#off-topic-tech
1 messages Β· Page 146 of 1
I have a feeling I am going to struggle with setting everything up. But that's part of the learning curve I guess π
What
The RTX server GPU's exist to provide compute not to play games
If you really want to play games on them properly, you need to use GPU-P with virtualization
for us normies, sure
Quadro cards can still game perfectly fine, just not as well as GeForce cards
There's a big difference between inability and inefficiency
Quadro has been retired
But despite that my point still stands. They are compute cards, not gaming cards. They are good for rendering and other hardware acceleration. They are not good for realtime rasterization for gaming
You can do things to make them work like that, but they aren't going to be as efficient as the gaming cards for gaming
GB100 has 24 ROPs, GB202 has 192.
You really don't want to use GB100 for anything that gets rasterized in ROPs.
Also no DirectX support it seems, only OpenCL and CUDA.
Going back to Ampere, GA100 still had 192 ROPs, still no DirectX, OpenGL or Vulkan support.
Satisfactory was playable on my laptop with Quadro M4000M
But not nice at very densely built area
Yeah, that was just about the insane price AI accelerator chips.
That they just cannot be used for gaming, as the HW doesn't support it.
The lower tiers use the same chips as consumer cards.
Just with the Studio drivers, which are made to be correct.
While Game Ready drivers are made to be fast, fuck correctness as long as it somewhat works.
Hmm I have Steam installed on this PC, I could install 3D mark too and try Time Spy
Actually usefull to compare it to the 4060, as the new PC should arrive maybe next week
The GPU on this desktop
So same chip as in 2070 Super, 2080 and 2080 Super.
But using even more cut version than 2070 Super.
With lower clocks.
Yeah, big assemblies with few thousand parts are wayyyy more smoother to rotate in 3D view in SolidWorks with RTX4090 at home.
Downclocked RTX 2070 using bigger die.
Yeah, just little slower clocks than 2070.
Otherwise same amount of functional units, just using harvested TU104 instead of full TU106.
thinkcentre was boxed in bubblewrap.
a bit dusty on the inside.
Wanted to check CPU, but I am a bit hesistant with the cooler.
There are these 3 screws, but when you unscrew them.
They start to make a clicking sound (you can't unscrew them anymore) but they are still connected somehow.
(not my unit btw, just screenshot from la internet)
You need to remove that fan first.
Which has plastic quick disconnect on bottom right.
And the actual heat sink needs to be then pulled towards image bottom before lifting.
Maybe I should boot it up first to make sure it works. Just incase of a dispute, then they can't blame me for opening it all up π€£
Looks like my GPU might be dusty as well, boosted to 1625MHz for like 20 secs from the starts and at the end of 2nd graphics test of TimeSpy it was down to 1200MHz and 90Β°C GPU temp
Oh yeah it's bad π«£
But I have solution
Just make sure to fix the fan blades in place
Setup jumpscare
Are there (servers) racks, for small factor devices that are adjustable.
Oh yeah there is
In the sense, what if one has a few small devices (tiny PC, switch, NAS etc). Is there a way to store them and have some cable maangement?
Ah, something like this.
https://youtu.be/y1GCIwLm3is look how cute he is
The time has come for the MINI homelab!
Special thanks to DeskPi for sending over their mini racks! Note: they did not pay me to make this video nor have any input into the contents.
The DeskPi RackMate T0 Pi 5 cluster giveaway is over, and a winner has been contacted. I have not yet heard back, but will update this post with further informati...
(The rack)
But this thickness is a pretty new standard, hope you have some 3D printers if you go that route
Else there is indeed small racks, like 6U that can be wall mounted, pretty cool too
I'm ok with a small rack personally
Yeah small racks are plenty enough for most ppls, im just that type of guy who do stupid stuff
yeah i've known guys that just HAD to get ahoild of a huge rack
3dmark.com/compare/spy/54064105/spy/54062608# Before and after dusting
Interesting, all the improvement came from the GPU, CPU is basically identical
Both runs are today on same computer, just cleaned cooler.
Score is about the same as my i7-4790K / GTX 1080 Ti from 2017
I know - I would've expected it to be more uniform
Thermal throttling is a bitch.
The CPU haetsink was clean compared to other parts
Baldur, going to order some ethernet cables.
Cheap ones should be fine?
Or is it worth to spend some extra money on "quality ones".
Wait, I give screenshot of "cheap" vs "quality" (How I perceive em).
If filling the spec, shouldn't be any difference.
The only real difference is how the locking clip is protected or not.
Still same material as the actual insulation.
Just extra looks material on top.
isn't woven* tiny more durable?
The cheap one has better conductors.
So the "more expensive" has put the money into looks, the cheap one into functionality.
Of those two.
You want stranded for patch cables.
single conductor is for wall installations.
The insulation mainly matters for weather, or smoke production in case of fire.
So first doesn't matter for indoor use.
Let me check what stranded translates to for me language.
That right one.
tp link switch
And left is single conductor "solid".
Holy shit, some of these cables have what the site calls: fast tracing.
You shine light on one end of the cable, and the other end lights up. So you can track them in the office etc.
So just optical fiber added.
Avoid flat cables and AWG >26
Should be flexible core, 6a.
26 is already stretching it
That /7 in the AWG 27/7 means 7 strands, totalling AWG 27 size, where smaller number is thicker conductor.
omg if I ever get to run patch for a new building I am doing this
Or in general adding cables to our server room
Ah I didn't notice that.
Just looked for flexible core that is 6a
and I read awg is some american measurement
so I kidna ignored it π€·ββοΈ
Does it matter anyhow for what I am doing?
Conductor size, american equivalent of mm^2 cross-section numbers.
Where smaller number is thicker conductor.
Important for POE, somewhat important for 10+G speeds.
Where that AWG27 is common patch cord conductor size, but anything higher number from that isn't usable.
And AWG23 or 24 is common installation cable size, but you can get patch cords with these thicker conductors.
But mainly the question is: Do you trust the store to only sell cables that actually fulfill the spec they say they do.
If not needing good POE support.
No, I don't trust the store.
It's like a big warehouse that sells* all sorts of cables. Not a specialist.
Your install doesn't have regulatory requirements to only use LSZH (Low Smoke, Zero Halogen) cables for fire safety, so all insulators are ok.
And after that it comes just into that spec and physical charasteristics like the protection of the lock clip.
I'll look for better store.
Yeah, but depending on such operation if they sell absolute anything from China, or reputable stuff that will actually fulfill the spec it says it does.
Even when that latter can be very cheap and from China.
Some barely have any info on them.
Others have more. But I don't know where they are made.
The second on says Certified with rapport.
That 2-3e is completely reasonable price for spec filling 0.25m Cat6 cable.
Without extras.
AWG is American Wire Gauge, but it's still internationally used nomenclature
But then a faker can take bigger profit at same price by not filling the spec and using cheaper cablestock.
I'd set the limit at AWG 26 for basic home Ethernet
go-to rec in a different tech server that I'm in are the monoprice cat6a cables
Nothing amazing, but it's actually decent and you know what you get
Monoprice is US operation.
