#off-topic-tech
1 messages ยท Page 163 of 1
You kept doing that during this whole conversation multiple times, ignoring previous parts of conversation and returning to your starting arguments.
low INT high WIS
I tried linux after you guys insisted on it to backup my claims and all of them got verified
same for firefox
low int, low wis, low strength, low dex type of build
how wouldi work out, i dont trust gyms
I have a -6 contitution penalty at the moment
have you ever heard of these things called "confirmation bias" and "self-fulfilling prophecy"?
what are his claims on linux
low charisma too cuz after five minutes of talking to you, people want nothing else but to strangle you to death
He, as total Linux noob, tried to install and use ArchLinux.
With Ubuntu he couldn't find how to install software in the GUI side.
and gaming performance sure as hell isnt 1/3 of what it would be on windows
From distro repos.
steam deck would not be a product if the performance was that bad
but see, what if all you play is Valorant, Fortnite, Destiny 2 and TFT?
pretty sure all of these wont even launch with their kernel level anticheats
not sure if they all use one
but you get the idea
none of those work on Linux, exactly
Fortnite and Destiny 2 could work if the devs weren't lazy fucks (Easy Anti Cheat has Linux support, but needs to be enabled by the game's devs)
TFT and Valorant both use Riot's Vanguard which Riot has made very clear will never receive Linux support
gemini.google/overview/gemini-in-chrome/?hl=en yippee
Partly because on Linux, they cannot protect their anticheat from edited and recompiled kernel.
also cuz it's too much effort for too little gain
And IIRC they can not easily have anywhere near same level of "anti-cheat" as on Windows Kernel driver.
blatant lie.
the nanosecond they consider it profitable, they will support linux
riot is too greedy not to
thats the issue
everything would work on linux if its devs cared to implement it, which a lot of times for big games they dont
ive used it in the past, the gaming experience is just not that amazing
Point was to just include the specific external bits Linux version, and allow Proton do the rest.
Not do native Linux.
well yeah - but that's assuming it will ever be profitable for Riot to cater to linux users
which was part of their statement, the "never" was an exaggeration on my behalf
im not saying it will.
but IF, then they would literally break the laws of nature to implement it at an instance.
riot games is so greedy that i cant put it into fitting words
So glad I got this monitor now...
mfs will always have an xkcd for every scenario
i hate that gif
dident try... i did
the wiki has answers
Get 30% discount in the GVGMALL with my "SKAG" code!
Windows 11 Pro ($22): https://www.gvgmall.com/gvg/SKAG1
Windows 10 Pro ($15): https://www.gvgmall.com/gvg/SKAG
Windows 10 Home ($14): https://www.gvgmall.com/gvg/SKAGH
Windows 11 Home($20): https://www.gvgmall.com/gvg/SKAG1H
Office 2019 ($46): https://www.gvgmall.com/gvg/SKAG9
Today I br...
linx is slower in gaming?
Install and USE.
You installed it, and then tried to replace your Windows system 1:1, weren't satisfied in very short order, and removed.
(for nvidia it is, because they refuse to open source drivers)
i used ubuntu so this is completely irelivant
no its not, you used something that isnt made for gaming and then got upset its not good for gaming.
why is this lawnmower so bad at trimming my hedge!!!!
IIRC with Ubuntu you never found how to install stuff from the distro package repository.
ubuntu has a GUI way of installing sofware
From Ubuntu maintained repository of precompiled programs, drivers and other packages.
yea the store with nothing good in it
No, that was different thing.
Point was that there is that distro repo where even those Nvidia GPU drivers live in.
Not the Nvidias own installer script.
You succesfully AFAIK used the Arch version, but not the Ubuntu version.
On Ubuntu you only used the flatpack software "store", not the distro repo.
i used ubuntu as much as i could with the limited software available aka browsers and thats it
Like I said, he never found how to install distro repo stuff, only flatpac stuff.
none of them work and whe few apps that do are absolutely terible
a lot of stuff you can also just install with a .deb from the internet
not that its the best way, but you can
And as that flatpak (finally correct spelling) stuff is much more limited and is for just certain user facing software in general, the availability was lacking.
As most of things still live in the distro .deb repo.
i probably would be able to use ubuntu a bit more if the computer im tryign to use it on wasent constantly in shambles because i keep getting scammed on replacement hardware

And back then he just wouldn't accept there was another software source in there besides the one he had found.
And I don't use Ubuntu and their own documentation is almost completely missing, so I couldn't provide good instructions where that actual distro repo GUI was.
i know what you mean and its not what you think
theres 2 updaters
one of them being the nvidia driver updater
and the other being the app updater
the actual os updates are just merged in the app updater for no reasson
thats bad, why?
and then all the broken custom drivers are in the driver updater
Because both are just packages in the distro repo?
There is NO difference between them.
And the actual Nvidia drivers ALSO live there, no idea what that Nvidia driver updater actually was.
cause its hidden and nerver tells you there in there
me when all it is sudo apt update && apt upgrade to update all apps and nvidia drivers
so you wouldnt look for "updates" when trying to update things?
is what what you are saying?
wdym hidden, it shows up when you look for updates
look where???
the search bar?
there isent one
there legit is
there is lmao
theres one in the settings but the os updates arent in there because it would make too much common sense
its so joever
"there is no search"
Type to search
i have never used ubuntu, but let me make a wild guess
its this icon, and then it shows the search
ok thanks for embeding that discord
yea that opens that menu
but that dosent change the fact that the search bar is an app search
and you can type update in the search bar?
But basically with Ubuntu, MagicZ never found how to install distro packages.
And then IIRC tries to use the Nvidia site driver install script to install GPU drivers, instead of the distro packages.
it wont just magically find the update INSIDE the app
you mean it cant find this?
And just complained how limited the selection inside that flatpak store was.
And then gave up.
But I'm off to bed now.
i complained that the provided command did fuck all ๐คทโโ๏ธ and then stopped caring because the preinstalled one was good enough
good enough is a stretch considering it barely did basic dispaly driver things
you called it that
its good enough comparitively when the stardard for a dispaly driver gets set to as much as 3 seconds of display latency by default
not because its any good compared to anything competent
thats not even a thing
375 WATTS??????????????
this has to have been an absolute titan back then, i think its performance is like 2 times that of the r7 300 series which were released 2 years later
anyway im skeptical, even though we're talking about emulated games from 2010 will these old ass cards with 1gb vram be enough
Because that Nvidias script was meant for older Ubuntu versions.
Right command would have been something like "apt install nvidia-gpu-open".
Or doing the same thing via the distro repo GUI which I tried to get you to find.
Ok, wrong name, and Ubuntu also now has ubuntu-drivers tool (and GUI version) to autoselect and install.
I found Ubuntu Servers documentation, which is good.
While the Ubuntus consumer documentation was almost nonexistent.
They have official documentation for server? ๐
Never even looked...
There's so much in the way of user generated tutorials and how-tos
UwUntu > Ubuntu
Because good official documentation is much more complete, gives all the various options, and tells the differences in the ways to do things etc.
man is a thing ๐
We should have never stopped pounding RTFM into peoples head heh
Different use case.
In this case for example: https://documentation.ubuntu.com/server/how-to/graphics/install-nvidia-drivers/
This page shows how to install the NVIDIA drivers from the command line, using either the ubuntu-drivers tool (recommended), or APT. NVIDIA drivers releases: We package two types of NVIDIA drivers:...
That used to be a common response on IRC... and they were right. It's all literally right there
Yeah that's where you use a tutorial/howto
It has worked for me every single time with stuff like that.
Yeah, but to use man, you first need to know which command you want to use.
Other complex systmes you aren't familiar with that are documented even are best left to a howto since there are a lot of variables involved with which software you can use for each function
like with a mail server...
And the previous time I tried to find Ubuntu help, was to locate that distro repo/apt GUI for MagicZ.
There isn't even any one way to do that...
Just the name and where to find it.
Which in Debian documentation is part of basic install documentations "next steps" part.
Telling how to start finding and installing rest of programs you might want.
digital ocean has tons of great stuff... as far as documentation for ubuntu
well not so much docs but showing you how at least
Doesn't take much documentation to beat what I cut my teeth with lol
My first "real" computer had DOS 2.11 which had a physical manual that was good enough I guess. There wasn't that much to DOS itself so there wasn't a ton to document. You did learn about things like batch files.
And the commands all had a /?, not as good as man but similar.
