off-topic-tech

[dire igloo]

This is the guy who made the AM5 motherboard spreadsheet, idk where he gets his info from but I 100% trust him.

[twin dew]

https://arstechnica.com/tech-policy/2025/09/taiwan-pressured-to-move-50-of-chip-production-to-us-or-lose-protection/

[pure karma]

thats not happening

[thin trout]

They can't do it because if they bring half of their production and especially their current nodes to the US, US will just take it and walk away. Their word is worthless

but having unique and valuable production is leverage

Their government controls that stuff, and they know that it's an existential issue for the entire country

[jagged snow]

Hey y'all, I've got a really odd tech problem

I have a video encoded in a format that nothing seems to recognise. No software I have on my pc is able to do anything with it. On my android phone I can play it, but it's clearly in some obscure format that requires software decoding as I can't play it back in realtime. It's also at an absurd bitrate, 750MB for 3 minutes of video. All I can discern from the metadata is that it's in an mp4 wrapper.

(I've tried opening it in windows media player, vlc, and handbrake)

Source for the file was that I asked a very non-techie friend to grab a video of something for me and I think he used the AMD equivalent of shadowplay, but clearly with some very odd configuration.

[thin trout]

It might be a broken mp4, they don't work if saving is interrupted

[jagged snow]

Could be?
Not sure why I'd be able to play it back at all on my phone then.

[thin trout]

yeah i dunno

[mental oriole]

hmm

[pure karma]

i had some obs videos like that when i selected simply too low amounts of comphresion but thats likely not the issue here

[jagged snow]

Yeah, it's not just the high bitrate

[mental oriole]

can blender open it? 🙂

[twin dew]

Check with some program that just reads the codec information, as one option:
https://github.com/MediaArea/MediaInfo?tab=readme-ov-file

FFMpeg can also do it.

[jagged snow]

👍 I do have ffmpeg cli installed

[mental oriole]

Probably a specific codec under the mp4 standard that isnt installed by normal plugins.

[twin dew]

https://stackoverflow.com/questions/2869281/how-to-determine-video-codec-of-a-file-with-ffmpeg

[jagged snow]

That's my guess as well

[mental oriole]

On linux you can install additional codex for vlc via the command line.

dunno about windows though.

[twin dew]

So in theory "ffmpeg -i file"

[jagged snow]

Seems that you're correct about it being an interrupted save

[pure karma]

makes sense then

if whatever software your using (and as far as i know none of the ones you named do) check when the last actual frame happened and set the end frame there then it just wont play

[jagged snow]

Except my phone's video player for some reason, and that's very clearly dropping frames and playing slowly

[pure karma]

its probably trying to check them in real time or something

[thin trout]

It's why i record in mkv and repackage afterwards. Mkv is tolerant to interruption, mp4 bricks your last 3 hours of video

[jagged snow]

Yeah, this guy just gothis first pc a little while ago and is not tech savvy

So I was impressed he managed to save a video at all

[soft bloom]

Why would placing down SF BP trigger shortlived massive spije in dram read/write, so that after 5ish seconds bandwidth would remain at about half of that spike at 6 write 3 read?

[jagged snow]

Constructing even one new item is fairly memory intensive and doing dozens or hundreds at once is even more so, especially if they're not batched

Each new machine has to be added to the references for factory calculation and their states must also be loaded into memory

I'd have to guess at internal implementation details to say any more than that

[thin trout]

The way that buildings are simulated is a bit.. weird. I'm not sure exactly

but it's kind of like for example when you make a belt, it's added to the end of a belt array. The more belts you have, the longer it takes and the more traffic.

building belts and pipes on my biggest save looks like this

and when i build certain things, i drop from 200fps to 2fps

it seems like it's tracked on a per building basis, so for example mk.5 and mk.6 belt will have their own counts

4m foundations are extremely laggy for me, 2m is still snappy

[jagged snow]

Yep

My guess is that there's an ieterable list of all production items that must be checked every factory loop

And that you may not even be appending, but inserting items in some cases when building

It may even take multiple passes over some tasks

[night girder]

Do we want to know the 🍝 code? 🤣

[jagged snow]

heh

I'm sure it's interesting

[night girder]

There is so much going on in SF that it's not as simple as we think.

[jagged snow]

Oh, absolutely

[night girder]

If you followed the patchnotes, you know things are sometimes weird. And the fixes are even weirder probably.

[jagged snow]

It's an insanely complex system

[thin trout]

It's surprisingly efficient, but this edge case where you have a lot of one specific building seems like an avoidable snag. One that doesn't really affect most people because they don't build degenerately enough

[jagged snow]

And they have to have done some absolutely wild optimizations behind the scenes
I'd wager there are some really obscure heuristic approaches

[jagged snow]

*now that non-factory buildables behave differently

[thin trout]

Foundations take up a pretty crazy amount of RAM. I think we just need bigger foundations, like 32x32x4m

[jagged snow]

Lol
Or some sort of bundling approach

[thin trout]

yeah, just so that we don't store essentially the same information 1000 or 100,000 times over to make a giant plane of foundations.

In that case it could be simplified to only a little bit more data than a single foundation.

[jagged snow]

Create a tileable surface texture and neighboring foundations can essentially be treated as one object with center coordinates and some dimensions and you just tile the correct texture across those surfaces in a shader

[night girder]

I think I might be going off topic for tech, but for a while I was wondering why SF implemented the belt system as is. IF you know that coal miner produces 100 coal. And on the other end is a machine using 100 coal. The items don't need to be on the belt. Just make consumer machine use 100 coal every minute regardless of items on belt, and let the belt just be visual show off.

[jagged snow]

That's what it essentially does now, actually!

[night girder]

No it doesn't. Because I had some hard bugs with belt items.

[jagged snow]

The items you visually see aren't 1:1 lockstep with the simulations of what belt items are actually doing iirc

[night girder]

Very obscure, but sometimes my coal dissapeared somehow 🤷‍♂️

MK3 belt. Doesn't matter. But this made me realize, it's not a simple calculation as I though it to be.

[thin trout]

visuals don't match the sim 1:1 but there is still stuff going on with belt sim

[jagged snow]

Yes

There is still travel time, for example

[night girder]

Yeah who cares?

[thin trout]

i do

[night girder]

If machine produces 100. Let the other consume 100.

[jagged snow]

I do as well

I quite like the solution they have

[night girder]

The distance doesn't matter. The flowrate is same.

[jagged snow]

It does have its issues

But what it enables is factories that are imperfectly balanced to work intuatively

If you're producing 50 and using 100, it's easy to see and understand what is happening

And the logic for that sort of scenario would get more complex with your solution

[night girder]

How does distance impact flowrate of simple setup 🤷‍♂️
You have to wait 5 minutes for items to arrive. But once it's going, rate stays the same. Unless machine goes down.
But then the consumer should consume less since production is less.

I am saying that I think, you could seperate visuals from simple calculations in background, and avoid some bugs.

[jagged snow]

Especially when multiple consumers were present

[night girder]

My proposal would not be as realistic. But it made me think why they opted for that.

[jagged snow]

My best guess at how it's handled right now is that each contiguous segment of belt is treated as one, and an item is tagged with an arrival timestamp when it's placed on the belt and then a recieve action is triggered when the "travel time" is elapsed

[umbral marten]

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

[night girder]

Imagine all the calculations being made in the background. Just to let items travel over a belt. And keep it in sync.

[thin trout]

there are too many problems like how to deal with multiple inputs/outputs and different travel times between them

[night girder]

Shapez 2 also had issues with flowrate and items on a belt.

Not 100% sure, should check, but I think they disconnected the two.

So that if an item does something visual wonky (it falls off the belt) the consumer still gets 100% altough technically an item fell off because of bug and consumer should get like 99%.

Then again, in shapez 2, you cannot take items of the belt. Which we can in SF. Oh, I see now. I get it why they did it 😄

[jagged snow]

😄

[night girder]

Else people just yonk stuff off the belt, and 100% still arrives. Sounds like a dupe method lol.

[soft bloom]

ok ~10Gbps times 5s is huge amount of data to move around, much more then entire save

[jagged snow]

The entire save is serialized and minimized

[jagged snow]

Also, if you're using hwinfo that polls momentary bandwidth once every two seconds

[soft bloom]

thing is, I just placed down 4 assemblers with some decorations
it's not a big BP

[thin trout]

Games have gigabytes per second of memory traffic just to send the commands to draw the game world

would be much more if we didn't have 100MB caches to catch 90 percent of the memory accesses

[soft bloom]

yes
but the thing started with "I see spike"
relative to what was before placing down BP - maybe 1/20 of that spike's height
that's unusual

[jagged snow]

Well if we didn't have modern caches we'd more or less need to switch to fast sram or hbm

[thin trout]

Yeah, i think the game juggles the data around in some way that takes time and data which depends on how many of that building you have in the world.

So if you have 4000 assemblers it might take 10,000x longer than if you have 1

[thin trout]

modern caches are fast sram 😄

and hbm has 5-10x more latency, games are mostly latency bound

[jagged snow]

Well yea, I meant fast off-die sram that we could even semi-affordably do in a decent amount

[thin trout]

we're just getting to interconnect designs that make the idea of having an off-die cache not completely garbo

[jagged snow]

Even if it was still on-package and in a seperate die

Or stacked like amd is doing, but that has its issues

[thin trout]

When Intel did the off-die SRAM cache on the same package, they had 50ns of latency. That's more latency than overclocked DDR4 on the same CPU - but a lot better than spec 2133 cl15 DDR4.

AMD has 97MB of cache accessible to one core now with 8-9ns latency.

