off-topic-tech

[silk hemlock]

I appreciate the MSI only has 1yr warranty compared to the 2yr on the gigabyte

not that that matters much in Australia

[twin dew]

That Katana 15 has the CPU of the G6 and the GPU of the G6X.
With less height (2.5cm), and less heatsink fin area.
So I would expect that the Katana has significantly worse cooling at same sound level as the Gigabytes.

But at the gaming laptop level of sound, the sound quality matters much.

[silk hemlock]

oh I thought they all had the same gpu

[twin dew]

Same GPU, at different power limits.

Which then directly cause different maximum performance.

[silk hemlock]

doesn't worse cooling tend to affect lifespan though?

[twin dew]

Laptops mostly run everything at high temperatures and then modulate the cooling, as the noise is what matters more for the comfort of the user.

[silk hemlock]

I suppose headphones could mitigate the noise fairly well

[twin dew]

Also additional point I just noticed.
The G6X has 180W AC adapter, MSI Katana has 200W, MSI Thins have 120W.
That sets the long-term sustained max power the CPU and GPU combined can pull.

In short term if going over, the laptop will use battery to fill in.

G6 has 150W

[silk hemlock]

huh so someone using the thin might start draining the battery even while plugged in? wild

[twin dew]

Any of them.
75W CPU and 115W CPU on the G6 for example.

So total of 190W + VRM losses, + other components, with 150W PSU.

260W for the G6X with 180W PSU for the same before extras.

Only 210W with 200W PSU for the MSI Katana 15.

[silk hemlock]

I'd never considered laptops losing power while plugged in

[twin dew]

Sized with expectation that only one or the other is used at same time, often same for cooling.

[silk hemlock]

huh coool, well thanks for the help xD I thikn I have some good options to preset to her

[twin dew]

So same level overall for all on build quality, keyboard, display (but 15" 16:9 vs 16" 16:10) etc.
And then performance & cooling differences.

With the performance of the Slim 15 MSIs being pretty nerfed for the GPU side, with the max power cut from 105W to just 45W.
With Katana 15 and G6X with the full 105W 4060, G6 with 75W 4060 and the Slims with 45W 4060.

CPUs between the Katana 15, G6 and G6X are very similar, with higher power limit on the G6X with 4 extra E-core (6+8 vs. 6+4 on the rest), same max boost but the G6X might benefit from the higher power limits it has for the CPU.
But then the AC adapter limits also come in if stressing both CPU and GPU at same time.

And without hearing them, I would expect that the cooling performance would go G6X > G6 > Katana 15, at same heat load and sound level.

But that all are going to be annoying whiners at load, like almost all laptops.

[edgy hazel]

Man just buy a laptop. Why we gotta keep having long ass 2 person conversations about laptops.

[twin dew]

Because laptop cooling works by spinning fans at insane speed to pass enough air through too small heatsink finstacks.

[edgy hazel]

https://www.bleepingcomputer.com/news/security/telegram-zero-day-allowed-sending-malicious-android-apks-as-videos/
Lol

[twin dew]

So you get both lot of flow-noise and then the fan itself.
With the Gigabytes having much more space for air to both get to the fan intake, and for getting out via the heatsinks.

Gigabyte G6 bottom as first, MSI Katana 15 bottom as second:

[silk hemlock]

that's generally how talk about tech goes

[twin dew]

Because laptop makers like to do laptop selection as hard as possible, and there isn't any one source that tries to even review one model per generation per chassis...

And when you have two laptops with identical official specs, one can be great and the other absolutely horrible.

[silk hemlock]

one of the many reasons I hate them

[twin dew]

Like the MSI Slim 15 4060 (Mobile) @ 45W vs 4060 (Mobile) @ 105W, second is about 60% faster for the raw GPU calculating performance.

And not going into the hidden display, cooling, keyboard and touchpad quality etc. things.

And then if that specific model has bad display hinges that will break the chassis in few years etc. that just cannot be known when the model is being sold.

So many laptops where I have had to epoxy the hinge screw locations back onto the chassis on or similar as the manufacturer has used too little plastic for the mounting location surroundings.

[willow pike]

https://arstechnica.com/gadgets/2024/07/intel-has-finally-tracked-down-the-problem-making-13th-and-14th-gen-cpus-crash/

But if you're already noticing stability problems, Tom's Hardware reports that "the bug causes irreversible degradation of the impacted processors" and that the fix will not be able to reverse the damage that has already happened.

[twin dew]

And so basically Intel should do easy RMA replacement for ALL 13700, 13900, 14700 and 14900 CPUs in use.
But we shall see.

Because even the ones that don't show symptoms right now are already damaged.

But couldn't use iwd instead of wpa_supplicant for the Pi 4, as the iwd just didn't switch from 2.4GHz to 5GHz. 😭

[grim panther]

RMA? lol. You should be happy if they change the microcode, so the bug will not show up because your CPU is crippled by 20% performance. They did the same with meltdown and spectre.

[willow pike]

people are gonna be posting "hey why is my game crashing" for years to come

[latent crest]

um it cant be that bad i mean my game did something much worse to me back on my old laptop

here try and figure out what caused this

[night girder]

#off-topic-tech message

I think they will.

Intel is delivering a microcode patch which addresses the root cause of exposure to elevated voltages. We are continuing validation to ensure that scenarios of instability reported to Intel regarding its Core 13th/14th Gen desktop processors are addressed. Intel is currently targeting mid-August for patch release to partners following full validation.

[edgy hazel]

Iirc gallium will stain copper but not eat it as it does aluminium

At least the copper plate on my old AIO was stained to shit when I used liquid metal

And I couldn't get it off

[latent crest]

yes gallium eats other metals and or makes them stupid brittle

[edgy hazel]

no

if it would, gallium based liquid metal would be unusable

[latent crest]

if i remember right its aluminum for the most part it will eat through. any way you put it though it dangors

[maiden coyote]

love the new wallpaper, what theme is that?

[latent crest]

???

[maiden coyote]

having a IHS with liquid metal on the silicon die either isn't possible, or cost too much so now you have it between the chip and heatsink with a nice little foam barrier to keep it contained.

[night girder]

tripping on acid.

[edgy hazel]

very cool. Hope you enjoy it

yes

[twin dew]

Ok, seems that the firmware files I have for the Pi4:s WLAN chip from base Debian are more limited than the firmware files in the Raspberry Pi OS.

Need to work on that tomorrow...

[night girder]

Enable child lock 😉

don't give em sudo 😛 or admin account

[charred pewter]

sometimes... you know ... it makes sense

[mental oriole]

sudo give me all da power

[night girder]

I tell me mom stories about scams all the time to scare her, it works 😂

[edgy hazel]

VPS: ordered.

4 bucks/month

2 cores, 2GB ram, 100GB storage

Gonna be my firewall 😊

But my ISP doesn't allow me to use it as a tunnel to hide my public IP, nor am I able to look into their logs

[languid gulch]

LOL "the new 15900K: 4 cores, 4 threads, 4.0GHz, $650"

[sage pier]

15100ks 1 core 2 threads 7GHz, $600

[wanton orchid]

using less economic cores have a cost you know

[pure karma]

honestly 7GHZ 8 Core would actually probably be decent

i think the problem comes down to trying to shove down economy cores instead of throwing actual improvements on the table

[languid gulch]

https://www.youtube.com/watch?v=OVdmK1UGzGs

so confirmed oxidation

[tough owl]

good job intel

[languid gulch]

J2C mentioned that some people are seeing 1.6v on startup which could cause instant degradation

[tough owl]

1.6V is insane for a cpu

even when extreme OCing

[languid gulch]

combine that with the oxidation, i can 100% see why they fried

[languid gulch]

can't wait for the oxidation to affect 14th gen too

i noticed how they made sure not to address that at all

[tough owl]

cant let their flagship cpu be under fire

when it struggles against amd

[languid gulch]

i'm betting we'll finally see an admission once 15th gen is out

[tough owl]

they should be sued for this

14th gen has been out for how long?

7 months now?

and 13900k for 2 years

https://www.youtube.com/watch?v=-cMTtRIrJOk now this is the type of youtube video I like to see

[languid gulch]

the cpu orders from server farms & datacenters are gonna be wild

[tough owl]

amd better take this opportunity to perfect zen(6?)