(mostly all)
I miss the sounds of extra clicky hard drives, floppy drives and a modem connecting
The big BBS in this area had 40 dial in lines and I saw his modem rack... pretty insane lol
haha one of the old galactiboards
One slot, bunch of modems. They had a box variant too that housed inteneral style modems.
We had networks... worldwide. Just not the internet. It was stuff like FIDONet or ECHOnet
And it's so simple and silly by modern standards but... we had QWK... https://en.wikipedia.org/wiki/QWK_(file_format)
QWK is a file-based offline mail reader format that was popular among bulletin board system (BBS) users, especially users of FidoNet and other networks that generated large volumes of mail. QWK was originally developed by Mark "Sparky" Herring in 1987 for systems running the popular PCBoard bulletin board system, but it was later adapted for oth...
Just 8 serial port card.
As each port only actually needs 3 (SW flow control) or 5 pins (HW flow control)
you would d/l a new packet that had all the new messages since the last one, you could then clear the phone line and read/replied to messages then connect and upload your packet that would get spread aroudn the net
"just 8 serial port card"
keep in mind the era pal lol
ISP's weren't even a thing, they weren't stocking up on them for sure, these were hobbyists slapping them in average machines at home
Point was just that that card isn't really something "special", just implementation of common thing 8x on same card, with then one connector to split that out, as most of the 25-pin connectors at other end are not actually needed.
And even on the 9-pin COM port, half the pins are not really needed.
It also wasn't super common for somebody to happen to have a T1 or better to the home
and they may not have been... wired
General explanation for others, so everyone doesn't have to try to look what that thing actually is.
unless i had one in front of me again i couldn't tell you
the pins physically being present means absolutely nothing
All the pins on that 9-pin version have uses, they just aren't important.
And that 25-pin version has more than 9 pins in use, but again the functions for most of the pins aren't critical.
The number of people that cares about any of that was about 0 heh
*cared
Like most products we were just buying something that solved a problem
In fact even doing that kind of research into the hows would have taken way too much effort
since... it didn't matter at all
And I was just trying to make quick explanation about what that specific card is and how it can work for people who don't know details about how modems were connected.
Mostly what those things were doing was allowing you to have enugh IRQ's to even had that many modems really
that was the "magic" part
6.25 pins per serial port in that card side connector.
So almost certainly serial with HW flow control (SW flow control really wasn't viable back then anyways), with possibly the Ring Indicator pin too connected as the sixth pin.
that grid can all be modems... 256 lines
plus the local conole could login directly
What was amazing is you could have a crapton of dialin lines on very VERY modest hardware, even a 386
TBBS was even more lean
MBBS had a huge draw though, MajorMUD. MUDs were big and even though that was kind of a "generic" one is was fun. Drew tons of users.
Loooks cosy 
Yeah most I ever saw lit up was 40
It was also interestinb being able to just spy on somebody (or everybody) at will.
And this will really rustle somebodies jimmies. You could even see passwords right there in plain text in their account details
Very security, much protection.
Well, Unix only got hashed password in 1985-1990 depending on main version.
Plain text before that.
Well to be fair we had remote console access limited so even if an operators account got compromised they couldn't access that stuff. Anything touchy you had to be physically at the console. And if somebody is sitting there you have bigger problems than the passwords lol
Like who is this, why are they at my console...
why are they suddenly leaking all their radiator fluid...
Ah, no, passwd could contain hashed passwords even before password shadowing was added.
That was only done for the reason that passwd needed to be global readable...
Sorry, mistake on my part.
oh ok
Point was to be able to make that actual hash containing file only root readable.
I'd have people ask for the forum pw for my minecraft server and i mean i know it's them
but im like uhhh i have NO way to see that... i can manually do a reset via email or set something and you change it
Oh another cool thing with the BBS thing was the get togethers. The big boards would have regular meetups. Go bowling, minigolf or whatever. Everybody was within a reasonable range of each other to meet up.
And... back then that's where the nerdgirls were ๐
Oh OH
Man.... they just had to shut up
Just let me have more VRAM
It is so bad to give option for less VRAM...
But yeah, shouldn't have commented.
I got bored
Looked at a window
It has nice sunlight scattered from the nearby building
While focused on far
Placed 2 fingers close together, near the eye
And made a tiny hole until it appeared round despite it's actual shape
A little further (smaller) - and now i can tiny black circles moving around
Not a microscope, but still reminds of how much is missed in plain sight
Description incorrect: swirl the index finger for that hole
Actually now I'm concerned if i should be seeing so many of them
Also reminder of how our consciousness removes everything mundane and uninterestingvl: those tiny circles, hair...
And focuses on just what is directly in line of sight

walk in to swril the index finger for that hole
not even gon... im just... yeah i'll be out in the garden lol
theres nothing i hate more than games that take 47 decades to launch up
nothing worst than going to play a game and havign to sit through 10 mintues fo useless mumbo jumbo before you can actually play the game
"We wanted to give buyers the option of cheaper variant, which is perfectly viable, if they are willing to use lower than maximum textures, in some current and future games."
That's what I hate about large Minecraft mod packs + decent textures takes way longer than it seems like it should.

i just had the misfortune of looking at x870 mainboards.
Asus has gone completely mad.
they offer only 2 Sata ports on 450โฌ mainboards
when did mainboards become more expensive than high end CPUs
my current asus strix mainboard was 200 and i thought that was already quite bad
Asus used to be expensive but at least have actually good products.
But this gens Am5 boards seem to get beaten by almost all competitors at a lower price
same time as everything else chip shortage as an excuse for the increase, inflation as a excuse for the price staying
its the same for storage any well known nvme drive from 8 years ago is still the same price now
you have to be kidding
dear god whats the matter with you
ssds have been going down in price almost monthly over that period
if that was the case my pockets would be filled with 2TB nvme drives
but no outside of very few there still 400$
adata drives cant even live for a week
thats been reported loads that there drives are awful
have some samsung in that case
why are you just making shit up thats clearly not true
what site is that
well, my gpu went and died on me
must be nice being in europe
i dont think its different for the states
i bought the 4tb version 2 months ago for like 240
yea i paid 200 each for my SN850X's and they were 50% off
where do you live?
762948 Samsung 990 EVO Plus 4TB Samsung V NAND TLC NAND PCIe Gen 4 x4 and PCIe Gen... 1 $249.99 $249.99
That was in Janruary
canada so the land of inflation
because if its outside eu or states, stuff is likely just more expensive
its canadian dollar man...
dont just say 200 without mentioning its fucking canadian
i love how easy it is to defeat you with a simple problem
Well at any rate the drive is great for me so far ๐
In USD that converts well, less than I paid but that makes sense since prices always fall
defeat?
Just look at the time... it's game:30
no but seriously call me dumb all you want idc but atleast act smarter and do things right in that case
checking for curency is like step one of comparing pricing
i love this
uses windows 7 sounds
i wonder where the audio files are located
found them
using old windows sounds in new windows is tight
i do miss the old "chunky tech noises"
where there was a audible sound of the system doing something productive
Banana Boy #3 overlooking the projects
automated fidget toys that sound like old tech
Wha... Why?
People forgetting different currencies exist...
Including MagicZ.
Also, welcome to PC advice hell:
- not knowing where the other person lives, so your local recs don't apply to others
- dealing with MagicZ
100 USD = 137 CAD
It's not even that, dude just threw the name Samsung in the ring as if the logo on the sticker means anything about quality.
990 Evo Plus is a good Gen 4 drive, just not worth its price
And the USD price is without VAT, CAD price includes it.
It was meant to be the expensive drive.
MagicZ complained how 2TB drives cost $400 (CAD).
Then one example of sub $100 (USD) drive, then another example of Samsung drive.
With Exceria Plus G3 and NM710 costing around 100โฌ, there's zero reason to spend 130โฌ on a 990 Evo Plus
Then MagicZ showing the price inside Canada.
For that expensive drive.
It was about price comparison between countries in the end.
Not about what to buy.
MagicZ is just incompetent, lol
Looks at specific models instead of setting loose filters and sorting by price
Skill issue
Yeah, but point was that in Canada, electronics are more expensive than in US when comparing just numbers, without currency comparisons and including VAT.
MagicZ was still wrong about the prices even inside Canada.
As the expensive drive was less than 300 CAD, and not over 400CAD minimum for 2TB drive.