Still available in Europe tho
I just never bothered with price comparisons
I'm using old cables from work that we replaced
We have ceased our EU-based operations but continue to serve you across the pond.
https://www.monoprice.com/pages/MP_EU
Some stuff on Amazon.de and Amazon.co.uk, but at highish prices.
"Made in China" for the Skarkoon 
and no clue about AWG. (website doesn't share that information).
Basically all of them will be made in China.
made in China means jack shit nowadays
If there's no info on cable thickness, just avoid it
If it's not mentioned, assume worst case
Yeah, even the most expensive one I can find.
It doesn't matter where it is made, only if there was proper QC done by the company getting that manufacturing done at contract plant or their own.
To prevent manufacturing line shenanigans.
But for example one of the Finlands biggest internet electronics retailers has Goobay, DeltaCo and their own brands for sale for network cables.
Oh I know where I can search. Let me check tweakers.
never heard of Goobay. (sounds weird)
After applying filters:
2.326 uitvoeringen - Pagina 1 van 59
it's Ubiquiti
But probably not that many options for those 90 degree turned connectors.
Still way too expensive for the length.
But for professional install might be worth it.
In some special cases.
I don't want 90 degree at all π
Yeah, but as partial explanation for the high price, as it is very specialty product from enterprise solution provider.
10 centimeters rofl.
This seems something ...
awg 26, 4 euro, cat 6a, 0.5meters
and shielded
Specsheet.
goodbay cable are good here though
it's basic standard but no bullshit either
way better than a lot of other shit here imo
cables are always a pain
especially display cables
too many variations by too many brands and its dam near impossible to find the same cable ever again if you ever want another one of the same type
That is US & Canada problem π
are there regulations or something elsewhere? or are there just less brands and options in general
Because you have done insane race to bottom where only the cheapest fake crap gets sold.
And almost every other place than Amazon is dead too.
every time i need a DP cable its an entire adventure
no stores localy store them anymore because its "old-ewaste" most online results are adapters unless you go on amazon and thats because amazon is just 500 options with most being crap knockoffs or just garbage
and there arent really any people testing cables in the mainstream
Aaand it's time to call my electrician buddy... overloading one of the circuits to this room :/
lets have a moment of sillence for my laptop having to cope with me insisting on playing satisfactory with no cooling because i dont want to disturb anyone with a jet engine
how its holding 70C is beyond me
this is like every colective thing and optimisation i have learned over the last 4 years combining together to make this happen
12fps with framegen and dlss on isent great but dammit its enough and actually playable and in pure silence too
15w cpu is wild
its either 100C at 115W or 70C at 15W for only a 50% performance lost not to mention going from max fan speed to idle fan speed
and plus its ram bottlenecked on everything so you never really notice the effects of the lower clocks π€·ββοΈ
i feel old caring about efficiency but really all i care about is still having my hearing in 4 years
You know what keeps tripping me up? TDP.
35 watts TDP, doesn't mean it will consume 35 watts?
its 35W once settled but it boosts higher
Depends on age.
Today no, and it is just number for OEM machine heatsink sizing.
like my 10700F a 65W cpu draws 200W for a decent while to keep clocks higher
Because I know the TDP is 35 (that's on intel specs). But I want to know power consumption π¦
my 13700K dosent because idk i guess it literally cant turbo
i have nerver seen it do it if it actually has
Max depends on the BIOS/UEFI config or code.
The max boost power and max boost time variables.
So the PL1 power limit by default is 35W on that CPU.
Yes.
And that PL2 can be over 100W.
And in some cases MB can set that PL2 turbo time to infinite.
YEP π₯
Don't know what they are for that specific CPU by default.
is that healthy?
by a lot sometimes
Or what Lenovo might have set them.
I have something to measure powerconsumption of connected devices if I want too π
however the more aggresive generally the shorter the duration
like mine maybe lasts 5 seconds
whatever you do MagicZ, don't listen to baldur.
Don't try to set PL2 to infinite on MB 
its a locked chip
Point was that many Intel MB have done it by default for long time.
π€·ββοΈ
Good, because I can see it happen next week: "My laptop fried and I don't know what happened" π€£
been abusing the power plans to control Voltage and shit lmao
Is that healthy for CPU? I don't think so.
What if somethign is wrong in system (processes hanging etc, or in loop, ...) and it keeps boosting for no reason?
Just makes the CPU boost to thermal limit.
Or to voltage and frequency limits.
it goes for the set time or until a limiter is hit
Ah, that makes sense. Infinite time. But still limited by max temperature ofc.
Two power limits, higher power turbo time limit, current limit, temperature limit, voltage limit, frequency limit.
Just removing the first two does nothing to the others.
But some 13th and 14th gen MBs were also removing that current limit.
And setting both PL1 and PL2 to 1000+W by default.
I've seen my Macbook sometimes getting "locked" by a process with high CPU usage.
Imagine having nothing to throttle CPU and it keeps ramping up 
or people creating infinite loops or other stupid shit in code ...
167Amps... 
Lately I've been pondering of taking a electronics course or something.
The wattage number must be based on the psu or something.
Because assuming 220V or sth, and the power supply unit is at 208W.
You'd get about 1A.
asking myself the same thing
No
Watts / Volt = Amps
Then wtf is that about?
They all wrong?
I keep forgetting, so I am googling a bit π€£
I am only asking, because lately, I have to work a lot with electronics stuff.
Big thick cable allows for higher amperage.
cpu pins => tiny af => they be melting (if 167Amps)
And I want to train my brain to remember what is what.
ahh then ig my 13700K is molten magma then
Stuff like that.
and my 10700F is the center of the sun
Thats why there are multiple power pins
and most of the sound engineers I meet lately, all have strong background in electronics.
A cpu rated at 65W pulling 208W is gonna melt.
Those numbers are not showing true numbers.
Design vs. actual.
Same die is used in 35W CPUs and 250W CPUs from Intel.
Where the "Spec" for those 250W ones is actually much lower.
I'm happy to be proven wrong, but 1.28V and pulling 208W does not sound like correct numbers.
Thats just Intel boosting π
Sure there are multiple power pins, but we'd have to be at a lot of power pins then... :))
For that 8th gen:
Spec is that MB must be able to deliver 193A for 127W TDP CPUs, 104A for 35W CPUs.
a high amperage requires a thicker gauge cable to prevent the cables resistance from heating it up; a high voltage requires beefier circuitry at either end of the cable and more protection from arcing
162.5A
Alright. Thanks 
I'm probably wrong then :)))
#PEAK so called PATH so called TRACING
For 13th and 14gh gen, Intel recommendation was to set certain "current" limiter to 400A as max.
Not actually measuring current.
very thick cable?
Intelβ’οΈ
for like 5 or 10 seconds or something. though motherboards would just ignore this limit and go maximum constantly
For 12th gen, 150W TDP CPUs are specced to use up to 280A of current.
125W TDP CPUs are specced to use up to 240A current.
(the requirements to MB VRMs).
