🔌│tech

[gusty kiln]

Ultra vs high

Is upscaling on for hwu?

[astral haven]

obviously not

[vagrant pagoda]

they got a bigger spread on 1080high too

[astral haven]

yea my bad

[vagrant pagoda]

smh cant even screenshot without the ui XD

[astral haven]

speak for yourself

[vagrant pagoda]

XD

i didnt even see it lol

i didnt shift fullscreen so it was only part of my display

[astral haven]

and i didnt even fullscreen because im omega lazy

anyways point is even if they were the same fps, we could still say they are gpu limited if we know that the gpu is hitting 99.9% utilization

[ashen spindle]

the game seems to like tight timings but the HUB video doesnt show timings anywhere

[vagrant pagoda]

yeah but he linked gns memory test

and on gns gpu test at high/1080 4080/4090 got same fps

[astral haven]

ion

i believe you are confusing "limited" with "bound"

[gusty kiln]

7200cl34 for 13th/12th gen intel

[astral haven]

bound means no matter how much you improve other factors, x bounds your fps and prevents it from increasing

limited means improving it will increase fps and nothing more

[ashen spindle]

cl34 is the least important one almost lol

needs the rest

[gusty kiln]

Looks like 34-45-45-115

If they are only using xmp

7200 is showing more scaling compared to the 6000 that GN used

[vagrant pagoda]

ive never heard it used like that

bound and limited are the same lol

used interchangably

[astral haven]

🤷 and now you have a chance to learn

[vagrant pagoda]

i mean

you saying it dosent make it a standard

[astral haven]

people use literally incorrectly quite frequently, doesnt make them correct either

[vagrant pagoda]

also

both of your definitions mean the same thing

if something is bound

then making that faster will still increase fps?

[astral haven]

if you think that is true, you misunderstood what i meant (and im pretty sure i stated it precisely enough to have said what i meant)

[vagrant pagoda]

no i dont

because its something you pulled out of your ass lol

[astral haven]

correct
but nothing else will be able to meaningfully make it faster (if it is bound)

[vagrant pagoda]

show me somewhere that backs your useage lol

because ive never seen anyone use them differently

they mean the same thing

[astral haven]

https://www.google.com/search?q="cpu+and+gpu+bound"
https://www.google.com/search?q="cpu+and+gpu+limited"

observe the differences

a thing can be cpu and gpu limited on its own, but most of the folks talking about being bound are talking about telling the difference

[vagrant pagoda]

none of those use both differently

they both mean bottlenecking

[astral haven]

thats implying that bottlenecking is a unique point in the performance space

bottlenecking is a massive set of things

varying from the negligable impact on performance to the extreme impacts on performance

in fact its possible to have a bottleneck that has zero impact on performance

[vagrant pagoda]

again find me someone else that differentiates limited and bound for cpu/gpu/memory

[vagrant pagoda]

HWU and GN use them interchanably

[vagrant pagoda]

thats just limited

[astral haven]

and ian pretty carefully tore down their objectivity

[vagrant pagoda]

?

[astral haven]

they dont provide rational reviews, just (useful) subjective ones

they just pretend to be more rational than linus while making fewer errors

(and less obvious ones too)

[vagrant pagoda]

that has nothing to do with thme using terms

[astral haven]

https://www.intel.com/content/www/us/en/developer/articles/technical/locate-and-resolve-cpu-gpu-bottlenecks.html
heres a nice technical definition

[vagrant pagoda]

that only uses bound

[astral haven]

no it just means that they are likely to make errors like literal vs figurative

[vagrant pagoda]

you cant understand english or youre ignoring my request

find me someone else that differentiates limited and bound for cpu/gpu/memory

dont link something useing one term

some thing use BOTH and using them to mean different things

[astral haven]

read

[vagrant pagoda]

limited is used one time

[astral haven]

if you ctrl f to limited, you can find a section on how gpu bottlenecks can be cpu limited

[vagrant pagoda]

and isnt used to descibe something different than bound

also

that is using limited opposite of what you said

as its saying both gpu and cpu can improve

[astral haven]

exactly!