4TB NVMe in Canada, starting $300
https://ca.pcpartpicker.com/products/internal-hard-drive/#sort=price&A=3800000000000,24000000000000&D=1
Choose Storage
Umbranocturna didn't realize different people live in different countries
MagicZ didn't realize CAD is worth less
But in Internet, need to always remember that LOT of countries use Dollars, not only US.
And that US prices are without VAT, while almost all other countries include VAT, often significant amounts of VAT.
Highend gen 4 drive, <150 CAD for 2TB
https://ca.pcpartpicker.com/product/BqP8TW/patriot-viper-vp4300-lite-2-tb-m2-2280-pcie-40-x4-nvme-solid-state-drive-vp4300l2tbm28h
I think Canadian prices also exclude VAT
Not sure tho
I remember at some point, 7800x3d was so much cheaper in Germany compared to US, that you didn't even have to account for VAT
Before or after currency conversion?
After
German price including 19% tax converted to USD, still lower than cheapest option in the US
Ah, forgot how bad the availability was in US for long time when demand outstripped supply and everyone scalped.
That was during peak CPU craze, shortly before 9800X3D released while 7800x3d prices skyrocketed
But like, not peak pricing, it was during one of these windows where 7800x3d pricing was actually tolerable
Like around $420
9800x3d is at a reasonable price rn, but mainbords arent.
I need x870 chipset for the increased usb Ports
I might upgrade to a x3d chip down the road
Also, X870 is just single-chip solution with same amount of USB as B650.
X870E is double-chip solution like X670.
Both force inclusion of USB4-chip, which raises the prices.
And I can relate to the USB port thing, the fact my new monitor has a 4 port hub baked in saved me lol
and I still might, at some point, need to add a card
Hubs nah, pcie card perhaps.
And no i do want the current gen to be on the safer side for buying future am5 cpus.
I looked at a bunch of b boards
Quality doesnt look that good honestly, evwn with x870 quality doesnt look good untill the 350โฌ+ range
touching grass?
Go outside, touch grass. Go inside, touch ass.
outside the grass is so green that certain green man might be hiding in it.
and then I will feed grass
But anyways, X870 is actually B850E (B850 + USB4).
And X870 doesn't exist, only X870E (2x B850 + USB4).
IMHO complete fuckup from AMD.
And seems that the actual chipset chip used for all the AM5 "chipsets" is exactly same Promontory 21.
So B650, B650E, B850 and X870 use single unit of exactly same die.
And then X670, X670E and X870E use two of that die.
Just that the other requirements for the board to be branded with one of those "chipset" names changed.
Well tzst kinda sucks
The only benefit of X870 is USB 4
Just look at B650 LiveMixer IO spam
- How do you determine quality?
- What makes you say that 600 series chipsets won't be able to handle future releases?
The issue why early AM4 MBs don't have late CPU support is just BIOS/UEFI Flash chip size.
Not enough space to support all AM4 CPUs at same time.
A lot of people seriously underestimate the quality of AM5 boards while overestimating the strain put on them by their usage
The B650M-HDV, a somewhat simple $110 board, turned out to be a top three XOC board for AM5 (before 800 released) - only losing to B650E Tachyon and X670E Gene
iirc some mainboard manufacturers still made it work somehow
Yeah, separate versions for early and late CPUs IIRC.
Or stripping their own stuff to minimum.
Just to drive this point home:
It takes a good amount of user effort and manual tuning for a 7800X3D's power draw to reach the triple digits
In TPU's testing, the maximum it got to was 91W
https://www.techpowerup.com/review/amd-ryzen-7-7800x3d/23.html
But basically for AM5, you select a board that has the connectivity you want.
Because outside the absolute lowest end ones, all of them can run even 7950X/9950X at default power limits all the time.
And even some of those lowest end ones can do it.
Where limitation is purely VRM cooling.
And damn, 9800X3D is actually super power hungry, coming in at a worst case consumption of a measly 157W
https://www.techpowerup.com/review/amd-ryzen-7-9800x3d/23.html
And with 7800X3D or 9800X3D, all of them can do that at full tilt without any power limit.
Ok, so not possibly that 9800X3D unlimited in all loads.
Without top-down fan or some other active cooling for the VRM, on the boards without VRM heatsinks.
Damn, they actually got a 9950x to draw 300W
But in general AM5 boards have very overdone VRMs and VRM cooling.,
For reference
Only the absolute lowest end boards with separate chip VRM without any cooling can have issues with 100+W power draws if not using top-down cooler.
HUB testing B650 VRMs:
(Almost) all early B650 are fine.
https://www.youtube.com/watch?v=DTFUa60ozKY
Later B650:s differ:
https://www.youtube.com/watch?v=naX-DnKekCM
Latest, most tested are fine:
https://www.youtube.com/watch?v=pUUVW7wgR3s
For 7950X.
And that B650M-HDV/M.2 mentioned earlier still uses integrated power stages, even when it doesn't have VRM heatsinks.
And not separate chips (two MOSFETs and driver chip per stage).
im fairly sure all the full atx b650 can handle any am5 cpu
b650m is model specific
Yes, that am4 issue with early gens not having the easiest time is why i was looking at x870
Some do have very crappy VRMs with separate chips (much more heat) and no VRM heatsinks.
Both on full-ATX and smaller sizes.
Early boards all can do it, as AMD had mandated that all boards needed to be able to run 7950X, but that was eased later.
i think its just the primes
which should be avoided anyway
you can put 700W thru the 9950X
Issue with an usb pcie card is that i already might need an sata card.
For some reason am5 mainbaords are starved for sata ports
Because the chip only supports 2 per die.
isn't there an Intel chip that allows for more ports?
Ine of the strix 870 models has only 2 ports
So MB can have 0+4, 2+2 or 4+0.
I do want an x3d chip, intel cant compete with those
no i mean iirc there's a chip on some boards that allows for more ports
but it's intel, so they're not about to hand that over to amd
And any single-chipset AM5 board is starved for PCIe basically.
When MB makers want to do ton of M.2 slots.
why do you even need that many sata slots
Single Promontory 21:
I have 2 hhd and 2 sata ssds and 2 m.2 ssds
what capacity
8tb, 4tb, 500gb, 1tb, 1tb, 4tb
And on dual-chipset board (X670, X670E, X870E), one block of 4x PCIe 4.0 goes to the other chipset.
yea i have 3x M.2, 2x spinning rust, and could do 2x sata ssd if i wanted to
make a NAS ๐
I will when i have more money.
But thats not happening this or next year
i want an M.2 NAS where the drives are just out in the open and sitting like they're in a toaster
Not going to happen with M.2.
But that is the whole point of E1.S and E1.L
um you already have the storage
oh i know, i just want the chaos of it
thats 80% of the cost
And E3.S and E3.L now too.
Nah thats not worth it
If im gonna do a nas, im gonna do it right with redundancy and all
But that was the point in suggesting going with X670(E) board, as those have much more lanes from the double chipset, so more IO.
If you don't need USB4 and/or Thunderbolt 3.
Because X870 is just B650E+USB4 and X870E is just X670E+USB4.
Where that USB4 controller usually uses 4 lanes of PCIe either from CPU or chipset.
Ill look at them, i dont need usb4. I just need a good number with without having to daisy chain through a hub
And B850 is basically B650E.
Tiny difference in the PCIe 5.0 requirements between those on MB level.
Ah, no B850 is identical to B650 non-E.
The original version, before AMD allowed completely dropping PCIe 5.0 support in cheaper MBs later.
As if you have any PCIe 5.0, then the MB needs to use more expensive PCB material.
So what ur saying is that amd cooked mainboard prices
Part of the reason why there weren't any "cheap" AM5 boards originally was that added expense from needing to have at least one PCIe 5.0 slot, normal or M.2.
That was later dropped.
And the price of the better materials has also come down as usage has gone up.
Now that they aren't just used on server market etc.
But currently MB makers are in competition on who can have the most M.2 slots.
And both SATA and more importantly normal PCIe slots suffer because of that.
Even when it is trivial to convert normal PCIe to M.2, but hard to convert M.2 to PCIe.
Take an old PC, put some of your drives in there and turn it into a NAS
Damn, TIL B650 LiveMixer only has two SATA ports
BUT
And the accompanying AM5 CPU image:
Oh btw, that reminds me - do y'all know this spreadsheet?
https://docs.google.com/spreadsheets/u/1/d/1NQHkDEcgDPm34Mns3C93K6SJoBnua-x9O-y_6hv8sPs/htmlview#
I love the work that Thrip does for this community
i have 3x VR sensors, VR headset, mouse, keyboard, headset, xbox dongle, flash drive, and external drive sucking up all my usb ports
How many of those would still work fine with a hub?