[jagged snow]

Yeah, iirc modern hbm is around gddr5 bandwidths with worse latency than ddr3

[jagged snow]

Yup

[thin trout]

hbm has much higher bandwidth, much higher capacity, but mediocre latency

[jagged snow]

Yep

And it's on a seperate die in the same package, so you have the same issues there

[thin trout]

they're coming out soon w/ 64GB @ 2TB/s per stack, and superchips can have like 8 stacks. So 512GB of RAM at 16TB/s, and extremely efficient per bit of capacity and BW.

but yeah it's like 100ns tier latency

[jagged snow]

Yep, that tracks

I work with MCUs a lot for work and I'm totally spoiled by two-cycle access to my entire memory space

[thin trout]

h aha

[jagged snow]

It's a funny thing to have your registers only be one cycle faster in terms of access time than your mem

[thin trout]

Even L1 is not that fast nowadays, but frequencies are high enough that 3-4 cycles isn't bad at all. 0.7 nanoseconds

[jagged snow]

Yep

Well iirc the actual L1 is that fast but to increase capacity without balooning transistor count they tend to do some funky stuff on the load in/out from registers

[thin trout]

yeah it's more important to an extent to have a decent capacity than it is to drive the latency down further and then miss it and have to wait 3-4ns for L2 for your super critical data which couldn't wait an extra 0.2ns.

[jagged snow]

Yep

The fewer dead cyles you can have on the alu waiting for something to get loaded into the accumulator the better

And cutting an average of one cycle per 5 by having more l1 is worth it

[inland onyx]

https://tenor.com/view/its-working-star-wars-anakin-anakin-skywalker-gif-5187168

[safe trench]

i love linux
later im going to chage those ugly panels

[thin trout]

These are my IMC bandwidth sensors when i'm running my big save.

As it got bigger it got increasingly memory reliant, and it's starting to become bandwidth bound rather than just latency.

The bandwidth actually drops sharply though when i build or delete and the game micro freezes, which suggests that it's not mem BW bound, but maybe latency or some other part of the cpu

[jagged snow]

Or that there are just a lot of highly sequential operations

[thin trout]

this is what happens when i drop 100 foundations in 1 click, and the game struggles to process them

[jagged snow]

Very interesting

[thin trout]

that Read dip is deleting 108 foundations

Autosave drops memory traffic to nearly zero while it runs

[soft bloom]

in past we discussed with Baldur that some limits in UE5 cause performance drop once it detects high VRAM usage
I think, after some updates this year, things got worse, and now I see similar symptoms from merely using Blueprint

on top of that, fps drops gradually, and RHIT can get much much much worse
if before the worst I saw was 80ms, now it can get over 120ms

[thin trout]

RHIT?

[pure karma]

how much ram usage are you averaging out of curiosity?

dosent have to be that specific

[thin trout]

Check performance in a fresh save. It's night and day even with a "normal" developed world.

My save consumes 13GB at load, but it's down to 6GB when running for a while

[pure karma]

i presume thats game only?

[jagged snow]

Render hardware interface

[thin trout]

it just reads the same things in memory over and over and over again, even within a 1 second period, because the game/sim loops are using much more than 100MB of RAM

yes, game only

[pure karma]

then we have similar usage saves lol

[thin trout]

The memory activity is fairly normal given the scope. Factorio behaves similarly, and some other games like Stellaris

[pure karma]

but the diffrence is i have so much memory bloat that im running out of ram with 32GB

[soft bloom]

I get stutters from just turning around

[jagged snow]

Windows has gotten a little bad with memory management

[thin trout]

I'm cutting it too close, but i'm holding off on an upgrade to 64-128GB of RAM until zen 6 x3d is out. Need to figure out the details of what it can run, what RAM configs and chips accomodate that best, if that involves CUDIMM or not etc.

[thin trout]

that's GPU bound

[soft bloom]

game takes around 6G, there's plenty free ram

[thin trout]

I expect Satisfactory to benefit enormously from zen6 if some of the likely changes are in

it's parallel enough that having 12c instead of 8c in a CCX can boost performance

+50% L3 along with that would be enormous

frequency boost is always good

interconnect and memory controller? among the top scalers for sure

I'm really curious how my performance is going to change and how the performance profile is going to change (e.g. is it going to become more or less memory bound) when i actually turn most of this save on.

One thing that has surprised me is that the game becomes really heavy on the GPU when building a lot of things. I can be GPU bound on 1080p with a 5090 using some reduced settings, no GI. Changing resolution though has almost no impact on the FPS.

[soft bloom]

this is how it was: #1313883446757691452 message

[thin trout]

i think you are severely gpu bound there

[soft bloom]

and this is 'normal' DRAM usage #1313883446757691452 message

[jagged snow]

Or CPU is very delayed in sending draw calls but that's not what it looks like to me

[soft bloom]

if having under 60fps while CPU can support many more - yes, definitely
problem is that I know it can smoothly, but at some point performance drops
it was concluded that culprit is that UE5 does some weird shenanigans once VRAM usage is too high, and that can be caused by just having another app using GPU, like browser, or Discrod

[thin trout]

You're running out of VRAM there, and your system is writing to RAM via the pci-e BUS and CPU to accomodate it, which seems to be wrecking performance.

As a guide, never ever ever exceed 90% of vram capacity, and preferably 80%. 7.6 is way too high for 8GB and well past the point where your system is desperately trying to allocate mem elsewhere and tripping over its own feet

I've seen it firsthand many times

[soft bloom]

and when perf does nosedive, biggest impact isn't in fewer frames - there's significant input delay, because, as Baldur said, UE5 moves video memory from VRAM to RAM which we see on these DRAM bandwidth graph which causes delays etc

[thin trout]

It's not really a UE5 thing, that's the graphics driver and OS

[soft bloom]

funny bit is that I think i tested with different texture resolution and it didn't help or made it worse.
Lumen is higher contributor

[thin trout]

Daniel Owens has some videos on youtube with 8GB vs 16GB gpu's, and you can see often that even though the 8GB card never goes above say 7.5GB of VRAM, it starts to have severe performance problems and increases in RAM usage while the 16GB variant which is otherwise identical does not have either of those. Having video of both in the same computer side-by-side with full stats is pretty crystal clear.

[soft bloom]

eh, I am just rephrasing what Baldur said, it's all in the thread

also, somehow, only SF has this drastic hit, TTP2 doesn't, even though both use UE5 (but it could be different minor versions if that can make a difference)

[thin trout]

It probably doesn't try to use quite as much VRAM

Like, when you see 7.5GB / 8 GB used, the game can actually be internally using 12GB of video memory, and you just don't have a direct measurement of the spillover into system RAM.

While another game, 7.5GB could be 7.7GB, and not trigger severe problems. Maybe a bit of a framerate drop or microstutters.

[soft bloom]

this is what it looks like when I just loaded save after restarting game
there's some bandwidth, VRAM is high but not too much, framerate is fine-ish

[soft bloom]

eh, I would expect it to use more with more detailed models and textures. but that's speculation, I didn't measure

and these normal stats. smooth

and now something weird happened without any noticeable VRAM usage change

[thin trout]

Yeah i'm basically 100% sure that you're looking at video memory spillover into system RAM

[soft bloom]

question is - why it doesn't happen on loading into save?

getting worse btw

[thin trout]

you're close to the threshold, so it's not like that all of the time, but when the game loads in some more or shuffles data around it doesn't fit any more

[soft bloom]

and RHIT reflects it

[thin trout]

the GPU time is just getting murdered because it's not in VRAM any more

[rustic panther]

What GPU is it, out of curiosity?

[thin trout]

if your VRAM usage read 6.8 / 8 GB, maybe not, but at over 7,5 it's virtually guaranteed

[soft bloom]

closed Chrome

[rustic panther]

I'm guessing a 1070 :v

[soft bloom]

then turned around, and it's again this
despite lower VRAM

[soft bloom]

you won't guess

[rustic panther]

I already did 😎

[soft bloom]

not even same vendor

[thin trout]

r9 390 8gb

[rustic panther]

Ah I see :V

[soft bloom]

performance based guess would be wrong since I undervolt and underpower

and you don't know resolution

[rustic panther]

I saw "no dynamic res" and 8gb so I guessed ol' reliable

But the 390 was possible :V

Anyway, carry on

[thin trout]

so what is it

[soft bloom]

that spike is slightly after loading into same save with low texture quality
still using 6.5GB somehow

[soft bloom]

5700xt

btw this is me placing down 5x5x5 BP of 24 assemblers
that tiny spike appeared with delay
And is nowhere near original one that started this conversation

so... it really shouldn't affect bandwidth

[thin trout]

It's normal to see spikes and troughs in memory traffic, especially since your Y axis is compressed to only 3 GB/s. Mine is 0-70.

You're not fully CPU bound, so you will be becoming more or less CPU bound with everything that you do, and that causes a lot of variance too.

what isn't normal is the sustained increase in RAM used, and read/write traffic which coincides with GPU frametime getting doubled.

6000 is just the easy oneclick point. Basically every CPU can do 6200-6400 with uclk=memclk mode, and probably more do 6600 than get stuck at 6000

@night girder https://www.youtube.com/watch?v=bmoN6D1roXM

[night girder]

So reddit was right

Will give it a watch in a few.

[thin trout]

He wasn't able to kill another cpu or confirm the issue, but agrees with what i said about it

[dire igloo]

You're the first one to tell me that DDR5 speeds above 6000 are stable in 1:1 operation

[pure karma]

same but because of cost

[thin trout]

they're not just stable but standard for basically every overclocker since 2023. Only not recommended for people who don't know or want to know anything about overclocking, and only use 1 click profile.

[pure karma]

ram is still similarly priced than it was during ddr4 rn because of all the inflating and random bullshit of recent times

[thin trout]

https://old.reddit.com/r/overclocking/comments/18z4rm9/some_fresh_zen4_ramif_overclock_scaling_data/

^some of my testing for example

[dire igloo]

Yeah, I know they exist.
But if you want stable, 6400 is a coinflip and anything higher too unreliable.
Although you could switched over to 1:2

Anyways, sleepy time, good night y'all

[thin trout]

It depends on the CPU for high uclk yes, but 6200 is achieved by like >95% of CPU's, and 6400 by maybe half.

[night girder]

have you watched it yourself?

[thin trout]

it hasn't been out that long, i skimmed 😛

[night girder]

He mentions X3D but it does seem to mainly affect 9000 series.

and specific asrock.