[astral haven]

i think you misunderstood what i meant by nothing more
i meant there was no more meaning you could attribute to it, not that no other factors could improve fps

so like the car being on fire means that you can tell it is going to have problems and nothing more

it could have few or many problems but you cant exactly tell what problems it definetly will have

if limited then increase fps by changing it (and maybe something else)
if bound then cant increase fps without changing it

[gusty kiln]

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

[astral haven]

how long is this gonna last lel

[gusty kiln]

[astral haven]

last gpu bench had the 1060 outperform the 2080ti

[ashen spindle]

looks like cinEbench-cores wont do well in this version, time to call it amd sponsored

[astral haven]

avx512 should definetly give amd a lead in anything rendering

amds avx512 is pretty fantastic for the lack of extra hardware they put into it

[ashen spindle]

not just that but it's avx2 minimum and the e cores suck at even the vectorized instructions they have in comparison

[astral haven]

i mean these arent independent

also id be shocked if 2 extra or even 4 extra p cores would offset the loss of 16 e cores

like on the one hand sure a single e core is worse than anything else we have
but 4 e cores vs 1 p core?

4 e threads vs 2 p threads?

taken to the extreme, 40e vs 8p?

[ashen spindle]

no clue but I imagine funny numbers and that we see it used on zen5 release if it works that well in their favour again xD

[astral haven]

wdym again?

amds better density has had them keep the advantage this whole time no?

[ashen spindle]

intel removed it in favour of "real world" benchmarks and I thought amd also didnt show it for zen4 but maybe I'm mistaken

[astral haven]

sure?

cinebench is an awful predictor of performance outside of rendering apps?
what is new?

[ashen spindle]

I mean its not new, hence it will just be funny if we see it shown again after it was not used is all

[astral haven]

possible
i suspect they both removed it due to community backlash not really out of any particular malice?

[raw zephyr]

i can get the pg27aqdm for 930€ brand new, really considering it

[ashen spindle]

oh yeah its not malice its the usual marking to just put whats most beneficial into the showcase, it's just a funny cycle nothing more

[gusty kiln]

Wished they removed power play in the v2

Might have gotten it like 10grams lighter

[hardy idol]

honestly i like the new superlight because no double clicking but idk if i like the sound tbh

[rough osprey]

easy to remove yourself

[storm wasp]

Real

[astral haven]

awesome, managed to get my surface at work as an extended display "wirelessly"

[vagrant pagoda]

lol

[astral haven]

touch just works!
its a bit laggy when you are moving around large windows
kinda wish also that the size inside of this settings window of the monitor matched the scaling

5 is a 12 inch display and 6 is a 10ish inch display lol

they are barely different sizes

oops no 13.5 vs 10.1

also another big sad, after getting it working on one surface I tried to add a second one as another external monitor, however unfortunately 1 wireless display maximum 😦

[gusty kiln]

Nice tetris

https://youtu.be/asdZCxNc1OM?si=H3Yy9TghSjuNBaZ6

always enjoy this guy's videos

[astral haven]

whats worse. a tree or a patch of concrete?

[vagrant pagoda]

https://youtu.be/u6ALyhykcZI?si=bkwjCnnuroXzHmnP

[gusty kiln]

The more interesting test

[vagrant pagoda]

a 3 year newer cpu did better?

shocking

[final flame]

cores aint everything chief

same with clock speed

[ashen spindle]

3 years? my guy that's 6 years old

[pale sigil]

@shadow minnow i hate this

its happening more and more this year

[vagrant pagoda]

like 3.8 all core vs almost 5ghz

[ashen spindle]

oh wait thats not the 13900k, nvm just 3 years then

but yeah, zen1 was not amazing after a while

its a great start not a long living thing

[gusty kiln]

10900k was 10 core though

So not half

[ashen spindle]

read this, mistook that name

and zen3 is like 3 gens after a do-or-die uArch

[gusty kiln]

10900k was nice before they fucked up the 11900k

[ashen spindle]

ahh the good old days of trying to get bexin to get zen3 and not the fucked up rocket lake

[reef patrol]

But Pleasant. Apple good.