Everything but the VR-headset and possibly external drives.
Well, the externals would work too, but might affect speed.
Hm, yeah. Depends on hub and whether the drives are USB powered
idk about latency on controller, mouse n keeb if you were a high level comp gamer
And in that case, easy to get internal to external USB plate for USB2, as often MBs have three internal USB2.0 headers today.
Each which is two ports.
rn if i plug it all in, it's fine. the second i plug in my external M.2, it gets a bit jumpy
how can i stop this bullshit from running in the background?
Why are you trying to delete that file?
Or did you delete that file and leave the program that uses it on autostartup?
trying to remove synpase completly from my pc
Gonna have a look at that later. Nice
theres also one for am4
Also a couple more for HEDT and LGA1851
Tho the AM4 one isn't made by Thrip iirc
why did the guy have to make the collums too small for the text.
thats actuallz painfull to look at
it should be vertical, idk why it isn't
you can just join the Discord server he linked and ask him
there's an entire channel specifically for motherboards
just to make sure, here's the correct link:
https://docs.google.com/spreadsheets/d/1NQHkDEcgDPm34Mns3C93K6SJoBnua-x9O-y_6hv8sPs/edit?gid=2064683589#gid=2064683589
@main locust can u please add dank memer bot? and make a bot cmds channel for it?
#off-topic-media already exists?
Thought you might enjoy my latest... screenshot.
mmmmmh pixels ๐คค
no thanks.
Thinking about these as secondary machines for light use. In case Copilotplus or the underlying NPU ever become useful.
YAY microcenter bought a place here
I keep hoping they'll take over the still vacant Fry's location here.
sadly our old fry's electronics was abandoned so long ago it's probably haunted if it's still standing
no some random cat is just sitting there
Here the only brick and mortar competition to Microcenter would be Best Buy.
Have you heard about subpixel art?
the term rings a bell but no idea
oh no, before that it was an Incredible Universe, it's definitely haunted
plus it's right near the Superstition Freeway
Nice hitting ~90 FPS with FSR in MS Flight Sim with the details up pretty high
Was worred about pushing the pixels on this display on that one but it works fine
i'm actually kinda disappointed that MSFS garbage fps didn't last too long
Does anyone know if itโs possible to create categories for YouTube subscriptions? Like entertainment and research being separated?
This is probably the oldest electronics in my house. It's 25 years old (maybe even more) and still working perfectly after all those years: โค๏ธ
Correction: It's 30 years old and my dad bought it while visiting New York
I still have and sometimes use an analog meter I built from a kit when I was 16 o.O
It's nice seeing something that lasts even longer than Toyota Corolla lol
*the older models that is
New Toyota Corolla can't be compared to the model from 2002
Good ole electromechanical occupations class in high school
that was my vocational 3 out of 4 years
before digital i could work on things like dishwashers, clothes washers/dryers...
Too bad I didn't ever use the HVAC stuff, forgot most all that over the years
I could still do basic stuff... but to fill or recharge I'd def need to look up some numbers... ambient temp matters
Back when I was in elementary school, we had "technical culture" classes and everyone had to learn how to wire a plug or a wall outlet. The teacher even dumped an old car engine on his desk and showed us how everything works
I highly doubt they teach those things to kids today
Probably not, unfortunately
Home economics used to be a thing too, and it needs to come back.
Learn some basic finance stuff... how to sew a button on a shirt. How to actually cook....
The teacher had a small box with a breaker and each student brought him a wire with a plug on one end and a lightbulb on the other end. If you triggered the breaker, you failed the lesson lol
Like when you graduate high school you should understand things like loans and interest for sure, clearly a LOT of people were clueless
they signed off on loans and got in over their head...
Tried to find how old my microware is, but Panasonic didn't mark it in the type plate.
Old enough to have been manufactured in UK, and the serial starts with U90, which might mean 1990 as manufacturing year.
Bought used.
You can create plasma with a grape... in a microwave
We had a 15-year old zanussi microwave until my dad accidentally damaged it by forgetting to put food inside and turning it on empty lol
Old microwaves might be useful to be from time to time... but for the transformer heh
This thing is very nice, as it is button based, in a way that actually works.
And allows you to select two different power settings and times, then start.
Yeah that damages them pretty fast
the magnetron degrades over time regardless tho
These days a new microwave might not even last long enouhg for somebody to notice but if you've owned one long enough to realize it's not as fast as it used to be
Not really a problem, as I lot of what I do is intentionally on lower settings.
Very often 1 minute on max, then 2-10 minutes on very low power.
Also don't play with microwave transformers or magnetrons (friendly PSA)
Gets nice even temperature.
The fact those waves could maybe cook food was discovered by them... kinda cooking food
dude had a chocolate bar in his shirt pocket and was tanding in front of a magnetron and noticed the chocolate melted hahaha
Yes, getting new microwave would be prudent in many ways, but when none of the current microwaves offer same kind of ease of use and utility...
My grandmother owned one microwave ever. Had the dial timer on it haha
Which is very usable?
She never used one for much though... she was old school and cooked
It was usable during it's life there
she had it i guess maybe couple decades
Todays microwaves absolutely fail on usability, and most people don't even try to time or adjust power with them.
Can't find information online when Radioshack 22-163 was released on the market. Dad actually things he is wrong and the multimeter might even be 40 years old!
THAT microwave I could repair lol
He needs to ask his friend because he rememebers visiting New York in 1988 when he bought the radioshack multimeter
And the highest end model of that lineup could do three setting sets...
While that middle model can only do two.
I think most microwaves are around 800 wats and cooking recipes assume everyone is using 800 wats
I have one of those
I can also lower the wattage on my microwave if needed
And usually people end up with hot outside and cold inside.
Or dried up outside.
Usually it's stashed in a box but I break it out sometimes for the memories
But point is, when they cannot even set time, and just set something and manually stop, the UI has completely failed.
Sooo yeah... 37 years old. Guess I did a good job with the solder joints ๐
I have samsung microwave atm and I like the rotary dial since it's intuitive and easy to use
You can change the timer with it and power if needed
Oh my grandmother hated our fancy microwaves when she'd come out here
well how do i just add 2 minutes i just want to add 2 minutes!
There was literally a button dedicated for "add 1 minute"...
On samsung microwave, there is a default time and you can increase/decrease it with a rotary dial
Much easier than pressing buttons
Looks like I'll be finishing the mailbox tomorrow, def gonna be raining today
Need to find that old sony radio stashed somewhere. It's probably even older than radioshack multimeter lol
Too bad EU is shutting down FM so you won't be able to use it anymore
Even sat radio is on borrowed time at this point
Not even sure when the last time I listened to FM was
Most of EU has already implemented DAB (digital audio broadcasting) and my parents own a car with DAB receiver:
Dark blue = regular DAB service
Finland in turn had it in testing, then shut down and continued to use FM.
Oh interesting
Most us recipies/products assume 1100 or 1200 watta
Probably in USA. I think 800 watts is standard in Europe
Microwave are pretty cool
This afternoons agenda is fixing a minisplit thatโs freezes over when and of the other two are in use
Just went to check that the concrete anchor for the new mailbox post was all set properly
that post will snap before it gets knocked outta the ground lol
Thrilling wifi speed
whoa slow down there
My microwave is 1700w input. 1000w output
What country?
In the US 1700 watts in would mean you're almost at the breakers limit
Well I take that back you can have a 20 amp circtuit as well
That 120v can be a problem at times...
damned EA app sitting there using 3GB of memory for WHY
Kade is stateside iirc
had a special edition one of those ๐คฃ great car
definitely too short for me now, but a very fun car
With W10 reaching end of life and my refusal to use W11, can i just get an anti-virus software for W10 for a few years and call it good?
Not recommended, although possible
Your os will still have unpatched security vulnerabilities
I just dont wanna deal with the increased spyware, changed look and other bullshit of w11. Most of which couldโve been an update to w10
Resistance is futile! You will be assimilated.
That borg cube represents future Windows ๐
Seriously though you will. I get where you are coming from but ultimately you'll have to. I say rip the bandaid off and get it over with heh
There are also some scripts you can run that disable all of the telemetry among other things.
I didn't bother... but it's a thing so
No.
mmmk ๐
It's just a matter of time though.
I know people that should still be on Windows 95 still...
I can assure you they are not lol
Ditto for the 98 ones... XP...