So they fucked up somewhere

Like if it was AMD, then it would be on all boards. Unless the boards already compensate and counter the issue somehow.

Well, atleast they talk to GN. So we have something 🤷‍♂️

[pure karma]

how are you eve nmonitoring bandwidth on hwinfo

i have literally nothing at all about bandwidth

[thin trout]

It's a feature on the raphael / granite ridge i/o die

they report the actual read/write BW to sensor

[night girder]

didn't know we had this added.

should check it out.

[visual tree]

My biggest nightmare has happened. One of the city power substations exploded during the night and several districts (not mine fortunately) had power blackout. Electricity was restored later but then several blackouts happened after the power has been restored

And that's probably because there are too many new residential buildings being built and the infrastructure can't keep up with the demand

I knew this would eventually happen....

Heard two explosions (most likely substation or multiple of them blowing up) 7 hours ago

[twin dew]

Unless you are on Intel IGP/GPU, that RHIT is still very high.
If on Nvidia, probably broken GPU driver install that needs DDU cleanup and reinstall.

Or if you did hit those VRAM limiters, then it is expected.

[twin dew]

Was GN intentionally misinterpreting or just accidentally when they just seem to care about average TDC and EDC...
EDC limit on that specific board was lowered by 10A...

[languid gulch]

he did say they weren't able to replicate it

maybe they missed something important

[twin dew]

And point with any kind of voltage or current limit are the max spikes, not the average.

So talking multiple minutes how the average hasn't changed completely misses the point.

[languid gulch]

yup

but if they're missing data, then yea, they're gonna miss the point

[twin dew]

Point was that they missed the point before starting the testing, and collected wrong data.

[languid gulch]

i do wanna know how long it actually takes to char a cpu like that

[twin dew]

For that specific section.

[languid gulch]

yup, bad data, mostly pointless video

[twin dew]

Need to see the rest.

[languid gulch]

really the only relevant info is that the burning is still happening

would be interesting to have more industry data from other mobo maufacturers to see what their failure/burn rates are

[twin dew]

But basically the incomplete "training"/knowledge of self-taught Steve shows up in many videos in the details.
But much better still than most.
Finnish "rockstar OC & tech person" Sampsa Kurri is just horrible in comparison.

Who only got where he is because he had shitton of money to burn on breaking OC HW when that scene was starting.

[languid gulch]

i tend to agree, but also the people who are experts at the real digging are probably the ones doing it professionally at one of the companies

[twin dew]

He still has horrible actual comprehension what is going on, just lot of personal experience how to get things working.

Meaning Kurri

[languid gulch]

i do appreciate that he gets eyeballs on issues like this, for consumers

oh i mean steve

not kurri

[twin dew]

Yeah, just that in Finnish language sphere, Kurri is somehow tech god for many, just because he had the money to buy and break lot of HW to get OC records.

[languid gulch]

limited language support for consumers like in Finnish has gotta suck

[twin dew]

Basically with Kurri, he can do stuff, but he has no idea why that stuff works. And doesn't seem to really to have any interest in finding out why it works.

[languid gulch]

yeesh

so not really investigating anything, just screwing around

[twin dew]

Not even trying to do anything investigative.
Just started with world record level OC stuff way back, and then started also acting as reviewer, but most of those reviews are technically still crap.
Even after 20+ years doing that.
And any technical explanations he puts out for world news "something is weird" stuff is usually just bad.

[languid gulch]

so kinda like JTC but way more crap

[twin dew]

With very good world record OC skills, but gotten via throwing money and time at the problem to experimentally get that.

[soft bloom]

well, exploding electrical equipment now can have some link to natural disasters.
I think it was couple years ago when some city saw many explosions and weird lights before eq hit

one would expect some automated failsafe to keep equipment safe from damage in case of overload

[modern swift]

What are people opinions of LCD steam deck is currently 20% and

I am unsure if the price delta of $380AUD is worth it

[night girder]

GN was just doing his damn best job with the information and tools he got?

Whenever they found "breadcrumbs" as he would say, they verified with engineers or secret sources.
They replied often it was not an issue. Then GN moved on with investigation.
They found nothing and threw it all in a video int he hopes maybe someone else spots somehing.

And if you skimmed through the video as someone else did in here, ... , you might missed the part where internel and external sensors were of by I think 0.6 or something.

[night girder]

Man, watch full video. Wtf. They analyze freaking seconds. Not averages.

[twin dew]

For voltages later.
For EDC and TDC they check the set limit, and then measure average in steady state test.

I was commenting that specific section, where they completely miss the point even in test setup phase.

[twin dew]

VOLTAGES

My commentary was purely on that TDC & EDC point much earlier.

[night girder]

Got it. I read "mostly pointless video" somewhere in here 🤨 Which I disagree with. Fully.

[twin dew]

Not my comment.

I was just stating that that specific section completely missed the point.

[night girder]

Also he is comparing bios version ffs.

How do you compare it, without taking an average.

[dire igloo]

If you have an issue and an almost certain explanation, your testing should try to verify that number 1 candidate for the source of the problem

[twin dew]

By those max values mostly etc.
Average isn't damaging, spikes can be.
And that 10A drop in spike max (if limiting works perfectly) can be significant.

[dire igloo]

By using percentiles or extremes

[night girder]

So Assrock said it wasn't TDC, EDC etc. So they look what the bios updates do.
Like I mentioned before, it's frustrating we know nothing. So now community has to do some sort of backwards engineering of the patchnotes,
and check with sensors what happened.

Is it the best? No would be better if assrock told us. But I don't see anyone else doing it.

[dire igloo]

🫏🐓

[twin dew]

AsRock specifically said that problem was TDC and EDC being higher than with other makers as the only cause.
Which isn't true, but it might have been part of the cause.

Cause for some of the cases, not all.

[night girder]

But they didn't change much.

The limit they lowered does jack. As he explains. (PPT)

[twin dew]

Then GN tests that averages didn't change, and blames AsRock of lying.
Without testing if the max spikes changed or not.

When they tested completely different thing compared to what AsRock said (averages vs. limit)

[night girder]

If the max changed. The average would too? It would be more precise to include those to avoid these comments.

[twin dew]

That was my complain.

[twin dew]

When the spikes are very short, impossible to capture with software monitoring.

[night girder]

it's not software I think? It's external. not 100% on those tests.

[twin dew]

That whole EDC/TDC part was software.

[dire igloo]

If the spike is short enough, not really.

[night girder]

So ... EDC/TDC would spike (according to your theory) without seeing spikes further in Vcore VSOC etc when they do those tests ...

[twin dew]

Would have needed to get voltage at those caps, and voltage at socket, and then try to correlate from the voltage changes between those two to try to determine how current changed, as that would cause more droop at socket compared to cap.

[twin dew]

They only measured the voltage at the VRM cap, not at socket.

Because their probes couldn't catch to the socket caps.

[dire igloo]

If you look at 1s of usage and there are two 5ms spikes, that's only 1% of the entire measurement.
And if there's a 100W difference for the spikes, that's only 2W change for the average

[night girder]

I think GN did a good job nonetheless.

[dire igloo]

haven't watched the video, just pulled some numbers out of my ass

[night girder]

They even verified these dips.

[twin dew]

Yeah, just was about that specific section where they completely missed the point by going with averages vs. spike maxes.

[languid gulch]

it's also possible that like with the 13/14th gen intel shitshow, the monitoring could be lying

[night girder]

was is miliseconds. So it's not that just ran the test and didn't do their homework by looking close at data.

[twin dew]

Rest was done properly, even if they didn't find a smoking gun.

But that specific section, where they then directly blamed AsRock as lying was just completely beside the point and testing apples to oranges.

[night girder]

Just frustrating from consumer perspective, GN is all we have on this.

The whole hardware industry is either blaming someone else, covering their ass or tigh lipped.

AMD threw Assrock under the bus with a comment. Didn't help us. And assrock is lying out of their ...

[twin dew]

AsRock might be lying, or there might have been actual problem with the current limits being higher.
But because GN tested completely different thing, their data for that specific section didn't test what was needed to determine if there was any change.

[night girder]

They lying. Like they said; it's not this. Then bios update that changes those values.

then 5 more bios updates in a few months.

"But nothing is going on ... "

"It's not PBO." - "Disable PBO" 🤣

[languid gulch]

it's also just annoying to have an update of "nothing so far"

[twin dew]

AsRock and AMD specifically have talked about too high current limits that AsRock was using causing issues.

And then AsRock in GN interview blamed those as the ONLY issue.

[night girder]

Update by who? Asrock? Can't put that on GN. Not his responsibility. Bro claims he lost money investigating this shit.

[languid gulch]

GN's update

[night girder]

He bought someone murderboard (loss of money) and then spent hours paying his whole crew testing a board in the dark.

[languid gulch]

yup

[night girder]

He did his part 🫡

[dire igloo]

Can easily be CYA statement.

Or PBO isn't the root cause but disabling it can still reduce the likelihood of it occuring

[night girder]

AMD said "a vendor" or somethign cryptic 😉

[twin dew]

Yes.

[languid gulch]

all i know is i'm glad i don't have any asrock boards

[night girder]

I posted the article yesterday. I talked with aeryn about it. I said it before. And GN confirmed my hunch.
They just throwing bios updates at us and see what sticks.

[twin dew]

But point was, that the sections in GN video that are invalid, because of mixing average to spike max:
Test: ASRock's "Fix", Test: CPU PPT Comparison, Test: TDC, EDC, PPT Limit Changes

[dire igloo]

He also bought a dead/burnt am5 board that people online had already investigated to be caused by user error.
GN testing showed: it was user error

[night girder]

I wouldn't dismiss the data. But you do you Baldur.
There is always room for improvement when reviewing.

[dire igloo]

Not every issue is one that needs investigation

[night girder]

And this isn't even reviewing but troubleshooting 🤣

[twin dew]

Data is good, for what they measured.
But they didn't measure what would have needed to be measured, and then used that invalid for purpose data to make invalid conclusions.

For those specific sections.

[night girder]

Maybe, or he is right. And made the right call and moved on.