[shadow minnow]

Surely this one is not on Apple though, Apple is far from perfect but this is based on devs, they may choose to use functions that were not available in a 7 year old device if not more

[gusty kiln]

Seems like every other month this happens

I do have one unfortunately

I'll check for updates when I get home

I'll get a netgear next time

Or what's the router meta?

[astral haven]

unifi, pfsense, asus still

cause asus is using a version of wrt which is very frequently updated

like the other vendors have these bugs we just never learn about them until all the routers have been taken over

[ashen spindle]

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

[gusty kiln]

just got home and my router didn't have a new firmware update

[wheat plume]

It looks like the release of stable Android 14 has been delayed by a month.

There were rumors that Android 14 might be released today, however, that didn't happen. According to Mishaal Rahman, the Android 14 source code release has been delayed until next month.

Google had pre-planned to release the new version of the Android OS on September 5, 2023, and all OEMs expected it to happen today.

Now, OEMs are being told that the vulnerabilities detailed in the Android 14 security patch list will be released on October 4, 2023, so that might be the new system release date. This date coincides with the Made by Google event date, where the Pixel 8 and Pixel Watch 2 are expected to be unveiled.

[fair gull]

why the formatting

[wheat plume]

Was already there when I copied it

[fair gull]

unsure how that justifies you just pasting it as is

[wheat plume]

Android 14 delay ;_;

I'm sad

[fair gull]

im not. not like i really care about a new phone os

[pale sigil]

https://tenor.com/view/oh-no-top-gear-jeremy-clarkson-no-one-cares-gif-18925814

[crisp granite]

it's android

it'll suck anyway

[astral haven]

you say that as if its not the least sucky option

[crisp granite]

I'd rather a nokia 3310 with basic ass call functionality than an android phone

[astral haven]

why dont you have a 3310?

[fair gull]

lana if youd put the (for me completely unreasonable) passion you have for always NEEDing to run the latest phone OS into other aspects of your life

then maybe

just maybe

you know

[crisp granite]

haven't bothered to buy one

[astral haven]

so what do you have instead?

[crisp granite]

that and I feel it won't fit in my pockets
just because of how thick it is

[fair gull]

i really wish i still had mine

was the shit back then

[crisp granite]

iPhone SE 3

it does everything I need without shitting the bed every couple days like my old android phone did

[astral haven]

se 3 is 2023?

i have the 22 se

and its complete garbage

the camera is atrocious, slow to launch and fails to sync to the cloud

[crisp granite]

2022

[astral haven]

ah so you somehow have the same phone as me and yet get somehow a vastly better experience from it?

or was the android you had previously just as much worse vs the se as modern android phones are better?

[crisp granite]

somehow
either that or I care about completely different things

[crisp granite]

I mean...I'm comparing apples to apples as much as possible
the android phone was absolutely unusable even a year after launch and one OS iteration later
became a space heater in my pocket overnight

my SE3 is still as good if not better than day one

[astral haven]

i would guess that if you spent the same amount on an android phone as you spent on an iphone that you would have a similar experience, with the option to have much better experience

the one thing is with android you do in fact have the option to do more stuff, and if you in fact do do that stuff, it will slow down the phone

the custom camera apps that i installed have a ton more shutter lag as the tradeoff for taking better photos for example

[crisp granite]

did absolutely nothing unless OS updates count

I kept it bone stock because I really...just need my phone to do phone things

[fair gull]

os updates on underpowered SoCs are the death of cheaper android phones

[crisp granite]

call, sext, occasional discord shitpost

[crisp granite]

this was a samsung that was top of the line during it's day

I...didn't cheap out

[astral haven]

model?