[languid gulch]

god i hope so

[tough owl]

www.tweaktown.com/news/99439/amd-ryzen-9-9950x-pricing-leaked-499-9900x-for-399-9700x-299-and-9600x-229/index.html

this pricing doesnt look bad? I dont remember ryzen 3000,5000,7000 though

[jagged snow]

That's actually a price cut

7950x was at 550 if memory serves

I'm really really curious to see what happens with 9000x3d launch

That's when I'm planning to upgrade, current rig will be nas + a few other light server duties

[tough owl]

im honeslty happy with my ryzen 5600+rx7600

[jagged snow]

Oh wow, 7950x on newegg is 520 currently, on sale from 595

[jagged snow]

I have a 5600x + 6700xt atm

Fine for gaming but it's not keeping up with some productivity workloads

[tough owl]

my macbook does all my heavy lifting

m2 pro is really nice

[jagged snow]

I don't think a mac would keep up with a lot of what I want to do? I hate the OS too much to use it anyways

[tough owl]

I know its not for everyone but I really dig how its Unix based

[charred pewter]

even with a decked out m3 ... photoshop still runs like shit

(i blame adobe)

(adobe blames apple)

[tough owl]

i blame adobe

PS code has to be awful since the start of Creative Cloud

[charred pewter]

yes

photoshop had supported paths (vectors lines shapes) for a long time, and it used to have a way to export layers as SVG ... they silently took that out because "Photoshop is a raster program" ... yet the export feature worked fine before. They now say "Buy and use Illustrator for that"

🖕 adobe

[tough owl]

silly adobe

why make one app when you can make two

and make double the profit

[charred pewter]

hehe its like they are a corporate business or something

but that was the classic "remove features that work fine, and put them in something else to pay for" move

i still cannot believe how they still have not fixed the random blurry text in buttons and dialog box labels when you are on a non-4k+ monitor (it might be blurry on those resolutions too, but its harder to tell)

[tough owl]

using it on a 1080p monitor kinda sucks

text is blurry sometimes I agree

cs6 my beloved

[charred pewter]

yah, i have fits trying to use photoshop on highdef res though, cause it makes images all 1 to 1 pixel, and its impossible to see without constantly using the zoom tools

because "Photoshop is not a web graphics tool" ... "You need to buy this other program we have for that"

aye, i loved the days of just buying a working photoshop version that worked, did it all, worked

[tough owl]

does cs6 still work on windows11

[charred pewter]

it should ...

almost everything still works on whatever newer windows

macOS on the other hand ... hell no

i dont even think photoshop 2023 works on macOS 14

[tough owl]

at least cs6 on macos was 64bit

[charred pewter]

i miss my cs3

it still runs fine on my snow leopard machine, and i dare say it runs 'faster' too ... even on that old laptop lol

[tough owl]

I was younger than 5 when cs3 was released

[charred pewter]

for those who use the short-urls from google .... they killin it (cyberinsider.com/google-url-shortener-service-set-to-sunset-in-august-2025/)

[tough owl]

I seriously doubt it costs google that much to keep goo.gl links

[charred pewter]

🤷‍♂️

amazed they aint just saying 'it will cost 1 penny per 100 clicks' ... like whatever shit they doin with recaptcha enterprise

[tough owl]

or just slap an ad in the redirect to make money 🤑

[jagged snow]

It's not a resolution thing, it's a display scaling issue

[charred pewter]

ive yet to come across any other app that has an issue rendering their dialog boxes and controls clearly ... but according to adobe, its an apple bug

but honestly at this point, with how bad the macos is getting .... i wouldnt discount it

[jagged snow]

Autodesk software has the same issue

[charred pewter]

dang

[tough owl]

I use it for work. I have no idea how its been so bad for so long

[silk terrace]

For all the adobe haters: I think Affinity still has a 50% off sale going for a couple days/weeks(?) 👀

[twin dew]

But got the Pi4 to work on proper WiFi regulatory domain with IWD after adding the Raspberry Pi OS repos and switching to using the firmware blobs there.
Just had to add support for Debian baseline kernel to their initramfs generation scripts.

[sharp oasis]

https://www.tomshardware.com/tech-industry/new-scale-tool-enables-cuda-applications-to-run-on-amd-gpus
Zluda part two

[twin dew]

So just going to replace currently known unstable CPUs for the people who hear about that latest message on Intel Forums.
Not even full press release, just forum post...

Ah, sorry, Intel Forums for incomplete version and some specific SubReddit for the expanded one...

[willow pike]

this is the biggest pc fuckup I can remember since the 1.13ghz pentium 3

[edgy hazel]

Just graduated with a bachelor in misspelling sudo 😎

sudp
suo
dudo

[wanton orchid]

sodu here

suod

[edgy hazel]

sudop

[dire igloo]

every Linux admin's favorite puzzlegame - sduoku

[charred pewter]

sewdoh

[rustic panther]

Suso

[charred pewter]

sus

[tough owl]

https://www.youtube.com/watch?v=X4Akj5qF-3Q I know where my next paycheck is going

[jagged snow]

Woah

[tough owl]

people have gotten it to send messages over 103 miles

meshtastic.org/docs/overview/range-tests/

[jagged snow]

I could see these being super awesome for a lot of stuff I do actually

[tough owl]

you should try it

[jagged snow]

I'll look into it more on lunch

[tough owl]

a lot of it is simple esp32 stuff

want a windows&linux port

so I have a pc case im going to sell on ebay. I got the pc case originally from bestbuy delivered to my door in the box that it comes from retail with you think that is adequate to potentially ship it across the country.

[gilded helm]

People do that with stuff I've bought off eBay all the time, I've never thought to complain

[tough owl]

i just know from secret shopper stuff from GN and LMG they dont like when the case is shipped just with the case box, especially since my case has a psu and aoi preinstalled

[dire igloo]

If the case is properly protected inside the box, I see no issue

[tough owl]

hard styrofoam 👍

[dire igloo]

PSU won't move and AIO needs the CPU block tied down securely

[twin dew]

For professional seller, I would expect another box around the case box for whole system.
For consumer reselling not so much.

The case box with the included styrofoams is much better than most consumers can do otherwise.

[tough owl]

it has something in place from the factory to keep it in place. its an nzxt h1 v2

im just glad buyer has to pay shipping

that thing weights like 6.7KG

[twin dew]

DPkg::Post-Invoke {"! test -f /var/run/reboot-required && /usr/sbin/needrestart -p 2>&1 || exec touch /var/run/reboot-required || true"; };
The stuff you copy and then modify to be one-liner inside the Invoke instead of separate shell script...

So each late evening the various Linux systems get update lists, then upgrade any packages they can, then needrestart checks if any old files are still in use and sets that flag-file, which the automatic upgrade system part checks after midnight and reboots the computer if necessary.
Without that it only got set for critical updates by the unattended-upgrades itself.

[tough owl]

I love windows, only when im in windows does my the motherboard hardshutoff the system when the cpu is overheating, it doesnt overheat on linux

[jagged snow]

CPU overheating is a hardware issue though

[tough owl]

maybe it is?

[tough owl]

itll hardlockup only when im on windows and the mobo will boot into a bios menu saying cpu overheat

but on linux when stressed it doesnt

[jagged snow]

Yeah that is definitely hardware

Even with a power virus that chip shouldn't be getting that hot

[tough owl]

dont tell me that

do I just need to remount my cooler, cuz its using the stock one

[jagged snow]

Oh maybe

Stock paste as well?

[tough owl]

no

[jagged snow]

What paste?

[tough owl]

arctic paste

uhh let me find the exact one

[jagged snow]

mx4 or 6 probably

4 has a bit more longevity

[tough owl]

its 4

[jagged snow]

What's the workload right now?

[tough owl]

cinebench all core

[jagged snow]

Hmm

[tough owl]

i was playing fortnite and it kicked me out too

[jagged snow]

Yeah, it's pinned at tjmax and you don't have pbo enabled

I'd go for a remount for sure, then maybe tweak that fan curve

[tough owl]

Its at full blast in the BIOS

[jagged snow]

Those stock coolers kinda suck but they're usually able to keep them under tjmax

[jagged snow]

What's ambient/case temp?

[tough owl]

I have no idea

Maybe 76F?

[jagged snow]

Yeah, try a remount

[tough owl]

After some wiggling to get the cooler off that's what the paste looks lile

And the cooler

after repaste

maybe sidepanel off helps?

nope

ambient is 83 according to my food thermometer

[willow pike]

well it's not hitting thermal limit anymore but that's concerningly hot

[tough owl]

i love no ac with outside temp being 95F

[willow pike]

you will notice that your CPU is now pulling more power and running cooler

compare CPU package power and CPU core power before / after

the repaste job helped it out, and it can now run faster

[jagged snow]

Not really for stock cooler

Those things have negligable surface area

[willow pike]

yeah i missed this was under cinebench

this is running as expected

@tough owl these are the things to pay attention to, they determine how hard the CPU can run itself, so long as it isn't hitting any limits (power, current, thermals) it will run as fast as it can

and in that first example it was 100% thermal limit, not overheating but not able to run as fast as it wants to

after repaste it can go harder

a bigger and better cooler will let it run a little faster but mostly you would buy it to run much quieter

also, if you take the side panel off and the temperature doesn't change, it might just be running faster / pulling more power, so it still might have an effect

[jagged snow]

If you take the side panel off and the temperature doens't change it means you have sufficient case airflow

[tough owl]

i have a peerless assassin at home with no mounting hardware

https://a.co/d/anvdBaY

[willow pike]

that would be a lot better

[tough owl]

problem is I am 300 miles from home

[willow pike]

what happened to the mounting hardware?