Is that from the cyberbug dlc?
Because I don't recognize that place.
(or is that the "endgame" tower?
13th and 14th gen document was moved behind login-wall in last fall.
first 2 minutes of the game
I recognized the game by the yellow circle with exclamation mark in it
Ahh, makes sense I don't remember it π€£
and then I saw "Araska" π€£
maan
you really do need DLSS/AA to play this game
both XeSS and FSR 3.1 look like garbage : (
mmm.
And so for 12th gen Intel desktop CPUs:
For 150W TDP CPUs, the default PL2 is 241W.
For 125W TDP CPUs, the default PL2 is 241W.
For 65W TDP CPUs, the default PL2 is 202W.
For 35W TDP CPUs, the default PL2 is 106W.
Gonna get the dlc and rerun once I get my new pc up and running...
I got hit by π© thermals in my new case so I just ended up getting a new different one that can fit an AIO (first time doing that)... and it also got room for airflow .
Niiice. Besides thermal issues.
But for that i5-8500T, PL1 should be 35W and PL2 44W by default, with 1 second Tau by default in hardware, and recommendation for MB maker to set it to 28 seconds.
Yeah, wasn't worried about it. Just curious π
Thermals are on me for doing shitty (quick) research.
don't forget the 550W PL4
Not sure what type of build you making. But was the case the issue? Or not good enough fans?
Case too small to put any decent sized fans into it. I could only slap 2 60mm ones next to an intake.
The new one I can put like 2 120/ and 3 140mm all over the place (and still fit an noctua 360 AIO for the CPU)
latest intel CEO is a venture capitalist with a history of running companies into the ground and the press just says he's "well versed in semiconductors"
Also was an sff build, still doing it, but the case is marginally larger and also fits things a bit better in regards to airflow.
Did you select first case on the looks? π
NR200 my beloved
mainly, but the new one also looks good (lian li A3)
Lian Li gang!
Probably running companies into ground by design, while maximizing short term returns and debt in the bankrupcy when it finally dies after all money has been pumped out.
More and more people have Lian Li in here.
Pretty normal venture capitalist way to do stuff.
Would have bought it first if I found it before the other case... But I fanboyed back to fractal so It's on me π€£
I do not recommend the Ridge (altho it looks nice. Airflow in it is π© like no tomorrow. I might keep the case and turn it into a small little NAS .)
he was already on the intel board and left dissatisfied that they weren't making deep enough cuts. he's gonna take a fucking hatchet to it
Buy something, pay for it by borrowing done by the bought company.
Do sale-and-leasebacks for anything the company owns at outrageous terms to another company you own to get all the money out.
Let it stagger until death, pumping as much money out as you can.
Then let it die under the lease and interest payments.
And I don't really have a budget cap so spending the extra on another case is not the end of me.
you can try to return fractal case. Or sell it on 2nd hand market. Or keep it as spare.
Like I said, probably gonna turn it into a NAS π
It works fine it it does not do any work, or has a weaker cpu with less power draw.
You can always try to provide extra airflow π€£
I already did π€£
Modding is always an option :))
I'll get some pics when I get the new case sometime next week π
when I have my project done. I am going to install opnsense on M920Q.
Nice, and good luck π
Yeah but once I got that running. I might look into some sort of tiny server racks for it and my switch π
Fun fact.
I ran my messagebus to try bench my thermals, but my message bus was so optimized even with 100% cpu thermals never spiked.
They spiked more when I swapped to discord π€£
It feels like I am at 0.01% on completing the home network I have in my mind.
I know that feeling.
Haha, Electron 
It's like at work.
We're like 95% done, but the last 5% feels like an eternity.
The same kind of applies when starting a project π
I got my whole project up and running in Electron. Started to refactor to svelte + tauri. In the meantime I am learning a bit of rust for the back-end.
Not a fan of rust syntax, but cool either way π
For a application to run locally, so not a web application, Electron isn't the way to go.
Atleast, that's my experience and opinion.
Rust syntax?
was there already discussion about ThreatInteractive recent video?
Not a fan on how it looks when coding it, but I'd take it any day over java 
It's also probably me being stuck in c++land.
pub fn
let const? idk. I hate let
just u32 variable = value
I would have preferred that.
But can't change that now π
Guy yells at clouds how he is the only one in the world who can code, news at 11.
I don't know, so far I worked in java, javascript, typescript, python, ... then all the frameworks like angular, spring, electron, and for web I know html and css, sass (scss). And despite all that, Rust was a bit of a leap for me.
I haven't coded that much in Rust atm. So hard to judge atm. Mostly is front-end, which is typescript, html and scss atm.
but to get access to operating system, I had to dig into back-end and learn rust. Else I wouldn't be learning it π€£
Fair game!
I also tried a bit of rust hacking, for a local AI backend I did not really click with how it worked, but tbh I never really gave it a true shot.
I should probably try do something more of an actual project in rust before sticking with my "it's bad" opinion.
Not having anything against rust, but my first impressions have been so-so, so I'm avoiding it for now π
I am not sure, my first judgement was also; "wtf is this?"
the match entry {} within a function is pretty weeird to me.
Where return?
you don't need to give return.
last line is always return value if there is ->
So Ok(paths) is returned.
What if I want it to not return?
hmm
which makes sense, why give a return statement, if you can see by function it returns something.
What if I want an early return?
like
if(cond) early out
I assume you can still do that?
function a(): string { return "" } - the "return" is useless, since you know :string is the return type, so you know function returns something.
you can still do return for values if you want.
I do prefer explicit to show intent.
it's just not needed and I believe some opinions on the internet discourage typing out return if it's not needed.
It's like
(void)printf(args)
in C.
Showing I know it returns a value, but I'm not interetsted in the return value here.
I think the point he makes is more that nobody thinks to do it the way he says it can be done
and the drama around celebrities (DF)
I mean, I don't know. the return keyword at the end of a method is pretty useless. There is some logic to it. But I also agree with you about making code explicit.
Fair argument π
also for the "map_err(...)?" function, the question mark states that something could return there (a error). Then it breaks out returns the error.
.
after reading more about it, it seems more complex than that 
Not surprised 
fn do_the_thing(i: i32) -> Result<i32, Error> { let i = match halves_if_even(i) { Ok(i) => i, Err(e) => return Err(e), }; }
does the same as
fn do_the_thing(i: i32) -> Result<i32, Error> { let i = halves_if_even(i)?; }
Behold 6900XT OCF on crack
but the ultimate question is.... does it work?
Yes
why did you do this anyway? just didn't come with a cooler or what?

48c at 410w is nuts
this sounds a lot like when i was tempted to just strap the 2 nzxt stock fans i have sitting on my shelf to the bottom of my laptop so it can actually run at full clocks
but obviously there arent any fan headers in sight so it wouldent work
i doubt there is just magically a fan header to usb-c adapter
best i can do, it still reached 100C with 6 120mm fans pushing air directly onto the heatsink... nothing can redeem this disgrace of a cooling system
did manage to go from 3000 to 11000 on the cpu score tho lmao
altho it was kinda fun seeing how far i could push it
you can have a fuck load of stuff being string in your function, and you could return void, this is uselessly implicit
I know a lot about pyrotechnics. And by a lot I could muddle my way through enough of what's going on to turn myself Vader crispy...