"I'm NEVER updating"
Switch to linux
Staying on 10 is really quite a bad call long term
If you hate 11 that much, linux is a far, far better choice than 10
But then i may not be able to use all of my programs and games
Eventually you won't be able to use them on 10 either...
And besides, compatibility layers are quite good and linux support is growing
You kinda want everything here and you're unwilling to compromise
You're not going to get an answer that you're perfectly happy with
why the fuck does archlinux team not want to land 6.14.8
Until you wanna play multiplayer with kernel anticheats 
Yeah for the like 5 games that that includes
heh I just love that age ole fight!
I'm so tired of it
PUBG, Escape from Tarkov, Tom Clancy's Rainbow Six Siege, Call of Duty, Destiny 2, League of Legends, Valorant, fortnite, rocket league
its actually closer to 10!
eac is not kernel-level\
Still doesent work lol
also eac does require a kernel level driver
use a virtural box with windows 11 unactivated
microsoft cant spy on you enough to be sellable
is easy to set up
for games:
use linux wine
its werid
but it works
so i just did my first "big clean" on the new pc & made some discoveries
the intake filters work really well
bottom one is easy to clean, side panel one is absolute fucking garbage, so much so that i may trim it down & attach it to the fan bracket insert thingy on it
i found a mobo screw on top of the 140mm rear fan, wedged & stuck itself in the back mesh cutout
then i discovered 5 mobo screws were never installed ๐คฃ
so, thankfully i have a pc screw kit that came in handy for that
As much as soldered memory makes me vomit, I did pull the trigger on 2 of these.
As lightly as they'll be used on a desk as secondary machines, they should last as long as these Sandy/Haswell machines I have kicking about.
Me with 7
ngl, I'm happy I switched to 10 and fwiw, 11 ain't that bad either
And anyone who truly cares about telemetry/spyware should migrate to Linux anyway - and if that brings downsides with it, you gotta weigh your options
And basically any "spying" in Win11 is also in Win10.
Except that Recall, which in turn needs CoPilot+ computer, so limited to some newest laptops currently.
EAC depends on the game devs.
It generally supports Linux but has to be enabled by the developer of the game.
For Fortnite and Destiny 2 for example, Linux-support of EAC is disabled
Which also means that unlike Vanguard, EAC isn't a guaranteed nope for Linux
I'm looking forward to tinkering with Copilot+ once I get those Costco laptops. Safely on secondary machines.
There are a few usecases that just aren't possible to do with Linux and many more that require extra steps.
For example, me doing LoL esports broadcasts can't be done on Linux since LoL can't run in Linux because of Vanguard.
On the other hand, image editing with Photoshop can't be done on Linux due to Creative Cloud not working (or at least not working properly), but you can use Gimp as an alternative - it's just not as good of a tool as Photoshop
Rocket League actually works in Linux, natively even - however, none of the other titles do
protondb.com/app/252950
I might take the plungle this year
Every year im getting closer to doing it.
I use linux occasionaly on my laptop, so desktop use isnt that far off for me
What can be the reason for wifi router to not display it's web ui upon trying to reach it in local network?
Just endlessly loading in browser until connection timeout
Please don't paste large blocks of useless LLM generated answers.
When you didn't even seem to read through the stuff yourself and remove things that are in no way related to the question.
But internet works fine on same computer?
Or nothing works?
Yeah, internet works
Trying to connect by IP, or by DNS name?
Then most likely local DNS cache has bad data.
If Windows, use ipconfig /flushdns and try again.
Or just look what the default gateway is, and try to connect by IP to that.
Hmm
I ran resolvetls' command but get basically same result
It refuses connection
Oh wait
I am dumb
It doesn't refuse, but same behavior: endless loading
Basically, I want to change restart time, which is daily, so the request to restart probably won't change anything ๐
Then the http-server part of the router might be down, and next thing to try is that power reset of the router.
Yeah, i started reaching router right after it re-powered ๐
Takes time for the right parts to get up.
But this is how it behvaes for the last... 30 minutes?
Ok, then too long.
Got a vpn enabled?
Disabled. Never works with vpn on (except one time? I think, somehow...)
With vpn on it doesn't connect at all
So, thier support's chat link creates new tab about:blank
Nice
Very likely script link that doesn't work with ctrl+click?
On vpns.
I used nord nvp for some time because its cheap.
Switched to mullvad last month.
Night and day difference in responsiveness and speed
So much better in every way
check HTTP and HTTPS
check port
check IP
So, after not finding my model, got a troubleshooting article that basically says: power on, check dhcp, disable proxy, if nothing works - reset router
Navigating from mobile
Basically that for every router.
I wish I was sure that i got a decently recent settings dump
What was this in reference to?
Screenful of bad ChatGPT "technical advice"
That then got deleted.
posted by KEKw in relation to their problem I presume
No, newbie
Ah
Someone with still the new to the server tag
Wouldn't have made sense for it to be KEKw anyway
And i was about to ask chatgpt a random unrelated question
Bees
If I remember correctly, they don't fly at night, and... During eclipse
Ta least I think i remember a line about this in Destin's video
But they do fly in rooms during day,
And navigate their beehives (which I guess is more walking? No idea if there's typically space or need to fly).
But what is the difference? I want to dig a bit into that. Like, would that means there's some way to see a a random bee in a room without danger of it flying?
Also had some random large wasp visit balcony couple days ago. It really liked one of the dark corners.
Chatgpt can do well with trouble shooting, but it needs to be the person with the issue, not someone unrelated
Ehhhhhhh, ChatGPT is really good at recreating existing patterns.
If that fixes your issue, good - but that probably means it was a generic issue.
Anything more in-depth can actually be worse off with LLM troubleshooting because the model might be too unfamiliar with the issue and can't judge whether the advice it's giving could end up being harmful
As an example: search for "hallucinate" in this channel
Have recently loaded a 950 page programming manual for a controller into ChatGPT.. It gave me an answer quicker than Adobe Acrobat Reader would load the file..
But I wouldnt ask ChatGPT to produce a piece of software function, without prior knowledge to how that function should somewhat look like - because it will just give you a whatever answer, as to what Fireworker2000 say - in an existing pattern which can be totally unrelated to what youre searching for..
And oh boy do I see a lot of colleagues just blindly trusting the programming examples that ChatGPT prompts them, without checking up if the function would work in real life..
Take that Windows fanboys, Mac has NEXT LEVEL gaming now.
But in general, if someone wants LLM "answer", they can do the prompt themselves and don't need someone else to paste crap here.
a lot of issues are generic
Can't wait for security researchers to find backdoors in LLM generated code that nobody bothered to check and nobody is qualified to analyze
and even if it does not have the solution, it sometimes points you in a direction to look into
I know, that's why first level support exists
the new models that search the web can also be a bit more specific, wich is nice
But can still hallucinate even then.
Where the synopsis they put out and what the referenced sites actually contain are two completely different things.
And yet big corps fires the employees who have a slight chance to actually find these backdoors.. :D
But if your AI generated result tells you to buy some specific network card cuz it has a specific spec, then you should not simply by that thing and first triple check if it does actually have the spec and if that spec will actually help with fixing your issue
I'll just be over here watching the world burn ๐
Yeah, the web where a lot of low level content is becoming more and more AI generated
sure, but if it works, it works.
if it dont, i go look for myself
And due to it being AI generated, it looks a lot more in-depth and better researched than the boilerplate releasenote article-spam
My go-to example for this:
I asked Copilot which ITIL category would be the best fit for providing technological infrastructure for stage production
It told me "event management" which is an obviously wrong answer - event management in ITIL means system events, checking logs and monitoring infrastructure.
It did a great job at recognizing a pattern, it just failed because the pattern did not apply
But I made my uBlock hide those distracting new tags \o/
Need to later make it more robust and generic.
What new tags?
Laptop doesn't boot
Looks like yet another update tried to put something into/boot and that notification about it having no space left (on /boot) was more important...
So i went to uefi menu to boot from usb, and was surprised to find tpm2.0 setting. Is that thing this old?
tpm is nothing recent. just the requirement
Those after some peoples names, that they have been able to select from stuff servers they are on provide.
TPM 2.0 first spec version was release in 2014.
saw systems with intel 6th or 7th gen that had a TPM 2.0
Yeah, I was just trying to find the first CPU generations from Intel and AMD that had the firmware-TPM 2.0.
And was used in separate chips from 2016 or so onwards in actual computers.
So for Intel, Skylake and newer have inbuilt TPM 2.0 (100-series chipsets and newer?)
