off-topic-tech

[dire igloo]

like a device between PC and wall outlet?

[tame gulch]

Displays power usage across home total, I just turned it off and it subtracted about that much

[dire igloo]

ah

[granite otter]

my whole system is probably around 350-400W while gaming

[dire igloo]

works decent-ish as a rough estimate

but it counts CPU, GPU, PSU, rest of system

(PSU meaning the inefficiency of power conversion)

not to mention all other peripherals and so on that are shut off with it

[tame gulch]

I think satisfactory pushes my pc the furthers out of all my games, so I should be good to run most things, and the optimisation update which I hope is coming down the line will make it even easier

[granite otter]

decent PSU will be close to 90% efficient these days

cheap ones might barely do 80%

[tame gulch]

Where do I view psu on task manager?

[granite otter]

its the power supply and its not visible anywhere

[dire igloo]

if you want to test your PC's cooling and/or power delivery capabilities, run Prime95 (small fft) and Furmark together for like 15m

[tame gulch]

Ah

[dire igloo]

you don't, you can't

[granite otter]

PSU=power supply unit 😉

[dire igloo]

that's why smart meter is a bad metric for detailed measuring

task manager also isn't ideal - use HWinfo

[granite otter]

at least you get the total with losses

[tame gulch]

Tbh I’m just looking the total drawn, which the smart meter is pretty decent at

[dire igloo]

which is entirely meaningless if you wanna troubleshoot GPU and CPU temps

[dire igloo]

and I literally just explained that that's a bad metric to use

[tame gulch]

Why?

[dire igloo]

#off-topic-tech message

[tame gulch]

Im not talking about just the gpu

I mean like the entire pc

[granite otter]

the total power draw is the important thing either way

its what you pay for 😛

[tame gulch]

And I know that the gpu is some fraction of the total, so as long as that’s low then the gpu power is low

[granite otter]

and most of that power is dissipated as heat as well

[tame gulch]

Yeah, and as long as it’s not literally melting/boiling stuff it’s mostly fine right?

[granite otter]

well sure 😛

[dire igloo]

if only your PC draws 350W watts from the wall and you have a 700W Gold rated PSU, that means that about 35W don't actually make it to the components.
And as I just explained, those 350W include other parts that aren't actually relevant to the discussion of component temps

[charred willow]

how many watts is the PSU in question?

[granite otter]

how long it will last is a different question if its riding on the max all the time

[dire igloo]

the only power not dissipated as heat is converted to either light (LEDs) or movement (fans).
So it's safe to say that >90% of all power drawn is converted to heat

[dire igloo]

and that's why it's a bad metric to use

[dire igloo]

weren't you having uncomfortably high temps?

[granite otter]

for sure...

[tame gulch]

nah i was just wondering why is was 80c when normal is like 60c for other games

[granite otter]

probably because the GPU is fully utilized

but a 20C difference is weird to me

all my games will see my GPU at roughly same temp

[charred willow]

or different parts of the gpu are more/less utilized by each individual application. true for cpus in a different way, but its not just cores bound together by magic smoke

[dire igloo]

hence this #off-topic-tech message

[rustic panther]

80c is perfectly fine AFAIK

[charred willow]

80c is plenty perfectly fine. might throttle a bit up there, depends on the card, mfg settings, and if its already throttling to be at 80c

[tough owl]

80c is cool for a laptop too

[maiden coyote]

Indeed, whatever law of thermodynamics.. energy can not be created or destroyed, only converted from one form to another

Your computer does zero work, so all energy consumed by the computer must be converted into heat

[dire igloo]

or light or movement - which there is some form of

[maiden coyote]

I think semiconductors actuality transfer the majority of it to ir light like you said

[dire igloo]

isn't IR light just heat?

[maiden coyote]

No xD

It is once its absorbed by something

But until then it's just light.

[twin dew]

When there is will, there is a way, even for something completely bonkers...
https://github.com/codyd51/uefirc

[maiden coyote]

That goes into black body radiation which says that everything produces light according to its temperature

[twin dew]

Should I use this?
This should not exist.

[maiden coyote]

So a glowing piece of metal glows because it's hot, the color that it glows at corresponds to its temperature. On blue LED's they "cheat" because the electrons jump from a very high energy level to a much lower one releasing blue light without actually being 10000K

[twin dew]

Well, in all LEDs

[maiden coyote]

and everything else that's a semiconductor like your cpu.

[twin dew]

Also fluorecent lights etc.

[maiden coyote]

Same thing applies. Light = energy released same as heat and while the two are very connected they are not the same

[twin dew]

Different way in those to LEDs, but still not direct temperature based emissions.

Where as incadescent lights work via pure heat.

[maiden coyote]

But if you have heat, then you have light. It's the same concept with electricity and magnetism

[dire igloo]

why's my PC dark then?

(joke question, I know it's cuz it's just not hot enough)

[twin dew]

I like my stuff to not have black body radiation in visible light spectrum.

[winged valley]

Anyone here know if it's possible to pull a panic log off of an ipad/iphone while the device is powered off? (Battery appears to not be charging, thus can't enter into DFU)

All Things Repair discord is slow to respond at these hours

[winged valley]

On an unrelated note

I'm upgrading my boot drive from a 2tb sn580 to a 2tb 990 pro. Aside from raw speed, the biggest difference is no dram vs dram

I don't expect this to be a super noticible upgrade, but I want more storage space

I was given a 200 amazon card so decided why not

[dire igloo]

The difference in its purpose of being a boot drive is none

[dire igloo]

Because you can almost get 4TB for that money

[winged valley]

I needed to get other things too not just the ssd

[dire igloo]

990 Pro pricing sucks ass

You can get MSI M482 for $100 from MSI directly
That's what you're competing with.
DRAM vs HMB is a non-factor for NVMe OS use

[winged valley]

Have you not seen the sales its on

[dire igloo]

Actual sales or just pseudo-"sales"?

[winged valley]

Nah the 990 Pro was like almost $300 for 2TB for a long while but the past 1 or 2 months the price on all skus has been way down

[dire igloo]

"way down" as in "competitive with the actually sensible 2tb competition" or as in "not $300"?

Cuz you can get high end 4tb for like $230

[winged valley]

Somewhat competetive

Okay show me a high end 4tb nvme for 230

[dire igloo]

and competitive 2tb pricing is anything below $130

[dire igloo]

https://pcpartpicker.com/product/rGjRsY/transcend-250s-4-tb-m2-2280-pcie-40-x4-nvme-solid-state-drive-ts4tmte250s

MP44 is $233

Cras C910 is only $200, but that's more midrange

[winged valley]

I have never heard of Transcend in my life and I've heard of a lot of weird brands

Cras???

[dire igloo]

I would mention VP4300 Lite, but that one's been swapped to QLC

[dire igloo]

SSD family of Klevv, Hynix-affiliated brand

[winged valley]

So what its just another drive with a phison controller and hynix nand or something

[dire igloo]

Sounds like a you problem tbh, fairly popular in Europe.
Uses SM2264 which is a Phison E18 competitor and also has DRAM

[dire igloo]

Innogrit actually for the C910, IG5220
Not plagued by the IG5236 issues

And yeah, Hynix of course

And MP44 uses E27T+Bics6

Samsung is just too expensive to be recommendable. The only exception I've seen this year is 1TB 990 Evo costing 66.48€ in France

[dire igloo]

You can just set a vendor preference for Amazon, then scroll through the options:
https://pcpartpicker.com/products/internal-hard-drive/#sort=price&A=2500000000000,24000000000000&D=1

And you can use this to check the internals as well as potential part swaps:
https://docs.google.com/spreadsheets/d/1B27_j9NDPU3cNlj2HKcrfpJKHkOf-Oi1DbuuQva2gT4/htmlview#gid=0

[meager turret]

hey folks, have a question...
I want to play with some mode (start from scratch is really long)
Is there any way to download update8 instead of 1.0 on Epic?

[low scaffold]

Hello there fellow Pioneer!
You're currently chatting in off-topic channels.

Please note, the entire off-topic category is reserved for not Satisfactory related chattery. To comply with your FICSIT Inc. contract, please make use of the appropriate channels in the Satisfactory category. Thank you!

Command executed by: fireworker2000

[tough owl]

I love nationwide carrier outages

[twin dew]

And seems my Radeon 6600 would need repasting.
Hotspot hitting near 100C when main GPU is just below 70C...

IIRC originally 15C difference.

Bought 2024-02-10, with 3 year warranty.

[twin dew]

So I need to make it into manufacturers issue at some point, via the shop it was bought from.

[sage pier]

Verizon?

Is it with 5g or lte or both?

Cause im on lte rn and its fine

[tough owl]

Both

No SMS or calls

I didn't have LTE or 5g up until 10 minutes ago

And it's gone again

[jagged snow]

Yeah, ars ran an article on it

AT&T is down as well

[tough owl]

Not for me 💀

Back for some people

[sage pier]

It was down for like 10min for me

[tough owl]

wccftech.com/dimensity-9400-first-chipset-to-cross-3-million-in-antutu/ this is impressive, I wonder if it had active cooling or what device it was in.

even if theyre cheating in benchmarks

[sand saddle]

Until they tell you that anything below 115 is within spec

[twin dew]

Will see.

[twin dew]

The plus side of having to deal with a store with competent technical people and not direct with the GPU manufacturer.

[languid gulch]

"so long as it doesn't melt or delaminate, it's fine!"

[soft bloom]

Curve settings? Additional heatsinks? Underclocking?

[twin dew]

At stock with Zero RPM mode disabled.

At idle both are same.
But under load it turns into that 30+C difference.

Classic symptoms of paste migration causing bad partial contact.

[soft bloom]

Did you check at max rpm, if the difference will be negligible?

[twin dew]

As that difference has also grown over time.

[soft bloom]

Oh. Not great

I can say that i have had difference on my radeon gpu, and making curve for higher area, and setting lower power budget helped (you saw results)

In default it can easily go over 80C

[twin dew]

Yeah, of course more cooling and less heat helps.
But still indication that the paste is going bad in between GPU die and heatsink.

Very common with GPUs.