I didn't have small enough ties
I've got turbo cooling for mine
If you wanted to, you could do somewhat of the same thing, design a custom shroud and 3d print it
I need something similarish to make a fume hood
however i do not have a 3d printer so I'll just fabricate something π
You could probably get by smaller fans. Maybe 92mm rather than 120's if you cared about looks
Shopvac with a hose sucking the fumes, and another hose going outside lol
I don't really need to completely remove it just move it away from my mouth and nose and dissipate it
it's solder fumes...not good to breathe up close
a 3d printer would be super handy though since i'll be using quite a few 125B style project boxes... wouldn't even need to drill holes if I were printing them
Could also do stuff that'd be hard to do otherwise like the round "disk" shape of the old Fuzzface guitar pedals...
my dad said it'll put hair on your chest
I'm planning on it
Tbh I may stick with the 120mm's for just maximum overall flow rate. Even if it makes the card longer
I added the entire Buffy series to the Jellyfin server.... because why not
If I do go with smaller might as well just use the stock shroud
And if anything
I wanna fabricate it from metal.
Where's the pyro?
Have you ever made fireworks experiences along the lines of "holy fucking shit, please never let this happen ever again"?
I've had that twice at least
niiice π
Bro, this synology log "center" sucks.
It should be the place to find logs. But then you have seperate logs in other services like cloud etc. Which you cannot find in the log center.
So you gotta open all applications you use and check logs manual one by one. No unified place in log center π€¦ββοΈ
I think it's about time I start crack this synology a bit to get shit done.
Log Center offers an "easy" solution for gathering and displaying log messages from network devices.
We are going to completely ban pyrotechnics here this year and only allow people to buy harmless pyrotechnics like party poppers and cake fountains + rockets (which aren't exactly "harmless"
Too many children lost their finger or hand using military-grade firecrackers
Not sure if this was an actual firecracker or an improvised version.... (most likely improvised) https://youtube.com/shorts/QTEWqyoPkPo?si=Z8Jvx-8HhrULx-Nu
And no, this video was not edited btw
Germany only allows pyrotechnics to be sold if they passed strict government quality control - and only very few types of fireworks are permitted to be sold/burnt year round, most are limited to new year's eve
"ohne BAM kein Bumm"
I am shocked that the government here allowed the sale of firecrackers which can seriously hurt you, not to mention parents buying kids these
BAM is short for "Bundesanstalt fΓΌr Materialforschung und -prΓΌfung", the government institution that tests and certifies fireworks for sale.
Every piece of firework being sold legally in Germany has to have a BAM number on it
Firecrackers are legal in Germany too, but they're 18+ and can only be purchased in the three days before new year - four days if one of them is a sunday
And it's only legal to burn them on new year's eve
Same here although animal right activists might ask government to completely ban virtually all pyrotechnics. They already managed to ban most of them
Wouldn't be surprised if they ban professional firework displays too
eh, doubtful
I just wish that drone shows would become more popular
tho imo, private fireworks should be banned. both for public safety and environmental protection
Yeah, we still allow people here to buy rockets for fireworks
It's funny how the government thinks firecrackers are dangerous and should be banned while rockets shouldn't. Rockets can be even more dangerous if you don't know what you are doing
There are a lot of crazy kid around my house and I'm thankful not a single rocket was accidentally fired on my house
Especially when there is wind and things could easily go wrong
rockets have the same regulations as firecrackers here
and for all of them, there are limits
We heavily regulated firecrackers while letting rockets mostly unregulated
class 1 fireworks can be sold all year round to people age 12 and up
class 2 fireworks can only be sold leading up to new year and are 18+
class 3 and up need a special pyrotechnics license
I have to admit, I love the rocket battery boxes though
You have to follow safety guidelines though or you might endanger yourself and everyone around
oh, yeah, another thing: every piece of firework sold in germany has to come with instructions on the box
most importantly safety instructions
Like, put the battery box far away from buildings and once you are done with the fireworks, soak it in water and then throw it away next day
I've had two of those tip over and fire sideways.
one from uneven ground and lots of wind, the other from recoil and me forgetting to fold out its stabilizing feet
yeah, fireworks cleanup safety is smth I wish people were more educated about
Luckily, mine was larger and heavier so it was hard for the wind to tip it over
But the small battery boxes can easily tip like you mentioned
I've seen the family dog react to fireworks and I gotta be honest, it was absolutely heartbreaking
ever since I saw him cowering in the most distant corner of the house, whimpering and shivering, I have refused to buy and light fireworks with a loud bang
That's usually the reason we have banned most firecrackers here
for wildlife, it's even worse
Ever heard of carbide barrel shooting? People here in rurar areas do that all the time
Somewhere in central europe, apologise if this is someone elses video I just got emailed it.
Dont really need explosive propellant for that. Camp fire and sealed barrel with some water will do. When the lid bending lock gives out, is t0 π
Abandoned quarry was the place for that. The camp fire kinda spreads out too on the launch π
Hope nobody got trapped by a falling barrel 
halp I'm stuck 
Great graphics driver Nvidiot
Very coolπ
Did that for no joke 5 minutes straight just cycling displays
i literally had time to WALK up and down the stairs to get my phone
Oh great its doing it again
how tf did satisfactory survive lmao
its the white window on the left
fix : get AMD next time
yea but they got a lot of very enticing OC headroom π€
who/what is "they" in this case ?
blame yourself
il start with AM6 maybe
cpu killing speedrun sounds enticing
its the only pc part i havent had a failure with
the only PC parts i havent had failures with are : GPU,CPU,case
i took storage off the list today and my case feet gave out under the weight sooooooo
i think thats everything
HMR is cool and all, but nto when it starts to duplicate my files and also all the code 
Everytime HMR kicks in, it creates another +page.svelte instead of replacing. Normal behavior would be one "drop" started event from only 1 +page.svelte page.
@pure karma so I added more zip ties

honestly at a glace you could pass this off as a real card
With some time I'm designing a proper shroud for it
So I can make it a real card
Ish
still better than my cable management so thats a pass from me
I tryed to keep it coherent lmao
Making the shroud is gonna be tough
As I wanna make it semi ducted
So there's very minimal heatsink that's getting wasted
Or not actually receiving full airflow
But that's gonna add a lot of thickness and bulk
And I'd need to test it and see how it holds up as I want there to be as minimal restriction in airflow as possible
Wile keeping somewhat pleasing to the eye
Anyone seen good (netflix) shows lately?
Does your gpu have extra PWM connect headers?
My 3090 oc has so I can technically connect case fans directly to the GPU.
From the looks of your screenshot it does look like it, but hard to see π
Sometimes I spend 30 minutes on Netflix browsing for movies/shows I like but give up in the end π
I found something to watch, thanks anyway π
Need to check again if there is something interesting this time π
Didn't have to be netflix, just something good to watch,
Giving "Dark Winds" a go atm. Has 3 seasons. Some sort of detective/thriller with native americans.