[dire igloo]

Having worked for over 2 years in 1st level support, I can confirm:
That's how troubleshooting works sometimes.

Try potential solutions until one is successful, then find out why that solved the problem and tackle the root cause

[night girder]

Or he needs to get more into detail for TDC/EDC/PPT.

[dire igloo]

Accuracy vs Precision

They collected great data but were off target

[night girder]

🤦‍♂️

[twin dew]

And all those voltage measurements were on target.

Point was just that those specific sections in the video were off target in what was measured compared to what the data was supposed to look at.

[night girder]

on target? There was 0.6 difference. If you mean external vs internal.

[dire igloo]

Exactly my point

They hit exactly what they were aiming for, but they aimed somewhere that wasn't the actual target

[twin dew]

In AVERAGE, by spot SOFTWARE MEASUREMENT.
When the possible problem would have been current spike max currents causing electromigration.

Where that happens exponentially faster at higher temperatures and higher currents.

Where there is some magical line where that ramp happens, depending on exact silicon tech and design.

[night girder]

ah sorry, I was off.

[twin dew]

Again, voltages.

[night girder]

Those are just to see how the measurements are different.

It's like a sanity test to see how much difference there is between internal and external. And to check their probes.

[twin dew]

I'm not talking about those later voltage sections.
Or earlier base line, just the section from 11:37 to 17:15.

[twin dew]

As expected, because they probed at VRM output, and the internal measurement is after MB losses, socket losses and CPU internal losses.

[dire igloo]

TLDR

ASRock: "shit breaks when current spikes exceed X value, so we put a lid on those spikes"
GN: "currents didn't really change on average, ASRock are lying"

[night girder]

I don't get it. How you believe something van spike somewhere down the line in the system, wouldn't show up when checking VSOC/VCORE and VDO.
How can something spike, without the input also spiking?

[twin dew]

Current spike would actually cause voltage droop.

[night girder]

Or anything. They see nothing to worry about in their further tests.

[twin dew]

But as there is buffering, short spikes don't show up at VRM end.

[night girder]

The drops are normal. The spikes are normal. They sleep. They shutdown. They do all sorts of things with system.
ANd everytime; the voltage seems to do what is to be expected. Nothing unexpected.

[twin dew]

Buffering from all the caps.

Basically would have needed to measure voltage at CPU substrate filtering cap, compared to VRM cap, then chart the droop between those, to see how current is changing.

At very high sample interval.

As at high current, there can be 0.2+V drop between the VRM output cap, and what the CPU die gets.

[night girder]

that burns a CPU?

[twin dew]

No.
But just as example why monitoring at VRM cap leg will show higher voltages than the internal measurements.
And why short current spikes don't show up at that VRM cap leg at all.

[night girder]

Ok.

[twin dew]

As voltage changes.

Basically the on-substrate and behind socket caps fill that in already, and the VRM caps also fill in with delay, smoothing the voltage by providing the current, giving time for the VRM to ramp up for more current delivery.

In case of permanent current draw change, for short spike, it just gets filled in.

Which is why you would need to monitor that voltage change between as near to CPU die as possible, and the far voltage, with higher interval than software monitoring can provide.

And which is why BuildZoid probes at socket caps most of the time, and would like to probe at the CPU substrate, but that would mean too much work to modify a CPU for his purposes.

Which he did for some older CPU.

[night girder]

Yeah, funny GN contacted BuildZoid for this video for the probing.

[twin dew]

And for the purposes of that voltage section, the probing was suitable and much easier to do.
Just not as accurate values compared to what CPU dies see, but that isn't that important.

So question is what GN told buildzoid they wanted to measure and determine.

[night girder]

Or what BuildZoid told GN 🤷‍♂️ Why you people shitting on GN today.

A year ago it was Linus 🤣

[twin dew]

Again, only shitting about the average vs. spike max thing in that specific section.

Not the whole video.

[night girder]

Yeah, you are. True. It's ok B.

I agree with what you said about that part 🙂

But I can think of a few reasons GN did what they did. But I cannot speak for them.

[twin dew]

Point was that they made major error in test design phase for that single section.
And ended up in that apples to oranges situation.

[night girder]

With this I disagree.

[twin dew]

.

[night girder]

I read that. I don't agree.

[twin dew]

That is the same reason why Intel now forces the current limits to be respected with 13th and 14th gen CPUs.

When on those Intel Default Settings.

[night girder]

Asrock isn't limiting shit. And GN didn't say they lied. The just stated the resutls from test I think?
And if they said ASsrock lied, it was about free shipping of brokken mobo's.

GN explains removing a 1000 limit doesn't mean it will run at 1000 for example.

And Asrock is lying. We all know it.

[twin dew]

Yes, but they didn't test if the TDC or EDC max values changed or not.
They just tested that the average current, with snapshot software monitoring, didn't change in steady state.

When the problem would be current spikes in changing load.

Like the voltage testing they did with CineBench going on and off.

Basically would have needed to get current value analogues for that same test between BIOS versions.

[night girder]

Ok, then you forgot about the whole video. Dismiss it. Don't mean I have to. Or others.

[twin dew]

AGAIN!, I'm saying that the voltage measurements and tests are valid.
But that the specific part wasn't, because of testing average in steady state, and not max in changing state.

[night girder]

Baldur — 12:21 PM
Point was that they made major error in test design phase for that single section.
And ended up in that apples to oranges situation.

This sounds very different to me. Major error. Single section. Apples to oranges. Sounds like a claim for invalidating their whole video.

[twin dew]

And that that specific voltage test would have been what should have been tested for current spikes too, but they didn't instrument the system in way they could get those current analogues.

[night girder]

I don't know. You could always contact GN to ask why they did what they did 🤣
I think I know; time constraint. Money. Can't test everything into depth (which he mentions).

[twin dew]

Section from 11:37 to 17:15.
Containing parts:
Test: ASRock's "Fix"
Test: CPU PPT Comparison
Test: TDC, EDC, PPT Limit Changes

Was invalid, because they tested for average current, in steady state, with SW monitoring.
When they should have tested for spike current max, in changing load, with HW monitoring.

[night girder]

Go wide, see what hits. Then go deep.

[twin dew]

Basically: How high does the current spike, when you go from idle to max all-core, before PPT, thermal limits etc. start to apply.

Was there change in that between the BIOS versions or not.

Rest of the video was good for test setups, data collected and interpretation.

[night girder]

comment section of the video is interesting.

TDC and EDC aren’t there to protect the CPU, but to protect the motherboard, they’re there so that the CPU doesn’t try to suck in so much power that it pulls the voltage rails down.

[twin dew]

For most part yes.
But like I said, too high spike current can cause much faster electromigration for the spike duration.

[night girder]

I would bet a significant amount of money that the issue is that there’s either some fault in VRM routing or the component selection that is fine in nominal values, but fails when the tolerances stack up together. This could cause spikes during load transitions, which can easily kill the CPU.

The problem with this is that not only you need a board that killed, you need all those parameters to “align” again, but most of those depend on temperature, some of them have hysteresis effects… in other words, a big fricking mess.

[twin dew]

And that is talking about voltage spikes, and is possible.

But this was JUST about what AsRock and AMD said, vs. what GN measured, which was apples to oranges.

[night girder]

That's from someone who claimed to design zen 2 industrial motherboards.

[twin dew]

And Intel had to stop MB makers from disabling current limiting, because it was part of cause for the dead 13th and 14th gen CPU.

[twin dew]

So not active silicon design person.

[night girder]

so if EDC/TDC protect motherboard, I don't see it burning CPU's.

[night girder]

Good enough to talk to Steve about it and get reply.

[twin dew]

Yes, but my point was, there is some semiconductor process & die design specific point, after which the extra current can cause major issues from exponentially faster electromigration.
Which is point that only semiconductor die designers care about.
In three-way interdependancy with voltage, temperature and current.

And AsRock had on several of their MBs set TDC and EDC 10+A higher than any other MB maker.
Don't remember the exact numbers.

And then lowered them.

To be same as rest of the pack.

Yes, their main point is to stay in MB VRM current delivery limits.

[night girder]

And people think ASrock is clueless and don't know what they doing 🤷‍♂️
"Throwing shit at the wall and see what sticks.." GN said it. Other journalist said it. Everyone with a Asrock board said it 🤣

[twin dew]

But Intel (and now AMD) have had to rein MB makers setting those limits too high, back down.

Intel in general with those Intel Default Settings.
AMD with AsRock on AM5.

[night girder]

It's not because Asrock pushes out bios update that disabled PBO, PBO was the culprit.
They seem to do this "disabling/lowering stuff" shit untill problem goes away.
Without knowing the actual cause of the problem.

[languid gulch]

it's a liability BS thing. so long as nobody's too specific about declaring specs or standards, nobody gets liability thrown their way in court when something goes to shit

[twin dew]

Probably because that current limit used to be so high that it didn't matter, until latest nodes in use.

[night girder]

And probably also that. And if boards break and there is new bios every month, and user missed bios, "aaah, out of warranty should have bios update" even if the bios update is a shallow box of nothing fixes.

[twin dew]

Current before issues in the active silicon getting damaged purely from it.

Without excessive voltage.

[night girder]

And they lied about paying for the sending broken boards to em.

[languid gulch]

and you can also post a bios update, then declare "oopsies, that's actually a beta one, how'd that get out there?"

[night girder]

well, there goes their credibility out of the window.

They will get it back in a few years if they manage to get some future hardware stable

[pure karma]

Probably realized how expensive it would end up being

[twin dew]

And my electric drill gave up the ghost...

[twin dew]

Seems to be just the trigger unit...

[twin dew]

And I might just have been an idiot and not pressed down hard enough on the trigger, or tuned the limiter to minimum instead of maximum...

Because the marking on that limiter shows how much it limits, not the max speed allowed...
So need to be turned towards the thin end of the line, not the fat...

Yeah, works...

Ok, the commutator brushes don't get proper contact all the time anymore.
Need to see if just cleaning the housings helps, if they are just getting stuck from debris, as I don't have replacements right now.