[tough owl]

¯_(ツ)_/¯

[dire igloo]

https://www.amazon.com/dp/B0D1CGL7D1

[languid gulch]

just weld it on, it'll be fine

https://www.youtube.com/watch?v=nXRxWm9y3QQ

[tough owl]

thas good

[tough owl]

do I need this cooler for a 5600 or can I just buy thishttps://a.co/d/dLOBvQk

[jagged snow]

For a 5600 basically anything will work

65w is pretty darn tiny for most of today's machines

[tough owl]

I got two hours of logging with hwinfo

and no issues it looks like so

maybe I just need to cool down ambient temp and be fine

https://a.co/d/37FTMjQ I wonder how good this would be

[languid gulch]

i have a 5600X with the assassin 120SE dual tower & it's definitely overkill

but i also got it to have a leisurely existence until i want something a bit monstrous

[tough owl]

https://a.co/d/1s8uDj2 like how is this only $30

[jagged snow]

Because good air cooling is now a commodity and people only buy noctua out of blind loyalty 😉

[languid gulch]

yup

[tough owl]

even an all white one is still $30

[languid gulch]

also air coolers should all be about that price given that they're slightly refined chunks of copper & aluminum

[jagged snow]

In all seriousness, the biggest drawback to that thermalright tower is that it has an aluminum coldplate which offers slightly worse thermal conductivity, but it's still plenty for any cpu under 200w

[tough owl]

meanwhile corsair

[jagged snow]

You're also buying the brand name with corsair

[tough owl]

its probably made in the same factory as thermalright

[jagged snow]

Maybe 😄

Corsair might have higher quality fans than that thermalright cooler too

[tough owl]

how many assassins do you want?

thermalright: Yes

[languid gulch]

$33 for the 120SE is what they're all competing with

[tough owl]

the 90 Sse is only 29

[jagged snow]

90mm cooler is really only suitable for stuff at that 60w range and below

[tough owl]

even dual tower?

[languid gulch]

yea, 90mm fan ones are basically stock cooler straight replacements, not upgrades

[jagged snow]

That's pretty much the same fin area as single 120

[tough owl]

their most expensive cooler is a $74 360 aoi

https://pcpartpicker.com/product/htw7YJ/thermalright-aro-m14o-736-cfm-cpu-cooler-aro-m14o who let teenage engineering design a cpu cooler

[jagged snow]

Assuming the fin stacks are the same depth and equal fin spacing, dual 90 is 162 cm^2 and single 12 is 144 cm^2

[tough owl]

ill probably see if I have the am4 mounting hardware on the cpu cooler I have at home so I dont have to spend money on something I already own

plus its working fine 85 ambient is not helping

[jagged snow]

Yeah, don't think you need to worry about that

[languid gulch]

i have one of bequiet's 92mm coolers because i was trying my hardest to make a Dell case work, & it was a lot narrower than the standard ITX cases

oshkosh airshow going on today

[tough owl]

https://www.youtube.com/watch?v=SATYQyIcimM

[twin dew]

And having the normal power limits off and the CPU on unlimited turbo (PPT at insane value to get that 8-9% PPT) isn't really what it was designed for.

But that paste was also just gone from the heat in the pics.
Lot of the thermal pastes on the market just cannot take 90+C temps.

Often starting to degrade at as low as 60-70C.

[twin dew]

So disable the PBO.
That isn't meant to be used with stock cooler.

If you want, you can manually tune the limits to more sane value to what the stock cooler can still handle, but above the non-PBO.

You should be hitting the PPT limit and the Thermal limit at about the same time when cooler limited to not get issues.

You have that PPT limit for power, two current limits (TDC and EDC), then voltage & frequency limits from the turbo table, and then that thermal limit is supposed to be the last backstop to not fry the CPU.

Enabling PBO with stock cooler makes the first three so high that they don't limit, and the turbo-table max cannot be reached on multicore before the CPU fries with stock cooler.
Usually the throttling should keep it at the limit without shutdowns, but the difference at your current settings is so high that it doesn't work properly.

For the "doesn't happen in Linux" part, have you checked what the CPU boosts to under Linux.
In some cases you get limited by default to base clocks max without tweaking, when ACPI tables contain wrong values that Windows ignores but Linux doesn't.

Or by scaling governor or scaling driver selections and settings.

[twin dew]

Hope AMD doesn't need full silicon respin and that it was just minor process issues.

(Silicon level change, die revision)

[twin dew]

But that would mean that either the turbo-table pre-creation testing and the final validation has been insufficient (easier to fix).
Or that the silicon just cannot meet the published spec in large numbers (silicon revision or spec change downwards needed).

[twin dew]

Time to go against the instructions again.
And try to get ZoneMinder to work with Lighttpd instead of Apache2.

[dire igloo]

At $30, get Frozen A620 - and for a 5600, even a basic AX120 works (tho I am a fan of BA120 personally)

[twin dew]

Because Noctua coolers cost a lot compared to raw performance?

I still use them, but they aren't for everyone.

At any time there is usually some specific cooler that can get the same noise-to-thermal performance for 1/2 to 1/3rd the cost.

That is the good side of Noctuas, you can be almost certain that they will keep going, and will be supported for new sockets etc. for very long time.

[twin dew]

So target is to get ZoneMinder with FFMpeg on Lighttpd (with OpenSSL and PHP-FPM) and MariaDB working...

[languid gulch]

with the current shitshow at intel right now, i wonder if they're going to dial it back a bit on 15th gen & lower the power usage, & give coolers a chance

[twin dew]

Ah, it was FirewallD limitation that you cannot have both MAC and IP based source or destination filters in same rule, not NFTables limit.
Might be limitation in IPTables so FirewallD cannot support that either.

[wanton orchid]

I agree so much with this

[twin dew]

Yup, iptables for IPs, ebtables for MACs in the legacy toolkit.

[wanton orchid]

tbh I don't know why people still claim A is 1/3 cheaper than B
when B last twice as long with stable performance

[twin dew]

There is reason why all my desktops use Noctua coolers.
But need to give both alternatives to others for them to decide themselves.

[languid gulch]

just realized all 3 PCs i've built have different coolers on them 🤣

same case tho

[twin dew]

Two U12S and one U14S

[languid gulch]

AMD stock, be quiet pure rock slim, assassin 120SE

& none of them have ever had any issues

only part failure i've dealt with is that PSU this week 🤣

oh & some crap ram

would blue screen my system at stock speeds 🤣

[dire igloo]

You always gotta calculate longevity against price.
Lasting twice as long is nice but ultimately still a worse deal at triple the price.
The only reason not to do this is to be less wasteful.

Also, the main reason why air coolers fail is the fans - and they are pretty cheap to replace

NH-D15 costs $110 according to PCPP.
Dark Rock Pro 5 is $80
Thermalright PS120SE is $36

Noctua isn't worth it in today's time, even if we account for the support and longevity

Also, DBB is much better for longevity than Noctua's oil pressurized bearings

[twin dew]

But ball bearings are loudish and get louder as they age.

When Noctuas sleeve bearings can do over 10 years of almost constant use.
But no idea how much start-stop they can take, that is the time when most bearing wear happens.

(Sleeve bearing with centering magnet for startup and enhanced lubricant circulation features).
Sleeve with the circulation features is sold usually as Fluid Dynamic or Hydro Dynamic.
And then Corsair sells theirs with that magnet as "Magnetic Levitation", when those aren't.

Real magnetic levitation are very different, and don't take to start & stop well.

[tough owl]

I think pbo is off already though

[twin dew]

No, you have insanely high PPT set

[tough owl]

Ah

I'll have to look in my BIOS then

[twin dew]

Probably the 4095 W or so that many MBs set.

[tough owl]

Because I swear I turned pbo off

[twin dew]

No, not that high.... lets see..

Probably set to 1024W?

Or around that.

88W reading as 8% of the limit would be limit at 1100W.

[tough owl]

What would it be under on my MSI mobo

[twin dew]

What MB?

HWINFO main side should show the current limits:

[tough owl]

://us-store.msi.com/B550-A-PRO

[twin dew]

Still not providing any BIOS documentation, so no idea...

Graphical mode, OC on left, Precision Boost Overdrive should be the topmost option.
Overclocking\Advanced CPU Configuration in "text" mode?

But to need to go trawling various forums for pics of the BIOS insides...

When everyone else has at least the launch-time settings and screens documented...

[willow pike]

if your motherboard is cranking the power limit by default I'm gonna do some war crimes

sick of this shit

"oh yeah, building a PC is the best, you just gotta use an arcane ritual to access a scary looking screen and mess with a bunch of weird acronym settings to get it to work properly"

[tough owl]

I'll look after workv

[twin dew]

Well, that has been what almost all Intel MBs have been doing for years.

[willow pike]

yeah, just hadn't seen that they were cranking amd so much too

though i do know they were burning out X3D parts for a month or so

[twin dew]

All parts.
Just that the X3Ds were affected much quicker.

There had been non-X3Ds dying for months at that point, just wasn't noticed overall.

[willow pike]

it was quite the surprise when raptor lake turned out to have a defect, i was fully convinced it was more motherboard shitfest

[twin dew]

Both

[willow pike]

yeah, both

[tough owl]

https://tenor.com/view/both-agree-yes-good-gif-4591420

Why blame one when you can blame both

[willow pike]

motherboards were so shit it almost covered up intel's own problems

[twin dew]

Ok, so lighttpd is now working with HTTPS (TLS1.3 only), PHP-FPM and digest-auth...

[wanton orchid]

your lighthttpd is going to be less light

[twin dew]

Just name of a http-server program, same as apache or nginx

More limited than those two.

[wanton orchid]

I know that
but it's intended to be lighter still

[twin dew]

But good enough for this use.

But when my two actual possible programs I want to use are "Motion" or "ZoneMinder".
First would be just console application (but no easy way to look at the stored recordings), but want to check that ZoneMinder first, that is PHP, but is much more featured for network NVR.

Nothing else is stock in Debian, and most of them were only available as Docker images.

And then those images just will contain some form of http server running PHP or Node.JS etc.

In configs optimized for tens to hundreds of cameras on high-end hardware.