George RR martin is one of the producers.
But I doubt he actually did anything, he probably just said they could put his name on it π€£
Or gave them some money he got from the GOTS show.
thats honestly such a good modification to do
I should check Max app too. Haven't opened it in a while
I saw Reacher S2 the last few days. S1 was good. But S2 wasn't as strong.
Especially the cameran who shot the action scenes fucked up.
Example; you see Reacher kick was his left feet to break the knees of his opponent. But the camera always goes to a weird place, never actually showing it. And the edits are pretty bad you notice it. So instead of showing the foot hitting the knee, it would go to the arms or something totally unrelated.
It did that so many times that it annoyed me π€£
I do not I'll be soldering an adapter to it
At least you can watch tv with popcorns and beer
. My stomach is still messed up from stomach flu I had
Ahh fair π Good luck tho.
RN I just plug it into my motherboard and let it work
No popcorns here tonight 
So far I'm working on making the ducting layer between the fans and the heatsink
I need to buy a popcorn machine. Don't like the silicone bowl for use in microwave
you mean a microwave?
And part of me wants to get a custom PCB for fan power delivery just for the sake of looks and cleanliness
To be fair most brands do not add the extra pwm:s Asus is one of the few that actually does.
Need something like this
I'd say pass for now and maybe do it later when you've gotten the other stuff done.
I prefer my ways of making popcorn π€£ (and no need to buy machine).
My way is no popcorn.
π
Still haven't found a way to make that buttery flavor like in cinemas
Tried using a pan but doesn't work tbh
our cinema popcorn doesn't taste very buttery
That's the latest stage RN I need to figure out how to model the shroud. And I suck at cad. I'm only half okay with blender
they add a lot of sugar though
Once, I bought caramel popcorn. It tastes like shit lol
Hope it's tasty. Not a big fan of bagged popcorn tbh
Haven't tried any of their other flavors. But so far, they have been subtle.
It's pretty balanced for my taste buds.
I like microwave popcorn but afraid of buying them because they cause cancer
I think they banned packages that contained cancerous chemicals but I am not sure...
autodesk fusion is free for personal use
As Djatlov said in the Chernobyl series..
Not great, Not terrible..
Tinkercad is okay for folks who don't have solid understanding of 3D works, and can get your basic stuff made for the 3D Printer also
I'm trying my best. I don't have a from grasp on modeling at all teally
But I want to make this project as it would be hella cool
Nah but I feel like Tinkercad is still somewhat understandable to work with, as its mostly "just" drag and drop the objects into the canvas..
I think I somehow blocked YT's message about AD blocker violating something, hehe
@dire igloo No I start yelling like Kylo Ren "more MOOORE"
yeeeeeeep
especially because all the 3D modeling software you can actually use for 3D printing sucks compared to regular modeling tools
So I came up with a better idea. Insta if of making circular cut outs for each fan it's one big frame with pins where the fan screw holes are that just attaches to the gpus screw holes.
Simpler
that's deeply sus
unless it's rt up the arse
oh that's 7800xt i read as 7900xt
I can't read
Not surprised becuase AMD numbers are all over the place I don't even know what is better at this point 
The number that starts with Nvidia is better 

I love how hackable macOS sometimes can be. For windows, you need to spend a day to just mute update notications. Or use group policies. Or hack regedit.
On mac, you just set the dates to next year. Or to a 3000, whatever π
hairy ...
Absolutely no animal hair in my home
sheds all over
not surprised cuz those random ass YT benchmark videos are hella questionable
Carpet
I hope it's not standing on the carpet while it's on. the static electricity will be π
Oh it's plasti-dip. Very splotchily applied.
with RDNA 4 having pretty impressive ML IMO
would there be a way to make DLSS run on AMD cards ?
it'll have some kind of DRM in it
checks for driver signatures or something
conversely FSR4 should run on nvidia once the source gets released
ada lovelace supports FP8 so it should run on geforce 4000 one day
Since when can we manage updates?
I think it went well. The chipset driver updates.
25.3.1
You can still download "amd_chipset_software_7.02.13.148.exe" if someone wouldn't trust it.
oh dear god not this again... Twitch is being sued for 200 Nonillion Dollars
how long you wanna bet this shit lasts
its the same strategy of a infinitely multiplying fine from the google one from a little while back
please dont tell me that we will have an RTX Titan Lackwell edition
bet it'll be $5999
9999999
RTX Titan AI SUPER IT SUPER2
missing rops
LMAO TRUE
MSRP
you forgot the 2 12v power inputs for 2x more burning power
Missing Some ROP's Probably (AKA MSRP)
GB100 only has 24 ROPs, so you'll need DLSS gigaperformance mode (240p to 4K) to not overwork them
Just ditched Chrome and moved back to Firefox because of their ad-blocking shit...
Ublock origin is no longer supported while the lite version which is supported causes long youtube loading
Fifefox? I ditched it because of their privacy policy shenanigans a few weeks back...
Guess I'll move to Opera if something like that happens to me....
What's the engine behind Opera?
I know that Opera GX is Chromium, but idk about Opera itself
Lemme guess, Librewolf now?
For now...
I think they used Presto in the past and moved to Blink recently
There is no harm in installing it so I might as well test it atm to see how it compares to firefox
opera is litteraly the single worst browser you could choose from
edge and chrome are better than this hellhole
yeah firefox do sure love to have PR fumbles, but im so used to it, they arent evil, but stupid
uhhh yeah there is harm to install it π
You make it sound like it's some kind of malware lol
i mean
Guess I'll stay to Firefox then
doxxing your exact location without your consent is pretty malware like
Oh, that's a valid point
Opera or Opera GX?
and like, entire proven goverment backdoors in some of their other products too
Ain't GX Chromium based?
think so
Edge is unironically my go-to Chromium browser.
Not my daily driver (that's still Firefox) but if I need Chromium for anything, edge it is
So it literally can't be an Opera reskin unless asablic is wrong about the engine
yeah fair, i personally prefer brave for chromium, even tho their crypto shenanigans can be a turn off
It's all stuff that I need to install
Edge is built into windows already, so I can just use that if I have to
didnt ment reskin like that my bad, i ment that the general privacy intrusive things are the same on both browsers, but not the entire backend
And I can't pivot to Linux cuz I'm doing LoL broadcast work
ah rip
Correction: Migration to Blink from Presto was done around 2013, not recently
@visual tree @dire igloo things you might find interesting of why i hate opera /gx that much
yeah point 2,3 and 4 are preeeetty itchy
Uninstalling Opera: - 10.000 social credits 
I've been aware of most of this, just not that it applied to any and all opera browsers
some technically dont, but still using their products while knowing what they're doing is pretty wild
its like trusting a known scammer with your money in their other buisness, it doesnt make sense
Noted: don't fart near an air compressor while it's filling the bottles
https://www.youtube.com/shorts/XyjtV1IXOhA
Visit www.Firedeptcoffee.com for great coffee and gear!
