budget RAM and modify the SPD to report the lower speed.
@SonictheAnonymousHedgehog The obligations are for devices with certain specs. If RAM has to be underclocked significantly to remain stable, they aren't going to give it to an OEM that demanded a higher speed ("I thought I said I wanted to have 1600 MT/s!")
 
12:24 AM
I've looked at Amazon reviews for WD hard drives, and seen repeated complaints in them that they're OEM drives and WD refuses to honor the warranty. Which is why I always buy parts straight from the manufacturer (e.g. Western Digital Store), so I know I'm getting a retail part.
 
@forestdistrustsStackExchange chinese motherboard makers will make custom boards if there's demand
 
...and had no problems with them
 
@SonictheAnonymousHedgehog The "portable" WD drivers are indeed drives that failed QA. They are measurably less reliable. Try running it in any sort of high-use/high-stress environment and see.
It's horrible isn't it.
 
@forestdistrustsStackExchange Prebuilds are not affected. DIY boxen are - but you literally can't get any GPU at the moment
 
@JourneymanGeek Thankfully I bought my GPU rig a while back. :P
 
12:26 AM
Anyway, the reason why you can find GPUs and AMD CPUs in OEM PCs isn't because they're hoarding or anything - they just have negotiated contracts that were made before the shortage
 
Yeah that makes sense.
 
@forestdistrustsStackExchange right now, I can go out and get a motherboard with a cheap celeron and an older chipset + 8 PCIe x16 (mechanical) x1 (electrical)
 
I try to buy enterprise drives when I can, for reliability (WD RE). They're expensive, but they're worth it. Three WD RE drives in RAID5 is a lot safer than if they were WD Red.
 
Imagine something like this, with a modern chipset and chip, drop the PCIe slots... and just stack/tile em up
maybe even have ethernet strung across a multi 'tile' motherboard
 
It's easier if you get 1U servers, since then you can stack them more easily.
(With a cheap server rack to hold them)
 
12:29 AM
@forestdistrustsStackExchange ... rack?
 
Or 4Us if you really need lots of GPUs.
 
Most of these setups are basically cheap angle iron shelving
 
@JourneymanGeek Yeah. It's like a shelf that fits servers of specific units.
So you can have a rack that you can just slide the servers into and plug them in in the back.
 
Oh I know what a rack is
 
Some even have built in ethernet switches, although usually you put the switch into a slot.
 
12:30 AM
most of these setups don't care
 
aliexpress.com/item/1005002533899949.html is a good example of a case
Actually I seem to remember someone linking some ex-mining FPGA PCIe modules... so you could in theory build a FPGA cracker on the cheap if you knew how to code em
 
Rob
@JourneymanGeek Have you seen a Cryptomining Motherboard:
 
@Rob I literally linked 2 in this convo
 
Back in the days those were called hash cracking rigs. :/
 
12:32 AM
@forestdistrustsStackExchange Well if crypto crashes, you can build really nice ones cheap now :D
 
Heh I didn't even think of that.
 
Rob
Those have far fewer slots.
 
It's about time I upgraded mine. The GPUs are years old.
 
@Rob those are cheaper
@forestdistrustsStackExchange uhhhh.... :D
weren't we talking about the inability to get modern GPUs? :D
 
Even mid-end ones? It's cheaper to buy mid-high than high-end for hash cracking.
Is the shortage more than just the latest and greatest?
 
12:34 AM
Yup
I had folks complain they couldn't get 2000 or even 1000 series nvidia stuff
 
I mean, if you do, good on you
 
Well, my rig should be competitive for a few years more, at least. :(
 
what're you running now?
 
Given that I bought it non-anonymously, I don't really want to say. But they're 2012-2013 high-end AMDs. Quite a bit weaker than even modern Nvidias.
 
12:36 AM
You might be able to get ex-datacenter cards (I'm brainfarting on the names) that would give you a good boost
 
Don't datacenters usually use crap like ASpeed?
 
....
ebay.com/itm/… for example
@forestdistrustsStackExchange oh, for GPGPU stuff
 
you don't need the video out, just the compute
 
Rob
@JourneymanGeek You are correct, what I suggested saves a lot of money and space.
 