[maiden coyote]

Frigate is pretty popular too

I gave up on using any kind of open source nvr mostly because they all seem too buggy for primetime

[twin dew]

But have gone to that Docker Only route at some point.

They used to also support bare metal installations, but not anymore.

[maiden coyote]

And docker is really cool, but it's an extra layer for bugs /:

[twin dew]

And the internal components aren't then under Debian Security Team umbrella etc.

And updating the image seems to be real pain when new version comes out for any reason, like security bug in some of the ancillary components needed.

[maiden coyote]

My docker vm and truenas vm died yesterday.. I might just need to start looking for a new motherboard

[twin dew]

So when ZoneMinder mainline Debian Stable package, why go for Docker?

[maiden coyote]

Docker vm logged something about rcu stall and cpu frozen. Truenas had nothing in the log, but was stuck at 100% cpu

Because docker is stupid easy to deploy, and very difficult to make a container run under the user you'd like if the creator coded it for a uid of 1000

[twin dew]

Yeah... the current Debian Stable ZoneMinder package was packaged on same day as the same release was released on ZoneMinder GitHub.

[twin dew]

(For Debian Unstable, then moved later to Stable)

[night girder]

News I read today about docker:

In 2018, a security issue was discovered where an attacker could bypass AuthZ plugins using a specially crafted API request. This could lead to unauthorized actions, including privilege escalation. Although this issue was fixed in Docker Engine v18.09.1 in January 2019, the fix was not carried forward to later versions, resulting in a regression.

Regression: The fix was not included in Docker Engine v19.03 or newer versions. This was identified in April 2024 and patches were released for the affected versions on July 23, 2024. The issue was assigned CVE-2024-41110.

[twin dew]

What regression testing...

[night girder]

source: www.docker.com/blog/docker-security-advisory-docker-engine-authz-plugin/

So after 5 years, they discovered they forgot to push security fix to production 😂

[tough owl]

💀

[twin dew]

Regression testing is including (semi)automatic testing for any known old security holes for any new version built.

To make sure this doesn't happen

[night girder]

Yeah, but when something gets fixed. The testers need to add this testcase (automated or not).

a testframework isn't magic. It doesn't automatically know what to test.

So if fix got lost, testers were in the dark, and it's gone into a void. Wonder how they realized after 5 years it's missing.

[gilded helm]

Claims of some sort of dynamic power allocation, interesting.

www.notebookcheck.net/Anker-Prime-6-Port-250W-new-desktop-charger-could-launch-soon.867616.0.html

[mental oriole]

[languid gulch]

i can't wait to see the videos of reviewers & tech people melting 13th/14th gen CPUs for the lolz because of how worthless they'll be

[gilded helm]

If it's going to mess up laptops, I wonder how much that'll cost Intel? Then you're killing the whole device, basically. As it is know, someone on the Cyberpunk server was pissed off because even though they're getting their money back for their CPU, they're still stuck with their motherboard and they're not sure what to do.

[wanton orchid]

I'm smiling devious over this kind of situation

Intel will have lasting payback

[languid gulch]

i kinda hate to say it, but me too

it'd be nice if qualcomm could take advantage of this & become a 3rd cpu maker in the desktop space

[wanton orchid]

[tough owl]

thats a good joke

theyd have to invent a new socket for cpu and I dont think they want to

[languid gulch]

pull an AMD, steal their socket 🤣

oh sorry, "reverse engineer"

[tough owl]

id be fine with that though

[languid gulch]

same honestly

just get a 3rd competitor in there

[tough owl]

or have ampere make desktop consumer chips

[languid gulch]

that'd be weird

nvidia showing up in the cpu world

[tough owl]

ublock origin is starting to struggle with blocking ads on youtube

[languid gulch]

yea, saw that today

[tough owl]

oh well

Good thing I use grayjay

Beats me

The app still works

[languid gulch]

i believe it's called "enshittification" now

[languid gulch]

google's gif search has especially gone to shit the last couple weeks i've noticed

[gilded helm]

Brave browser has had no struggles with YouTube in my experience

[twin dew]

And completely change how they do firmware, drivers, etc.

MB makers not bothering to redo defaults again:
https://arstechnica.com/security/2024/07/secure-boot-is-completely-compromised-on-200-models-from-5-big-device-makers/

And just using the example secure boot keys from the base UEFI provider, when those should be model specific if possible, and at least manufacturer specific.

Asus is also kind of fail in that they seem to be using same PK key since at least over 10 years ago.
My Z87 MB (Intel 4th gen) and my AM5 board use the same key...

Not just strings, byte-by-byte comparison.

[twin dew]

"Why make it harder on ourselves when we can just make sure the private key never leaks"

[pure karma]

even as a owner of one i am too lol

[pure karma]

dident they make a cpu tho?

its not a consumer one but i tough i heard sometime that they did managed to make one

[twin dew]

They have made multiple different ARM CPUs.
Including the SOC in Nintendo Switch.

Out since 2014.

No, even earlier, that was when Denver based SOCs came out, first seems to have been ARM11 based APX 2500 in 2008?

Most seem to be branded under the Tegra name for the whole SOCs.

The ones used for the Nvidia GPU servers aren't really branded, just code-names and models, as you cannot get those except when you buy the whole server.
When the earlier ones were sold as just chips.

But AFAIK Nvidia is just taking the stock ARM core and adding their own accelerator blocks to the common bus.

[pure karma]

so what the Switches CPU is a nvidia chip?

[twin dew]

Yes, to get the Nvidia GPU.

[pure karma]

well it sure is one impresive cpu for what it is

[twin dew]

Nvidia was also pretty desperate to sell them AFAIK, so Nintendo got pretty good deal.

Not really,. the CPU is crap for the time, even for mobile chips.

The GPU was great.

[pure karma]

yea but its better than the equivalent laptop chips i have personally tried

[twin dew]

Switch launched in 2017.
The Tegra X1 was launched in 2015.
It uses CPU cores from 2012, when the replacement was launched by ARM in 2016, at very low clocks.

[twin dew]

Completely different OS?
So cannot really be directly compared.

[pure karma]

yea true but still has very solid performance

[twin dew]

Very heavy GPU bias too, when most cheap laptop chips also have crappy IGP.

Because those are only expected to do 2D things.

[pure karma]

yea that was the choke i found the most when comparing it to anythign else mobile really is that it almost always had like 4-10x the framerates BECAUSE of that gpu

[languid gulch]

might not be too weird to see i guess

since now you can have a pc with AMD cpu/gpu, & intel cpu/gpu now

nvidia cpu/gpu might be next on the menu

[soft bloom]

if you really want to get rid of ads - best option is to block their domains

also, NoScript

but also - let nice sites that you enjoy get few pennies from you watching ad

[edgy hazel]

No thank you. I'd rather not support an industry which researches how to best manipulate a person into buying stuff.

[soft bloom]

everything has its cost

[wanton orchid]

and this specifically have a cost on all humanity, it does not makes any progress

it is regression
and that regression type is called corruption

[charred pewter]

haha, spyware makers get hacked 😄

[jagged snow]

And if there is an alternative, I'm happy to pay for it
I donate to mozilla because I value the privacy I get from firefox(and thunderbird to a lesser extent) so much

But so many services don't have an alternative

[midnight osprey]

Hey guys, what can I do to resolve this what looks like light bleeding?

That should simply be black on green.

[twin dew]

Is the text image or actual text?
And is the scaling for display & browser at 100% or something else?

[wanton orchid]

^ looks like nasty scaling performed by monitor

I really still think monitor should enforce integer scaling

they should never assume what data is screened, it's not their job, that's the job of the graphic driver and processing

[midnight osprey]

text, I tried 100% no difference

[midnight osprey]

maybe settings on the monitor would fix it?

I have 3 monitors, happens poorly like the image shown on one, less on another, then not noticeable on my expensive one

hmmm, monitor settings fixed it

These settings here. I had both monitors on cinema mode.

all sorted now

[edgy hazel]

You vs the guy she told you not to worry about.
(Coming next week. Hyped af)

[languid gulch]

do they have heat pipe cooling yet?

[edgy hazel]

heatpipe would die way too fast

[languid gulch]

really?

i assume that's a 3d printer

[edgy hazel]

"Copper/water heat pipes have a copper envelope, use water as the working fluid and typically operate in the temperature range of 20 to 150 °C"

PETG prints at 230-250C

[languid gulch]

ah

molten salt heat pipes then 🤣

[edgy hazel]

no hot enough for that xD

[wanton orchid]

since what update firefox added background noise to pictures

blank picture makes blank background noise

full blank picture :

wait no even blank have it but even less

magnified

and this is the pattern of the pattern

[languid gulch]

AI prevention stuff?

[wanton orchid]

without the picture being zoomed-in in firefox (then magnified, then zoomed-in without conversion (1x1 to 5x5 pixels)

I caught an horizontal pattern

but hard to find a vertical one

[wanton orchid]

avlues always between 31-35 (5 level)

[stray badger]

Maybe an addon added it

Privacy badger or amth

[languid gulch]

i'm wondering if it's one of those copyright traps

stick that behind every image, so that if/when an AI scraper comes along, they can examine the layers & find that background & claim copyright on it

[wanton orchid]

it's not in the image directly, it's on firefox background

so it's only useful for in screenshot or screenrecorders

and I'm wondering if it's leaking some info

[languid gulch]

rumoured ring bus flaw in 15th gen

[twin dew]

How would the one in 13th and 14th gen that Intel didn't know about get fixed in 15th gen?
Running it too close to the absolute max VCore it can take?