[soft bloom]

Also i put small heatsinks on backplate where memory is, which is darn close to chip, so maybe that also helped to cool backplate overall...

[soft bloom]

Sure. And to prevent paste degradation (after repasting): "more cooling and less heat" 🙂

[twin dew]

But I can easily ask for warranty at some point, and if unsuccesful, just changing the paste to TPM.

[twin dew]

No, just using right material.
Most normal pastes don't really work for GPU.

And the base cause is uneven and too low pressure heatsink to gpu contact.

Where the unit to unit differences cause large effects.

But the point was, I could easily fix the issue myself for 10e permanently.

But doing that would reward the GPU manufacturer.
Point of the RMA is to make them pay as much as possible, so they would start using better, but little more costly, thermal interface material between the die and GPU by default.

And not the cheapest paste they can find that seems to do the job in large enough number of cases.

Just replacing it myself would be the easy option for me too.

[soft bloom]

How does quality of paste application affect its longevity?

[twin dew]

Not much the application way.
Just the paste chemistry, pressure and the surfaces.

[soft bloom]

Well, pressure would be part of paste application in my book, and surface affects pressure.
Aka - end state of paste after you assembled gpu.
Cause to me it seems like what only matters is how much heat it conducted.
Pressure and even contact can affect local behaviour: how much het goes through one region over the other.
And, probably, if some area has worse contact - it will heat up more.

I remember Steven made some experiments with pressure a few years back, but I didn't follow on more thorough analysis

[twin dew]

Less pressure, more the distance between the die and heatsink change on heat cycles.
And the pressure with GPU mount is determined by the flat spring used.

The screws are made to bottom out before that is not relevant anymore.

And then the various tolerances stack up to determine the actual pressure in that unit.

Only way to change that would be to do some washer mod between the flat spring and the screw heads, or the springs on the screws and the PCB if not using a flat spring and just individual coil springs.

And that is dangerous with direct die, as you can crack the die with too much pressure or uneven pressure.

[languid gulch]

i'm still kinda impressed with my 6800XT. 78C hotspot on SF max settings

tho that might be my 5600X bottlenecking it

[pure karma]

last time i checked hotspots i saw 97 and then proceded to close HWinfo and nerver look at hotspots again

[sage pier]

I used to see 115c

[twin dew]

Hadn't been monitoring, but needed to check FurMark for reference to another user and noticed the current mismatch.

[sage pier]

I use 3dmark's timespy

[soft bloom]

Nothing little tweaking can't improve.

Ok, honestly, i think companies should start sharing temps and power for underclocking.
Loss is negligible while savings are noticable.
Would really like to see shift towards energy efficiency

[gilded helm]

Consumers already have that power and it's pretty easy. It's just that nobody cares. If they release with lower default power targets, they'll just get bad reviews.

[languid gulch]

oh definitely. i did try overclocking it, but all i got was like 3-5% improvement for like 15% more power consumption

[edgy hazel]

I'm on like 80-something

[languid gulch]

i may have gotten sick of some of the weird ramping up & down of my cpu fans, so i just cranked my case fans up

[pure karma]

just checked im at 58-63C on the hotspot at max settings

[languid gulch]

i'm also kinda pushing it at 1440UW/144Hz

[pure karma]

im at 1440P 165HZ but im barely getting 70fps and my 1% lows are probably as low as 20fps because build mode is unstable AF now

[languid gulch]

oh btw get ready for silicon prices to spike

Sibelco shut down because of the hurricane

it's one of the purest quartz mines in the world

[stray badger]

we truly are a very effective species

[soft bloom]

not asking to change target power targets, but run benchmarks at them and publish them, so it will also be compared.

[pure karma]

another day another dollar

[pure karma]

the problem with that is people including myself dont care about power efficiency unless it also comes with a significant performance gain

and that really has shown with ryzen 9000 so far

[soft bloom]

cost of power trippled for me in last year, so it's like a yet another gaming subscription

[tame gulch]

definitely dust, cleaned it off and now we chilling at 50c

[rustic panther]

Nice!

[tame gulch]

tho the sinks do look like this if i want to afk satisfactory while doing anything else lmao

[low scaffold]

Hello there fellow Pioneer!
You're currently chatting in off-topic channels.

Please note, the entire off-topic category is reserved for not Satisfactory related chattery. To comply with your FICSIT Inc. contract, please make use of the appropriate channels in the Satisfactory category. Thank you!

Command executed by: hergarian

[tame gulch]

bruh i mention it once while its in relation to my pc temp lmao

@edgy hazel

[edgy hazel]

😡

[pure karma]

the satisfactory awesome sink has nothing to do with pc temps

[tame gulch]

This was talking about how i had to make the game pretty much look like that to afk without making my temp skyrocket. It did, in fact, have something to do with my pc temp.

[soft bloom]

Ok so i tested that "gentle heat" doesn't damage PCM plastic bottle (pot of water in a pot of water)
at the very least that gives me method of carrying heat from gas stove to myself for the time when it will be the only utility available

[winged valley]

Satis is heavily CPU bound and CPU puts out a lot of heat in it

[cunning glacier]

Surprised at how well it runs on max settings in 4k vs my 4090 tho

Game has great LOD to compensate for huge map

Literally just stresses cpu and I run 7950x3d and it handles the load well

Maybe 40% max

Still a shit ton but not melting apart

[cunning glacier]

Probably dust issue

Tbh

[tame gulch]

@cunning glacier

[twin dew]

"Nice"... Intel has moved their CPU documentation/datasheets behind login wall in last week or so.
The document that contains the actual max memory speeds for their CPUs etc.
When they just tout the "Up To" for the best case scenario elsewhere.

12th gen version is still open.
13th/14th gen wants Azure Login.

[soft bloom]

BS

why on earth they do that for not shady reasons?

[twin dew]

Very shady, when those docs have been public for long time already, and the older versions still are.

Just all the revisions for 13th and 14th gen are behind login wall now.
Even the older version that was 13th gen only.

[languid gulch]

why provide something for free when you can paywall it

[twin dew]

If it had been behind login wall always, then that would have been that
If they had moved all of them behind a wall, not just the latest problematic CPU models, then that might have been legit reason.
But that they only moved those two... Somethings not right.

And when they don't publicly otherwise say some specs that are actually relevant, like those RAM maximums depending on MB RAM slots, DIMM amounts and ranks...

Then it becomes shady as fuck.

[glossy glacier]

Its still available though?
https://www.intel.com/content/www/us/en/products/docs/processors/core/core-technical-resources.html works fine for me

[twin dew]

There is/was HTML version.

Lets see if that contains the same stuff or not.

Yes.

Page 112 of the Volume 1.

But the html version under edc.intel.com is now behind login wall, but only for the 13th and 13th+14th gen versions.

12th gen:
https://edc.intel.com/content/www/us/en/design/ipla/software-development-platforms/client/platforms/alder-lake-desktop/12th-generation-intel-core-processors-datasheet-volume-1-of-2/010/processor-sku-support-matrix/

13th & 14th gen:
https://edc.intel.com/content/www/us/en/design/products/platforms/details/raptor-lake-s/13th-generation-core-processors-datasheet-volume-1-of-2/processor-sku-support-matrix/

The links are still in Google, as they used to work.

[dire igloo]

remember to not look at total CPU load, but per-core load

[twin dew]

Remember to set the game as a game in Game Bar.
Not on by default for Satisfactory.
And the game just loves the cache-CCD, but is running on the frequency-CCD by default.

[pure karma]

thats fun until you realise how disapointing multicore load spreading has goten where its more common to have one core at 100% and the others at 5% than to have all cores within a 10% or so utilization diffrence of eachother because there actually all contributing

[dire igloo]

that's definitely a scheduler issue - and can be influenced by many factors

however, it's a clear cut indicator that a CPU with more cores isn't gonna benefit you meaningfully (example: going 5600X to 5950X)

[pure karma]

well yea no but it does on older games

but the amount of care and effort put into good optimisation and load spreading has gone down significantly

[glossy glacier]

spreading single-thread tasks over multiple cores is not good.
way to much performance lost by context-switching, cache misses and moving data

spreading actual different threads over multiple cores would be neat, wouldn't it using-the-two-best-cores-for-everything-windows

[dire igloo]

I can see why it's beneficial on paper to have two single-core tasks run on two cores instead of one, but idk how that'd mess with things - and whether the performance gain outweighs the performance cost of scheduling

[rustic panther]

Also, race conditions

[dire igloo]

https://tenor.com/view/peter-griffin-hoyoverse-genshin-impact-family-guy-mihoyo-gif-3332330561976006572

race condition

[mental oriole]

[mental oriole]

My custom game engine can get 100% utilization on all cores.
It's a skill issue

By 100% it's more like 95% max because kernel will always have highest priority.

[twin dew]

Busy-waiting and spin-locking all the time?

[mental oriole]

Processing 60 messages/ second to 400k subscribers (update logic)

in debug mode so performance can probably be better 🙃

[cunning glacier]

sowy uninstalled the bloat long ago

[twin dew]

Needed with dual-CCD X3D:s.

Because that is what works with the X3D driver to sleep the frequency-CCD and force games to the cache-CCD.
Or you have to use something like Project Lasso etc. to do it yourself.

[cunning glacier]

Ive never had issues with it so why even bother

[jagged snow]

If you have a dual-ccd x3d chip you're leaving a lot of performance on the table without game bar

[cunning glacier]

I have it disabled lol

Im not missing any preformance because there is nothing that loads my cpu enough because my 4090 is the bottleneck

lol

[jagged snow]

That is very untrue

[twin dew]

Basically you get absolutely no benefit from the cache-CCD without either manual tuning or using the Game Bar.

[jagged snow]

You're likely introducing stuttering and/or drastically hurting your 1% lows, even if average frametimes are still acceptable

[cunning glacier]

I just got it for the hell of it im not really looking for preformance

I still play quake

I dont even play present aaa games

i just like building cool puters with beefed specs

[twin dew]

And various simulation, turn-based, factory games etc. are what benefit from the cache-CCD the most.

Not FPS games.
But some like the latest Jedi thingie did benefit significantly.