Thought this video was a joke but apparently it's a bad idea to fart near an air compressor
Correction: it's bad if you fart a lot near the air compressor which has a filter that hasn't been replaced in a long time
What i hate about the 12pin connector is if its on a power supply, the power for that connector is restricted for that connector. So if you dont have an nvidia gpu, you cant use it.
Modular PSU, problem solved.
my cable is in the box heh
i would not buy those though lmao
it should be possible to go from the 16pin connector to 8 pin connectors
What I hate is people so excited for new frame gen crap
hows about we have some optimized games instead
frame gen can eat my ass
Frame gen can kiss my ass
so we don't need that bullshit... hows about that instead
"oh AMD need 4x MFG to compete!" literal psychotic statement
FFS I try to avoid frame gen if at all possible
I dont mind ray tracing but dlss, frame gen and any other ai bullshit needs to go
I want a faster card so I can get better FPS... not so I can get better fake frames.
i hate how path tracing is used as a bludgeon by nvidia to sell 5090s and they need frame gen for it to work
If they removed the tensor cores and added more cuda and rt cores, real performance would be so much better
the problem is not the actual connector, its how the power is received on the GPU afterward
But first, the need to make it not so power hungry
That is a comically untrue statement
Its the gpu. Wtf does a gpu need 600watts?
All the baked in AI stuff leads me to think they aren't moving away from the fake frames any time soon... or ever probably
4K performance DLSS with 4x MFG running at 80fps. so that's 20FPS quadrupled, from a 1080p input
A more accurate version is "They did not put appropriate mitigations for a very poorly designed connector on the card"
im speaking from the Nvidia tracing documentation, so yeah idk what you're saying
The gpu is the problem and the connector is the piss poor excuse of a solution
I hate both
8 pins don't melt
You're misquoting the facts
Exactly
The issue isn't "Perfect connector, bad gpu"
Isnt the 8pin rated for 300 but is restricted to 150?
It's "Bad connector, imperfect gpu that doesn't fix the connectors problems"
Its bad gpu and connector
so I've looked into this and tensor cores are a marketing sleight of hand; they're not completely separate from CUDA cores, they're extensions of the CUDA cores
The pin spec is little over 350W for the better pins of that type for the 8-pin, with the derating for 8-pins in one.
Coming soon.. GPU With Bronze, Silver, Gold, Platinum rated Pins
the connector is perfectly capable of handling 600w alone, its how the GPU afterward is connected to the power that is the problem, heres an exemple
Point is that the connector is designed to the ragged edge of the parts absolute max spec.
For the new one.
Which makes it unsuitable for actual consumer use.
You're still missing the point
You always have a safety margin incase shit goes wrong. 12pin does not
Or do like EV chargers.. Water cool them damn cables.. What could possibly go wrong with that :P
basicly for 4000 and 5000 gpus, the power connector is just afterward merged between all cables, so imagine if half of the physical cables fails, then the GPU cannot know about it and still ask for the same amount of power, which the cables cannot handle and therefor burns
And that kind of merge all has been done with the previous GPUs on per-connector basis for long time already.
Without major problems, because the connectors had ton of safety margin.
That the new one is missing.
That doesn't mean there are no problems with the connector
That just means the gpu doesn't have mitigations for a bad connector design
... Again
For the 3rd tkme
oh i didnt say that i liked that connector dw, just that its nvidia the main problem here
The connector needs to die and gpu power draw needs to drop.
My 7900XTX uses 300watts at most
my 7900XTX has two 8 pins
Same
The main problem is still the connector(that nvidia designed), not the cards with insufficient load balancing and safety features for a crappy connector
No, main problem is that the connector was designed wrong.
And then not doing extra stuff to do active current balancing makes it worse.
oh i never said that the connector was perfect, far from it, just look at the burning 4090 because the connector wasnt fully plugged in
Nvidia decided the safety margin on the connector should be minimal, then designed GPUs with no current balancing
all to save a few cents on copper
which is the only reason I can think of for doing it
ehhh i'll be real, the connector is a singular point of failure, but basicly all the time it just works well, like the amount of burning GPU is still very low, doesnt mean that its good, but that its mostly an NVIDIA problem :/
Switched to smaller pins.
And then sized the connector to ragged edge of what the pins are specced for as absolute max.
Going against any normal design principles for consumer connector.
Where all the previous power connectors had "insane" safety margins between the computer spec, and the part absolute max spec.
the failure rates on these connectors are a legitimate health and safety concern
I gotta say... PedalPCB's printed boards are pretty good quality
This. I wish people would take the time to optimize their shit.
(people in this case big corpos)
I improved parts of my code by 3.3% by removing inheritance today 
Call an ambulance
The plan is to use it until something like the 9070 xt or 5080 is in stock at a good price, then give the 5070 to my sister.
Does it come included a fire extinguisher? :P
I got one, don't worry
U know your PC rig is right, when the light dims when u turn it on :)
I had a monitor power supply apparently go so bad that when I brought it to a computer store it tripped two separate Earth Leakage Breaker Circuits
So either the wiring in that store was weird or it caused a massive short circuit
Trying to isolate a problem I have. Anyone have ideas?
My oldest monitor (10 years old) sometimes flickers (happens 2-3 a day). Not the power that cuts out I think, but it just goes black for tiny second (like a reset), the power led stays the same.
The monitor is connect like this: MONITOR => HDMI => KINGSTON HUB => USB-C => MACBOOK PRO.
Sometimes, when I move the hub, I get the same behavior. So the HDMI connection of the HUB is a bit wonky.
But when this happens, the HUB isn't moving. I am just coding or watching YT ...
So it's either:
- HUB
- HDMI cable (easy to check, should try other cable).
- Monitor
- Macbook fucking it up somehow?
Just remember this...always. The green wire goes all around.
Try what happens if you wiggle that USB-C connector?
Nothing happens when I wiggle the USB-C connecter at macbook side.
Why do I always get this flaky as hell behaviors with my hardware, that are so hard to track down because they happen "at random".
Ok, so not that connection then.
That was the most fiddly connector of the bunch under most "load" so wanted to check it first.
The reason I ask is, I don't want to just buy another monitor, or invest in better hub, without knowing the cause π
Maybe I buy monitor and hub, and it turns out my macbook is dying π€£
I have an extra monitor for my work laptop, everything through USB-C
I have found out that the USB-C cable is very very fragile - And assume its breaking the wires inside, but last I bought some rather expensive USB-C cable for this monitor (Roughly $60) after having rolled the cable up 20-30 times, the monitor began again to blink, I changed the cable - and problem was solved..
Anyhow .. you have a lot of steps, which all can produce fault..
But it does sound like a broken cable
hey, i currently have an almost full 1TB ssd card and i have 1 more free slot for a SSD, should i get another 1 Tb or a 2TB?
im just wondering since mixing for example 2x8gb RAM sticks with 2x16GB RAM sticks isnt such a good idea either
Mixing capacities of storage doesn't matter in the same way, what are you storing?