[languid gulch]

it basically wouldn't

they'd probably make the binning stricter, & really stick to keeping the top performers to be i9s

it sounds like they just tried to overvolt everything to be a higher tier than the design really allowed for

pump the voltage, screw the longevity, hopefully we don't piss off enough of our customers to lose them

and they might try to pull off the same thing with 15th gen if AMD scares them enough, which it seems like it did last gen

they design these things years in advance, & they can't just turn on a dime

i suspect the only way this stops is when intel goes to a new platform

[twin dew]

And have painted themselves in the corner with the pumping of the stock OC way up during the 14nm decade.

As the node changes don't give them enough headroom to tone that down and still give improvements from previous gen in max performance.

[languid gulch]

they're fucked because they advertised themselves, and for a time, showed, their performance numbers. it will be a true disaster if they drop performance going to a new generation

[twin dew]

Like AMD is doing now with 7000 to 9000 desktop side for example.

[languid gulch]

we kinda new that LGA1700 was going to be supporting 15th gen, but no hints as to after. i wonder what kind of panic is going on behind the scenes if they were planning on having it support 16th gen on the same architecture

what worries me is what caused the oxidation issue

the voltage thing is just something they can turn down a bit

[twin dew]

Mistakes on the manufacturing line when something was changed in the process, and was fixed near end of 2023.

[languid gulch]

supposedly

[twin dew]

That one is very simple thing.

[languid gulch]

we haven't seen them specify batch numbers for others to verify it

[twin dew]

Because Intel doesn't want to replace all those CPUs.

[languid gulch]

yup

they only specified that there was oxidation in reference to 13th gen, with 14th gen having just as many failures

i have 0 trust in PR statements

if it was just overvolting frying the chips, there would have been at least a dip in 14th gen failure rates, but so far they haven't said that's the case, & as wendell pointed out, 14th gen is failing at at least the same rate

[twin dew]

But that danger for oxidation in vias is known thing in the industry based on videos made about the rumors by people in the know.

The question is why Intel fucked that up, not why it can happen.

[languid gulch]

that's why i worry about it

this kind of failure is one of those "ok, this is our one thing we do, and now it's all fucked up"

[twin dew]

And why Intel fucked it up is MBAitis.

Same as all the other Intel problems in 10+ years in all their products.
Which has been contant.

Their NICs have gone from "only thing you should use" to "fucked up shit" etc.

Even on enterprise side.

Their WLAN chips issues I already talked here.

[languid gulch]

they're toyota and they just ruined every single camry & Rav4 they've made for the last 3 years, & the next 3 years

that could be an outright killer

the top 10 car brands make up 47.2% of that market. intel has 70% of their market on their own

[dire igloo]

Now they're saying that apparently ALL 13th and 14th gen 65+W units are affected
https://www.techpowerup.com/324963/intel-will-not-recall-failing-13th-and-14th-gen-cpus

[twin dew]

Part inertia, part anticompetitive practices, part TSMC production line limits for AMD chips.

[languid gulch]

oh wow 2hrs ago

have there been reports of the CPUs taking other components with them when they die?

[dire igloo]

I wonder if this is related
https://www.techpowerup.com/318867/german-court-prohibits-intel-processor-sales-amid-patent-dispute

[languid gulch]

ouchie

huh. did intel try (and fail) to copy R2's IP?

[dire igloo]

They copied it for 10th-12th gen, got caught, changed shit for 13th and 14th and now those are fucked

[languid gulch]

LOL maybe this is why there's rumors 15th gen will be ditching hyperthreading. literally can't handle it

lol. so can't sell 10th-12th cuz of infringement, & 13th-15th gen just kills itself

looks like maybe going on their own for that stuff was a bad idea

[twin dew]

The R2 seems to more like patent troll in that case, and they lost in US before going for it again in Germany.,

[dire igloo]

10th-12th depends on ruling, this was only for Germany. US courts ruled in favor of Intel.
They're also back on sale here, must've found an agreement of some sort.

Anyways, it was just a weird coincidence that Intel got an IP claim for 10th to 12th gen power circuitry and then had power issues on 13th and 14th gen where they changed things up

[twin dew]

Expectation would be that Intel would try to redesign the stuff under court when then US case was filed.

To limit their liability if US court case had gone the other way.

[night girder]

So intel confirmed, the microcode patch will not solve issues of CPU's that already are damaged. It's a preventive patch. Which makes sense ofc.

[languid gulch]

"shit's fucked y'all" 🤣

[night girder]

The CPU's are damaged physically because of the high voltage

[languid gulch]

as J2C pointed out, it could happen on startup

[twin dew]

And at same time not taking any responsibility, only taking RMAs with fight, during the normal warranty period if you can prove instability...

[languid gulch]

and probably a tax write off for the whole thing

also who wants to randomly trust a mobo that may or may not have the updated bios

[twin dew]

There have been cases of MBs with 2+ year old BIOS versions being sold.

[dire igloo]

yeah, that's what I've heard. any damage done is permanent

[twin dew]

Well, that is always the case with overvoltage or any other cause for degradation.

[night girder]

It's what intel says themselves. Or atleast doesn't deny.

Citing unnamed sources, Tom’s Hardware reports that any degradation of the processor is irreversible, and an Intel spokesperson did not deny that when we asked.

And, Intel confirms, too-high voltages aren’t the only reason some of these chips are failing. Intel spokesperson Thomas Hannaford confirms it’s a primary cause, but the company is still investigating. Intel community manager Lex Hoyos also revealed some instability reports can be traced back to an oxidization manufacturing issue that was fixed at an unspecified date last year.

[dire igloo]

Yup. This would've only been reversible if the issue was some wonky code or smth that just caused stuff to use extra processing time or smth

You can't fix a hardware issue with a software patch

[twin dew]

Sometimes you can mask it, but not when the issue caused permanent damage that will keep getting worse even at the lowered voltages.

If the issue had been too low voltages because of fuckup, then that could have been fixed.

Intel is just trying to triage the CPUs already damaged so that they last until after the warranty is over.

And to not damage new ones.

[languid gulch]

wendell getting sources to confirm that it's happening even on the "as tame as we can run it" servers is also really bad news

that to me sounds like even undervolting won't solve it

[twin dew]

Those were probably getting damaged more, as the BIOSes weren't undervolting the CPUs like the consumer MBs were, and the max boost voltages weren't limited more compared to consumer MBs, just no extra boosting.

Point is that most consumer MBs were undervolting the CPU VCore via too low AC Loadline settings while under load.

But in general I meant that IF the issue had been firmware doing undervolting, and not overvolting like it was, then software fix would be enough.

But it wasn't.
Intel was overvolting, consumer MB makers were undervolting.

[languid gulch]

i wonder if this is going to impact intel GPUs

[twin dew]

The MB maker undervolting was fixed with those newest BIOSes, but the Intel overvolting isn't fixed yet.

Why would it?
Completely different team, and produced on TSMC, not Intel fabs.

[languid gulch]

maybe if they could swap out the orders?

[twin dew]

You don't move chips between node techs.
That needs almost full redesign of the physical layout.

Which was the point of old Intel Tick-Tock.
Move known working arch to new node, then design next arch for that now tested node.

So you only had the Arch changes or node changes as problem points each time.

[languid gulch]

yea good point

[twin dew]

Also: TSMC doesn't have anywhere near the necessary fab capacity.

[languid gulch]

how dare they not have triple their capacity tomorrow

[twin dew]

I would expect TSMC to have more current and "next" gen capacity than Intel has.
But there is no slack in TSMC fabs of the 6nm and down nodes of theirs.

[twin dew]

Source 3 part has something from the oxidation thing.

Ah, minecraft server 14900K degrading rapidly with very low temps, but that is the optimum "constant max boost with max voltages" situation.

30% failed in 2 months of use for some server provider...
With TVB disabled, down to 5% or so.

Disabling Thermal Velocity Boost causes the CPUs to drop 200MHz from max clocks when cool.

On SuperMicro blade server platforms.

And the specific data center has stopped selling new rent contracts for those blades.

83C absolute momentary max, 50-60C normal under load.

[cyan crescent]

Im building a pc and i am doing a custom water loop. The final product will have 2 360 rads, a cpu water block and gpu water block. I am missing the gpu water block and 1 rad. Im tempted to build the loop i can right now and finish it when the last 2 parts come in. Should i wait or make the cpu loop and add the gpu later?

[edgy hazel]

Depending on the wait.

[cyan crescent]

Its an order from ekwb that im waiting on to be delivered

[edgy hazel]

If you add the GPU later you'd most likely drain the loop or at least a bit of it.

[cyan crescent]

I know

Its a soft tubing build. Should be kinda easy

Its my first custom loop

[willow pike]

you'll forgive me if i don't believe anything that comes out of MLID

this feels like him making shit up to jump on the intel flaws bandwagon

the ring bus being cooked is a theory that originated from twitter months ago, and here he is posting it as if it's inside intel information

he's also been caught out posting outright falsehoods several times

doesn't mean it's not true, but does mean anything he says doesn't amount to squat

[twin dew]

As if he has confirmation for the suspected issue.
That would match to both the symptoms and the Intel official outpour with limiting the VCore.

But if that specific part is true, the "oxidation issue" is pretty small in actual delivered CPU units.

If it is just from one fab for part of production for few months.