[cunning glacier]

Sounds like a minmaxxing issue

[twin dew]

Getting the performance you paid for, in most cases by just leaving that Game Bar enabled?
And in specific cases having to add one extra tick into it while the game is running?

Pretty small thing in my mind.

[cunning glacier]

It is but I already disabled that stuff, so i dont really care about it

Didnt buy the cpu for preformance, just for fun

I did buy my 4090 for fun too even though i still play games that are well dated by now

[leaden kraken]

what cpu u got

[cunning glacier]

7950x3d 16 core

[leaden kraken]

that ur saying "the 4090 is the bottleneck"

the 4090 is not the bottleneck

[cunning glacier]

depends on what you do though

[leaden kraken]

true also depends on game

[cunning glacier]

i dont run software and shit that stresses either

[jagged snow]

That is not a blanket statement that can be made with any degree of accuracy

[leaden kraken]

maybe in like cyberpunk yeah but if u play shooters, as u said u play quake, thats more on the cpu

[cunning glacier]

yeah but like

2% xdddddddd

[leaden kraken]

i mean tf would u get if the 4090 is the bottleneck?

rtx a6000? or whtv that like 40k$ one is called

[cunning glacier]

threadripper

[leaden kraken]

im ultra confused now lmao

[twin dew]

Up to 100% in some games when they run on the Cache-CCD and not the Frequency-CCD.
And up to about 20% in AAA games, outside the ones like that Jedi Survivor that needs the Cache-CCD to not stutter.

[cunning glacier]

i
dont
play
aaa

🫠

[twin dew]

And the AAA is the least affected segment?

But that even in that part, the gains aren't "2%"

[cunning glacier]

half the games I play are server bound, im literally playing eve online rn

[leaden kraken]

tbh a 7800x3d could probably have been better unless ur also using ur pc for multitasking

[cunning glacier]

i mean like unless you count like running multi accounts on games then

kind of ish

but thats more ram issue

[leaden kraken]

what ram u got?

[cunning glacier]

48gb dominator @6000mhz

[leaden kraken]

5k?

[cunning glacier]

i have a nxzt b650e and it only allows 6k

[leaden kraken]

ah 6k

[cunning glacier]

I got the board because its pretty but its more of an apple product with how its designed than anything

[leaden kraken]

yeah was about to say 5k is too fast to be ddr 4 but its like rly dogshit ddr 5

why 48 tho thats such an akward number lmao

is it like 12 gb sticks?

[cunning glacier]

24x2

i mean there is no difference

you can have whatever size ram

[leaden kraken]

theres a small difference

2 sticks is a bit faster thn 4 sticks

[cunning glacier]

in that case

i mean the size xd

[leaden kraken]

but thats some really minimal nerdy difference that a casual player wont notice lmao

i mean size doesnt matter (i hope lmao)

[cunning glacier]

it was 300$ at the time and I already was like 5000$ knee deep so i didnt want to make the build more expensive

[leaden kraken]

if u only use like 40 gb theres no reason to upgread to 64 gb, the only upgradability is the speed and cl rating
if u use all 48 gb thn u have to upgrade

[leaden kraken]

fair

[cunning glacier]

I would go for 256gb of ddr5 at some point

because I thought it would be funny

I was gonna try doing 2 4090s but then I remember SLI is kind of dead and only community managed now and i did watercooling so its means +1 waterblock

only good reason is if im running vr and there is community driven SLI for VR

[leaden kraken]

jesus 😭

buy me a 4090 ffs im having to survive on a 1050 ti lmao

[cunning glacier]

buy 1080ti

[leaden kraken]

no.

[cunning glacier]

:3

[leaden kraken]

actually its 2 1050 ti-s

but 1 of them is mostly idle cus i play cs mostly lmfao

[cunning glacier]

hey its still somehow good

[leaden kraken]

1080 not worth

its basicly a 3060 that consumes like 2x the power

[cunning glacier]

I thought it was still a very much viable card and now its 200$

[leaden kraken]

and doesnt have new dlss or ray tracing

[cunning glacier]

dlss bad only good is ray tracing

[leaden kraken]

dlss is not bad....

also id rather go amd

[cunning glacier]

its very on edge, I think

[leaden kraken]

probably gonna get a 7700 xt

[cunning glacier]

like only good in some cases

[leaden kraken]

fsr>>>dlss

[cunning glacier]

others you notice how much shittier makes games look

FSR yeah kinda better

I can agree on that

[leaden kraken]

im using fsr rn

cause its avalible on most gpu-s

[sage pier]

dlss is better. fsr is just more available to most gpus

[dire igloo]

DLSS looks and performs better than FSR at equivalent settings

XeSS is also widely available, gives less of a performance boost than FSR but maintains more of the image quality - also, it seems to be the only upscaling tech that doesn't visually remove items from Satisfactory conveyor belts

tho for the sake of the discussion: don't get a GTX 1080 unless it's dirt cheap

[pure karma]

people only recomend it for the Vram there is not real reason to get it other than that because its power efficiency is awful

[cunning glacier]

tbh if on budget you just run that or buy refurb 30 series

i dont mind getting refurb cards since sometimes they have small issue you can fix or something that isnt huge like coil whine

[languid gulch]

it's fun to turn all the settings on SF down just to see what your metrics are

i think it's like 1C above idle 🤣

still sucking down 12gb vram tho

[gilded helm]

Jealous, I can't get Satis to use more than 5GB

[languid gulch]

oh i've maxxed it out

[gilded helm]

You've got way more oomph than my laptop though, the game is probably deciding it can't actually manage all the things it could otherwise put in VRAM.

[languid gulch]

oh yea laptop ain't gonna compete 🤣

[pure karma]

one of my regrets is not going up to 64GB of ram when i upgraded

[tough owl]

I've never maxed 32gb

[tribal kraken]

Couple of my assemblies with SolidWorks give low RAM warnings with 32, can work around them with reducing non needed detail but also more RAM saves time to not spending time on that

[rustic panther]

That's a professional usecase though

[tribal kraken]

Makes no difference to me, workshop PC has 32GB, home PC 64GB. Use the both for same work, but no gaming with the workshop PC, which I'm on atm.

[pure karma]

i have many times

[stray badger]

me neither

[soft bloom]

is it odd that whenever i need to annotate object that has other objects without any specific indexes/labls assigned to them of limited size, i still use Iterable even though I don't actually expect it to be infinite?
Like, i don't care if it's list, tuple, array, set etc - i just need to iterate over items.

okay, i have the wildest question: what if i implement object presents itslf as having infinite length?

[mental oriole]

[soft bloom]

ok, so it can be done, but will raise error on execution.
it needs some type checker to be detected before execution

[mental oriole]

Just cast everything to list?

[soft bloom]

i don't like that idea. first of all because list is mutable and costly to create

but mainly because it doesn't have to be list

i like iterators. but i don't see how to properly annotate most common kind of them - finite iterator

Sized Iterable is the closest i can get.
but having Sized means i need to provide method that measures size and thus - goes over iteration.
Mostly it's not a problem but could be sometimes and i am not used to think about size that often

[mental oriole]

You're using python...

It's slow by default (:

Use type checking, and last time I checked you can just enumerate all of those with for x in mycontainer

If you have maps/dict in there its nore tricky.

[soft bloom]

well, exactly.
but it's not about implementation in first place, as about annotating type

cause using tuple/list everywhere is exhausting
i prefer for and yield

[mortal stirrup]

@dusk scroll I'm guessing you're in NA, how much does the 7800 go for over there? I get the feeling it's probably more expensive where I live

[gilded helm]

I will say, don't sleep on Alder Lake. I think I've seen absolutely bonkers deals on the 12700K.

[dusk scroll]

@mortal stirrup you guessed wrong, I'm from the EU 😛

[mortal stirrup]

Ah lol

Sale prices there are probably similar to NA though right?

[dusk scroll]

@gilded helm no point if gaming is the main focus, the 7800X3D beats the 12700k by far

[mortal stirrup]

I do a whole lotta stuff on my computer but usually gaming yeah

[gilded helm]

Oh, there's no dispute that the 7800 X3D is the better CPU. But price is a factor.

[dusk scroll]

@mortal stirrup usually you just substitute the dollar symbol with the euro symbol, yes 😉

@gilded helm you have to consider the price of the whole platform. Also I'd happily pay 200$ extra to get ~40% more performance in CPU bound game scenarios.

[gilded helm]

The 12900K was $210 on Amazon (USA) some time in the past few months. That's the deal I'm thinking of.

[dusk scroll]

for productivity work sure, for games nah, I'd get the X3D

[mortal stirrup]

It's definitely not that cheap in my country xd

Seems to be around 357 USD on Amazon rn

[gilded helm]

Right now according to PC Part Picker in the USA, 7800 X3D is $575. There's no disagreement that it's a fantastic CPU. But that's really, really expensive. And maybe that's worth it for a serious Satis player, and I'd totally understand. I'd have one now if I weren't living a more mobile lifestyle, I'm stuck with gaming laptops as my option.

[dusk scroll]

@gilded helm that's why I would wait one month and buy it used once people upgrade to the 9080X3D

[mortal stirrup]

I checked on Amazon and it's $450 rn

The 7800X3D that is

[dusk scroll]

yeah it's 450€ new here in my country

[mortal stirrup]

Sweet, rare instance of PC components not being ridiculously overpriced here

[gilded helm]

We don't even know that the 9800 X3D is going to be that good or create big discounts on the 7800 X3D. For all we know 7800 X3D prices will go up. All it would take is AMD continuing their habit of stupid day 1 pricing on the 9800 X3D and it being a marginal improvement over the 7800 X3D.

[mortal stirrup]

Yeah I don't intend on waiting for AMD to do anything

The only thing I might wait for is a holiday so that I can rebuild everything in peace

[gilded helm]

7800 X3D prices already have gone up as it is.

Meanwhile the 12700KF is actually $166 according to PCP. You need to put together the whole platform cost, obviously, but that's cheap enough that it's not worth immediately dismissing as an option for buyers in the USA.

5800 X3D is $449, in comparison.

[dusk scroll]

I'd only buy the 12900k/12700k for a good deal with the intend of overclocking and pairing it with fast RAM AND my workload also included productivity work. For just gaming, X3D all the way. Easier to set up, doesn't require fast RAM, less power consumption, more performance in most games.