12:37 AM
@forestdistrustsStackExchange talking to real security types hurts my head sometimes.
 
Those GPGPUs usually have a lot of GDDR5 memory, which is expensive when you don't need all that much memory at all, just a lot of cores. And memory-hard hashes aren't good on GPUs even if they do have enough memory due to how GPU memory bus contention works (which is why bcrypt, which only requires 4 KiB of memory, is inefficient on GPUs).
 
@forestdistrustsStackExchange they do - but if you shop around, they're the cheapest way to get a ton of processing power, more so than 'gaming' cards right now
 
hm
If they're also fairly energy efficient, that might be good.
 
@forestdistrustsStackExchange compared to the older AMDs, possibly!
And those are the high end stuff, there's lower end teslas, though I don't know how they compare to your AMDs
 
I think those have fewer cores, but really complex/powerful cores.
Each individual core is surely way faster and more capable than mine.
 
12:41 AM
and larger, less efficient process sizes.
(and some folks buy em, and do a few minor tweaks to run games :D )
 
I should learn VHDL or something and just use hand-optimized FPGAs.
I wish HDLs were easier to learn. :(
It's on my backlog of things to do, but...
 
lol
that never gets shorter does it
 
It gets shorter when I'm high enough to do it.
But usually no lol
 
With some motherboards, if you're crazy enough, you can fit 4 of those (there's m.2 ones) on a carrier board into an x16...
can't find anyone who has tried it tho
 
Especially since the ROI on switching to FPGAs would be quite low. Learning VHDL (or SystemVerilog, which I think is prettier) is one of those "just for fun but not practical" tasks.
 
12:45 AM
@forestdistrustsStackExchange Well - even if you got em cheap, it would be one off, unless you were actively developing an FPGA password cracker
 
The idea would be to optimize it for each campaign.
But that kind of micro-optimization would only be worthwhile if I already knew an HDL.
I don't even compile my own GPU shaders as it is. :D
"Learn OpenCL" - Yet another list on my backlog.
Along with OpenCV.
 
heh
Apparently work has courses in ansible and a few other things which I'm working through
that is on my backlog
 
OpenCL would be pretty useless for me. But with OpenCV, maybe I could make something that would automatically recognize and shoot cops. :^)
 
._.
@forestdistrustsStackExchange I really don't want to give a free trip to an FBI agent
 
"So why are you in prison?" "Oh, I just made a machine to shoot cops but my kid walked through wearing blue. I really wish I spent more time reading OpenCV docs"
 
12:50 AM
I don't see that happening.
and lets not talk about shooting cops please.
 
lol ok
What about shooting nerf guns at cops to annoy them? :)
(nerf guns are toy guns with foam bullets)
 
Some can't tell the difference between a toy and a real one.
I know what one is.
 
That's why they have red tips. But yeah even then they're generally not bright enough to tell the difference. They're looking more at the color of your skin than the color of your gun's tip.
 
I feel like police firearm training exercises would become significantly more funny if they used nerf guns.
 
(Or orange tips)
lmao
Anything capable of recognizing and aiming at a cop could shoot anything, even nerf guns. Or maybe a TASER. Or phaser, to get that "pew pew" sound. :^)
 
12:53 AM
Friend of mine is about to graduate from the police academy. His last major hurdle is getting tazed.
 
Sounds like you need to choose your friends more carefully.
 
They advise that you wear a diaper before you get hit, apparently. That's a yikes from me.
 
lol
@Spevacus I mean, all it does is disrupt your central nervious system
 
No, it does a lot more. It can kill.
The myth that they're safe was perpetuated to allow police to keep using them.
 
@forestdistrustsStackExchange which can then kill you
 
12:55 AM
Someone should make a self-defense gun that uses computer vision and a 500mW green laser.
 
It was in reference to needing diapers
 
@forestdistrustsStackExchange so just blind people?
 
Lol
 
12:56 AM
"Do not look at law enforcement with remaining eye"
 
lol I was thinking about self defense from cops. Not for cops.
 