Part of production in as that some getting scrapped.

When Intel hasn't publicly stated more than that the issue was resolved by end of 2023.

Most of the video was crap speculation from him based on those lines, some more, and general internet stuff.

[willow pike]

it's definitely true that intel aren't saying enough

[twin dew]

Which is why I posted only that single pic.

[willow pike]

yeah that's fair

[twin dew]

Basically the guy goes way too far in speculation done by him, based on scraps. And then presents that speculation as facts he got via inside sources.

Leading to very bad hit-miss rate, when he can have actual inside sources.

[willow pike]

one time, someone sent him an obviously fake tip and he reported it as insider information

[twin dew]

The first entry would be the least likely to be from actual Intel source, second might be questionable too, third is probably real IMHO.

Just from some low level tech with not much overall understanding of the tech.

[willow pike]

I'm not gonna feel comfortable recommending them even after the microcode fix, I'd need a solid year of no further failures first

[twin dew]

Small issues with the clean-room air circulation and purification system (HVAC but not HVAC) can easily cause issues in the produced wafers that wouldn't be noticed until the dies got into the binning phase of production.
And the timeline would match that pretty well.

And at that point you would have at least months worth of affected wafers in the pipeline.

And if you don't shut the lines down immediately, you easily get second months worth.
And if you do shut down the lines, you don't get months worth of wafers at all.

[willow pike]

yeah I read that the via degradation was from HVAC which is wild, those fabs need beyond surgical levels of clean

[twin dew]

AFAIK there wasn't possible cause for the process issue "known" before this, which was the point I posted that screenshot.

Just that SOMETHING had happened and some amount of Intel CPUs produced before end of 2023 had extra oxidation in vias.

So the question in the air was, if Intel had fucked up the base process, and everything produced until the cutoff was affected, or something else.

So if that specific comment is true, it is pretty small affected population, from non-base-process cause.

But IF that is the case, Intel could easily just recall all the affected batches now if they wanted.
But they are still "lets minimize our immediate costs" mode.

[willow pike]

they should, but as you say, probably won't

it would announce the problem to more people

[twin dew]

As the lack of proper easy RMA for the unstable CPUs etc. show.

[willow pike]

and Dell/HP/acer etc would be big mad

[twin dew]

The OEMs were informed about the issue near change of the year.
Which is the source for the leak about the issue now.

[willow pike]

lmao

[twin dew]

For the oxidation.

[willow pike]

lord what a mess

this is what you get when you source from one company for decades

[twin dew]

Probably along the lines "We had fuckup, we shall do easier CPU RMA replacements for you if you get CPUs that don't work right".

But didn't tell more to the OEMs at that point based on the leaks.

No recall for them either etc.

But for consumers, even right now the Intel RMA side still seems to make people with unstable CPUs to jump through ridiculous level of hoops to prove that their CPU is actually unstable.

[willow pike]

gotta put a bad taste in their mouths

[twin dew]

For OEMs it has been very easy for some time already.

[willow pike]

when the xbox 360 was having its red ring of death, MS rolled out the red carpet for replacements

this, on the other hand, is a superb way to alienate customers

[twin dew]

Leading to for example that Wendell provided "All our 13900Ks got replaced with 14900Ks by Intel" etc.

But then that didn't fix anything, just postponed the return of the crashing.

[willow pike]

still the same die yeah

[twin dew]

But that was much earlier this year.

But Intel seems to have been getting lot of OEM & big customer RMAs for their CPUs for long time, and had lowered the standards for that.

For the 13th and 14th gen.

[willow pike]

on the consumer side I think games should include more info in error messages after shader comp crashes, just detect if the user has a problem CPU and tell them they should replace it

[twin dew]

But somehow the information hadn't moved from the RMA side to the design side?

[willow pike]

I feel like a lot of people are blissfully unaware of this problem and just assuming games are unstable these days

[twin dew]

One game now prompts about that on startup.

[willow pike]

yeah I saw that but wasn't sure if it was real

it should be

[twin dew]

That dev is one of those that has been talking to Wendell and/or GN.

[willow pike]

hell yeah

[twin dew]

Nvidia had that in their GPU driver patch notes several months ago already.

[willow pike]

imagine if fortnite did that

[willow pike]

amazing

[twin dew]

But not with prompt, just if you read to the end of the notes to "Known Issues" section.

About the specific out of VRAM error message, that has been affecting Satisfactory users too since last year with U8 experimental.

Had been happening for over 6 months by the point of that first time inclusion into Nvidia driver notes.

About the cause being unstable Intel CPUs.

But this issue has been slowly catching speed since 6+ months ago when first finger pointing to Intel started.

And was happening for at least 6 months before that, in last summer.

The specific crashing game issues around Unreal Engine games etc. from this "current" thing.

[willow pike]

yeah I've been keeping tabs on it since the rumbling started

it's so much worse than I feared

[twin dew]

Need to try to kill my firewall today with UEFI and BMC firmware updates.

[tribal kraken]

Wonder if my first 13900k was effected by that. Bought it 20.10.2022. Lasted abouth 6 months until games started crashing and couldn't reinstall windows

[twin dew]

Probably.

[tribal kraken]

Just have to hope that the replacement works for some more time in future. Can't RMA it as it's delidded. Wouldn't have done that if the issue was known then

[charred pewter]

i have a 'spare' 12600k on the shelf ... i was not impressed with it compared to my 14 i have (for some reason games slogged cpu-bound on the 12600)

so im hoping my 14 survives

[soft bloom]

he did it again!

[edgy hazel]

huh?

[stray badger]

Seems like it

[twin dew]

Took longer than expected, little over an hour...

Partly because I hadn't made all necessary preparations in time, partly because I forgot to completely remove power in another and spent 10+ minutes waiting to see if I would get picture back after UEFI update, while the system rebooted couple of times.

[visual tree]

I think I finally eliminated mold in my room from ever appearing again....

Replaced cabinet and wardrobe back panel which is made from plywood with moisture-resistant hdf board

And removed plywood cornice from the ceiling

I should probably get a dehumidifier soon and finally end this nightmare

Lesson of the day: don't use plywood anywhere!

[languid gulch]

i live where humidity doesn't exist, so it's fine here 🤣

[edgy hazel]

Why is the air so wet and hot I want to die

[tough owl]

I don't have AC

Its consistently 85F in my room

[vast tundra]

What do I do if my fan hub dosent have enough ports I need to connect 3 more

[languid gulch]

hope you have an extra connection on your PSU to attach another fan hub

i'm not a fan of using splitters off of a hub to attach more fans, but connect the most important fans first, hunt down another hub for later

[vast tundra]

So do I order a different hub then

[languid gulch]

check the power usage totals, & then decide if you can get away with using 1 big hub, or if you need to have 2, with each on a separate power connector from the psu

[dire igloo]

Can you daisychain?

[soft bloom]

NightHawk releases video of Sky Cooling Fabric

[vast tundra]

Idk I have to see

I can tell u in like 10 hours

[dire igloo]

Already watching it

My brain just can't stop this King of cotton chicken candy nuggets

[silk terrace]

To the tune of the Turtles theme song

[soft bloom]

i'm just thinking, if we can break such an used-to thing as 'albedo cannot be above 100%', than maybe there's a way to cheat on energy preservation?

[tough owl]

ryzen ai 300 has no support for sodimm ram

[twin dew]

Probably has, just not at the "wanted" speeds?
Or is it LPDDR5 only?

[tough owl]

lpddr5 only

[twin dew]

Ok, then.

[tough owl]

i assume, im just watching the ltt video

[twin dew]

No, the spec pages list DDR5 and LPDDR5X.

[tough owl]

just give us lpcamm2 amd 🙏

[twin dew]

Max Memory Speed: 4x2R DDR5-5600, LPDDR5x-7500

So can do 4x SODIMM at 5600MT/s.

[tough owl]

AMD's tech spec pages have been wrong before

i wouldn't be surprised if it got copied from last gen and the important fields updated

[twin dew]

Yes, but until we get more concrete source, that is what we have.
LTT is step down.

[tough owl]

also I don't like how they tested just "apple macbook pro m3"

like which m3 theres 3 skus

[twin dew]

And if that really is for 4x 2R DIMM, then that 5600MT/s is very high.

But LPDDR5(X) will almost certainly be the one you will get almost all the laptops with.
SODIMMs are almost certainly not be used by anyone, and that (LP)CAMM2 will be used in the few laptops with replaceable memory.

[tough owl]

pricing for these new laptops is nearly in gaming laptop territory

at least from asus

[twin dew]

Because those models have shitton of RAM for laptop, very high end screen for laptop, and bleeding edge CPU.

As expected from the initial supply with very limited CPU amounts.

[tough owl]

32gb should be standard now with the highest end sku cpus if the direction is soldered ram

[twin dew]

But still isn't, and even 8GB is still being sold, with 16GB as the "standard".

And launch products for thing like this will be the expensive models.

Look again in 6 months.

Early adopter "tax" is real thing.

[tough owl]

https://www.bestbuy.com/site/6581899.p?skuId=6581899

[twin dew]

Can both get higher profits from the people who just have to have it, and at same time empty the inventory of older stuff without losses from too high discounts.

[tough owl]

i mean yea thats fair given 8945HS laptop is currently at $999

[twin dew]

With only 16GB of RAM, otherwise seems to be same.