[mortal stirrup]

I need to get familiar with overclocking

[dusk scroll]

and that is coming from someone who has had an intel system for the last years with 12900k -> 13900k -> 13900ks custom builds with max OC and delid

[mortal stirrup]

Haven't overclocked anything other than my old GTX 1060 ages ago when I wanted to squeeze out a few more frames

[gilded helm]

Most stuff is pushed nearly as hard as it can be out of the box these days anyway. Gains are there to be had, but it's honestly not a big deal to ignore it.

[dusk scroll]

@mortal stirrup it's not that complicated if you are happy with ~80% of max OC. Squeezing out the last bits is where it gets super tiresome and time consuming. For modern Nvidia GPUs undervolting is actually the meta. 100W less power for ~7% less performance

[mortal stirrup]

Is that just to save on power?

[gilded helm]

Yeah, undervolting for stock performance is the way to go. Same performance (or even slightly better), and less power and more longevity.

[dusk scroll]

@mortal stirrup less power, less heat, less noise

my 4090 with @875mv consumes 100W less and it basically quiet while gaming

[mortal stirrup]

How could performance become better?

[gilded helm]

The GPU is running cooler, so it can use dynamic boost more.

[dusk scroll]

@mortal stirrup avoid thermal throttling

[mortal stirrup]

Hm

[dusk scroll]

modern GPUs and CPUs include TVB (thermal velocity boost) algorithms

if you hit TJmax it downclocks

[mortal stirrup]

Maybe I will look into that then, I don't really mind the extra power draw/noise/whatever on stock voltages but if there is a performance gain to be had then why not

[dusk scroll]

that's by the way the only method to overclock an 5080X3D/7080X3D after they voltage locked them

[gilded helm]

Undervolting is also good for longevity. High voltages are a way bigger killer of silicon than high temps.

If you ever get a laptop, it's also a way to squeeze some more battery life.

[dusk scroll]

@mortal stirrup make 2 profiles in MSI Afterburner for your 3080, one for undervolt and one for overclock

875mV has been the sweet spot since my 1080. Worked for 3080 and my 4090 too:

almost same performance, but way quieter and less power

[mortal stirrup]

I reckon I'll get it set up after I complete the upgrade, seems that's pretty much gonna be a clean slate for my setup due to almost everything being replaced

I appreciate the advice from y'all

[dusk scroll]

can't wait for the 600W memes about the 5090

when in reality it will be more efficient FPS/W wise than the 3000/4000 series

[gilded helm]

Steve from GN made a really good point about the supposed 600W rumor. It turned out to be bogus with Lovelace too, and it stemmed from confusion about a 600W number being communicated to have coolers be designed for up to that much.

[dusk scroll]

yeah that was for the 4090, with the 5090 I expect it to be actually 600W this time

I enjoyed having an oversized cooler on my 4090 due to that miscommunication though, never had such a fast and quiet card

[gilded helm]

Fundamentally all these architectures need to be designed for laptop anyway. So even if they push the desktop versions crazy hard, the sweet spot in their power curve will always be toward 100-200W.

[dusk scroll]

it's more in the 250W ballpark for the 4090

after 350W you see little gains

flashing a 600W BIOS on a 4090 to see 2% performance improvements at best in synthetic benchmarks... yeah no thanks

[gilded helm]

The top chip in each generation theoretically can be designed for all out power, due to being too big to go in laptops in any circumstance. So it's always possible the 5090's underlying chip might have a 400W sweet spot or something bonkers. But that's be some weird inter-achitecture power scaling.

[languid gulch]

new idea: stuffing my pc into the room's A/C return duct

[twin dew]

https://www.youtube.com/watch?v=xIHg-PPUZnk

[twin dew]

But at least Vulcan has modern guidance system that can actually self-compensate and isn't just flying preset.

[stuck nebula]

amd is going to change alot of stuff with 9000 x3d because 9000 series or am5 in general doesn't sells enough

they said that recently, if you are going to buy a new am5 system, wait for it

i'm still playing with my 7600x to be totally honest, i didin't felt the need to get a x3d yet

[vital prairie]

anyone a discord nerd, i just cant get my mic to work with discord

[twin dew]

Blocked in Windows privacy settings?

[stray badger]

I think my 7950x3d will be good for while

[edgy hazel]

I think my 5800x3d too 🙂

[gilded helm]

By the time those CPUs wear out, maybe we'll finally have a use for these NPUs that are insistently being included with new ones. And then you'll actually know what you're buying. 😛

[wanton orchid]

it's simple math : it's easy to do powerful npus
and it's marketing argument
so they all do it
after calculating wafer space / marketing advantage

simply do not evaluated a soc by the size of its npu

[stuck nebula]

it seems that... Frame Gen is working, this is with frame gen off, global illumunition 4(cinematic) + Full Lumen(r.Lumen.ScreenProbeGather.ReferenceMode=1)

[low scaffold]

Hello there fellow Pioneer!
You're currently chatting in off-topic channels.

Please note, the entire off-topic category is reserved for not Satisfactory related chattery. To comply with your FICSIT Inc. contract, please make use of the appropriate channels in the Satisfactory category. Thank you!

Command executed by: kosummer

[soft bloom]

if you need help: #1038092680493801533

[stuck nebula]

and with frame gen on,

but i couldn't get to hardware lumen work, it literally changes nothing on fps impact or gpu load

[stuck nebula]

why would i need help 😄

i'm demostrating that frame gen is actually working, i've been searching the version of fsr 3 since 1.0 but it seems we got the 3.1 one

[soft bloom]

and you are not reading purpose of this channel at all

[stuck nebula]

it's about UE5 tweaks on Satisfactory

[soft bloom]

the hand that will slap your face for not reading till the end of channel description is already beyond speed of sound so just relax and make sure i hear the sweet sound of smack

[stuck nebula]

omg, dude this is UE5 tweaking through engine files and demostrating features that is not in options that can be enabled through some text of config files

it's not neccesarily "satisfactory stuff"

[pure karma]

your using satisfactory for the example wich is a instant riot in here so just dont....

[wanton orchid]

you are talking about a satisfactory tweak by writing in satisfactory system files
for satisfactory feature purpose
how is that off topic ?
#satisfactory please thanks

[stuck nebula]

alright alright my bad, sorry for not typing in a proper tech channel that exists under satisfactory!

oh wait!

[soft bloom]

#open-a-ticket

[edgy hazel]

Why would i want to take a screenshot of my watch

Here is the time I guess

And I don't think I even have the option to set a custom bind for pressing both buttons at once. So it's just gonna be a screenshot key I will use so very often

[dire igloo]

eh, if it's general UE5 shenanigans that could be done on anything UE5, I'd say fair game
if it's specifically SF and you wouldn't do it on other UE5 games even if you could, that's a no

[gilded helm]

My recollection is moderators have asked advanced technical Satisfactory stuff to be in #math-and-meta

[twin dew]

The QA site serial misinformer strikes again with absolute garbage:

[gilded helm]

What in the heck is that insanity? I did leave the game running for over 28 hours, and I think that's the point at which you get a brief reset, with very clear on screen warnings.

[twin dew]

That is person who spews very random garbage on the QA site constantly.
That was just on the most insane end.

Sometimes with small nugged of sanity inside the completely weird stuff.

And the point is, if the system is actually stable, it can run for months on end without any problems.
And he was somehow mixing that needed DRAM refresh into needing user action for it to happen...

[gilded helm]

Yeah, the thing about needing particular memory is amusing

[twin dew]

Somehow mixing ECC into the mix, and no idea where that 11 hour limit is from...

[rustic panther]

It makes perfect sense though! Humans need sleep, so give your computer a little nap too! Some human beings have good memory so it makes sense that a computer also can have good memory!

Jokes aside I think someone needs to look into their PC (Or stop being 7)

[wanton orchid]

or stop thinking they know it all when they just did about a chatgpt prompt in their whole life about it

"but I asked on Twitter" kind

[mental oriole]

I swear LLMs are making people dumber 🤣

[rustic panther]

Nah, ChatGPT doesn't tend to hallucinate that badly

[wanton orchid]

either 7, seeking attention, or thinking they are worth "helping" everyone

[rustic panther]

This to me sounds like someone having a PC with pretty bad hardware issues or someone who's literally 7 years old getting a scare story from their parents to get them to turn off their PC at night

[mental oriole]

People dont turn their machines off?

[wanton orchid]

that would not necessarily translate to online "help" though

[wanton orchid]

unfortunately yes

[mental oriole]

💀

[rustic panther]

Yep

[wanton orchid]

Also worse since Windows does not really reset on shut down / power up

[rustic panther]

Oh their dad told them that, so it must be true!

[wanton orchid]

I've seen 150 days system runtime in task manager

[mental oriole]

Ofc. Daddy is always right...

[mental oriole]

Had 350+ on my laptop, but I never had issues with it. (Hibernation still counts on uptime somehow)

But I always shut down my tower when not in use, because power bills lol.

[wanton orchid]

[mental oriole]

[wanton orchid]

the problem is not people being dumb

the problem is people getting vetted while being dumb

the "it's not his fault" kind of stupidity
striving to prevent people from ever improving themselves

pushing the blames on random somebody else

and that without even asking sometimes

it pushes people into the mc syndrome

[mental oriole]

Yup.

[glossy glacier]

Hibernation is sleep-to-disk.
From the point of view of the OS it never got rebooted

[mental oriole]

I know that, just didnt expect the timer to still count :))

[soft bloom]

I would just ask them if they reboot smartphone as often

[glossy glacier]

Windows counts from the last kernel (re)boot. So all sleep modes and shutdown with fastboot enabled don't count

[stuck nebula]

woah, what kind of L3 Cache this guy's cpu have to store the whole game inside of it

he must be living 100 years ahead

[twin dew]

Where did you pull that L3 cache from?
SRAM doesn't need refreshing.
DRAM needs.

[stuck nebula]

i'm only focusing on the storing the data part

[twin dew]

Staying in memory?
Which applies both to RAM and cache?