Oh I saw this military crowd control device that shoots microwave radiation targeted at a location quite a ways away
 
Something a little safer than just blindly firing real ammunition at someone who points a gun at you (which is likely to result in your own spongification with bullets).
 
@Spevacus which has its own side effects
 
It's a quick jolt that's just enough to hurt.
Well, mahbe a lil more than that
 
12:57 AM
and the rumoured russian one that just messes you up from a distance
 
But yeah, side effects and all
 
@Spevacus No, those cause extreme burning pain.
 
@forestdistrustsStackExchange then his buddy shoots you. THEN you get charged if you survived.
 
They can also cause permanent cataracts.
@JourneymanGeek Not if they all get blinded in 0.5 seconds.
Recognizing eyes in a pixmap image is something even modern phones can do.
 
@forestdistrustsStackExchange Yeah... then they issue those laser safe sunglasses, and shoot anyone with a laser pointer.
 
12:58 AM
Mirror contacts, anyone?
 
Guess it wouldn't be very effective if it became popular. :P
 
Super sunglasses!
 
Upgrade from 500mW to 500kW? Enough to blow through a foot of steel lol
 
LOL
 
12:59 AM
Good luck making sunglasses for that!
 
@forestdistrustsStackExchange cue "we need bombers"
 
I'm sure 500kW could take out a hellfire missile pretty easily.
But now you're starting to shift from the budget of "paranoid freak with a CS degree" and "entire military defense system".
 
I mean we could also just make the Death Star. Can't protest if there's no planet to... Protest on.
 
.... you might have a point there...
And no police if there's no planet to police.
 
We're so smart.
 
1:01 AM
New anime title: "I blew up my planet because its police are racist. Can I live on yours?"
 
tbh if I could build a reliable 500KW laser, I'd wonder if you could use it for space to ground or ground to space energy transfer.
or even space to space. :D
 
Depends on the wavelength.
If it was ~511KeV (near gamma rays), sure.
But visual green? It'd just heat the atmosphere.
 
it would look cool
if it was blue...
ORBITAL ION CANNON <3
 
"Mommy, what are those lights in the sky?" "Just scintillation from our country's laser dear"
 
1:30 AM
Is a question about Linux kernel code and behavior better on Unix.SE or SO?
 
 
 
1 hour later…
2:43 AM
@forestdistrustsStackExchange My guess would be UNIX.se, but if you are talking about the code significantly in the question, it would also be suitable on SO (I think). But UNIX.SE seems more appropriate in any case.
 
 
2 hours later…
@Feeds we need an explanation why this didn't onebox ;p
 
5 messages moved to Chimney
 
5:11 AM
@MetaAndrewT. For whatever reason, the podcast blog posts never onebox, but other blog posts do.
 
I wonder if the content is different somehow
 
5:37 AM
@JourneymanGeek yes, it is missing an <h1> tag in the markup
 
@10Rep It's this one:
0
Q: In what circumstances will a process be in state X (dead)?

forestAccording to fs/proc/array.c:130, the following array defines various process states: /* * The task state array is a strange "bitmap" of * reasons to sleep. Thus "running" is zero, and * you can test for combinations of others with * simple bit tests. */ static const char * const task_state_...

So there is a bit of code, but it's not about the C language itself.
Like "what does this statement do" or anything.
 
@forestdistrustsStackExchange looks like the right place
 
5:57 AM
cool
Hope it's not too imposing. I just don't want to dig through the source and historical source and try to figure out the edge cases. Would be nice if someone on Unix already knew the answer.
 
Yeah, on SO you probably get into a debate why you wrote that code like that. You don't need that state.
 
Thing is, the TASK_DEAD state is probably needed, but internally.
 
yeah, I can see how the scheduler would need that.
 
I'm just having trouble comprehending in what case it would get stuck in that state and actually result in /proc/<pid>/status displaying it as "X (dead)".
 
I expect you could see it if ps runs, and the scheduler stops handling that exit code and gives ps a few slices of time to fetch the process status for the process that is dead.
 
6:09 AM
I'd guess that part of the scheduler isn't preemptable.
 