If you're storing games, my suggestion would be to buy something in the 2-4tb range, put all your files on there, and keep a minimal c drive
and leave out the one ssd i have rn or just put a second one with 2T in?
Honestly that depends on how it performs
What drive do you have currently?
there is only one reason to intentionally pick a smaller drive here and that reason is cost
I'd guess current prices are around $50 for 1TB and $110 for 2TB
oh, germany it is
A2000 not particularly great, but I guess it's workable for a main drive
But might be better served by a very quality gen4 or gen5 drive as boot and bulk
yeah im just trying to get more storage space for as little work as necesarry and as little money as needed yk
outside of avoiding ewaste, name one reason as to why
i can live with a slightly worse performance if i dont have to copy all data from one ssd to another new one
my current one isnt available anymore
especially gen5 has zero justification right now unless you're really hammering the drive constantly with lots of writes
might be better served to throw your money in a fire in that case
you shouldn't buy it anyway
you just keep your current one and add in a second one
you just add it in there, no need to move anything
I wanna add a second 4TB to this system...
nice
Easier to organize, default save locations will be easier to track down and depending on what board/platform they have there may only be one nvme slot... also game launch times are nice on gen4 compared to gen3
Wasn't arguing for it, just throwing it out there as an option
for a baseline 2TB drive, I recommend an NM620
does everything you need it to, comes in at 100β¬
these are some recommendations of the website just if u were wondering the prices in germany
that is ZERO reason to go for gen4 and especially not gen5
and the load times argument is debunked constantly by any TPU SSD benchmark that releases
i got SN850X's but there expensive as all heck
throwing it into the discussion as an option needs to have a justification
and there just isn't one
I'm in Germany myself, checking Geizhals rn and cross-referencing SSD specsheets
of those SSDs, I'd consider maybe two and even those need prior research
They had the 990 Evo Plus 4TB on sale when I got my parts so I jumped on it
yeah well as u know those recommendations are just crap most of the time
aight, my rec for 2TB is Lexar NM710 at 103β¬
excellent performance, great power efficiency (not that it matters) and very fair price
especially on websites like amazon and thomann or whatever where they say: other users also bought this and its the worst crap youve ever seen
recommendation for 4TB is actually the Lexar NM790 from the overview
well i dont think i need that much considering the price as well
there are a couple which are cheaper but they either come with bad durability or a quality gamble which isn't worth the savings
i see
btw, the very first message they sent in this convo mentioned "i have 1 more free slot for a SSD"
yeah im not quite sure how many i have left but def 1 more
im not that into tech and when i read my motherboard detail sheet on geizhals i was pretty confused i must admit but yeah
xD
anyways, NM620 is the baseline for 2TB, but 3β¬ more for NM710 is worth it
for 4TB, NM790 really is the best pick
these are the MB specsheets that I deal with
https://docs.google.com/spreadsheets/u/1/d/1NQHkDEcgDPm34Mns3C93K6SJoBnua-x9O-y_6hv8sPs/htmlview#
much appreciated
yw, happy to have helped
btw, this is where I look up SSD specs
https://docs.google.com/spreadsheets/d/1B27_j9NDPU3cNlj2HKcrfpJKHkOf-Oi1DbuuQva2gT4/htmlview#gid=0
that or the TechPowerUp database
and if neither of the two have the specs, it starts a wild hunt for advertised performance/specs and eliminating components based on their attributes
holy sht seems like you know what you are doing
will defenitly copy paste those links in my self notes on whatsapp for future use haha
Is there an adapter that turns a 12pin connector at the psu into a couple 8pin connectors?
storage is completely independent unless you use things like raid (storage array aggregation logic hardware and software) which you likely don't
the only issue you can run in is limited compatibility when multiple storage are plugged in because the motherboard does not always have straightforward layout
double check your motherboard + cpu support this ssd port
this case is surprisingly light once there's nothing in it
setting up for a triple pc transplant mixaround, and step 1 was gutting my old pc
also this is the first time i've completely gutted that tower down to the frame since i got it
it needed the deep clean π€£
there is no 12 pin connector?
there is 4,6,8,16,24
the 16 pin used to be 12 pin when it was nvidia exclusive, before it grew the extra 4 sense pins, so it's still called 12 pin sometimes
Officially now 12V-2x6, with the revisions, 12VHPWR before that.
ah yes people complained about 12hpwr so they changed it to 12v-2x6 and now no one complains about it
Name change was part of the revision that changed the sense pins to be shorter and power pins be longer.
In attempt to mitigate that partially inserted connector issue.
So that the sense pins would only mate once power pins were fully connected.
Another giant fuckup on why that wasn't done in the original version already...
And as they kept backwards compatibility, that isn't still assured, just more likely to happen.
still doesn't address how the spec has pretty much no room for error on power draw, and gpus don't sense if some pins have bad contact
if it had one of those fixed, it would probably be okay
with both you end up with no room for error and nothing to save you when you error
Lipstig on a pig, by just changing tolerances.
so what now it has even less power pins so it can catch on fire better π
il keep my 2x8 thank you
No?
Same amount, 6x 12V, 6x ground, and then 4 sense pins, two which are used for max power detection on GPU end, and two in theory would allow GPU to send information to PSU, but no-one uses those.
should sense the fucking power on each 12v pin
But that is GPU or PSU level thing, not connector thing.
That should in theory be done on ATX, EPS and PCIe connectors too, as they too have multiple lines of same voltage just bunched together.
it's so obscenely hard to get a new standard added to PCs and this is the first new one we get in years
Ignoring ATX 12VO?
Not even OEMs use it.
It was attempt to standardize what OEMs have been doing, but everyone has just continued to do their own thing.
when nvidia forced the standard through, they were able to cause they had an intel-level dominance of graphics cards
now uh. less so
i doubt board partners are happy about the return rate
and now for the first time in years there's a competitor good enough to get people to switch
it should at least keep 5v on psu main
I don't get the 12vo thing
it's again marketing bullshit corrupted
always "it's better because it's simpler to think about for investor" logic
nothing is actually simple nor should be at all
5V is almost completely unused today.
That is the least used rail after -12V.
I mean dropping 3.3 ? ok
dropping most 5v lines ? ok
making it only 12v no
And even 3.3V is not used lot.
Point is that today almost everything uses 12V already and downregulates on the device.
Even HDDs etc.
usb ? ssds ?
hdd controllers ?
if it's device regulated is mostly for stability reasons
because most people got the bad habits of buying cheapest psu in town
and these psu will never work to spec
Not really.
Regulating to whatever you actually need from highest voltage available is today very cheap, and not lot of chips use even 3.3V anymore, but lower voltages.
So you just take 12V and get it to exactly the right voltage your PCB needs, instead of middle steps like 5V or 3.3V.
it's more efficient
then I would prefer 24v
go all the way in
also a much smaller connector rather than the yuge 24 pin thing
Why 24V, just go to 48V straight, which is the "max" it can go easily with current regulation.
Yup.