[stuck nebula]

yeah, what i'm saying is that it's wrong from the start, the whole game dosn't load once and stay in the memory as he says

it depends on the size of L3 cache

[twin dew]

No?
L3 cache doesn't come anywhere determining that?

Memory usually refers to DRAM, not caches like L3 inside the CPU.
The base exe and DLLs etc. do get loaded to DRAM on launch.

[glossy glacier]

This. And then, when the CPU has to execute instructions these plus the data necessary gets loaded into the caches

[stuck nebula]

which what i was trying to refer to, the game dosn't get loaded once and stay in the memory, it keeps pulling data and if the L caches is not big enoguh to store the data memeory has to do the work, which he says the exact opposite

that's what i was trying to say, probably worded it wrong

[twin dew]

Stuff gets loaded into RAM before being loaded into CPU caches.

[stuck nebula]

yes?

i don't see me saying the otherwise

it's the exact reason why in a small games like factorio or other sim games which require a lot of nonstop calculations CPUs with bigger l3 cache give a stable and high fps

the whole computational data can be stored in the caches and memory dosn't have to keep sending it to cpu according to what cpu pulls

[soft bloom]

but it can be loaded to caches without ever meeting RAM

[stuck nebula]

DirectStorage?

no, that was for GPU

nvm

[twin dew]

Yeah, brain fart in way from me.
Because it probably has to hit the caches in most cases before getting to RAM.
Disk to cache to RAM.
And then back to cache on usage.

[stuck nebula]

anyways, i guess you got what i mean

[twin dew]

But anything DMA capable can send it direct to RAM.

Which is almost anything today.
Disk controller DMA has been thing since 90s.

[stuck nebula]

in a hypothetical world where your cpu caches are humongous, you can theoritically fit the whole data of the game in once and memory won't ahve to do anything afterwards

[twin dew]

So IDE disk controllers in chipsets etc. could send the data direct to RAM without CPU involvement.

[stuck nebula]

which i was semi joking to the guy's claim

[glossy glacier]

Considering the new EPYC X3Ds with like a gig of L3$, loading entire programs into cache isn't that far off 🙂

[stuck nebula]

yeah, but i guess we are still far from fitting entire games in it

unless the game is small enough XD

[glossy glacier]

Depends on the game
It could fit DOOM

[stuck nebula]

holy. 512mb of l3 cache?

AMD is not joking

[glossy glacier]

More

[twin dew]

96MB per CCD, 12 CCDs.

[stuck nebula]

how much that thing costs

[twin dew]

1152MB of L3 total for single socket.

[stuck nebula]

[glossy glacier]

Too Much for us mortals

[stuck nebula]

the right answer

[twin dew]

And then you can have two of those in single dual-socket MB.

About 5k€ per CPU in retail.

[glossy glacier]

Still want to try a dual-epyc system with numa scaling 0 once 😦

[twin dew]

CPUs with "only" 768MB of L3 are cheaper

[stuck nebula]

Well, eventually these systems will also end up in consumer products years later

[twin dew]

Epyc 9184X, 16 cores with 768MB of L3 cache.

[stuck nebula]

if we are lucky enough to stay alive since then and Intel finally fixes itself, maybe due to competition we may start seeing better cpus in the market

[twin dew]

With 4.2GHz max boost and 3.55GHz base clocks.

320W TDP.

[stuck nebula]

AMD will have to find a faster transfer system though

[glossy glacier]

CPU to CPU or Cache to RAM?

[twin dew]

Both are currently being limited by Infinity Bus.

[stuck nebula]

both, due to IF limitation

in this example, it's all IF that connetcs them inbetween

CCX TO CCX

[glossy glacier]

Right

I think rumor has it that Zen6 will improve this

[stuck nebula]

i think Intel is doing much better in that case

AMD gotta speed up or they might end up facing Intel soon

[stuck nebula]

let's hope so

[glossy glacier]

Well if you need one can just adjust the NUMA layout. Nuna-per-Socket(NPS) let's you scale from

• 0: All CPUs as One Domain
• 1: Every CPU is one Domain
• 2: Half a CPU as one Domain
• 4: Every CCD as one Domain
  If the Software/OS is NUMA aware it'll try not to cross it

[glossy glacier]

CCD to CCD btw. CCX was the 4-core sub division back in Zen2. Since Zen3 its one CCX per CCD

[stuck nebula]

you are correct, i keep confusing CCX with CCD to this day

i don't know why

i think both Intel and AMD might end up having to work together again under license like they did with x86 and x64

[twin dew]

No, we still have both.
Just that both have been 8 cores for long time.

So the CCX hasn't been talked much.

There are Epyc CPUs with two CCXes in one CCD.

[stuck nebula]

Intel's mesh link system is much better suited for new type of core stacking

[twin dew]

With Zen4c cores.

[stuck nebula]

what what is called, was it foveros or something?

[twin dew]

That is just die interlink substrate thingie.

[glossy glacier]

Was about to ask if is a thing (again) with ZenXc

[twin dew]

And in Zen3 and 4 cores it is there, but as there is only one CCX per CC-die, it isn't talked about.

[glossy glacier]

Let's see if the collaboration on UCIe ever bears fruit 😅

[stuck nebula]

collaboration is always nice tbh, setting up a specific standard together is always benefitial to whole market

because if you don't, we end up with stuff like CUDA

which totally limits the market

thanks nvidia, as always.

[glossy glacier]

https://tenor.com/view/thanks-steve-intel-gamers-nexus-thanks-steve-gif-2570818544130153718

[wanton orchid]

always did unfortunately
Intel memory stack is much less naive than amd one
its been mostly what Intel as been pushing for performance
(low l1 latencies, high cache hit, complex cache eviction mechanisms, low latency buffers, streaming data, etc...)
intel always match amd with much less direct cache space

[stuck nebula]

which much less efficiency though

[wanton orchid]

sometimes I dream about a ryzen core with Intel cache stack

[stuck nebula]

they always suck at that part

[stuck nebula]

that is also my dream

[wanton orchid]

I meant there is a huge gap
maybe amd can do some interconnect tradeoffs in this direction

[stuck nebula]

AMD will also have to change it's architecture sooner or later

[wanton orchid]

the ryzen core themselves have much less naive logic

[stuck nebula]

Homogenous tile system is the go imho

zen worked really nice up till now but we are starting to see it's limitations

with dual ccds, they tried to solve the core parking issue via software

which it seems is not the best solution

now they are talking about having x3d cache in both side of ccds

what will that add up to whole system? it will just get more expensive and it will change nothing

since we will still have to use 1 ccd at a time

it's an archtitecture limitation imho

[twin dew]

That Core parking is only needed because Windows scheduler is just bad...
And MS isn't willing to do complete rewrite to fix it, so both Intel and AMD have had to do weird stuff to interface with it.

When the same problems aren't happening on Linux, as it knows to not spread threads over multiple CCDs or keep switching them between P and E cores etc.

[glossy glacier]

complete rewrite to fix it for AMD
they did it for Intel P+E-cores

[stuck nebula]

i heard that Windows is planning to totally fix this issue with Win "12"

but i doubt, we are talking about MS

you know how they fix stuff

[twin dew]

They didn't, just added a cludge.
Like they did now for AMD with that Game Bar integration.

And the cludge was so bad that Intel added more silicon to work around the issue in newer CPUs.

[stuck nebula]

afaik, Intel has a dedicated hardware part where the parking been made through that

i don't remember what was it called

[glossy glacier]

both HW and kernel afaik

[stuck nebula]

AMD is totally software

[twin dew]

Thread Director.

[stuck nebula]

This

[twin dew]

Wasn't in the 12th gen.

[stuck nebula]

yes, they had to bring that because back then Windows was even worse with core parking

[twin dew]

And was added later as the MS Software cludge didn't work good enough.

[stuck nebula]

multi billion dolar company

XD

[twin dew]

And also changed how the threads go through the cores.
Used to be slowest to fastest in all cases, now some start at the fastest cores.

As the ramp up was too slow.

Intel for that Thread Director in latest laptop chips.

[stuck nebula]

Windows has to change some of it systems because the further we move, the more asynchronous the whole cpu is going to get

due to how important AI tasks started to become

i hope they don't stall too much

they are still doing good stuff though, DirectSR and DirectStorage is top technology imho

[languid gulch]

ok, got a weird one. i have my pc set to turn the screen off after 5 minutes, but always be on otherwise (no sleep/hibernate). sometimes i'll wake up the monitor and my windows will be thrown around & i have to put them back where they were. no signs of a gpu crash or any other driver issues

[glossy glacier]

Depending on firmware some monitors can disconnect when going to standby, causing windows to move the windows to still connected displays

[small bridge]

I developing a discord bot to upload and download saves and blueprints good or bad idea?

[mental oriole]

For learning sure, but it won't be usable on this server.

[soft bloom]

To store http headers as dict repr strings in FS as separate files
OR inside sqlite as pickled objects?
(not sure if it would be right to convert to JSON since that requires decoding to UTF-8 but then why is it stored as dict with bytes keys and values as list of bytes?...)

[mental oriole]

dict of strings

assuming you're using python requests api, it will convert to a valid header automatically

and a dict of strings you can just json.dumps/ loads to/from file

[soft bloom]

am not

the lib i use (Scrapy) stores them as dict[bytes, list[types]], so that's why i am not sure i should convert to str

maybe it's legacy of python2

i don't expect anything fancy in headers, but... don't want to catch these errors later

[mental oriole]

No clue then, never used scrapy.