You did check ps uses the same terminology to declare a process dead? Not that ps throws a bunch of states together to declare them "dead"
 
"X (dead)" is a specific process state. I think ps looks in /proc/<pid>/status where they're listed by letter and summary text (so "Z (zombie)" isn't it).
 
@forestdistrustsStackExchange yeah, that is why you need to ask on linux.se. You need someone that understands how the scheduler works.
 
grep ^State: /proc/*/status - I suppose that'd be identical to using ps for that purpose.
 
@forestdistrustsStackExchange would be my assumption as well, but that is it: an assumption.
 
6:12 AM
I'll check that part of the code. That'd rule out being able to see a process in state X due to a race condition with an unlucky preemption and a high-priority ps.
Weird, that says it's EXIT_DEAD and TASK_DEAD is I.
And elixir.bootlin.com/linux/v5.12.9/source/kernel/ptrace.c#L553 says "mark it as in the process of being reaped", for EXIT_DEAD.
Oh, one of those is used in tsk->exit_state and the other in tsk->state.
 
 
1 hour later…
7:31 AM
Looking at the tag-excerpt, it is probably not a good fit: "For questions about how political events influence the Stack Exchange network. For other questions about politics, consider posting your question on Politics Stack Exchange."
But looking at some other questions in this tag, not all of them are "about how political events influence the Stack Exchange network".
 
8:12 AM
@Martin I would be hesitant as well. I don't think it needs the politics tags. It is about policy. We're not asked to take a position on what they put on their profile is agreeable by the community.
 
Someone really wants us to buy their aluminum.
 
Well, since wood prices have basically become a meme...
 
We're all nerds here, so I think these spammers would be better off if they tried to sell us tritanium. :P
 
8:53 AM
@Rob yep, top priority during a world pandemic.
Finding new ways to die.
 
9:09 AM
Meh. Destroyed as nonsense.
 
Fine with me
 
9:51 AM
Hmmm nice message when I load that last smokey post
connection failure
 
Everything seems down ;)
 
Hmm network wide, but chat doesn't seem affected
 
> Fastly error: unknown domain: scifi.stackexchange.com.

Details: cache-ams21078-AMS
From the sff election page...
> Error 503 Service Unavailable
Service Unavailable

Guru Mediation:
Details: cache-ams21030-AMS 1623146027 561302274

Varnish cache server
From the hnq list :P
But hey, at least we can chat! :D
 
yeah I get all 503's now as well
What did Taryn do today? ;)
 
Yeah, same here ...
I bet it wasn't Taryn :P
Ominous... the church bells are starting to ring here now :P
 
9:59 AM
haha
Here the church bells are removed for restauration
otherwise we have one at 21:55 every day
 
That's a weird time for church bells.
I'm guessing this one is a funeral though. It's not a Wednesday so it can't be biddag or dankdag, and it's also not a national holiday or a Sunday XD
might be a wedding but I don't think they do bells for that.
 
Or it is just one minute early in alerting you its noon?
 
@Luuklag Nah, because they're still going on, it's been almost 3 minutes now ;)
 
ah okay
we are some sort of half way back online though
I could load a question
but flagging, or inbox is still bugged out
 
Well I still can't even load meta.stackeschange.com/questions, so there's no way I'll be able to load a single question either :P
 
10:04 AM
Never mind stupid lightshot
 
It is cached in AMS
 
Well I might as well unfollow StackStatus on Twitter... nothing there yet :P
 
I managed to pull up this spam Q
-5
Q: what are the benefits of home loan to first time home buyer?

user99479 Everyone’s idea of a dream home could also be different, but what’s common to each Indian is that they wish to have a house of their own. For many, it's quite just an asset; it's a source of pride, affection, security, and stability. this sense is further strengthened when it's your first house....

but flagging it won't work
 
doesn't load for me
 
Here neither.
 