[fair gull]

yeah which one

that doesnt sound like an S series phone

ive serviced S3 and S4 phones for ages even

[steep creek]

I’d be more inclined to believe if it was the exynos model but even then it shouldn’t lag that bad

[crisp granite]

S7

[astral haven]

lol that wasnt a year old in 2022

[crisp granite]

and I am comparing the phone a YEAR AFTER LAUNCH

to a year after the SE launch

the S7 was horrid a year after

[fair gull]

tbf the s7 was garbage for multiple reasons but lag i never noticed when handling any of my customers phones

[crisp granite]

the SE 3 is still going strong

[fair gull]

the most horrible thing imo was just the shit tier edge detection

[crisp granite]

everything within context carmo
I'm not an absolute dumbass

well, I am
but not as much as you think

[astral haven]

my s5 on current os update is currently as responsive as the se 23 except for in webpages and the shutter lag isnt great

but thats a fkn 9 year old phone

[fair gull]

oh yeah webpages on any old android phone suck

the browser just struggles with too many updates and not enough ram

[astral haven]

its not the browser itself that struggles

running stuff like the CRD website its snappy asf

but modern webpages are just too complex for that ancient soc to handle

[fair gull]

from what i encountered modern pages just loaded the ram up

and then the browser struggles cause it cant do shit

[astral haven]

again keep in mind tho that phones have progressed significantly since then
an iphone 6 was not a peachy experience a year on...

[crisp granite]

definitely
but from what I recall (this was a while ago)
it was no where near as bad

or maybe I got a lemon

[astral haven]

now sure a 2016 android definetly aged worse in terms of battery life vs 2016 apple, but thats because apple got sued for intentionally slowing down their phones to retain battery life

modern phones have met a happy medium at this point

the battery degrades as does all lithium, and the processing speed somewhat decreases slightly over time, but its mostly the same between different vendors nowdays

[astral haven]

tbh i doubt you got a lemon
but you are comparing a 2016 android to a 2022 apple
using a flagship apple soc

a fair comparison would be with a 2022 android that sacrifices the money spent on camera to get the flagship soc just like the se 22 did

[crisp granite]

there is no SE 2022

well, yes there is

wait now I'm confused

[astral haven]

the model name of my phone is iPhone SE

[crisp granite]

give me a moment

[astral haven]

thats it

the year of release was 2022

[crisp granite]

and I am not going to shell out money for a phone I'll never use

[astral haven]

sure

[crisp granite]

pretty content with one
good to know that this was very much a "shit was bad back then"

[astral haven]

but saying your phone is better than a 7 year old phone is a given

[crisp granite]

...from a "how do I feel a year out"
on the most basic crap

[astral haven]

the only spec that a 7 year old phone has a chance to beat your current phone is camera quality

[crisp granite]

like
phone is not a pocket warmer
very good

[astral haven]

yep

[crisp granite]

phone battery lasts at least a day

very good

[astral haven]

and my 2021 android phone meets those specs

and there are plenty of other android phones that also meet those specs

[crisp granite]

wonderful

[astral haven]

anyways point is there are a ton of qualifications that needed to be added to this statement that we found out in this conversation

[crisp granite]

and I thought I made it rather clear a few seconds later

[astral haven]

you specifically said you didnt

[crisp granite]

from perspective of "non tech enthusiast who needs functional phone for more than one OS iteration"

which I made clear

[crisp granite]

I'm glad to know that it was this and not what I thought originally

[astral haven]

again that statement still needs more qualifications

i know that we have come to an agreement, you just havent said it yet

[crisp granite]

we've come to an agreement?

[astral haven]

yes

[crisp granite]

news to me
whatever floats your boat, chief

[astral haven]

i understood that you agreed that comparing a phone from 8 years ago isnt representative of phones last year

[astral haven]

i know that we disagree on more stuff, but this at least i inferred from what you said

7 not 8 oops

[crisp granite]

and yet that isn't much of an excuse for "unusable a year later"

I've either got a lemon or phones really were that bad back then and I just got the worst of it

[astral haven]

you dont remember the iphone 6?