And for AMD, Ryzen (300-series chipsets and newer?), so basically any AM4.
Not completely sure if CPU or chipset that determines.
So 6th gen Intel and newer have fTPM 2.0.
Named as "clanTag" in the HTML/CSS.
IIRC you were the person in this channel who had one: POOR.
Or someone else, not going to disable to check.
Yeah, I have one
I just got confused cuz either you're using discord in a browser or there's ublock for the discord app
Both kinda weird
Smart
The app is just HTML in Chromium, so why not just run it in browser 
F
It doesn't boot from stick
Ubuntu 22, which is ae version installed there
f-ing discord doesn't send photos...
basically, all it display on a balck screen is a couple rows of pixels with only a few of them litten up to white, with some big and small gaps
I think it played some sort of animation?
idk, it seems like a bad state
another day another $1000 hole in my pocket
Dats some expensive pants you got then...
i see it more as a worthless currency
In Taiwan I sat down with ASRock on the Friday afternoon and questioned their motherboard team about the Ryzen 9000 series CPUs dying on ASRock motherboards, and they said they have just fixed this issue coming in Computex, and are rolling out bios updates.... Though who is to blame?
Sponsor : VIP-SCDkey 30% Code : BFTYC
Windows 11 Pro (21$) : ...
tl;dr asrock fuckup
anyone with ryzen 9000 update yo' bios
more motherboard bs
So too aggresive default PBO settings if you enable PBO.
If that is true.
But lot of MBs have set the default PBO on settings way up.
So I somewhat doubt that result.
except msmg is currently down
Microsoft does not want to to remove the bullshit
And then the person complains how CPUs should be stable and not OCd too much.
But forgetting he had personally enabled PBO...
(and no ntlite is not doing the same work)
Not aware that PBO is OC?
Yes, it was thought to be safe OC for most part.
But if it was always safe, it would be the default.
PBO is treated like XMP and EXPO so you get the same results of misinformation and people then in turn dont understand that its not a unlock to the full performance but rather an OC
Up to now it has been the temperature limits and voltage limits protecting the CPUs with PBO enabled.
As most higher end MBs have just set both current limits and the power limit through the roof.
1000-4096W power limit, and don't remember how high current limits.
wasnt it 1000A current limit and 4096W power limit ?
Depended on MB maker if the power limit was "just" 1000 or 1024W, or the absolute max possible to be set.
it still is
like i checked and even tho im on intel defaults anything not specified is still unlimited
so no wonder it runs like garbage
and cpuโs are still degrading
i see no reason why the exact same board makers wouldnโt be doing the exact same thing for AM5
Fyi they also get fried on non asrock mobos
9800x3d is just very power hungry
Saying that is a big disingenuous
it just happens that the issue shows more on asrock
Given Intel's track record
We are talking about amd
Also that ASRock mainboards have higher sale numbers
Exactly
Yeah
9800X3D is still quite tame when compared to Intel's top level flagship (14900KS)
It definetly is not "asrock bad womp womp now go buy this 300$ asus mobo" or whatever people say on reddit
it's power hungry compared to the rest of AM5
ahem, Asus OCP failure, ahem
compared to the 7800x3d, double the power, sometimes even more
(tho that only applied after the CPU was already dead)
Still very tame
160W isn't that crazy, just smth to keep in mind
It is crazy
Which tbf, can trip you up if you don't
Wait until you see 300+W 9950X or 400+W 14900ks
thats how much the 9900x consumes in most occasions
no wonder theyre frying themselves
If your PC can't handle a 100W CPU, how did you afford the 9800x3d in the first place?
What? they were talking about them literally frying themselves
I know
How is 100W enough for a CPU to fry itself
On any decently built MB nonetheless
looks more like a specific texture in the game that's being buggy, not your gpu
more a heat issue
the 5800x3d is fairly low power
still gets hot due to the cache
inverting the cores with the cache really does make a big difference
Not really
amd really pushed the "you shouldn't try to OC this" with the 5800X3D
Cpus can handle even 100C
not without downclocking
mine reaches up to 90c while not using that much power
its realy just the cache being harder to keep cool
Thats by design on amd
having the cache being an insulating layer on the 5800X3D doesn't help
Oh good. A new GPU isn't in my budget these days.
gpu buying these days is like buying a ferrari. if you have to ask the price, you can't afford it
All I know is that the xx70 is the series I aim for
Good balance of power and budget, because I'll never be able to afford a 90
gotta love how midrange went from $400 to $1200
I paid $1100 for mine but that's because it was the peak of the supply chain shortage and I already had every other component. So I was like "screw it, I'm not waiting 2 months for the last component's price to normalize."
i paid $600 for my 6600XT, i know the pain
i paid 140โฌ for my 2070 super
I don't remember how much I paid for my 6750XT
Grand Total: $343.03
Would have liked a bit more horsepower but I was on a tight budget
They're not at all
200-300w has been very common for a long time ๐๐ป
Sandy bridge through skylake quad cores did around 200w at OC. The later gens with more cores were specced at 253w, and almost every motherboard laughed at that and ran at 300-400w out of the box "in spec" since Intel considered that part of the spec as a suggestion rather than a law.
The reason for this (and i wouldn't say double power, but significantly elevated) is because there was a hard voltage cap on the 7800x3d. The top quarter of the boost curve was chopped off because the voltages used to run it would damage the vcache CCD.
That is no longer the case with the new design on the 9800x3d.
Zen5 x3d is actually more efficient at iso-voltage, it just has the option to boost higher using more voltage - and regardless, a single CCD doesn't use that much power. The voltage cap just made zen3 and 4 x3d CCD's use laptop CPU levels of power at most.
And 9800X3D is specced at 120W TDP, 162W PPT at stock.
And the issue was that if you enable PBO, which loosens those limits, the CPUs start to break, more so on AsRock MBs which raised the current limits more than other MB makers.
So the issue is mainly that with 9000-series Ryzens, PBO, which up to now has been pretty safe, is breaking CPUs on all MB makers MBs.
But with much higher amounts on AsRock MBs which had higher current limits (TDC, EDC) with PBO on than other MB makers.
When using AIO or water cooling.
Basically: Unlike previous AMD CPUs with PBO support, the 9000-series without power or current limits, with good enough cooling, can boost so much that it damages itself.
AMD then gave bad information to MB makers on what is the max allowed current with PBO on.
And AsRock set their MBs with good enough VRMs to use that absolute max current AMD said when user enabled PBO and didn't do manual tuning.
Other MB makers set the current limits lower.
And with the BIOS/UEFI update, AsRock set their MBs to only use the same max currents as other MB makers were using.
But that might be too high too, not just to same extent.
Previously safeish OC method (Enable PBO) isn't safe on 9000-series AMD CPUs.
The current limiters shouldn't be the issue. Almost all of the time, especially on the 9800x3d, the CPU sits under the silicon health limit rather than current
and that health limit is supposed to be tuned to prevent exactly those kinds of issues despite arbitrary current limits
Yeah, sorry, forgot to include that that was what AsRock said.
And I originally said that the explanation seems to be off.
But might also be that people are turning PBO on, and setting that FIT Scalar to 10x like lot of bad guides say.
you can override the health limit with scalar, yeah, but i recommend never to do that
unless you're going for world records and dont mind cpu degradation/death
but for a daily, not worth at all
But basically third of the video was the maker ranting how MBs should default to stable, how he wants stability and longevity for his work system.
And then he never denies turning PBO on...
most of the issues with zen in general have been with EXPO, which is automatic RAM overclocking. Boards set their own voltages (like SOC overvolts) for that, and those have been an issue
zen5 x3d's are quite vulnerable to instant death if vcore goes too high though. My friends have killed a few with manual overclocks at voltages which are surprisingly close to the daily operation voltages.
I suspect LLC voltage spikes are related. LLC adds voltage over the top of the CPU VID to try to counteract vdroop (voltage drop under higher current loads) - but when the current drops (such as a workload ending or stalling), that vdroop no longer happens and as a result the extra LLC voltage is added on top of the VID.
One of my friends was able to kill even a zen4 x3d by setting LLC to maximum and cycling a workload while otherwise fully within spec.
Yeah, AMD by default uses almost no Vdroop.
And just uses clock stretching (pause CPU while voltage too low) to deal with underruns.