[tough owl]

whats weird is that isnt an "on sale" price

thats just the normal listing price

Also I know intel has been doing this for a long time and now I guess amd is doing it too, but new cpu architecture is coming to mobile chips first?

[twin dew]

But mostly for laptops, below 1000 is "cheap crap" territory, 1000-2000 is the "proper laptops without dGPU" / "cheap crap gaming laptop" range, and proper gaming laptops start from 2000.

[twin dew]

Was supposed to be about simultaneous release of the 9000 desktop parts and the 300 AI laptop parts actual availability.

[tough owl]

well i mean in the sense reviews for mobile came out first. I know 9950x and I think 9900x got delayed

[twin dew]

Yes, you can get usable laptop in the below 1000 pricepoint, but usually at least the screen will be crappy one.

[tough owl]

I still think for most people a proper desktop tower and thin and light laptop is still the way to go rather than buying an expensive gaming laptop

mobile tax also exising(and nvidia laptop gpu tax)

[twin dew]

Yup.

[tough owl]

Baldur would you consider used buying 3.5in datacenter HDDS or for the same price getting half the storage and buying new drives

[twin dew]

Would depend on the source and exact type etc.

Would need to be new enough (not run more than 3 years), from trusted source, in type that is good and not something special etc.

Because you can fuck the SMART attributes up, or do something in refurbishment phase to cause damage etc.

As in intentionally modify the logged SMART stuff to hide things.

[tough owl]

I have an older Optiplex that has two drive cages in it and will be doing some sort of redundancy with the two disks. I was just curious since i'm limited to two drives if its worth the risk to buy used for more storage or play it safe with less storage and buy new

[twin dew]

Like I said, as with any used hardware, goes into specifics about the exact thing and exact source, I will not provide general answer than that in some cases the used can be good buy.

[tough owl]

Yea I need to do research on the exact drive I want and from the exact seller.

[twin dew]

And once you get it check the SMART.
Current drives log how long they have run, amount of powerons and head parks and max temp they have reached etc.

[tough owl]

and I wont buy from a seller that wont accept returns on HDDs

[twin dew]

Which was the point that unscrupulous seller could edit those to remove the signs that the drive has lived a hard life already.

And the point about the specific model is because some of the "Enterprise" models aren't any better than consumer drives.

And some can be SMR, some can be special host needing SMR, some can be SAS only, etc.

[tough owl]

yea I wont buy any SMR drives

nor sas stuff

another side note, if I wanted to take my database of imessage stuff and somehow import it into the android sms database how could I do that, if I wanted to

I ask because I have the entire imessage database on my mac and I can export it using this tool to .txt files for each phone number github.com/ReagentX/imessage-exporter. I can write a custom script to remove all of the details it gives about when a message was read and attachments etc

[midnight osprey]

hey guys, whats the best format here for preserving detail in 1080p 60 FPS videos?

[tough owl]

well some are audio containers and some are video containers

do you like mp4, mkv, mpeg?

[twin dew]

There are software made to do that specifically, but I would most be paid.

[twin dew]

And none of those are the actual video codec, just containers around it.

[midnight osprey]

i'm not worried about space it takes ... it's for a tutorial on youtube ... I have youtube premium but playpack controls are poor, need something thats editable

hmmmm?

[tough owl]

are you ripping a yt video?

[willow pike]

codecs go into containers

[midnight osprey]

yeah, using Freemake

[twin dew]

Container sets what it can contain, but most of those can do almost anything.

And that actual codec and settings for it set the quality.

[midnight osprey]

I: just don't wanna lose detail or FPS

[tough owl]

just rip using yt-dlp github.com/yt-dlp/yt-dlp

you can rip exactly what youtube has

[midnight osprey]

nice, thanks

[willow pike]

this is the way

[tough owl]

on the side of editing the android sms database?

[twin dew]

Yeah, you don't want to recode from Youtube, you just download the specific resolution and codec you want.

[twin dew]

Both getting the iMessages, and moving them to the Android.

Some are phone software, some are desktop software, not going to try to find the best one.

[tough owl]

well I have the imesssages using this github script github.com/ReagentX/imessage-exporter all in TXT files I just was wondering if someone knew of a tool to extract the raw android sms database and I could edit it manually and then put it back on my android so I can have the full message history with people

[twin dew]

play.google.com/store/apps/details?id=org.faked.isms2droid&hl=en <-one example of free app that says it can move imessages from PC backup to Android

[tough owl]

I dont have my iphone anymore and thats what the issue is

[twin dew]

But there are probably more, and haven't checked the permissions etc. on that.

[willow pike]

@midnight osprey containers: MP4, MKV etc. think of them as boxes you put stuff into, that stuff being the video, the audio, a thumbnail, chapter information, subtitles, different language tracks. but none of this changes the quality

codecs are the thing that do the job of encoding video: H264, H265, AV1, WEBM etc. they can all make a real nice looking 1080p60 video but newer ones can get more detail in smaller file sizes

[tough owl]

yea I dont wanna use it

[twin dew]

WebM is container for the various VP# video codecs.

[willow pike]

beans I could have swore that was a codec

in my defense I see webm or vp9 and my brain says no, bad

[twin dew]

It isn't general purpse AFAIK, but at least VP8 and VP9 for video, not sure about what audio codecs.

[midnight osprey]

very nice summary, thanks

[twin dew]

WebM defines the file container structure, video and audio formats. WebM files consist of video streams compressed with the VP8 or VP9 video codecs and audio streams compressed with the Vorbis or Opus audio codecs.
The WebM file structure is based on the Matroska container.

[willow pike]

it would take ages to really dig into the pros and cons of each codec, video codecs are some voodoo shit, but it helps to know what all these acronyms are a little bit

[willow pike]

webm is based on mkv?

huh.

[twin dew]

From the PR page.

[willow pike]

why not just use mk- y'know what, I suspect the answer is some dumb shit that will make me grow some grey hairs

[twin dew]

Well, taking good container format like MKV and using it as base for more limited version for www-pages, where if something says it supports that new "Web-Matroska" (WebM), it can play anything that can be contained.

Where as the Matroska itself doesn't limit anything.

[willow pike]

it's WEB MATROSKA

my brain is blown

[twin dew]

And no, I didn't know before this moment either.

[willow pike]

oh I'm keeping that bit of trivia up my sleeve

[midnight osprey]

yep

[tough owl]

looks like it works using the chat.db found on my macbook thank you!

only 74K messages to process

funny thing is that .db file is only 120MB ish

where as attachments got up to 18gigs, which I aint transfering

[twin dew]

Text compresses very well, compressed images don't compress anymore.

[tough owl]

imessages attachments will compress if you have data saver mode enabled. other than that they dont

[twin dew]

Or is that 120MB uncompressed?, also, text as itself doesn't use that much space even uncompressed as just text, without current type of formatting.

[tough owl]

I assume uncompressed

[twin dew]

I meant that if the attachment is for example .jpg, running it through lossless compression will often cause the file size go up, not down.

And even if it goes down, the difference is in few percent range.

[tough owl]

well its an sqlite database

and I dont think its encrypted which sorta kinda shocks me

[twin dew]

Where as raw text doesn't use much space anyhow (1 byte per character for most western ones), and compresses very well (75% reduction or more is easy.

[tough owl]

how can I tell if the sqlite database is compressed

[twin dew]

SQLite doesn't support compression by default.

[tough owl]

ffs

[twin dew]

Are you sure it is SQLite, and not just some Apples own thing with just .db extension.

Which isn't SQLite specific anyways.

[tough owl]

file permission issues

copied it elsewhere and now it works

[twin dew]

But also, remember that 120MB is 120+ million bytes.
So for ASCII, that would be 120+ million characters.

Uncompressed.

[tough owl]

that sounds about right

well I can encrypt it

but im surprised apple didnt do that on an OS level

givrn how much they care about "privacy"

looks like they actually delete the message from the DB if it was deleted, then again I have no idea what im looking at

[twin dew]

Predefined SQL clauses.
First one for example when triggered writes the ID of the deleted message into special table for syncing that that delete happened without retaining the actual message.

First is triggered automatically when a message type of data is deleted from specific table by something else.

Database internal automation, so that external program doesn't have to do stuff itself, but it gets triggered automatically.

Basically seems that the external DB user just sends "delete this message" command, and then all attachments etc. that belong to that message also get deleted, including from the drive (fourth one), and the IDs are added to the "these got deleted" sync table.

Also same for external just sending "delete this chat", with the database handling the minutea.

[midnight osprey]

I have used pot player in the past ... Might sound fussy but I want to just rewind 1 second at a time .. settings say it can do this but always rewinds to the last second that rounds with a multiple of 10 ... any playback apps with finer control?

[twin dew]

Would need one that doesn't do just keyframe level movements, but finer control.
That is pretty rare for consumer level playback software.

[midnight osprey]

omg, it works!

and in a free app too

[twin dew]

Basically most video codecs have keyframes (contain a new full frame), and then difference frames from the previous frames.

[midnight osprey]

I am so blind dude, turning off keyframes was in teh same page, lol

[twin dew]

And the amount of frames between keyframes is one of the things you can tune when encoding.

[midnight osprey]

used to be the dogs bollocks, not needed this kind of app in years

[twin dew]

Yeah, you can just jump to any keyframe, but for frames between, you need to go to previous keyframe and then go forward frame by frame.