[crisp granite]

not fondly, but it wasn't that bad

I don't remember anything fondly from back then

[astral haven]

i was a phone repairman back then

it was awful

eventually after the release of the iphone 7 we learned why the iphone 6 and 6s were so often such complete disasters

wow turns out that aircraft grade aluminum was the actual lowest grade most flexible stuff not that strong

🙀

i didnt have a phone in 2013 lel
first smartphone was the cheapest lg burner you could buy in a grocery store, circa 2014?

actually shockingly useable, however extremely limited

actually tbh besides the webpages and quality of the camera, and the lack of expandable storage, my experience is pretty shockingly similar tho its been quite a while

im very happy tho that plastic screens are omega dead

[steep creek]

My 6 was bent from the moment I unboxed it. Bluetooth among other things failed, dirt somehow made its way into the bezels and had to be replaced by the store lol

Been on iPhone since the 5s, the 6 was definitely the worst one for me. Others have been great though

[astral haven]

yea that time was not good for anyone

galaxy s6 was not that great,
Microusb???
in 2015??? come on...

worlds most durable microusb from what ive heard but that wasnt saying much

[steep creek]

Yeah micro was something else, I don’t know a single person who likes it lol

Just need something that plugs both ways and is somewhat durable

[fair gull]

i liked microusb

used it its prongs on the connector to score some paper

lol

[steep creek]

Hah fair

[astral haven]

were there any advantages over miniusb or usb c depending on the app?

[fair gull]

app as in application as in use case or as in software

[astral haven]

use case

[fair gull]

uh implementation cost

only 5 pins

[astral haven]

its more expensive than miniusb

[fair gull]

no need to jumble with usb3 and the whole spinny part

also the entire host/client system got done away with

a microusb device was generally considered the client unless the middle pin was connected in some way to enable OTG

dont need to support that though

[astral haven]

thats not true

[fair gull]

a usb-c device can be both the host and client

[astral haven]

minimum implementation for usb c 2.0 can be pretty barebones

[fair gull]

the spec discourages using USB-B as a host connector

same for using USB=A as a client connector

[astral haven]

thats an option sure, but you can (just using resistors afaik) force only being a client

[fair gull]

yes

but the spec itself doesnt discourage the use for both

[astral haven]

now sure it is more expensive
but i said vs usb c or miniusb?

[fair gull]

well micro was small enough

[astral haven]

yea and being able to use it as both has many benefits eg my phone can attach to my standard hub

and the same one that i use for my computer

[fair gull]

mini and micro cost the same but the footprint differs

usb-c just costs a ton more

[astral haven]

eh in bulk the pricing evens out

[fair gull]

no

[astral haven]

design is harder, but not insanely so

[fair gull]

a few cents can still make the difference

while sure a single connector isnt more expensive by much

when you buy a few million

thats possibly thousands or hundreds of thousands usd more

[astral haven]

oh definetly

[fair gull]

and i think usb-c is now cheaper than ever

[astral haven]

the question is will you spend that thousand less on warranty returns from the microusb breaking?

[fair gull]

what warranty returns lol

[astral haven]

or even your phone or email support staff having to respond to people with broken devices?

[fair gull]

ive never had a successful warranty claim on a broken port

usually the reply was its caused by the user

[astral haven]

my gmmk1's microusb port got flaky and they sent me a new board

i mean even if they say no, thats still support staff time

[fair gull]

well i can tell you samsung didnt fuck around and wanted hundreds of euros for repairs

but tbh usb-c also has issue in that regard although less

also depends on if you go SMD or THT though

[astral haven]

pretty significantly less tbh

[fair gull]

same as with microusb

well usb-c has 1 small advantage which is just a bit bigger anchor points for ground to solder into the pcb

[astral haven]

im pretty sure it depends more on how your casing is implemented

[fair gull]

micro and mini both just had 5 pins and 2 tiny ass case pins

[astral haven]

the high end shiz has the housing be the case just the same as microusb

samsung didnt use an off the shelf housing either

also btw you should guess how many pins are in a samsung s6 and s7 micro usb port
and im aware that those werent 3.0 type b ports, but they were also not 5 pins

[fair gull]

this was most microusb connectors

(this ones reverse mound but same concept)