10:05 AM
Details: cache-ams21040-AMS 1623146693 1214354366
 
Now it does no longer for me either
Stackstatus.net is also a useless site
The last maintainance wasn't even added
 
@rene you get a different lottery number each time you refresh. I put a script on it, want to claim as much numbers as possible to increase my chances of winning :P
 
I guess that is the phonenumber you can call
 
@rene good luck with that
The numbers only increment, so theyre not random
 
this is the main problem: fastly.com
 
Rob
> Error 503 Service Unavailable
Service Unavailable

Guru Mediation:
Details: cache-sea4458-SEA 1623146593 1339869351

Varnish cache server
 
June 08, 2021 10:01 UTC WARN ongoing CDN Performance Impact Investigating - We're currently investigating potential impact to performance with our CDN servic...
 
Yeah, a lot of internet is down and blaming fastly now :P
 
How come chat isn't affected?
 
No relation to fastly, I'd guess
 
10:18 AM
meh, never liked fastly
Wonder how fast they fix their bugs and issues.
 
If you look at their instances, only a few are affected according to their schedule
 
Seems something quite big.....
Reddit seems to be KO too....
 
Joy. Single point of failure yo
Xkcd is down too
 
Chat looks funny without CSS
 
Mobile looks normal
 
10:25 AM
@MadScientist I think I already loaded the CSS then
Fastly error: unknown domain: stackoverflow.com.

Details: cache-ams21049-AMS
And were back
or not
I/O error
Things are slowly getting back together
 
I almost forgot how to solve programming problems without Stack Overflow ...
 
I call it Tropen rooster and take the rest of the day off ...
 
@rene Sounds like a plan
 
Rob
10:41 AM
🆙🎉
 
quickly
 
the iOS app is still working :)
 
@MetaAndrewT. but not so fastly ;)
 
fastly status page now says issue is identified and fix being implemented.
 
10:51 AM
I can see Glorfindel's question, I just can't comment on it yet (an error has occured)
Same for upvoting
 
MSO works for me, MSE doesn't
 
Apparently widespread issue related to Fastly.
Multiple reports of sites down on google
image upload is KO too
just look at downdetector.com I guess
 
SU is up.
 
Yeah, it took half the internet with them. Big news sites, Reddit, Twitch ... the fun thing is that the news sites that do still work (like the Dutch NOS) are so slow in reporting it XD
 
MSE back for me as well
 
10:58 AM
Woohoo! Back to our normal procrastination routines ;)
 
Mse isn't but I have notifications
 
> The issue has been identified and a fix has been applied. Customers may experience increased origin load as global services return.
I wonder if what Nick tweeted was really worth the effort then, seems things happened almost at the same time?
 
@Glorfindel of course, it was made by Brian. :D
 
@Tinkeringbell they wouldn't know that at the time I suspect
 
@JourneymanGeek Oh, I know XD
 
11:05 AM
fix has been applied at fastly
 
@Glorfindel '(In case you were wondering: this question was posted via the officially deprecated Stack Exchange iOS app.)' 🤣
I do wonder how the alternative is... The android option I tried to the official app needs work.
 
@Tinkeringbell nos.nl/artikel/…
 
@Luuklag Yeah but that was after I had discussed it here in chat, went downstairs, and made and ate two tosti's.
On a scale of internet time, that's waaaay slow.
 
And their explanation by their "tech redactionist" is also meagerly at best
 
yeah. Then again, it's probably not easy to explain in a way that every average Dutch person can understand what's going on ;)
 
Rob
11:17 AM
They had an issue yesterday too: status.fastly.com/incidents/1glxxb8sf2zv
 
@Rob Took them about 8 hours to fix...
MSE is acting out for me once again
 
Rob
Wait a few minutes, they're switching back to F'ly:
 
11:30 AM
@Rob ..... that explain the issues I had yesterday. And here I was, blaming my ISP.
 
I think blaming your ISP is in general a good approach.
 
@rene or the crappy modem/router from your ISP, that's a common source of trouble for me
 
@rene don't worry, they are already on my top 10 enemies list.
 
@SPArcheon The "belastingdienst" made it to the #1 spot on my list today
 
11:47 AM
MSO/SO down for me now.
 
Probably switching that one back last? :P
 
@rene works here
 
Hmm, DNS caching?
 
always blame caching
 
and I'm back
 
00:00 - 12:0012:00 - 00:00

« first day (3961 days earlier)      last day (1055 days later) »