[soft bloom]

i wonder how is it by specification of HTTP headers... is it UTF-8 only?

ok, i have like 95% assurance it is

so probably legacy from python 2 days (pre unicode strings)

[wanton orchid]

bytes is simply more reliable, no need to interpret it

[soft bloom]

that"s what i think
but since my original intent was to put it into db as json field...
well, i went with encoding as utf-8

[wanton orchid]

but then you have the choice to encode as what you want when you want 👍

[jagged snow]

I believe I've received official permission from my landlord to properly network my house

Finally ditching wifi

Just have to get the ISP to move the fiber termination into the closet and off of the living room wall

[visual tree]

I remember the old days when there were ethernet cables running through the hallway hidden under the carpet

Well, before I started using powerline adapters

And recently before installing new flooring, we made a conduit and pulled cables through it so every room and floor is now connected with ethernet cables

[soft bloom]

and that reminds me that i need to find ISP before the are overloaded during winter power outage histeria

[visual tree]

In 2005 I had to use this nightmare, it was a pain to direct the antenna properly

[soft bloom]

notion that anbtena need to be "directed properly" is wild to be

[visual tree]

Yeah, I had so many issues playing battlefield 2 online and had to adjust the antenna all the time because my mother would move the adapter every time she cleaned the room lol

[visual tree]

Today, I don't even know why wifi routers have adjustable antenna. I can move them in all directions and still get a decent signal

[edgy hazel]

hole in walls

[visual tree]

Lol, we actually have makita and made a hole through ceiling

Bosch drill was too weak

Makita + sds drill bits are a perfect combo

[edgy hazel]

yeah but tbh I still got respect for bosch for what they did in ww2

[visual tree]

We had a small bosch drill though and bought makita hammer drill later. I don't think any small drill could pass throught a concrete as easily as a hammer drill

We have a decent bosch fridge with really good rubber seals and sometimes you need to apply a little bit of force when opening doors lol

Some guy here bought a large samsung fridge which got broken after warranty expired and learned the hard way certified repair center won't take a look at your fridge unless you live near the repair center which is dumb imo

Guess I'll be avoiding buying large appliances from samsung

[wanton orchid]

more like spysung

[visual tree]

Turns out repair center only repairs larger samsung appliances if they are under warranty....

Must be big fans of planned obsolence

[wanton orchid]

it's easier to update the software if you send it back soon

[visual tree]

I think they focus more on making the large appliances last as long as warranty and then decline repairs when the warranty expires so the customers have to buy new products (with upgraded spysung )

Although the same could be applied to almost any company today

[wanton orchid]

more or less

[languid gulch]

god i hate those

the handles always break

[twin dew]

Because that type of antennas have donut shaped signal.
So depending on desk, wall etc. mount they need to be adjusted.
And having them in little bit different orientations helps with the beam forming etc. when all clients aren't at the exact same level.

The single end connectec stick type.

[twin dew]

And higher the dipole antenna gain, flatter that pattern is in height, but is then wider.

[edgy hazel]

Me at donut shop

[pure karma]

the overly complicated donut shop

actually no this is technically the uh um flat donut macaronni earth theory by uhhh some satelite maker

[drifting kayak]

would this be the right channel to ask if anyone's done or found benchmarks with a 7800x3d on a late game save?

[dire igloo]

I don't have actual numbers, but the 3D vCache comes in super clutch

[twin dew]

Well, save that was 27MB in U8 and needed 40GB of RAM to load (Taros)
Now takes 20GB of RAM to load and is 23MB when resaved in 1.0.
6-8ms per frame for the CPU side processing on my 7800X3D.

9-12ms CPU processing at his max nuclear preproduction location.

This save is significantly above the UObject limit in U8.

[drifting kayak]

thank you, that's super useful info. might have to look for one on black friday

[pure karma]

84FPS minimum??? is it really that big of a diffrence because i know on my 8MB save before 1.0 i was allready down to atleast 35-40ms per frame on my 13700K

[twin dew]

For the CPU side yes.

[pure karma]

so a solid 8X performance diffrence or more.....

[hushed pivot]

Me, throwing this into the wind.

[dire igloo]

www.kickstarter.com/projects/908005844/jethro-v1-ultimate-chatgpt-powered-ai-mouse

[hushed pivot]

Idk, it seems easy enough to do without bolting AI onto it.

Just speech to text and show me the keys I need to press.

[soft bloom]

https://youtu.be/cu-QQaoSEac?si=fDIaJgSHHayml98b
I can't get over the fact how good this is
Especially the line at the top considering recent news about Rust in Linux

Something similar can be said about time when type annotations got introduced into Python

[mental oriole]

My argue against borrow checker in c++ is we have unique and shared/ weak ptr

But I'm watching the video now...

But I'm not against the concept.

[shadow cairn]

Hello! Greetings from the Phillipines! ^^

I'm building a new system with playing Satisfactory on higher graphics settings in mind.

I currently have an R5 2600X + 5700 XT - it's a 5-yo system.
I can play the game smoothly on 1080p with medium settings, but sometimes textures takes a bit to load and I hate it - an 8GB VRAM isn't enough.

I want to know what would run Satisfactory better.
A 7800X3D + 7800 XT or a 5700X3D + 4070 Ti SUPER?

I know you need a powerful GPU for games so I want the 4070 Ti SUPER.
But the 7800X3D is so good I want one too.
Sadly I can't afford to have both.

I already have the list of parts...

7800X3D + 7800 XT 16GB
B650 ATX + 32GB DDR5 6000 CL30 (10ns)

5700X3D + 4070 Ti SUPER 16GB
B550 ATX + 32GB DDR4 3600 CL16 (8.89ns)

• Montech Air 903 Base (will use as is)
• DeepCool AK500 (will apply Kryonaut and use 2x P12 Max for push-pull)
• Seasonic Focus GX 750W (Gold)
• 1TB Kingston Fury Renegade R|W: 7300|6000

It's just the CPU+GPU I can't decide on.

I'll be getting a 1440p monitor soon too.

Do I need the better CPU or the better GPU?
Also does the RAM latency matter?

[glossy glacier]

Get the 7800X3D combo imo

[dusk scroll]

@shadow cairn Satisfactory is more GPU bound, a 5700X3D should be fine performance wise. But why not buy a 7800X3D + DDR5 RAM while using your old GPU and then 1-2 months later upgrade your GPU when you have the money for a 4070 TI SUPER?

[winged valley]

Kryonaut is overkill for what you're going for

But if you want to spend the money on it (and take the risk) go for it

[jagged snow]

That is bad advice for this gsme

[dusk scroll]

@jagged snow what exactly is bad advice?

[winged valley]

mfw people buy rtx 40 series

[jagged snow]

Stating that satisfactory is "more GPU bound"

[dusk scroll]

@jagged snow a) that's not advice but a statement, b) that statement is true, you didn't even read what he wrote: "I'm building a new system with playing Satisfactory on higher graphics settings in mind." on higher settings Satsifactory is GPU bound, and that is coming from someone with a 4090 c) I specifically wrote - as advice - to get the 7800X3D and later upgrade to a better GPU

4090 at 3440x1440p with max settings hits 100% GPU utilization at 130-158 FPS btw on a small factory

[jagged snow]

You didn't ask what resolution or frame rate they were targeting

[dusk scroll]

"I'll be getting a 1440p monitor soon too."

please read what the OP wrote before you chime in and claim "bad advice"

[jagged snow]

I did read op post, just missed that.

Nonetheless, at 1440 a 7800xt is more than sufficient

[twin dew]

Kryonaut is not meant for long term use AFAIK.
It is optimized for liquid nitrogen OC benchmark runs.
Not for general usage.

[jagged snow]

When paired with a 7800x3d you will be GPU bound in most scenarios, but likely at 120fps or greater even on high/ultra

And CPU is always your limiting factor on large saves

[twin dew]

Well, for Satisfactory specifically.

[jagged snow]

Yeag

Op mentioned satis exclusively

[pure karma]

hes right about the gpu bound bit but yea no 1440 was completely missed

this game is only gpu intensive if you have lumen on

[shadow cairn]

I actually am on Kryonaut too with my 2600X 😅 5 years now and still haven't reapplied or replaced it. I could go for an MX-4 too if Kryonaut's overkill, thanks 🙂

[edgy hazel]

mx-6 is pretty nice

[shadow cairn]

I want 1440p not dropping below 60fps even with big factories

[pure karma]

then your stuck with AMD

more specifically X3D

[edgy hazel]

saucy

[jagged snow]

Then the 7800xt is more than you need
I'd go with the 7800x3d build you specced
No need for the extra GPU

[dusk scroll]

@shadow cairn MX-6 is out and very good, doesn't need to be reapplied that often. Kryonaut should be reapplied every 1-2 years and is more expensive for almost no benefit. I had both in multiple setups, would recommend MX-6 for regular users.

the 4070 TI Super is around 37% faster in 1440p scenarios. I'd lean towards NVIDIA just to be able to use DLSS, but that's up to you

[shadow cairn]

I didn't pay attention to my temps when I built my current system 5 years ago, but right now with Kryonaut my 2600X idles at 36-40°C and hits 65°C when playing Satisfactory.

23°C ambient temp

[dusk scroll]

@shadow cairn no need to panic, if the temps are good leave it. It's just that thermal grizzly products are made with high performance in mind. There is more pump out effect, hence why they need to be reapplied earlier for best performance.

I switched to liquid metal, no pump out at all. hehe

[shadow cairn]

The 1-2 years reapplication got me worried lol it's been 5 years and I haven't reapplied 😅

[dusk scroll]

just keep an eye on the temperature. If it's not overheating and you are fine with it, no need to change anything.

[dusk scroll]

@shadow cairn to your other question about RAM latency: the X3D CPUs are great at compensating slow RAM with their huge L3 cache. You will still benefit from faster RAM and especially tweaked timings, but generally speaking RAM speed does not matter as much with X3D CPUs as it does with other CPUs

[shadow cairn]

I don't even mess with timings I'll just set xmp/expo and forget about it 😅

I think I'll just go for the 7800X3D + 7800 XT

[dusk scroll]

@shadow cairn I personally would save money and then buy a NVIDIA card as DLSS is just better than FSR, but that's up to you. The 7800X3D + 7800 XT setup will work fine performance wise in 1440p if you are willing to use FSR / disable lumen in Satisfactory

[shadow cairn]

Do AMD cards handle lumen bad? 😮

[dusk scroll]

don't think so, it's just that it is very taxing. My GPU utilization goes from ~70% to 100% on my 4090 if I enable lumen

but it really adds a lot to the atmosphere as lightning otherwise is not great in Satisfactory. They can't precompute lightning given that the player reshapes the whole world.