Hmm... I might even have a spare set stashed somewhere that came with the machine...
Need to go find the instruction manual, should be in same bag if they exist.

[twin dew]

Didn't find, but I did find my other older worse drill, which works well enough for what I need it for right now.
So I can take the time to get some new brushes for the better drill.

And I had broken that ones wall connector during the storage...
So need to see about getting a new one, or finding those spare brushes, if they exist.

[twin dew]

And for final monologue bit, found old replacement wall connector for the cheap drill.
And contacted manufacturers spare parts support to ask for information to buy new brushes for the good drill.

[rustic panther]

@twin dew the Q6600 was a lovely cpu for its time though

That cpu combined with an 8800 GT was peak gaming in 2007

[night girder]

Ah good old classic. When suggesting Linux for gaming. Don't know why I even bothered bringing it up (different channel):

[pure karma]

i dont blame them

unless you only game and do web browsing linux still has a very high bar of entry and bad compatibility with software

thats how i came into linux when i used bazzite only for 2 weeks and it was fine but i dident do anything i knew wouldent work so its not fair either to say it was perfect if i only did the things i knew would work

[soft bloom]

Really depends on what software is in context.
There are lots of software that don't even consider windows as platform

All these fancy websites work on linux servers. Developed on linux. With tools written for linux.

[glossy glacier]

Not really anymore.
Very many distris ate very easy to use nowadays.
I think the biggest barrier is a mental one. It's not windows, so don't expect windows software to work. Use Linux software

[jagged snow]

Also, I think if you learned on linux you'd feel that macos had a very high barrier to entry

A gui is not inherantly easier to use than a command line interface, and as long as you have at least a basic understanding of how one works linux becomes quite easy to work with

[jagged snow]

@night girder to clarify: the concept of a GUI is not inherently easier to learn or better to use than a cli

In some ways, with things like man pages, cli can actually be easier in specific scenarios

[night girder]

Was just thinking about what you said.

[jagged snow]

Well by all means, share your thoughts

[glossy glacier]

Both GUI and CLI have their strengths and weaknesses

[jagged snow]

Oh of course

And for the things that most people do in their day to day lives, unless they really like markup syntax a gui is best for many tasks...
But modern linux can do pretty much everything in gui

And the stuff you can't, is arguably way easier than digging into windows registry or something

I also won't say that linux is better at everything, it has much weaker user management for enterprise users in a lot of ways

[soft bloom]

this should qualify as reaching vram limitations?
here's some DRAM read and write (that platoe is me alttabing)
yet there's no stutter, no lag, no terrible input delay
how

[jagged snow]

Because you could be dipping just a little into your system memory, that graph really doesn't tell us much

You could be storing just an extra 100MB in system memory with that setup

However, when you see huge performance hits it could be because you're having to push 4GB to system memory

[glossy glacier]

As much as I hate on windows, AD is great

[jagged snow]

Windows is just very good at being the default operating system

It has a lot of problems but it manages to work well for everyone from your grandma to secure fascilities and research institutions

It might not be ideal for any of them

But it works for all of them

[night girder]

Eh, don't like to give Microsoft a lot of credit. But they do have to keep in mind the different users for their OS.

Business, to my grandma, to power users, to gamers etc.

[night girder]

Why are ducks so popular? Found out there is such a thing as duckdb

a bit like H2 and SQLite db, serverless, broweser options etc.

[jagged snow]

Yep, I've heard of it

[night girder]

Good?

[jagged snow]

Never used it personally, so can't say
Somebody reccomended it as an alternative to mongodb

I use sqlite in almost anything I use that need a db

Or I build a custom system if needed

[maiden coyote]

I feel like old school mac os and linux were very close

[toxic dome]

Wasn't Mac a version of linux

[maiden coyote]

As far as cli usage goes, I'm not sure about current mac os.. we only have an iPad

Mac os was Unix based.. which linux was also based on

[toxic dome]

ah

what was old school mac like

[maiden coyote]

So navigating your way through cli is identical, but I think usage differs slightly especially with built in programs

[jagged snow]

Mac is based on the unix kernel, that's it

And unix is it's whole own world, it's not one unified os

Heck, even bsd started as unix

[twin dew]

Original Mac OS was purely GUI thing.
OS X and later (now branded as MacOS again) are based Mach and FreeBSD via NeXTSTEP.

[toxic dome]

i understood OS, GUI and Mac

anything else sounds like french 😭

[jagged snow]

FreeBSD is a version of BSD, which is short for Berkley Software Distribution. BSD refers to a group of Unix and Unix-like OSs that were originated by a team at UC Berkley back during the unix wars

They were meant as a non-proprietary but compatible competitor to Unix, which contained a significant amount of proprietary code. The first release of BSD was simply replacing what proprietary code remained, to make unix more open and free for people to use

[twin dew]

The other old implementation of Unix, with the other being AT&T Unix

As the original.

[jagged snow]

And Sun also had a version

A little later

[twin dew]

Those were forks from AT&T Unix via licensing.

[jagged snow]

Gotcha

[toxic dome]

Once I play satisfactory for an hour, I'll be an expert, we're good.

[twin dew]

Basically that AT&T Unix started it all, and then got forked into lot of manufacturer specific variants, and couple of university variants, one which was BSD.

[warm jay]

And when the Linux kernel come into that ?

[jagged snow]

Linux didn't come in until the 90s

It was originally written as a very simple rtos for one specific chip

[twin dew]

Linus wanted unix-like system, but nothing was available, and ended up coding one himself.

[jagged snow]

First release went out on the mailing list as basically just a curiosity, was only a few hundred lines

I say "only" because that's small by modern standards, but really is a lot of code

[warm jay]

Alr, I kinda forgot from when it came from

[twin dew]

AFAIK never as RT os, and Wikipedia doesn't say anything about anything like that either.
But as Unix-like kernel running on 386, as MINIX had very restrictive license.

[jagged snow]

Yeah, shouldn't have said RTOS as it wasn't RT scheduling
What I meant by that is just that it was a very simple scheduler and a basic shell

[twin dew]

Shell was just ported bash 1.08.

[jagged snow]

Yep, I'm aware

Basic as in "standard, not custom"

[maiden coyote]

I like how Google knows what roads are in private facilities

But that's kind of neat how long Unix has been around, and yet windows seems like the most popular os for enterprise for places we're people are inputting stuff into a computer

[jagged snow]

Windows got to be the default in the 95/xp days

And it's just held that position since

[night girder]

🫡

[pure karma]

depends how much you want to achieve and what for sure but yea i can see how a efficient command in certai nscenarios could be better than 5 sub menues stacked on top of one another

however my main problem is the lack of terminal documentation to go around

and when there is some its like you should already know 99% of this and if you dont fuck off

[pure karma]

thats fair but also i pay for most the software i use and when im paying for something i expect to be able to use it so "Just changing software" isent exactly entirely valid unless said software is free or easily replaceable

[jagged snow]

Are you familiar with man pages?

All standard linux commands ship with a built in manual that explains how to use them and what every option does

[pure karma]

like yea some things i changed easily for example i changed from task manager to mission center which is basically the same thing but more broken and less polished but in the end it achieves the same purpose to a level im happy with

[pure karma]

iv checked them out yea but i remember with ubuntu a lot of commands dident work and i kind of brushed them off after the basics

[jagged snow]

Gotcha

[pure karma]

and with bazzite i dident have any kind of reasson to go in the terminal because anything i would have needed terminal level control for would be way out of my experience level

the only thign that upset me with bazzite was not having the option for a centered task bar

[jagged snow]

Uh

[pure karma]

but i didn't care enough to try and make it myself

[jagged snow]

Bazzite runs plasma

There are so many plasma skins with centered taskbars

[pure karma]

i have no idea what plasma is

all i know is it wasent in the settings so as far as i was concerned as a normal user it dident exist as an option

[jagged snow]

Uh

So one of the coolest things with linux(KDE Plasma specifically) is that you can completely customize your interface very easily

Bazzite has an app for it(I think it's called the Bazaar or something) but it's pinned in the app menu by default, and lets you super easily install thousands of different skins for your desktop

[pure karma]

the bazzar is only a few months old but if i recall (since it had only been out for liek 2 weeks when i used it) its just essentially the microsoft store

[jagged snow]

Must not have been that one then

[pure karma]

i dident see anything customization related only app sugestions for various things

dosent really matter since i got rid of my bazzite install a few weeks ago since i had no use for it but i definitely want to return to it at some point

[soft bloom]

I turned on framegen and now I want to ask - why would some items on converyer belt look much more smeary then opthers?
bolt buckets seem fine, but rotors almost melt

could be just that blue color with no detail is much less notecable then rotor edges

[thin trout]

It depends on the contrast, angle, distance etc. But that looks far more awful than any framegen artifact i've ever seen in satisfactory. Is that FSR FG? What's the base framerate? Any upscaling?

[verbal raft]

is that there without FG ?

[bold tapir]

How do i pick a new phone?
I have experience with pc parts but when it comes to phones any review is advertisement. I just want my phone to have good Bluetooth for me to enjoy podcasts all around my flat

[verbal raft]

honestly doesnt look like FG artifacting
looks more like an upscaling artifact

or lumen

[verbal raft]

so more of a temporal artifact i belive

XeFG 36 to 72

1st is a real frame
2nd is generated

[soft bloom]

upscaling yes. fsr indeed

[soft bloom]

no

I don't think it was as bad when fg was first introduced to sf

[soft bloom]

does fps without fg qualify as 'base'?

[soft bloom]

lumen is on

[verbal raft]

mostly

[soft bloom]

otherwise no idea how to provide number

[verbal raft]

but FG's real base is half of whatever you get after FG

[verbal raft]

that wont help either

[soft bloom]

I was like 'ok what would it feel like to try reducing vram use by setting textures to low'

and also turned on fsr

[verbal raft]

FG increases VRAM usage tho btw

[soft bloom]

wires don't look... sharp, but they don't behave like wavelets either

[verbal raft]

is that a generated frame ?