2 small dinky smd anchor points

[astral haven]

yea but thats not at all what samsung used

you should guess how many fkn pins their microusb had

[fair gull]

idk only ever had the earlier microusb ones

didnt repair later models

[astral haven]

fkn 11 pins

[fair gull]

thats not on all their phones though

[astral haven]

s3 s4 s6 and s7 all had 11 pins along with the notes of those generations

[fair gull]

meanwhile their A series has 5 pins

[astral haven]

yea

and ofc the s5 with its micro b 3.0 connector has 10

[fair gull]

ngl i dont even recall an s3 with broken ports

most of the ones i recall are s7/s7 edge and a series phones

and usually from more younger folk

[astral haven]

yep

the s5 was also a bit of a problem

[fair gull]

although for the s6 and s7 phones we did have it was also ususally not broken as in ripped off

but

just the plastic ripped out

[shadow minnow]

My father had a 2017 MacBook Pro a few years ago, first gen with 4 usb c ports, and in like 5 months they got really loose and nothing would stay in, and I think either the whole pcb or the whole case had to be replaced, though covered by warranty

[astral haven]

yea if they designed the ports like modern usb c ports, where the port itself is integral to the frame of the phone we wouldnt have an issue

[astral haven]

yea apple tho is well known for making major engineering mistakes on their devices

[shadow minnow]

They claimed back then it was due to third party usb c adapter connectors being bad but never looked into it and if there ever was a genuine issue with third party cables

[fair gull]

probably the ports

unsure about the early macbooks with usb-c but the pros ive seen had individually replaceable usb-c ports

2 on 1 board

[astral haven]

yea apple claims a bunch of stuff but that was some of the more transparent BS
if you poorly attach the port to the case, and your port is not very structurally sound, then you cant rely on the case for the rigidity, and your port will break if used

[shadow minnow]

Looks like it was a common issue so probably ports for sure

Oh that’s interesting so in reality indeed only one module needs to be replaced

But looks like multiple people got told case has to be replaced too

[astral haven]

modern devices have the majority of the usb c port be part of the case of the device

photo credit ifixit ofc

tho i definetly prefer the framework usb c port mentality
$10 extender pcbs that you can replace with basically anything you want at any time

[hardy idol]

Guys if I can get metal finished really smoothly
Will it be worth it as a mousepad
I know there's been aluminium mouse pads

[astral haven]

the advantage of glass is that its pretty easy to start with a flat surface, and put enough craters in it for the mice to be able to see without really losing the flatness
metal seems like it would be really problematic for the sensors what with it being inherently reflective

[solemn remnant]

metal is really good for ball mice

[astral haven]

purely flat glass super duper doesnt work for light based mice

[hardy idol]

I promise it won't be reflective

But if you need something really smooth that tracks like ass

I could also polish it to mirror finish lmao

[astral haven]

by definition tho, any metal that is perfectly smooth will be a mirror finish
the question of how different roughing passes would affect tracking is an open question, you could run some experiments

[hardy idol]

Yes but not trackable even if it wasn't a metal

Right?

But I wouldn't get the metal smoothened by polishing it

[astral haven]

you could theoretically make a perfectly flat surface with the tracking as alternating materials
or you could put a perfectly smooth layer of something transparent that is extremely thin on top of a not smooth layer

[hardy idol]

It's a different method. Shall I do a test piece soon?
I'd have to order silicone feet I think so the metal plate doesn't slide around

Inb4 I start selling desks with a part milled out and filled in with a metal plate

[astral haven]

lel

[hardy idol]

Built in fucking mousepads

[vagrant pagoda]

i mean

you can have a flat but not smooth surface

[astral haven]

at a human level yes
at the physical level no

[vagrant pagoda]

loads of aerospace stuff is built like that

the surface is 80-90 perfectly smooth

with pits in it

[hardy idol]

Get glass skates on metal for a very nice sound experience

[vagrant pagoda]

they shot peen, cryo then grind them

[hardy idol]

Yes it's what i would do too

It's basically very advanced media blasting

[vagrant pagoda]

so they retain thin fluids better

[astral haven]

yes, but by definition thats not smooth...