[shadow cairn]

I watch Darren a lot and I want to play with lumen on too. He has a 3090 and plays 1440p with lumen on

[edgy hazel]

1440p 60 fps with a 7800xt is very easily doable

[dusk scroll]

yeah that's not a problem

[edgy hazel]

no need to spend extra for the nvidia name

[dusk scroll]

I personally aim for 120+ FPS

what does that have to do with "the nvidia name"?

[shadow cairn]

If I stick with 1080p is it overkill? Will I see bottlenecking?

[dusk scroll]

you get ~37% more performance and superior upscaling

the question is if he is willing to pay extra for that

@shadow cairn there is always bottlenecking, the question is just what bottlenecks you.

[edgy hazel]

nvidia costs more than amd in terms of fps/price

[shadow cairn]

7800X3D + 7800 XT on 1080p, what will bottleneck what?

[dusk scroll]

@edgy hazel that's true

[edgy hazel]

bottleneck is a stupid metric

[dire igloo]

where's that 37% coming from

[edgy hazel]

if you look for bottlenecks you will keep going up and up and up in price

[dusk scroll]

in what situation? That's such a unspecific question that it doesn't make much sense

[glossy glacier]

Always depends in the game
In satisfactory probably early game GPU, late game CPU

[edgy hazel]

https://tenor.com/view/made-it-up-not-true-gif-25155838

[dusk scroll]

@shadow cairn don't worry about bottlenecking, you will be fine with either a 7800 XT or a 4070 TI super in 1440p

[dire igloo]

also: XeSS is the best upscaling tech for Satisfactory. last time I had someone test it (U8), both FSR and DLSS removed items from Mk5 Conveyors

[dusk scroll]

especially if you are willing to play with a lower frame rate

[dire igloo]

Bottlenecks are not as simple as "X bottlenecks Y", and the term bottleneck is often misused and oversimplified.
Think of your computer like a factory. The CPU gets a bunch of parts on a truck from SSD and RAM, and builds them into a frame of the game. That then gets handed off to the GPU, which packages it up nicely and sends the new frame to the monitor on a truck. A higher resolution monitor acts like a bigger truck, and a higher refresh rate means the shipments leave faster. Both of these mean more work for the assembly line to keep up.
If you have a CPU bottleneck, the GPU is waiting for the CPU to build each frame. If you have a GPU bottleneck, the CPU is waiting for the GPU to finish packing the current frame and take the next one.
Because all parts of the line rarely work at exactly the same speed, you almost always have a bottleneck in your PC at a given time, whether it is CPU, GPU, memory, or storage.
Where the bottleneck is while gaming depends on the game situation in question, the resolution of your monitor, and the graphics settings you've selected. This means that different part selection and component balances make sense depending on what exactly you're doing.
For more information, watch this video: https://youtu.be/XwlEyOYcr1I

[dusk scroll]

@dire igloo any recent comparision to backup your statement? That doesn't align with my testing and DLSS vs XESS in general.

[edgy hazel]

also recommending a 4070ti super when the 7900xt exists in kinda sus

7900xt is almost 200$ cheaper

[dusk scroll]

@edgy hazel he specifially asked about 4070TI vs 7800XT

[edgy hazel]

you still recommended it no?

[dusk scroll]

I said I'd get it over the 7800XT because of DLSS and more performance

[dire igloo]

at same price, absolutely

[dusk scroll]

I'd pay extra just for DLSS

[dire igloo]

TFX

[dusk scroll]

what's 200$ over the span of 1-2 years until I upgrade

[dire igloo]

that's your personal preference, nothing you should let influence the advice you give

[edgy hazel]

DLSS is such a non selling point

[edgy hazel]

Bless you

[dusk scroll]

@dire igloo I made it clear that's my take and opinion, not my fault if you don't read what I wrote

[dire igloo]

that's almost 30 doner kebab, a good winter coat or idk what else

[edgy hazel]

God I wish I had 30 Döner rn

https://tenor.com/view/döner-gif-11672316089507059316

[dire igloo]

I know you did, I just rectified it - because ultimately, there's more to that decision than performance and upscaling

[edgy hazel]

But yeah Fireworker is right. DLSS isn't worth 30 Döner

[dusk scroll]

I think everything has been said and it's just arguing about opinions at this point. The OP can decide what he wants to buy for himself.

[dire igloo]

The right belt is moving at a constant speed, left is stationary. No downscaling used with either at 1440p
First is DLSS, second is XeSS - testing from April this year

[dusk scroll]

@dire igloo looks like an engine bug

needs retesting with 1.0

[dire igloo]

hence why I specified U8

[dusk scroll]

just for giggles, what belt speed was that tested at?

I'll do a quick test

[dire igloo]

those are mk5, so 780

let's specify belt speeds in Hertz: Mk5 is 13Hz

[winged valley]

Its fine for general use

[dire igloo]

degrades at high temp

[dusk scroll]

@dire igloo left picture DLSS, top belt stationary, bottom belt moving at 780u/min?

[twin dew]

Well, with my 7800X3D & RX6600 in Full-HD, my problem is the GPU in late game saves XD

[dire igloo]

bottom stationary, top 780

items disappear off moving belts

[dusk scroll]

yeah that's what I would have assumed, just read like it's the other way around

[glossy glacier]

perfect argument for the better CPU then ^^

[dusk scroll]

looks fine to me, all MK5 belts

[twin dew]

Better GPU you mean?
For me?

[dusk scroll]

ghosting / smearing is a common issue with DLSS, but not things disappearing from the screen. Must have been an engine bug.

[dire igloo]

note that 7600 isn't viable for Satisfactory for the lack of X3D, but should otherwise be similar to 5700X3D
www.techspot.com/review/2790-faster-gpu-vs-faster-cpu/

[glossy glacier]

nah,regarding krays' question

[dire igloo]

I'm assuming all of them are moving?

[dusk scroll]

yeah but some belts don't transport items as they are fed into machines. The screw belt is moving items at 780u/min though and nothing disappears.

[shadow cairn]

I'll be going with the 7800X3D + 7800 XT and play on 1080p while I save up for the 1440p monitor and also change my choice of thermal paste.
Thank you all for the responses @glossy glacier @dusk scroll @winged valley @jagged snow @twin dew @pure karma @edgy hazel @dire igloo 🙏 🫡

[dusk scroll]

@shadow cairn nice, good luck!

upgrading your monitor is always #1 priority 😄

[dire igloo]

yw

[maiden coyote]

Nivida cards are good for LLM's as well as their encoders. Not really something to consider unless you can see yourself getting into that

[dusk scroll]

I hope NVIDIA finally implements DisplayPort 2.1 in their 5000 gen and not 1.4 again like with the 4000 series. So disappointed that my 4090 only has DisplayPort 1.4

[glossy glacier]

LLMs yes cause CUDA
encoder go ARC 🙃

[maiden coyote]

Amd is garbage for anything AI or video encoder related.

Intel is the best yeah.

[dire igloo]

Blender renderings also benefit from CUDA

[tough owl]

AMDs AI performance is good. If you can get it working

[maiden coyote]

It's not even cuda though, you can run llm's on a 1080.. it's just.. 20 miuntes for an output a 2080 spits out in a second. I'm sure you can get it working on amd cards just abysmal performance

[tough owl]

And their encoders are decent. Again if you can get them working 💀

[dire igloo]

ROCm "works"-ish
https://cdn.discordapp.com/attachments/1177157277040513034/1178647656810160188/missing-image.png

[glossy glacier]

Linux + Native&ZLUDA

[tough owl]

Zluda is a PITA

[dire igloo]

https://cdn.discordapp.com/attachments/1177157277040513034/1177170200546770954/IMG_3702.png

[tough owl]

It works but its not amazing is what I meant

[dire igloo]

Encoder is solid actually - old testing tho
https://www.tomshardware.com/news/intel-arc-av1-encoder-dominates-nvenc

[maiden coyote]

I'm talking about amds encoder, not intels.. the jellyfin page summarizes encoder performance pretty well.. it's apple m > intel > nivida >>>>> amd

[dire igloo]

the testing includes AMF and Nvenc

and they' basically on par

[maiden coyote]

Well, setup a jellyfin host and try to get an amd card working on it. I don't have any experience with amd but when a rockpi is recommended over a radeon 7000 series for it I'll take notice

My info could also be outdated, as I haven't actually kept up on it

[dire igloo]

that sounds infinitely more niche than a hardware encoder - which itself is already not that common

[maiden coyote]

So if so, I apologize and I'll watch the video. Hardware encoders are needed if you want to game and stream

[tough owl]

That's easy

Just gotta use prop drivers

[dire igloo]

I just went off the chart in the article

[maiden coyote]

Then make it handle 10 4k streams like my little arc a380

[tough owl]

Which I refuse to use

[dire igloo]

not too big of a difference between AMF and Nvenc

[maiden coyote]

Oh, 4k at 60hz to.. good luck :p

That's fair the reason I hope intel keeps up with their gpus.. they might be okay at best for gaming, but the video encoders on them while niche, are some of the best

[glossy glacier]

ARC was a great first step. drivers are coming along great.
now they just need to continue

[maiden coyote]

Is not like intel has a habit of killing projects lol

[soft bloom]

Today i realised that curse of asyncio is that j forget about ProcessPoolExecutor

Kinda feels like i spend more time optimizing migration tool than moving my project forward with "tool that needed that migration"
But ideas of how to improve the way data is stored and migrated just keep coming...

And all i need it for is quicker/easier access to "did i cache that response"?

[soft bloom]

I found weird led lamp that has very large casing
It's benefit is high brightness ( at the cost of high power consuption, and size) and that it's impossible to create hard shadow.
In fcat, it creates that weird feeling that you can't block the light from it because you don't see any edges of shadow even when putting hand near the lamp - just so much scattering happens

[stray badger]

Ive had good results on mine

[maiden coyote]

What use case?

jellyfin.org/docs/general/administration/hardware-acceleration/amd/ towards the bottom

It seems they have added support for amd, but no AMF on linux

[stray badger]

AV1 from obs and DNxHR from resolve

[winged valley]

I just have an old quadro p500 for my jellyfin server

works great

I store transcodes in a ramdisk bc I'm insane though

Come to think of it, that's probably a bad idea bc of my transcode settings, I should change that

[soft bloom]

I'm trying to understand why would Scrapy split cache storages by spiders, if there's anyway a "fingerprint" at play, that takes a hash of 1 url 2 method 3 headers 4 body, aka everything

Turns out DBM solution doesn't do this... And it also doesn't care of request data (except for fingerprint)

Why am i not using it? i don't know what that is, and that definitely means that those using my tool wont know either. I hope they will get SQLite working, haha

[soft bloom]

alright, i got 0 fingerprint collisions from 120 000 cached requests(responses, but fingerprint is from request), so am pretty confident there' not need for anything else.