[thin trout]

you never said what base framerate is

[soft bloom]

hmmm, I definitely didn't see that smearing on rotors before FG (yesterday and day before)
but just runing fg off doesn't make em look articularly better. they seem to jump back and forth

[thin trout]

what about upscaling off

[soft bloom]

it seems to be 30(?)
when I turn camera it jumps to 48 🙂
no clue how that happens

[soft bloom]

eh, no clue

[thin trout]

yeah that is way too low to use fg

ideally get base framerate to 120, but 100 is okay

[soft bloom]

no fg
changed background

[thin trout]

that's a bad temporal effect of some kind, other than FG then. Likely upscaling or TAA.

Framerate is too low for temporal stuff to work properly.

[verbal raft]

what level of upscaling is that ?
and which upscaler ? (is that even FSR 3.1 ?)

[soft bloom]

performance (huh, forgot I set that yesterday, maybe that' the culprit?)
but not much better with balanced

[verbal raft]

could just be lumen too

[soft bloom]

SF only tells me that's FSR

[thin trout]

what target resolution?

and yeah, lumen is temporal too (and really bad in satisfactory)

[verbal raft]

asked wether its 3.1 cuz it thought you may be using optiscaler
which lets you switch out the version of FSR

[verbal raft]

1st real
2nd fake
FSR FG 36 to 72

footage:

[soft bloom]

so no upscaling here
they still jitter, but not melting

[verbal raft]

lumen still enabled here ?

[soft bloom]

I need it for aethtetics reasons
half my factories are passively lit by displays...

[thin trout]

reset to default settings maybe. You have a really bad setting somewhere and IDK what it even is

[soft bloom]

yes, at High

[thin trout]

but it looks like a broken temporal effect due to misconfiguration and/or too low framerate

[soft bloom]

how do you capture 2 frames one after another?

[verbal raft]

i use ICAT
it lets you go forward frame by frame

[soft bloom]

global illumination off

[verbal raft]

whats with the constructor too ?

[soft bloom]

could be just low textures making it looks extra smudgy?

[soft bloom]

dunno.
seems fine upclose

defaults set ultra textures lmao

maybe it does look sharper...

nah, still looks like rotors dublicate (it set FSR upscaling to balanced)

something weird going on

[soft bloom]

1440p

[thin trout]

do you accidentally have AFMF on or something (it's auto-on sometimes)

[soft bloom]

afmf?

[thin trout]

https://www.amd.com/en/products/software/adrenalin/afmf.html

[verbal raft]

heavily doubt it'd be captured by the snipping tool

and only relive can capture it
OBS, other recording tools cant

BTW i think ive found DLSS 4's secret juice
RCAS

okay its certainly a bit too much sharpening here
but there is actually a tad more detail with RCAS enabled than disabled

[soft bloom]

no idea if it would. first time I hear about it

[soft bloom]

not even an option

[tribal kraken]

Anti tree Harvester 🌲🌲🌲

[soft bloom]

but it doesn't harvest anti-trees!

also didn't know someone needs to harvest beavers

[tribal kraken]

I've operated 3 years Beaver harvester 😄

For some reason this manufacter chose to name the models after animals. Smallest harvester is Ponsse Beaver.

[visual tree]

Guess I won't be doing most of the work today

[mental oriole]

[wanton orchid]

it's called unreal engine

[foggy tangle]

hey guys im here for something lost

I lost my save (perma deleted) and have no way of recovery

[umbral marten]

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: fightorflight

[foggy tangle]

im running the 4th recovery app check to see if its thhere

[jagged snow]

Check the save backups folder

Make a post in #1038092680493801533 and we'll see if we can help you out

[foggy tangle]

not there for some reason

don't think how a forum might help. I'm nearly certain it's lost

[verbal raft]

this channel is not for SF related help

[jagged snow]

This is the wrong channel though, specifically for offtopic stuff

If you want more help with a satisfactory related problem, go to the right place

[foggy tangle]

it's not really about SF

well I did lose a save there but the game this is about is mc

[jagged snow]

Ah, I assumed it was satis because of the server

Thx for clarifying

Do youy have any windows snapshots or restore points?

[foggy tangle]

oh yea no like 1-2 years ago my SF save got recovered

[jagged snow]

Also, how did you lose the savefile?

[foggy tangle]

snapshots? If windows doesn't auto backup the last one was a year ago

[jagged snow]

You might have automatic restore points, windows usually takes them whenever you update

[foggy tangle]

It was my hardcore save, I was changing saves. Since renaming the file itself (adding numbers) doesn't show up the correct name in minecraft I thought the one file was the right one which it wasn't and deleted the wrong one

[foggy tangle]

ill check rq

[jagged snow]

Did you check your recycle bin?

Unless you did shift+del it should just be there

[foggy tangle]

the last one was 9/28 so no

[foggy tangle]

I have a stupid habit of shift + delete so nope.

[jagged snow]

I mean that would get you the file as of three days ago

Probably your best bet

[foggy tangle]

I've did alot during those 3 days (nearly everything necessary)

wait

well there's one actually yesteday

[jagged snow]

Awesome

[foggy tangle]

oh nvm I forgot that the dates are not named like that.

[jagged snow]

Hm?

[foggy tangle]

reference

hold on nvm I got confused the time I got confused

[jagged snow]

If you're in the states windows defaults to MM/DD/YYYY

[foggy tangle]

Yea its MM/DD

[jagged snow]

Be warned

System restore is technically only supposed to be for the os, so it doesn't always capture personal files

[foggy tangle]

another bet would be to try maybe getting my regions back via the xaeros minimap mod?

[jagged snow]

Before you try loading that, make sure you check and see if you have file history enabled

[foggy tangle]

Ill check

off

[jagged snow]

Dang

You can try system restore, but that will perform a bunch of writes on your disk and probably eliminate any chance you have of using file recovery software

So if it doesn't work, it means you're SOL

[foggy tangle]

SOL?

[jagged snow]

Straight Outta Luck

[foggy tangle]

thanks needed that.

[jagged snow]

Lol

Just try other stuff first until you're sure it doesn't work

[foggy tangle]

I wish I would've installed the autosave mod I was thinking for 3 days

[jagged snow]

It's just a very unlucky combination of things

Shift+del and no file history

[foggy tangle]

wait.

I have 2 save files named after my Hardcore save

and I had 2 files before the delete

ill check both just gotta know which is which this time.

[night girder]

@mental oriole Two weeks later. LinkedIN replied.

My question:

Their answer:

[jagged snow]

And please don't delete anything else

[mental oriole]

They still didnt reply to my ticket because it was an AI response 🤣

[foggy tangle]

didn’t check the date of the file tho :) I’ll do it after I get home gotta buy sm.
Assuming it’s the correct file I gotta thank you, I can swear I nearly lost hope till you cane

[foggy tangle]

Update: I need to buy the better version of the app because both the files are 100MB+

[jagged snow]

Oh lame

There should be other unpaid software that can do that without you needing to buy it

[foggy tangle]

I tried one but it didn't scan the same save file

[jagged snow]

Gotcha

[foggy tangle]

If you have one that gurantees a recovery of it im happy

tho im not sure which is which so I get the <100MB one first

[jagged snow]

Nobody can really guarantee file recovery

But I've only ever used PhotoRec and had a great experience with it
Doubt it'd work here though

[foggy tangle]

why do you doubt it

Tho I think it's obvious which is one is which.
37.5kb or 359MB?

[twin dew]

And if on SSD, basically nothing can recover files if TRIM was in use.

[jagged snow]

It's designed for specifically media files

And works by looking for headers/metadata

[jagged snow]

360MB seems a little large for an MC save but if it's been around for a bit that doesn't seem unreasonable

[foggy tangle]

I know, ill get the other file because my 1GB RAM brain somewhat recalls seeing a number while deleting

[foggy tangle]

what is a TRIM? Does it perma perma delete files?

As far as I know perma deleting just forgets the path to the file and waits to be overwritten

[jagged snow]

TRIM is a command that specifically tells an SSD that data blocks are not in use and can be fully erased

It allows the drive to go back to using those blocks for SLC mode caching or other things

[foggy tangle]

well thats unusual

[foggy tangle]

does it auto use TRIM or is it a manual interaction

[jagged snow]

If enabled, it's done automatically

[twin dew]

And on Windows, it is by default enabled on all drives that support it.

[foggy tangle]

NTFS DisableDeleteNotify = 0 (Allows TRIM operations to be sent to the storage device)
ReFS DisableDeleteNotify = 0 (Allows TRIM operations to be sent to the storage device

this is what I got back from fsutil behavior query DisableDeleteNotify

[twin dew]

Which is the point that on SSDs, old style data recovery software will find nothing, as SSD will just return nulls for any deleted files after that on by default TRIM is sent as part of the file delete by OS.

Even if the actual cells haven't been erased yet, just that the linking between LBA and cell was cleared.

Trying to undelete only works on CMR type HDDs, even SMR HDDs use TRIM.

[foggy tangle]

so it means that recovering it has a very low chance?

[foggy tangle]

I think my save is gone

[pure karma]

ehh for bedrock its fairly normal because of the higher render distances loading more chunks that then need to be saved

like my world was over a GB

[jagged snow]

Iirc that's not how mc saves work actually

I recall something about only saving a chunk once it's modified and otherwise just regenerating from the seed again on next load

Could be totally wrong though

[pure karma]

loading new chunks still adds loads of data to the save even when not modified so if it is true its only partially

[jagged snow]

Probably incorrect then

[steep zenith]

@feral walrus

[pure karma]

Yea no kidding

Sli, 4 sticks of ram, random bullshit everywhere, 463 drive bays you cant use

a cooler thats under compensating

But most importantly the random canadian doritos bag

[verbal raft]

AMD finally doing something first ? https://www.youtube.com/watch?v=VIU3xmxzuRk&t=4s

[night girder]

My monitor issue isstill interesting.

1. I turn my monitor off during the night for about 10 hours.
2. When I turn it on, the bottom of the screen is 1/3 or 1/5 glitching out. Posted some screenshots.
3. If I set refresh rate to something else it gets fixed. (software side)
4. If I wait for a few minutes it gets fixed. (doing nothing, grab a coffee)
5. When screen is turned on for longer period, no issue.