[final flame]

you want a matte surface, not a mirror finish

[hardy idol]

Glass mousepads also don't be see through so they're also not 100% smooth right

[astral haven]

and not just matte, you also want it to be irregular

[astral haven]

they are often blasted somehow

[hardy idol]

It will be trust me

[astral haven]

lasers, sand, who knows what

[hardy idol]

Water

It's funny but it's water

Super smooth

[vagrant pagoda]

depends on what youre specing for surfrace testing

[hardy idol]

Idk how exactly they do it

[astral haven]

yes
but on some level it must by definition not be smooth

[hardy idol]

Shall I just run a test soon and report back when needed

(I hope I can buy additional skates for my mouse)

[astral haven]

i mean whats the harm?

[vagrant pagoda]

yes but it can be prefectly smooth and in spec with the right tip size

[hardy idol]

I'll show pictures

[astral haven]

ion we are talking past each other
in spec for smoothness doesnt mean you cant get a better microscope and see the discontinuities

its smooth enough

[vagrant pagoda]

do you know the iso spec for surface roughness?

[astral haven]

do you think theres just 1?

[vagrant pagoda]

they are all specd for tip size

so it can be smooth at a 20 micrometer tip

and terrible for a 5 micrometer

[astral haven]

yea and if your spec is 2 angstrom, all of those are gonna get thrown away lel

[hardy idol]

Okay so

If I make my own mousepad

It'll be stainless steel probably

Surface roughness Ra 0.3 micrometer

Question is is that good enough

And if I can find a nice way to put them on your desk, (i don't want high ass feet)

Should I maybe just start selling

[astral haven]

3m has these adhesiveless "sticky" materials
probably insanely overkill for a mousepad nvm
you could probably just get some of that foamlike material that artisan uses?

hmm rounding off the edges of the pad probably wont be fun?

[hardy idol]

I mean I'd have to find it for sale first

[hardy idol]

Don't you think I got machinery to do that for me

Because I do luckily

[astral haven]

yea obviously, and im no machinist, but finishing the sides of mousepads seems to have been a major issue for years

[hardy idol]

I could cut them most shapes,

Fuck maybe ThursdayIndustries' first product is going to be a mousepad wtf

I always thought I was going to make some custom automotive parts but damn maybe

Yknow what
If you have ideas regarding size, design and requirements (and the feet but maybe I just glue on a silicone or latex pad)
DM me

Note that I can offer virtually any size. But if I were to stock a few, what would be a nice size

[hardy idol]

I think maybe the resident mouse expert has a thought on this and I wanna know actually
Yamato if you read this, please comment on this idea

[latent mural]

Like what? I have an idea to make something for a car as well

[astral haven]

Why have i done this

[vagrant pagoda]

lmao

you got more than me

only 6 here

[astral haven]

I still have an open port left
Also this is a 1lpc (Lenovo tiny)

Also technically you could probably match me, 7 is a surface pro connected via Miracast so it's a bit laggy

[vagrant pagoda]

i could do 8

[astral haven]

Yeah I am out of desk space at this point
To do 8 I would need to do a double row situation or something else
Unfortunately max of 1 via Miracast 😦

[vagrant pagoda]

i have 5x1 and 2 on top

but one isnt connected to my pc

[astral haven]

Ah funzo

Bet your PC is more than 1L lel

[vagrant pagoda]

very

but that isnt shit

you can get digital signage machines

[astral haven]

Sure but this was originally a normal office pc

[vagrant pagoda]

also

i love 9020s and they have dual dp and a vga

so mstx2+vga

[astral haven]

All these monitors were from users who didn't want secondary monitors or who moved to a different office and bought new monitors
#6 was from an exec who didn't like that the conference system was taking up room in her office so I'm "storing" it on my desk

[vagrant pagoda]

also

the newest dell micros have 4 dp

they have 3 built in

and you can get dp/hdmi/type-c/vga as the option

though i think they are slightly larger than a liter

but 13700t+32GB

[astral haven]

This is 4 minidp + dp+hdmi

The HDMI is the free, 6 attaches directly via usb

[latent mural]

Got a pic of the setup?