I agree for 48v was just thinking safety designs may not all be ready for 48v
24v is still pretty much template
48V is the "max" that is safe and allowed as such in regulations.
That was the point.
50V DC limit in parts of world, 75V DC limit in others.
IIRC EU is 50V AC, 75V DC max that consumers are allowed to handle.
Same reason why all light-hybrids use 48V, and why car industry has talked about switching to 48V for ages.
I meant safety circuit wise
not user wise
even usb power delivery can go up to 48V
12V was used because that was what cheap components could do.
Today 48+V is trivial.
For both PCBs, components, connectors etc.
like not risking too high spikes for capacitors
63V and 100V capacitors are standard?
10V, 16V, 35V, 50V, 63V, 100V, 160V and so on.
12V designs use 16V caps, 48V designs would almost certainly use 63V ones.
ok
then I'm for 48-50 (target-hard)
I even wrote draft for up to 50v baseline for a pc equipment network circuit
So I agree on paper
I was simply worrying most design would not work straight to 48v (phases etc)
I know too few about industry electronics
all I know is 12v is too low rn
and imo always have been
it was only useful for dirt cheap and simple components
like those dumb 6 cells batteries
nobody use that anymore
like in 50s
Lot of silicon FETs seem to be good only to 30-40V.
GaN FETs are good to 100-600V.
IGBTs seem to be from 300 to 1200V mostly for the high power ones.
the thing I hate the most is the power fighting drifting
like all 5v Chinese boards fails if provider under 5.1 actually
I see that a lot of motherboard are unstable if provided less than 12.05 or something
it's like negative margin
reliable circuit ? cheap psu
reliable psu ? cheap circuit
like no wonder it's lottery and then you get cheap psu + cheap circuit bad
the only way I know to prevent that from happening is by gating requirements with failsafe
making something break at 12.0 but everything of importance still work
designing psu that provide stable 12.2 but regularly fall below
or have circuit testing procedures at 11.9
adding an "boost mode (unrecommended)"
Chinese seem to like doing rectifiers with cheap standalone 0.6V diodes.
So you get 1.2V of voltage drop from that, so input 5V DC, and the actual circuit gets just 3.8V.
Even when rectifier isn't needed.
And also "spec" the whole device input to whatever is the highest spec voltage for a chip on the PCB, forgetting that kind of drops.
Why is AI so bad at generating regex jezus...
because it requires accuracy
And some understanding of the syntax.
Otherwise any regex will look like any other regex.
My query was: 10-20-2025 GMT +1 (European Standard Time). Give me two groups "10-20-2025 GMT +1 " and "(European Standard Time)" and it failed over and over again. For like straight minutes.
why not making it yourself
Solution was ^(.*?)\((.*?)\)$ well it gave me far more complex shit. But I just assumed this was something AI would be able to do.
dont you mixed the parenthesis ?
Especially when you just need to get anything before ( and everything after, with ( added/included in the second...
If all are actually in that format and no extra (:s.
Well, a lot of languages when you do string.split("(") they remove the "(" in the groups.
And you can just add that back in?
So then I get 10-20-2025 GMT + 1 in one group and European Standard Time) in second group.
Oh yeah with index + or - shit. Fun, and very verbose.
Or I just match a regex and be done with it.
And point was that you can do the same with regex too?
you said you wanted with parenthesis, that's all
I've been over it. I know how to split and get my ( back.
Point is, I don't want to spend more lines code for that.
what
I just want two groups straight out of the box sso I can do array[0] and array[1] and get what I need.
I don't mind if the () are gone.
I just don't want string) or (string ... either no () or both ().
also the second ? is not needed
if you need performance you can do [^(]* in first group
dropping the ? too
it will avoid backtracking
tbh regex can be learned as you need it, unless you are doing complicated logic
better spend 5min to read regex doc finding building blocks, than spending 5min trying to get an AI output which you dont really know it will work all the time or not
the key for regex is nesting and recursion
if you can work with that, you can work with regex, even if you dont have the whole regex vocabulary
(especially since the vocabulary change depending on regex engines)
cba with regex atm when I am focusing on front and backend stuff.
I have 3 regex expressions in my codebasee atm. I kinda know what they do but I wouldn't be able to come up with them myself.
Or if I have to figure it out myself, I would spend too much time on it. Rather focus on the important stuff.
It's just a means to an end
Regex is something devils devil invented..
Yeah, use that to test strings. The tests is pretty cool.
But I don't know if regex, SQL and some other languages are getting outdated. It sure feels like that.
Because this doesn't feel modern. π€£
Gonna build me one of those bad boys soon.
Guitar effects/amplifier prototyping board
Ground, +4.5v reference, +5v, +9v, -9v and even a charge pump to hi 18v for those extra grindy drive circuits.
six pin blocks to pop in up to six potentiometers...
Really only lacks baked in stereo ability but I won't be doing much outside of mono anyhows.
this is a form of art i have begun calling "locking in to create bullshit"
i did discover that the 4090, when doing nothing, is only drawing 7-20W
I sure love having the entire Three Stooges collection...
Makes great background content when I'm playing Civ or SF
i have a 10hr WWI documentary i'll play in the background if i'm bored while doing stuff
I should put a bunch of stuff like that on my jellyfin server
i used to watch stuff like that on the side before the history channel because the what the fuck is shit shit channel
The Science Channel was taken off my cable package many years ago and when it returned it had gone to crap too :/
Take your personal data back with Incogni! Use code standupmaths at the link below and get 60% off an annual plan: https://incogni.com/standupmaths
^.?$|^(..+?)\1+$
Props will be for sale here: https://www.ebay.co.uk/usr/standupmaths
Fantastic blog post by Illya Gerasymchuk. "Demystifying The Regular Expression That Checks If A Number Is Prim...
It's quite elegant actually, just super inefficient
"The Zen3 and Zen4 L3 ring fabric topology is replaced by a mesh topology which reduces latency and increases bandwidth." (Zen5)
Source: @rSkip
intriguing
when intel went mesh, latency went in the toilet
always assumed that meant mesh bad
should have realised it meant intel bad
And that real mesh is only on the biggest server CPUs, not the consumer ones which are still just single ring.
Even that multi-ring was only on server CPU dies.
even the 10 core xeon had shit latency compared to 8 core ring bus
remember when they tried to sell it in gaming systems
jee zus
Not gaming but HEDT, for those XEs.
Threadripper segment.
they were aimed at gamers, I remember it
they had a motherboard you could use a regular 8 core skylake in a HEDT system, losing memory channels and PCIe connectivity; the high core options were sold as an upgrade path
They were older gen in actuality (modified Skylake, so more corresponding to 7th gen Kaby Lake than 9th gen consumer chips), with significantly lower max frequencies.
Cascade Lake I mean.
They even used same MBs and chipsets as Kaby Lake-X and Skylake-X
But seems even Broadwell server CPUs had same level of core-to-core latency (40-50ns).
So might be optimization decision from Intel for the server CPU dies.
meanwhile hopping from 7820x to 9800x just to be safe (and have more lanes)