I am ready to buy new monitor once this dies, but I wonder what part of my monitor is broken?
I have a hunch it has something to do with cold hardware 🤔 And also something resomsible for refreshrate.

[twin dew]

Something in the monitor electronics driving that part of the monitor starting to fail, and need to heat up before working at high frequency.

[twin dew]

Commonly LCD-panel driving is spread into bands, and you have some of the lower down ones glitching.

[night girder]

That explains it. I had a hunch it had to do with heat. Wonder if I leave my monitor on, I "expand" it's life.

But then again, I could set this one to 60Hz and use it as reader monitor and buy new one.

[twin dew]

Depends if the degradation happens from thermal cycles, or just use time.

[night girder]

I am not sure, it's not possible that it comes from sending 144Hz over DP?
So when fidling around, I found out that my monitor was locked to 60Hz (in monitor menu).

And it was the DP cable. Replaced it with a HDMI and got 144Hz on monitor. But still the issue with the lower band of my monitor broken.

but I looked it up and I read that monitors should be able to receive higher refreshrate then set too.

[visual tree]

Can't wait for 10th of October = Tron: Ares + Little Nightmares 3 release

Although there is a pre-premiere on 8th of October here and I hope the movie won't suck....

Always thought pre-premiere was reserved for a specific group but turns out it's public and anyone can buy a ticket

[pure karma]

rare that you get to use 200% render scale at 60FPS nowadays but its actually feasible here at native no FG

actually decent optimisation for once

[verbal raft]

what game ?

[pure karma]

stray

not super new but still post enshitification

[fickle gust]

Chat fsr4 isn't working for me

Using 7900xtx any ideas

[verbal raft]

Wrong DLL ?

[fickle gust]

Nope I removed it and reinstalled it

[fickle gust]

I did verify files tho

It found something missing maybe it fucked with optiscaler?

[jagged snow]

After the repaste I'm much happier with cooler performance

It'll run at around 80w instead of the 50 or so I was seeing before

[pure karma]

yes

controbute to e-waste at all costs

Also i cant tell if i cooked with this move or not

i mean the AIO sure is cooking with those 45C liquid temps from the exhaust being blocked off but its kinda worth it i cant lie

i was trying to find a way to make my arc pc more interesting because its literally just a rgb less black box of boredom and then i remembered the random lego car i had holding my ethernet cable in place would absolutely kill on top of it

i also found that the arc branding mirrored is a absolute masterpiece

introducing the cralejni 0828 the ultimate gpu

arc is so solid already its impressive how little time it took intel to catch up for hte most part

i mean 2 years or so to go from potatoe igpu's that cant even do 720P20fps to stable dedicated GPU's with the best price to performance in the range is wild

[safe trench]

because sheeple

they want to be on the latest trends
if comp setups arent an example of that i dont know what is
minimalism is just rich people pretending to be poor

[safe trench]

also taught me that there is a machine shorter than linus sebastian

you know where this is going

i swear he had a body pillow or something

in one of the anti 5g painting videos for his house

pretty sure they guy that left recently whos last video was the firetruck sound system made his own video on why he left

expect
k95 > life!1!

[safe trench]

none of the sheep will buy it

because "less is more" and its "too big"

people want a 4xl from lttstore

i used to think 3xl is big enough for every single person

[languid gulch]

https://www.youtube.com/watch?v=5DzZhIs9bj8

[languid gulch]

i fall under "bigger than the average american" because of my height, but not insanely so, and i'm at XL for shirt sizing

also one clothing company's XL is someone else's L or XXL

[bronze jasper]

i got this one too. i never use the macro bank anymore though

i probably won't buy another corsair keyboard when this one needs replacing. corsair are patent trolls in the gaming realm of the worst kind. Controller patents.

want to put buttons on the back of a controller? you need to pay corsair for that right. they sued Valve for making the steam controller without permission

i want one of these new hall effect keyboards. see if i can make it last 10 years. i think i got my k95 abotu 6 years ago and the w key and r key are starting to need a little firmer pressure now an then.

it's so hard to find a keyboard with macro keys these days that i might just go for one with a 10key and buy a sepearate device for macro buttons

yeah logitech an corsair are the two brands that consistently add macro banks to their products. compact keyboards are really popular these days so everyone makes those i guess. i like my 10 key and page up buttons

i'm not buying anymore razer products lol. they're made to wreck in 3 years tops

thats if you're delicate with them

stream decks aren't so bad. they could act as a good macro bank and they got cute little screens

i've used 1000dpi mice before and didn't like them at all. i got a mx master 2s right now. i like the shape and durability of logitech mice and i don't think i'll try other brands again

if i ever see the new mx master 4 on sale i'll pick one up probably. i don't mind the low dpi on them. i'm not really a high dpi gamer

the haptic clicking is a really neat idea that i want to try out at least

"stream decks" are the name of what i'm thinking of for an extra macro bank. there's some that have status displays like stream lag meters and knobbies you can map to different things. and the buttons all got mini displays for custom icons

lol i got a steam deck too and i really like it. some games are just better with gamepads. i prefer mouse an keyboard but i've always been a chameleon iin that regard

playing silksong on the deck lately

i like the steam deck's joysticks a lot. xbox has always been a reliable one in the past. i dont like the switch pro controller because theres no analogs. The dualsense and dualshock 4 has wicked good gyros on it if you want to experiment with gyro aiming in games

steam controller's joysticks were good too but i could never get used to the touch pads

the gyros are some of the best available and they integrate with steam input very well

i wouldn't have thought race would be part of that at all until you mentioned it

i've never expected companies to be perfect. but there's one thing i hate. patent trolling. which i'm struggling with lately cause i really like nintendo and they're patent trolling with software patents these days.

probably. they lke major acquisitions in the gaming space. i don't think the AI situation has made them invest less in games. they're still making big bets on gaming today while AI is going on

mostly because he's been actively giving away his money through his and melinda's foundation though. people like bezos and musk don't do philanthropy

[full tide]

🎄

[bronze jasper]

starship isn't a finalized design. i don't think it'll ever be used to take a million people to mars though. that's just another hyperloop plan by musk

it is a badass rocket at least. saturn v is still my favorite

[verbal raft]

Still make sure you have the DLL with the int8 model, the DLL with the official model will not work

[bronze jasper]

i dont think starship is using the raptor 3 in production yet. they're still iterating the system and testing other aspects of it before scaling it up with more engines

[twin dew]

12th Raptor 3 just left McGregor testing range.

Sorry, 12th U3 variant. (Upper stage?)

[bronze jasper]

nice. once they dial in the r3 design they'll start mass producing them presumably

[twin dew]

But not being used yet.

At Starbase.

Next IFT-11 will still use v1(?) booster and v2 ship.
And IFT-12 seems to be planned to be v3 for both.

[twin dew]

Ok, should be v2 booster too.

[bronze jasper]

we weren't "made" to live in the arctic either but the inuit do it. people adapt.

i dont think we'll build a species saving outpost on either mars or the moon, but we will definately get bases and probably mining operations at them.

using mars for asteroid mining HQ would make sense

[verbal raft]

novideo experience

[night girder]

Remember a few days/weeks I posted about discord and their age identification asking for ID?

Discord got hacked And leaked a few of those ID's. Oh noes. Only a small portion.

[twin dew]

Not any IDs?
Only information provided to third party support by Discord, for users that used that third party support.

[night girder]

goverment ID images.

eg. driver license. pasport.

"from users who appealed an age determination"

[twin dew]

Ah, yeah, in the bottom, so some of the cases of those were handled by that third party support then, in case of disputes or like.

Yes, I skipped that part originally somehow.

Third party support got hacked, and whatever that company had access to was taken.
But some places were touting that lot more was taken earlier, in complete contrast to then provided list, which didn't include that addendum yet.

[night girder]

So for everyone who either thinks;

• People are too paranoid I trust my personal information with this tech company.
  or
• Tech companies are high level security stuff that never get hacked and careful handle our personal data.

You're wrong. This is one of the the reason people are pushing back against this.

[twin dew]

Example from yesterday: https://x.com/Pirat_Nation/status/1974265627838763033

Discord was breached two weeks ago.
Attackers accessed government ID photos, contact details, payment information, and more.
When the pic of list explicitly says that full credit card numbers or CCVs weren't included, only last 4 digits.

And IDs weren't mentioned at all.

[night girder]

I think legal department kicked in. And said; you have to mention that.

Or they took a risk assesment of between mentioning the ID's or not.

[twin dew]

Yeah, but that that users own message was basically complete nonsence compared to pic he included about message from Discord.

[night girder]

I think they really though for a bit if they can get away with not mentioning the id's lol.

[twin dew]

When it was Discords human support subcontractor company that was hacked, not Discord.
And the data was much more limited than what that user spouted.

[night girder]

not sure where the message comes from. Seems different layout.

[twin dew]

Email to affected user almost certainly.

[night girder]

probably mail.

Yeah. Let me check my mail.

[twin dew]

One who's ID wasn't in the dump.

[twin dew]

If you haven't interacted with Discords support in last X months, you aren't in.

[night girder]

Ah, sad for the people 🙁

But preventable. (By the company imo).

[twin dew]

Yeah, the support company is completely in the wrong, and Discord partly, by selecting to use that company.
But the damage is pretty limited.
Some mostly non-actionable information about anyone who interacted with that support in specific timeframe.
The actual harm is the dispute cases for age verification.

With leaked ID document images.

[night girder]

Uhu.

[twin dew]

And Discord is in the wrong that the email in the Twitter post didn't mention IDs at all.

Even if that specific users ID wasn't leaked.

But that Pirat_Nation basically didn't read the email either.

[night girder]

The name alone ...

Btw, I was thinking of upgrading from 1080p to 1440p. If anyone has monitor recommendations.
I leave out a lot of details, because if I ask for too much, people question everything 🤣
4k is a no go on this system.