(actually that's a terrible way to find asnwer on that questions, since the only possibility for collision was from job from one spider having same request params as from another spider, which i know for a fact is not possible since they have different url query params, and if there could ever been actual collision of fingerprint i wouldn't have oticied since it would just overwrite previous data)

[twin dew]

https://discord.com/channels/370472939054956546/1293461516628131941
Second time same idiot has not accepted information about his setup being unstable...
Or in this case that 7200MT/s RAM with AM5 is just bad idea, as it runs in 1:2 mode, causing worse performance than 6000MT/s until 7600+MT/s.

🤣

[languid gulch]

https://www.youtube.com/watch?v=WRI94x5MEMM

[twin dew]

Might have just been driver issue for the game crash in that case, based on very short private message he sent, but cannot check what the original crash report was since he deleted it.
And would have gone into GPU drivers once the RAM test was done if things didn't change.

[bitter willow]

“Profesionales” more like people who don’t know anything and will just tell him what he wants to hear

[twin dew]

And that some ready made machine was 7200MT/s RAM in AM5 system can just be marketing side decision, instead of 6000-6200MT/s RAM.

Which would have been faster in actual use.

[bitter willow]

Yep

Bigger number not always better

[twin dew]

In this case it might just have been broken GPU driver.
But testing without XMP on and RAM on slow would have been faster to test than start to diagnose the GPU side.

Which would have been my second step probably.

[bitter willow]

Yea but he’ll go to a repair tech and pay $100 only to be told he needs new ram

[twin dew]

Seems it wasn't RAM issue in this case.
But he wasn't willing to do the diagnostic step, and got hostile when I said the computer would perform faster with slower frequency.

[bitter willow]

Probable wasn’t a ram issue but if he didn’t want to do the diagnostic step it is his loss of money and time

[twin dew]

But the computer would still have performed better once we had found the actual issue, and then tuned the RAM to run at lower speed in 1:1 mode.

[soft bloom]

it happened!
#off-topic-media message
that looks like huge difference
and pcm is definitely thicc now

definitelly missed desired temperature... but if this is what is going to happen in winter - i would rather hold 25C bottle in my hand

[soft bloom]

i got idea on how to conenct sensors
i have some spare power cables from PC
so if i can find compatible connectors to put on sensors, it should be quite simply

[uncut ether]

@wispy heron you wanna continue here?

[wispy heron]

not much to continue i'd say, we're on the same page

[uncut ether]

👍

[winged valley]

I clowned on him in that post, so he deleted every message he sent in it. Then this also happened

[twin dew]

... 😆

The previous thread I mentioned: #1224240458662809651 message

Just bailed on that one.

[tough owl]

Hi loafs

[edgy hazel]

🍞

[winged valley]

Other servers have mpreg reaction, now our niche channel in a niche server has loaf reaction

[pure karma]

honestly at this point maybe just contact a server mod and get him out of here if he does it again

[twin dew]

Why?
Being bull-headed is no reason for ban when it doesn't affect most of the server.

I would have much more liked to get ban on person who yesterday said he had done things I had asked to check, when he hadn't.
And then tried to still say he hadn't lied to me.

Wasting my time in trying to find alternative reasons.

[pure karma]

not necesairly for that just if hes makign a bunch of posts for the same problem and then just not actually caring is a bit dumb

[twin dew]

Game not starting because one of the main asset files was corrupt.
Me telling to run file verification and if that passes compare file amounts.
And then he had uninstalled the game day before and started to reinstall, but never finished.

Well, he had verified the game files before the uninstall...

And somehow that counted in his head for my request almost day later.

He had originally the normal fatal shader compile error.
Which morphed into that broken asset file mid-thread.
So he had tried to deal with the shader compile one with uninstall-reinstall, failed on that and transformed into that broken asset file one.
But thinking that anything he had done before the uninstall somehow counted...

And that uninstall-reinstall was never mentioned in the thread before I took a look in my morning.

[tough owl]

wooohoooo recall is a dependency of file explorer in 24h2

[edgy hazel]

didn't they say it can be removed entirely?

[pure karma]

anyway i set a new record of having a driver related BSOD in less than one minute of launching the game

[tough owl]

https://www.youtube.com/watch?v=G9FRadIkkE0

[twin dew]

Might not have blocked you.
There seems to be more general Discord PM outtage at this time.

[winged valley]

Easy way to check let me try reacting to a message from him

Nope he blocked lmao

If someone blocks you and you put a reaction on their message, it removes it and shakes the screen

[twin dew]

@flint garnet That was almost certainly another warez version user still on 1.0.0.0.

He just cut the pic so the CL didn't show to confirm.

[mental oriole]

If that's true I'm perma linux.

However can't you just delete the recall exe...?

[gilded helm]

Heh heh, that was amazing to me when we had at least one person in the days before 1.0 asking whether satisfactory would use to Denuvo or anything

Insisting they were just curious

[twin dew]

The controversial Windows Recall is returning with added security features. Saving snapshots is now an opt-in feature. Secure Windows Hello sign ins are required to use those snapshots.
So will see what it does save when people dig into it when not enabling the snapshots...

[mental oriole]

I want zero snapshots because I'm not an idiot.

I don't want recall

Not gonna use it because bloat.

[twin dew]

Seems it can be uninstalled currently via the "Turn Windows Features On or Off" functionality.

[mental oriole]

I'm still on w10 so..

[tough owl]

I mean yes? But then file explorer breaks to the old not good one

Wrong reply whoops

[glossy glacier]

@restive gyro If your network is fast enough you just shift the bottelneck to something else.
With NVMe backed storage you can easily saturate 25Gbps Networks (or more, depending on PCIe version, drive speed and Volume layout)

[restive gyro]

hmm..

I've wanted to get that motherboard Linus made a video about, that 4 nvme NAS board.

welp

[glossy glacier]

full NVMe storage is pretty awesome but you trade capacity for speed when you are on a budget

[restive gyro]

honestly 8TB would be more than enough for personal use. 4 nvmes of 2TB each wouldn't be very expensive

[glossy glacier]

and a bunch of HDDs can saturate 10Gbps too on sequental workloads. IOPS of course not

[languid gulch]

it's also really nice to just socket storage instead of having to mount a drive & do the paranoid "oh god i hope i don't snap off something" dance

[glossy glacier]

you mean SATA/SAS or M.2?
cause i can see arguments both ways xD

[languid gulch]

i always get nervous when i have to move/swap my 2.5" drives that are mounted vertically on the backside of my mobo tray

those super thin power & data connectors worry me

doing my first nvme made me almost immediately run out & grab another

[glossy glacier]

yeah, i can see that
but if you have bays with a backplane it can be as easy or even easier than M.2

M.2 screws can get lost very fast

[languid gulch]

true

what really annoys me is having to play sata connector roulette & hope that you have the flat style instead of the 90° one that will absolutely snap the drive

[glossy glacier]

you mean the cable/connector?
depends on the case i guess

[languid gulch]

& psu

[glossy glacier]

oh, don't start with sata-power cables >.>
either they are way to long or just a bit to short to reach between drives

[languid gulch]

or the cables are just a bit too stiff, and you feel like something's gonna break if you try

[glossy glacier]

haven't had that problem myself tbf

[languid gulch]

i have 2x 3.5" drives under the shroud, and man it felt awful bending the power cables to make 2 in a row fit

[glossy glacier]

get a bigger case then 🙃

[languid gulch]

🤣

it's a mid tower, so plenty of room

i kinda want to get the extra drive mounts so everything but the psu is on the front side

[dusk scroll]

"plenty of room", "feels awful to bend over to make 2 in a row fit" what am I reading 😳

[glossy glacier]

tech-talk™️

[languid gulch]

my gpu is too big, but i could technically put 8 drives above the shroud. phanteks sells mounts where you can stuff a 3.5" on the top side, and sneak a 2.5" on the underside of each bracket

previous rat's nest for the drives, thankfully replaced. so a mid tower with room for 15 drives 🤣

[dusk scroll]

fractal torrent case?

[languid gulch]

Phanteks P400A

[dusk scroll]

ah yeah PSU at the bottom, should have seen. Looks kinda similar though from the inside

[tough owl]

Rip that SSD power cables. That looks so unsafe

[languid gulch]

yup, part of why i replaced them

swapped out both of those once M.2 prices came down

[tough owl]

And now there back up

I snagged a 2tb drive for cheap and now it's like $50 more for the same drive

[languid gulch]

yea, i grabbed 2x crucial P3 4TB drives when they were like $130 each

[tough owl]

I should've done that

[languid gulch]

not amazing performance, but i'm also still on AM4 so it couldn't take advantage of it anyway

[tough owl]

I'm speccing out an all SSD server currently but NAND is so expensive now

[languid gulch]

and prices are about to get way worse with that mine in NC being shut down

[glossy glacier]

i have a Antec P180 for my server. Plenty space for 3,5" drives and the trays are pretty neat

[tough owl]

Yippe back to looking at spinning rust I go

[languid gulch]

i did see an article about 60TB drives by 2028

LOL here's my first build in this case

everything's so tiny

[tough owl]

Spinning or what?

[languid gulch]

yea, 3.5"

[tough owl]

Those are going to be painful to rebuild arrays

60tb at 120MB/s 🥴

[languid gulch]

https://www.tomshardware.com/pc-components/storage/60tb-hard-drives-arriving-in-2028-according-to-industry-roadmap-hdd-capacity-forecast-to-double-in-four-years

[tough owl]

Please let them be SAS

Only

[dusk scroll]

looks clean! want me to trigger you?

[languid gulch]

i spoiled myself with an 18TB external, & idk if i'll get remotely close to filling it halfway before it mechanically fails