[midnight osprey]

hundreds of settings and the only one I will ever change 😄

well, jumping back just 1 second is fine

he sometimes jumped around MUCH more complicated versions of this too quickly and so hard to follow

[twin dew]

So if there are 1 keyframe and 9 frames between, then it will take 1 decode to show that keyframe, but to show the last frame before next keyframe will need 10 decodes.

[twin dew]

And if the ratio is 1:24, then it is 25 times more expensive to go to the last frame before keyframe than to keyframe, etc.

[midnight osprey]

cool

I have the computational power 🙂

RAWR!

[twin dew]

Just as an explanation why that "keyframe only" is a thing, and often the only one allowed.

For playback only software.

Video editing of course needs to allow for frame specific in all cases.

Most video playback isn't that critical, going frame by frame is not that common usecase.

[midnight osprey]

surprised its a free feature then

I wonder if Pot is open source?

looks to be!

[cyan crescent]

So im making a custom water loop and every fucking component has a leak somewhere. Kill me.

[edgy hazel]

is it EK?

[twin dew]

It was.

But also, what kind of fittings in use?
And where are the leaks happening?
From blocks themselves?
From between the block and the fitting?
From between the fitting and the tubing?

[edgy hazel]

A solution to every leak is just:
Screw it tighter

[twin dew]

Well, sometimes that is the cause of the leak, and then you are fucked.

When one of the parts is plastic.

Or if at least one side doesn't have a groove for the o-ring.

[twin dew]

Pi 4 stuff: Updated from base Bookworm 6.1 kernel to Raspberry Pi OS 6.6 kernel (to get all overlay files), but now the WLAN chip doesn't work properly with WPA3 anymore and had to disable the offload for that...

Ah, might be because of the kernel switch not overwriting files that the other kernel had included for the base device tree files.

Making a frankendistro is dangerous.

[twin dew]

Ok, might be that the specific module option to disable the WPA3 offload is the one causing the firmware blob warning.

Yup, so might be that the cargo-cult option is wrong...

[twin dew]

Yes, one or other of the capability flags disabled worked too...
What the specific bits was documented in the kernel source code.

And yes, it was cargo cult.
Just one of the two bits were needed, either worked but that was because one of them also disabled what the other disabled.
And only the WPA3 offload needed to be disabled, not the FW Supplicant.

[visual tree]

Just checked my old reviews on google and one of them wasn't posted because it apparently violated the policy lol

Posted how I had negative experience at a car wash because I paid for active foam which wasn't working and the brush was dirty

There are several other people who posted the exact same thing but their review wasn't taken down. This makes no sense lol

Unless Google algorithm messed up

[cyan crescent]

My new pc. Still waiting on the gpu water block and one more radiator

[languid gulch]

i hate to say it, but i'd have put a drain valve in one of those ports at the bottom of the reservoir

love the color scheme tho

[cyan crescent]

Thats one of the parts im also waiting on

[languid gulch]

yay!

also, nice call not putting any fans or anything too important at the bottom. if it's gonna eventually leak, i'd rather it dump out onto the desk & not get flung around by fans

how much of a pain was the wiring on the backside?

[cyan crescent]

Yes

[languid gulch]

🤣

oh tell me it's a modular PSU

[cyan crescent]

It is

[languid gulch]

ok good

cables just a bit long for the case?

[cyan crescent]

A bit.

Might see if i can do custom cables and make it nicer

Its a seasonic psu

[languid gulch]

nice 😄

be sure to hunt down the specific model when it comes to custom cables

not just the brand

things can go boom

[mental oriole]

ElectroBoom

[tough owl]

www.tomshardware.com/pc-components/cpus/intel-cpu-instability-crashing-bug-includes-65w-and-higher-skus-intel-says-damage-is-irreversible-no-planned-recall

ah thank you intel

[wanton orchid]

Intel wants you to know it hates you for being their client

[languid gulch]

has AMD ever done a recall?

[wanton orchid]

Last year

[languid gulch]

cpu?

i know the 7900xtx had some bad heat pipes

[wanton orchid]

for the damaged ones cpu yes

[languid gulch]

oh right

[wanton orchid]

tbh it should be pretty bad financially
I don't know how has Intel been thinking they can afford that

(by rushing cpu dev I mean)

[languid gulch]

i mean, those decision makers just get to resign & take their golden parachutes, why should they care

btw i almost mistyped it as parachites & i think that may be a better term for it

[willow pike]

man I just remembered, overclocking a pentium 4 to 1.7V would cause it to degrade. time is a flat circle

[dire igloo]

Was that stock config?

[willow pike]

when overclocking, so no

that this latest shit resembles such an old problem at stock this time is uncanny

[wanton orchid]

it's called public stock market

[languid gulch]

just realized that 11th gen intel probably doesn't look so bad now 🤣

[gilded helm]

Doesn't the 12900K still get recommended a decent amount if you can find it and a mobo for the right price?

[dire igloo]

There was a Microcenter bundle with it.
Other than that, not really.
12600K/KF is a common rec tho, 12400F too if priced well

[dire igloo]

LGA1200, so deprecated platform - 12th gen CPUs and AM5 became more viable tho

[languid gulch]

oh i know, but losing 2 cores officially vs having 2 or more cores drop dead on you almost sounds better to me

[dire igloo]

Except that there's been a whole lot of architectural improvements too.
For all I know, 12600KF beats 11900K while also costing less - and it's not affected by the current issues

[gilded helm]

Huh, 12900K is $315, 12700K is $234. Not terrible.

[dire igloo]

Remember the extra cost for PSU, mainboard and cooling that they bring with them

[gilded helm]

Right, you'd need a killer deal on the mobo.

[scarlet cave]

guys
is it normal that my 3200 MHz memory is working on 3200 MT/s
that shouldn't even be possible
is it a windows bug or sum
CPU Z says it is 1600 MHz but the memory minimum operational frequency is 2400 MHz
so like

DDR4 can't even go below 2133

https://tenor.com/view/emoji-gif-24664106

so visual glitch right ?

[willow pike]

1600mhz is 3200mt/s, that's DDR for you (double data rate)

2400mhz would be DDR4-4800

[tough owl]

https://media.discordapp.net/attachments/760302076608643102/1267606513917231195/image-56.png

[minor scarab]

first time I've used my satisfactory skills for something productive

routing pcbs is like connecting the belts in a factory...

and just like in satis i spent way too much time trying to make something tidy that i wont ever look at again

[minor scarab]

aand all the work was wasted because i fucked up

[minor scarab]

aand i did it all again

[languid gulch]

i love how all the tech channels are panic pushing the intel bios voltage limiters hard now

even before the microcode is out

[twin dew]

Like RyzenMan already said.
MT/s is what should be used, MHz is half that.
Just that lot of software and people use MHz when they should use MT/s.

3200MT/s (MegaTransfers per Second) is the highest JEDEC speed bin for DDR4.
Just that most sources talk about 3200MHz for that, when the actual frequency is half the MT/s for any DDR type.

.
.
And the AMD 9000-series "recall" and delay seems to have been pretty simple, IHS mismarkings labeling some of the Ryzen 5 and Ryzen 7 as Ryzen 9:
https://www.tomshardware.com/pc-components/cpus/amd-ryzen-9000-launch-delay-due-to-typo-ryzen-7-9700x-ryzen-5-9600x-confirmed

[dire igloo]

Ackchually, "Transfers" is not an SI unit.
You're still describing how often smth happens per second - and the unit for that is Hertz

[twin dew]

And the clock frequency is 1600MHz, just that there is bit transferred both on rising and falling edge of the clock.

Double Data Rate (DDR).

Why would it need to be SI unit to be used in specific niche?

All the actual spec documents for the DDR types talk about MHz for the clock and MT/s for the data rate.

And even QDR is a thing, where you move 4 bits per clock cycle.
Current GDDRs use that, RAMBUS RDRAM was QDR etc.

Just that marketing again got involved during the change from SDR to DDR and kept using MHz.

RDRAM marketing was even weirder, using Advertized Frequency that was usually double the actual frequency, half the MT/s (raw bandwidth), but in some specific speeds seems to have been just pulled out of their ass.

If the Wikipedia information is correct about that.

So yes, the data signal changes at up to double the rate the clock frequency, but it doesn't change at every point.

[dire igloo]

I read this as bambus (German for bamboo)

[languid gulch]

now i wanna see a tech company have spec they call DecaHertz per second

[willow pike]

this is cool as fuck

a practical application of satisfactory

[minor scarab]

actually, why use an ai auto routing plugin? write a program that converts PCB layouts to satisfactory save files and give them to satisfactory players they'll solve it for you for free

[willow pike]

intel quad pumped their FSB back in the day which made me do more math than I wanted to overclock it

[languid gulch]

i like how AMD's recall was because someone accidentally stamped the wrong model numbers on some

[willow pike]

people have done something like this! they gamified protein folding to help find new treatments

if we could harness the collective knowledge of every satisfactory and factorio player, we would surely build skynet

[languid gulch]

we'd probably just end up making amazon warehouses more hellish but efficient

[willow pike]

yeah

[edgy hazel]

[mental oriole]

when that reaches a week

[willow pike]

ah jaysus

[edgy hazel]

It exploded 🫢

[mental oriole]

[edgy hazel]

You know... Generally I'm against the death penalty. Its crude and doesn't help in the long therm in any better way.

BUT

whoever at creality had the BRIGHT IDEA to HOT GLUE the REMOVABLE CONNECTORS should be publicly executed on the town square