[dire igloo]

ahem, Cambridge Analytica, ahem

[mental oriole]

Because the compromised accounts on the support side asked to send photo of ID via email.

[dire igloo]

How do people constantly forget about Cambridge Analytica?

[dire igloo]

Depending on where your company HQ is or which country's users are affected, it can be legally mandatory to disclose the extent of a data breach

[night girder]

That samsung is 140 euro damn!

240HZ and good response time. Didn't look at other stats. And IPS.

25" tough.

[twin dew]

If the support company had gotten those IDs directly via email, and not via Discord, then Discord might not even have known about the IDs when the original emails were sent...

[night girder]

Wtf, even for contrast it stands strong.

[dire igloo]

Upper range for IPS

1k+ is impressive

[night girder]

140 euro upper range?

[mental oriole]

All I know is because my bro was affected by the hack (he did not send any photos or anything tho, so mostly unaffected) And they wanted things to be sent by email.

[night girder]

The samsung does seem to failw hen it comes to colors. "Fail", it's weak point.

[dire igloo]

No, a contrast of 1000:1 is the upper range for IPS panels

VA obviously goes higher, but for IPS, that's actually damn good

[twin dew]

Yeah, but point was that those then weren't even IDs used in the Discord verification inside client/site that were leaked, but the support company asking for ID via way that is completely unacceptable from security standpoint in any way or form (email).

[dire igloo]

140€ for a 1080p monitor from a couple years ago isn't too surprising tbh, but it's nice to see it perform so well.

idk the testing quality of tweakers tho, would like to have it confirmed by a better reviewer like Monitors Unboxed or rtings (latter ones didn't have it)

[mental oriole]

I know, those are two different things. Email has little to no security.

[night girder]

Problem is, I don't want 25". Else it might be a candidate.

[mental oriole]

What size you looking for?

[night girder]

I already folding by going 1440p from 1080p. I don't want to give up 27" and I don't want AOC again, I think.

[mental oriole]

Once you go big screen it's hard to go back 🙂

[night girder]

And I say I dont want AOC, because I bought this one in 2022 and I feel like 3 years isn't a lot for a monitor to die.

I haven't used it more or less than my older monitor which works fine.

[mental oriole]

3 years is nothing.

[night girder]

Right? And the lower band of the monitor is busted.

It needs a few minutes to wake up and start displaying.

But eventually, I fear that will not wake up and stay permanent dead. So new it is.

Might give this one to my mom or something 🤣

[mental oriole]

[night girder]

800 euro ... TN?

[mental oriole]

wtf

[night girder]

1080 24"

[mental oriole]

oh it's 540hz

[twin dew]

Taking the money from the people going for excessive refresh rate and wanting Asus for some reason.

[mental oriole]

don't...

[night girder]

still ...

I am playing a game right now, fully unlucked, full throttle, 99th% is 130FPS

am I dumb? I don't see it. Unless I had like godlike hardware.

[mental oriole]

240hz is the high bar for me.

But would easily settle at 144/ 160

[night girder]

My head; I have a glass that can hold 540 liters. But I have a bottle that can pour 144 liters. The glass is very expensive because it can hold 540 liters. But I can buy cheaper glass for 144 liters

[mental oriole]

yee pretty much

[night girder]

Thanks Diluted for being my sanity check.

[mental oriole]

[twin dew]

Basically for FPS type esports "professionals", for CS:Go etc.

[night girder]

Right... I still have opinions about that but I withhold 😄

[dire igloo]

Which ironically would often run equally as well if not better on an OLED

[night girder]

What resposne time is acceptable to gamers? 🤷‍♂️ (opinions?)

[twin dew]

144Hz is completely acceptable to me.
240Hz would be the max I might be in theory willing to pay for some extra.

Basically, how many games do you play where you get even that 144Hz with the settings you run it at.
Having monitor faster than the frame rate, with variable frame rate, doesn't basically give anything.

[dire igloo]

Anything above 10ms average is bad imo

iirc you'd want the pixels to be done responding before the next frame comes along.
Meaning: 1 divided by refresh rate

maybe you can find some info on that from the reviewers I mentioned earlier, they know more about monitors than I ever will

[mental oriole]

for 144hz its anything below 6.9ms (usually those monitors have 1-5ms ranges)
for 240 anything below 4.1ms (1-3ms range usually)

[night girder]

So 3ms for 160Hz isn't bad. 👌

Ugh, this is going to be a pain in the butt. Comparing all these monitors 🤣

[dire igloo]

Are you talking about advertised/specsheet response times there?
Cuz those are WILDLY misleading and far below what those monitors commonly achieve

[night girder]

ASUS ROG Strix XG27UCS has good colors, good white, good black. Low resposne time. Low energy usage. Or so it seems to me.

[twin dew]

Basically all with same spec have exact same panel.
With differences coming from the exact driver and firmware features offered.

[dire igloo]

Depends on the transitions they measured tbh

[twin dew]

The amount of different panels produced today is very limited.

[night girder]

So it scores really well. And average on response time.

[twin dew]

So backlight, driver electronics, firmware features, physical casing.

[dire igloo]

Uhhhh, not really.
There are a few outliers (like 28" 4K or most OLEDs) but generally speaking, there are always a couple panel models around for the commonly used resolutions/sizes/refresh rates

[night girder]

And not the best white value.

404.95 euro though 🙁 Yikes.

[dire igloo]

Also panel binning.
Lowend brands like Koorui often use panels that hot rejected for higher end product lines

[dire igloo]

spiritfader.github.io/Buying-Guide/monitors

[night girder]

Ah wait, black friday is soon mhh.

[twin dew]

Yeah, binning is a thing, but also leads to panel lottery when buying any monitor to some extent.

[dire igloo]

A list of known good monitors and reviews for them.
Starting at 27" QHD

[dire igloo]

Prime day is next week.
Black Friday only comes around late November

[night girder]

Did you see it? The one I posted is in the list you posted. 😏

[dire igloo]

Also: you may not get a deal on a good monitor come Black Friday

[twin dew]

But what are the options for for example 27" 1440p 144/160Hz IPS, except specific LG "Nano-IPS" panel?

[night girder]

What do I lose? Nothing but time.

[twin dew]

Now replaced with that 240Hz LG IPS panel at same size and resolution.

[night girder]

So if no deal. Nothing lost. If deal. money gained.

[dire igloo]

But you mentioned its price - so I gave you a list of several other good monitors

[night girder]

the other ones in same category are 2.

[dire igloo]

There may be a good deal on a good monitor now that will have ran out by then.
Stock may deplete more quickly on Black Friday cuz everyone will be buying.

Waiting for prime day is fine since it's only a week.
Waiting for BF imo not worth it

[night girder]

Seems to be price bracket is kinda the same. Besides Giga. Which is sus.

[twin dew]

And as LG seemed to be the last one producing gaming grade IPS panels, them dropping IPS and consentrating on OLED is just so bad...

[dire igloo]

Btw, there's only a single 28" 4K panel on the market, so variance across monitors with those specs is less extreme

[night girder]

I honestly don't need 4K 😒

Ah, yeah about the gigabyte.

[dire igloo]

Can have plenty explanations: older panel, less demand, less incentive for price gouging

[night girder]

That asus just scores really well. Got my interest.

[twin dew]

Both Samsung and LG have completely moved to OLED variants for future 😭

[dire igloo]

Innolux, AUO, BOE
Tho iirc the latter ones are facing punishment from US legislators for idk what

165-180 range is common now, 144 has died out

[night girder]

Lenevo has monitors..

Oh great review lol.

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

[dire igloo]

Lenovo R27qe is good, idk about pricing

[night girder]

Listen to this boys 😄 (review from Fireworker url)

[twin dew]

Several LG Nano-IPS panels were 144Hz originally, then "OC"d to 160Hz, which then came the norm, without any real change in the panel.

[dire igloo]

You don't listen to it, you just look at the data and draw your own conclusions.
There is no language barrier in numbers

[night girder]

Yeah, nobody has an interpretation of numbers that can be wrong.
You right. All the review videos should just show data. No voice. No opinions. No guidance.

[dire igloo]

Yup, and from what i can see, modern panels are 165Hz base and OC to 180

[night girder]

Just a powerslide.

Eh, I hate picking monitors.

Maybe not for a sunday.

[twin dew]

But that the last "premium" IPS developer (LG) stopped even production is just sad.
Because those chinese manufacturers were just copying best they could current and previous gen LG panel tech.

Because while OLED has that infinite contrast, it still has its own issues compared to IPS 😩

[dire igloo]

You can always draw bullshit conclusions, but if the data is good, the accuracy of your conclusions are up to you.

Tweakers could say anything about a monitor, but if you trust their data, you don't need their verdict

[night girder]

like burns?

[twin dew]

Brightness evenness, color evenness, even more as the panel ages etc.

[dire igloo]

For QD-OLED you got glare in bright environments

[night girder]

Why can't they make the ASUS ROG Strix XG27UCS Zwart ... but just 2K ... with same results

[twin dew]

Burnin is being "handled" by tracking individual pixel usage time and changing the driving voltages based on that based on average dimming from the use.
And running static from time to time.

[night girder]

Do they do this on purpose?

Viewing angle.

Tbh, I still think that ACS is solid. Altough being lower. The difference is minimal it seems.

[dire igloo]

Running it through a basic tweakers filter, R27QE is the 2nd cheapest known good monitor in Belgium and the cheapest in the Netherlands.

Only one cheaper in BE is Gigabyte GS27QA which performs well but lacks severely in ergonomic adjustability.

tweakers.net/monitors/vergelijken/#filter:TY1BCsIwEEXv8tcROkmaxhyg4MKVS3FR2iwirS1JEaH07s4oiKs_vHl_ZsOch5jbFMcBAUtO9wL1hZc5r8y60v_IEvsTewdSovbxnB4IlULhRZvGNeaCsIGsdpLPbkS4wjSGcNsVyNUkfJIaW5YPT92LZ-NtBTGsqf-ampxmR2tPnpO8PdpP1uSEG--EO980_GDf3w