With LLC and scalar and SOC at spec, i have not seen any reliable indication of problems - but those asrock boards are setting off alarm bells for sure. They may be doing something differently which is even within spec but breaking stuff. More likely i think is them setting something with e.g. LLC, Scalar or SOC voltage which is not fully in spec, and causing elevated vcore or SOC outside of tolerances. Both can kill quickly with the raphael/graniteridge SOC and with vcache CCD's.
I have my vcache CCD around 5.7ghz game clocks (~5.9ghz max) fully stable atm, with quite safe volts. No +voltage on the CCD, 1x scalar, and SOC is only +25mv over spec just for margin (EXPO is usually +100-200mv).
Also, 7000/9000 is first time AFAIK that AMD has VCore also inside the IOD.
And the 7000 VSOC deaths seem to have been from too large difference between VCore and VSOC inside the IOD.
I'm pretty sure that was just straight up too high SOC. Spec SOC on those parts was 1.05v, and motherboards from every vendor pumped it up to 1.4 - 1.5v if you tried to overclock memory (including EXPO).
Even before the deaths started i was manually intervening with that and making a scene because it seemed like a ridiculous and dangerous overvolt.
As hypothesis.
The failure rate was actually the same on x3d and non x3d, the x3d cpu's were just far more popular. If they're 80% of CPU's, they're 80% of failures.
In many cases it was IGP that started dying first.
Yeah, i disable and recommend disabling that too with high overvolts
if you're not using, it can be disabled. If you aren't using but need to use it after your graphics card dies or something, clear cmos turns it back on.
But point was, on AM4, that IGP uses VSOC.
On AM5, it uses VCore, as separate island inside IOD, rest of which works on VSOC.
As far as i know it's stepped down from vSOC
And VCore used to idle at around 0.5V, now raised to 0.7V.
And VSOC default is that 1.00 - 1.05V, and MBs were pumping that to 1.35V, and now pump it to 1.25V with 6000MT/s RAM set.
My board (x670e Carbon) tried to set the vSOC to 1.45v every time i entered the BIOS, even if i hadn't changed anything.
Not anymore.
That was AM4 IODs.
On the current AM5 IODs, it is straight up VCore, no difference from cores.
No separate regulation.
VCore is whatever is higher request from the cores and the IGP.
Do you have a source on that? Because it conflicts with the best information that i have
e.g.
So to protect my x3d, go into BIOS and check/lower vsoc?
Yes, that was wrong.
GN:s information was that it is from VCore.
And my own HWiNFO numbers support that.
vSOC will be 1.1v or less unless you have either raised it yourself, or turned on memory overclocks
1.4 can kill quickly. 1.3 will degrade/kill slowly.
The IGP voltage and core voltages go at lockstep.
interesting and weird if true ๐
Just enable expo, pbo should be off iirc
AMD was limited on IGP performance because of VSOC regulators on AM4.
And those went down in power on AM5.
EXPO is automatic memory overclocking, all motherboards tune vSOC in response to that - so if you haven't set it, the board has chosen an overvolt (and probably wildly excessively so)
As they don't need to account for IGP anymore.
EXPO is automatic memory overclocking, all motherboards tune vSOC and VDDP in response to that - so if you haven't set it, the board has chosen an overvolt (and probably wildly excessively so, although less excessively than before the widespread failures)
XMP/EXPO is historically much more dangerous than PBO for that reason.
The only CPU i've had die was due to an Asus Hero 8 overvolting some of the memory controller voltages too far when enabling a standard XMP; it killed them almost instantly, and the CPU never detected memory again with any memory or board. My CPU was not alone then, and it happened with even higher profile on the AM5 SOC failures with EXPO overvolts.
A lot of people don't understand that those failures happened because of overclocks with voltages massively raised over specs, but they all did. No CPU's failed at spec.
So with AM5, with current 7000 and 9000 IOD, that VDDCR_GFX is actually Bypass type from VDDCR, not DLDO from VDDCR_SOC.
No, cannot be sure.
Now it isn't at the exact same in HWiNFO.
But GN had gotten the information that it would be from VCore, not VSOC.
In their 7000-series failure videos.
I am not certain either way, but will keep it in mind! I remember there were some weird vcore pins.
I wonder if AMD enabled the existing DLDO in the silicon.
As those exists for each core in Zen4, and are used in the Epyc CPUs instead of bypass mode.
Might be the same for the IOD and they enabled it with AGESA update.
Or I remember the numbers wrong.
I believe that AMD actually uses the LDO's for both curve optimiser and voltage stabilisation (they call for a VID higher than the core wants, then step it down by a variable amount - that buys voltage headroom for faster boosting and reduces ripple)
but can't back that up with great sources ๐
For the CO stuff, you can make cores pull less current and even become unstable with -CO even though the Vcore VID and SVI3 voltage remains the same, in the same allcore workload.
But my IGP was now at 0.70-0.72V while driving two monitors.
IIRC it used to follow the VCore exactly, but cannot confirm anymore.
Not going to update to older AGESA to check.
I just don't use it. Don't need it, and the overclocking failures that we've seen localised happened at the iGPU. If you have it on, probably more caution with high overvolts is warranted
Checked back then when there was that mismatch between SkatterBencher pic you linked and GN.
And my first 7800X3D failed, with IGP going first, from just running 15 minutes on that too high VSOC when I set RAM to 6000, booted to Windows and checked VSOC, rebooted and set it much lower value.
Few months after.
Got it RMAd before it died, but IGP on the max HT frequency that worked was significantly less than without it, and that was going down steadily.
And the system wasn't in use yet, so not much lost.
Yeah i had to be extremely careful to never apply too much SOC. Like i said with the MSI boards, they would set SOC up to 1.45v if you literally just entered the BIOS. They wouldn't warn you on the save-and-exit screen. I was just being paranoid as hell and found that they were screwing everybody like that incidentally.
Nerdz
๐ค and proud of it.
These days being a nerd is cool...
When I was in school it wasn't so good...
Most of the metalheads were nerds of some flavor but we didn't get messed with because we would fight back.
an old receipt, note the dates 4 weeks before the first journalist reports and AMD statement ๐
i set a hard limit at 1.28 and wasn't even fully comfortable there when a lot of peeps were running 1.4 - 1.5, and MSI set it to 1.45 every time i entered the BIOS because my memory frequency was set to a non-spec multiplier. Paranoia and experience pays off when it comes to overvolts.
^Also 5 months after the platform came out - but basically nobody bought it because it was horrible value, DDR5 was expensive and vcache parts didn't exist yet, so there were few CPU's and therefore few failures
new platforms and parts are the most dangerous times to be overvolting things
that's one plus of AMD reusing their IOD from zen2 to 3, and 4 to 5 - we already know its limits and it's proven in battle.
now i wanna see what 2V would do
Also to this: "Scales" != safe.
Just means it still unlocks more potential performance.
Which baenz doesn't seem to be able to differentiate.
You can easily scale short-term while already degrading fast.
Just not instant death level yet.
But if you run the CPU at 1.3V SOC, theres no way to tell that was the cause if it dies under warranty time? (3 years)
For RMA to pass?
Not really no, but that's RMA fraud which is a crime
He does, smart guy, but a bit too gung-ho about running crazy volts sometimes IMO ๐
I can live with that crime
You don't need to censor MagicZ's name, we all know their OC skills
Der8auer is jeloux of this who person, who shall not be named :) who allegidly has fried more CPU than the OC Guru himself..
Too bad I did not get to keep the broken 13900K, would have made keyfob from it
Thats why the second one is now direct die cooled 5,5GHz all core waiting for degrage to start life as keyfob
IIRC Der8 killed a 7800x3d immediately when playing with it, applying 1.5vcore with 50c+ temps
๐
now i'm tempted to somehow convert my obsolete am4 cpu into my keyfob
stuff the battery & pcb under the IHS
figure out how to do buttons
ew nvm, forgot about all the pins
just uhhh... mold it in resin? or something
i was gonna say only 5.5 and then i remembered intel defaults tanks clocks
i havent killed a cpu
YET...
im just not gonna touch anything when i get the 9800x3d.
At best ill try a slight undervolt to reduce temps.
Surely it wont fry
I can get it to run 5.6GHz all core, can't remember how much voltage needed but its then right on thermal throttle limit even with direct die cooling. 5.5 for all core (4.5 e-cores) and 6.0 for single.
+300W power
If run on stock Intel 253W power, the direct die cooler can keep it around 70ยฐC
My god - here I have played Satisfactory on 59.934 Hz - where I could have had such a better experience running it on 59.994 Hz.. Such a waste of my life and time really
Intel being weird.
certified refresh rate moment
my 60hz is superior to yours
At this point it is worth to start a new save
Mem OC is worth it. Some pretty big gains
This is "lategame" but far from the heaviest world in existance. I'd like to rerun after my upcoming megaproject ๐
(geomean is relative to standard zen4 ccd, can just ignore that)
Several of the gains are also a lot higher than that since Asus defaulted an obscure memory addressing option to the wrong setting when selecting >=6000mt/s on zen5, which was a major performance loss for OC profile but not spec profile. I just did not rerun with that fixed yet. It's also not a maxed OC, just one that was decent and guaranteed to work for A/B comparison before i knew what the CPU could do.
im geting a 6k cl30 kit to begin with
there are some kits with slightly higher clocks and still good latency, but feel like its not worth the upcharge
Ye it's not really. The main difference though is setting timings manually - EXPO does not set most timings
Unfortunately not, and as a result they only get about 40% of the perf gain that memory overclocking can achieve
Only the primaries and refresh time.
And often that refresh time is set way up in the XMP/EXPO profile.
and how would i go by getting the other stuff right?
They usually set 4-7 timings. I manually set 25.
this is some real good data
im open to do it, but i do need something to follow along
For the most part it's as simple as set a tighter value and stability test, if it crashes then don't use it. If you can't make it error, it's good.
vcache really does make things less ram dependant but they clearly still love the lower latency
okay, but im not gonna be testing 4 different values for 25 settings.
and even then, i do need reference data to see by how much i should/can increase by
Buildzoid also published a set of timings which almost always work for any hynix 16gbit RAM, and they get most of the achievable performance. Almost double the gains that enabling 6000cl30 EXPO gives.
full memory profiles @ bottom
That tREFI at max is bad idea unless RAM temp is kept under 50C or so.
i never heard of hynix, im probably gonna buy this kit
discord plz
For which you need active cooling with DDR5.
If you click "open in browser" or w/e then it should be full size lossless image
Any airflow over the RAM, but that doesn't happen with tower coolers.
ye I got it
in my experience e.g. 500 rfc, 50,000 refi is pretty fine even hotter than that, but yeah if you have an issue you can drop it to e.g. 32,000
Spec is refreshing ~10% of the time. 500 rfc with 30k REFI is refreshing only 1.7% of the time, so still a huge improvement ๐
last time I attempted memory overclocking my hair went grey
so i would be using these for reference?
hynix make ram chips, you probably have hynix already
Yeah
Yeah, but too many places just telling to set that 65k tREFI, without saying it is very temperature sensitive.
ram is silicon lottery: ultra gambling edition, some kits go way faster than others
all hynix 16m/16a in kits that don't suck (so like 6000c30 bins) are solid ๐
ok but before ordering, how do i know if what im getting is hynix
Hynix were/are the only vendor making chips that can hit cl 30 @ 6000
Also, HWiNFO can now give most of that same information.
Without need for security vulnerable driver that ZenTimings installs and never removes.
there might be nanya or something now but rare, and still good
so whatever im getting is probably hynix
with those timings, yeah
It's just really time consuming. You can run RAM testers 2 days straight, and still the next odd game or software finds a way to crash to memory unstability
There is a reason for the noctua ๐
RAM and pci-e 5.0 nvme under it
so no real point with spending any extra, i dont care about looks
Yeah not really, although unlocked PMIC can be a minor benefit it's probably not worth using for a newbie. TridentZ has bad temperatures due to missing thermal pads
i got teamgroups because they're good in that respect
but most kits are fine
At that 30k tREFI as that was stable up to 80+C PMIC temp (external heating during testing)
IIRC 45k was what I tested, but then gave headroom.
And tRFC 450 didn't work.
Need to tighten some things back up, and loosen others, trying to get Memory Context Restore to work reliably.
Most cl32 and cl38 kits are Hynix too
Cl36 are primarily Samsung, but if they have a 1.25V spec, it's also Hynix
It should work if you have power down disabled
Could also just try and find a cheap 5600cl46 jedec kit if they're cheap and guaranteed Hynix
That disabled MCR too.
Point is, the computer boots reliably.
But it often does the full 60 second training, and not the restore.
Also, any Klevv kit is guaranteed Hynix, regardless of spec
You can shorten training with some of the bios config options to around 10sec. Unfortunately retrain is required sometimes, can be triggered by stuff like every X boots or ambient temperature change etc
but it should skip training like 90% of the time
I know.
Might also just be the long time between boots (several weeks usually at minimum).
Yeah, it invalidates after a few weeks i think
copy/pasting memory training is kind of a hack
that mostly works, but not always in cases like this
But have been trying loosening various settings to see if that makes it better.
Might just disable MCR and Power Down completely.
it helps a lot if you are restarting 200 times to measure and tune curve shaper or something ๐
For the tiny bit of less latency.
Talking about RAM made me think of this video again
https://youtube.com/shorts/0so1fe7DaKM
Think I've not seen single code 15 memory training after installing the waterblocks on them. Would like to have thermal camera to actually see how the chips itself stay cool. Not just SPD hub temp from HWinfo
For RAM, the people selling the RAM kits almost always don't make the chips themselves.
So it's like asking if you want a Zotac or a Palit rtx5070, it doesn't really matter much, it's an rtx5070 GPU and board/ram spec made by nvidia.
It's like that with RAM, you just try to figure out who sells kits which have hynix's chips in them and go with that.
In DDR4 it was one of the samsung chips that was godmode for overclocking. Tight timings, strong scaling with extra voltage (8gbit B die)
In DDR5 hynix has the lead, but there isn't much difference between chips as long as you don't get Micron (they have some god awful chips)
Only the single revision of single size.
Rest were just average to bad.
Yeah, they struck gold seemingly by accident ๐
For DDR5, every Hynix revision has been better than previous, for both die sizes.
Just didn't care and went for production price cuts instead.
And AFAIK that specific Samsung B-die was very varied in quality.
The another thing about the Hynix DDR5 dies is that they are very uniform instead.
I've used now Samsung 6000CL36 2x16GB kits from DDR5 launch and and later Hynix M-die 2x48GB kits from 6000CL30 and 6400CL32, with Intel Z690 (12900 and 13900K) and AMD x870 with 9800X3D and 9950X3D and 4090/5090 GPU. In overall have to say that playing in 4K with those it's really hard to say which Ram-kit is installed.
So for the B-die, needed lot of binning to get the good OC dies from the "barely works at stock 2666" dies.
With Hynix DDR5, you can basically buy any kit with those dies and can mostly expect them to clock etc. very well.
And not need to buy very expensive prebinned kits.
It depends mostly on the framerate that you're going to be playing at, and the circumstances. E.g. for a satisfactory world with 500k objects, RAM is hugely important and you're guaranteed to be CPU/RAM bound regardless of resolution
Yeah sure. Satisfactory would be fine at 60 fps but when you get used to 144 or 120 in shooters it feels bad
Hard disagree on the "fine at 60" part, but ultimately it's a matter of opinion ๐
going from framegen 80>160 up to 100>400 in worst case with zen5 and 5000 series was an insane improvement for me personally
dropping to 80 base, framegen artifacts are starting to show and the latency is getting bad too. Turning framegen off is much worse due to eye tracking blur and stroboscopic artifacts
For me around 90fps is enough in VR with Pimax K8X and with monitors too it's about there where it feels smooth enough
im running the game caped at 100, above is too much fluctiation.
and ever since i went all in on nuclear, the frames dip to shit anyway.
9800x3d should go a long way towards making it more stable
i find a very solid 120-240 to be pretty good, but for VR i would want better
part of why i didn't even try VR yet. If i can't play cyberpunk because 100fps is making me motionsick on a monitor, some 90-120hz headset isn't gonna have a chance
im waiting for vr to mature more (totaly not that i couldnt put money towards it rn)
For me VR is just uncomfortable, putting on/off the sweaty headset when going to WC etc. But really like sim racing with them with the depth immersion for braking points etc
Some other simulators would hugely benefit it from too
not sure if id like satisfactory in vr.
the constant flying around would probably not feel too graet
my upgrade will be just in time
so this hsould be fine, right? or under what name is vSoC in HWInfo?
It's not too bad, but lower is better
as i have EXPO enabled thats in your stated range
I know - the agonizing part of this video is that he doesn't.
And he's got a gigantic audience listening. Listening to him call a vengeance 5600cl36 kit "high quality"
Also, wasn't Hynix ddr4 also pretty good?