[astral haven]

hopefully this doesnt have any pii?

obviously a ton of cable mess bc im lazy

[vagrant pagoda]

https://www.youtube.com/watch?v=KQF4UzLPpr0

lol 800 vs 900 is like 6 smt parts

[astral haven]

thats 6 more parts than they usually put for differentiation purposes lel

[vagrant pagoda]

yeah the removed the sample memory

but left the passives on the back

[magic rampart]

https://www.youtube.com/watch?v=l2D7nDd7-do

prob posted already here

14hr ago

[gusty kiln]

@vagrant pagoda https://www.reddit.com/r/Amd/comments/16awybk/5800x3d_starfield_3200cl14xmp4000cl14_optimized/

about 12-13% increase from the ram OC

[astral haven]

fyi thats not just a ram oc, thats also a cpu oc, since IF bs

[vagrant pagoda]

not every cpu will do 2000 flck

[astral haven]

sure but if hes running desynced on the 3200 thats throwing a ton more performance out the window than desyncing at 4000

[astral haven]

https://www.uploadvr.com/meta-quest-pro-2-lg-partnership-2025/
true high end meta headset coming 2025!

[hardy idol]

I mean I actually don't think it's too much of a problem

[astral haven]

well people still mess up the edges of their mousepads, you may not make the same mistakes, but it doesnt seem obvious that metal wouldnt bring additional issues

[hardy idol]

It needs to be smooth

I have to see though how I will do it

A nice rounded edge won't hurt you

[astral haven]

Yep but you also cant have the edge raised up at all, tho since it's all metal I suppose there's slot you can do that isn't easy or possible with fabric

[hardy idol]

I mean you want rigidity, but you also want it to not be super thick

Lots of thick mousepads are giant so you can line them up with the edge of your table

I've pretty much already decided to use latex as non slip surface.
I'll have a sheet of latex (you can get them super thin) glued to the bottom
Then I want the top edges to be well rounded
And the bottom edge should be a little less rounded imo I'll see how I can achieve that difference

[astral haven]

I was more thinking of the issue of making the edges smooth with a repeatable process
CNC milling it into a rounded shape would avoid any issues
But the cheapest and fastest way to round metal is to fold it

[hardy idol]

The thing is for the metal
Too thick is unpleasant
Too thin is hard to round

[astral haven]

And folding would make problems

[hardy idol]

No no

You're thinking wrong

There's way better methods

Looo at this for instance

Beautiful round edge

[astral haven]

That doesn't look like an automateable process...

[hardy idol]

It's literally

I throw it in

Wait

Take it out

[astral haven]

Yea I suppose on the scales of boutique products like this that's fine

[hardy idol]

Oh I won't produce much I think

But I will sell them I guess
I really want to give it a go

I want others to give it a go and we'll have an experience

If it works, good.
If it doesn't, we have to improve or reach conclusion it's a bad idea

[astral haven]

Yea
I'm an artisan xsoft user
Somehow doubt metal is a good next option

[hardy idol]

Same

But if we don't try

I see it as a challenge

There used to be aluminium mousepads though

So I'm hopeful

And it doesn't have to beat glass or anything

I want it to be a good option that's durable and easy to clean

[mild dome]

I used to have a metal mousepad, they just stuck a layer of plastic above the metal plate.

[hardy idol]

Above?

For tracking?

That's an odd choice

[old marlin]

I actually have an aluminum pad

That is just a piece of anodized aluminum with a rubber backing

A steel pad could be interesting as well

Polished steel pad + sapphire skates

[hardy idol]

Is stainless steel okay for you?

I can make you custom mousepads to try

I think I would trust you most for testing, as you're the mousing guru here

[old marlin]

I don't have sapphire skates lol

[hardy idol]

My idea is surface treated stainless steel (rougness Ra 0.3 micrometer)
thin latex bottom for grip

[hardy idol]

I'm afraid of glass on metal tbh. Afraid for the sound

I feel like ptfe will probably bethe best option

[halcyon moss]

more PFAS

i heard its real good for you

[old marlin]

You want polished stainless and sapphire for friction properties

Ptfe would suck on polished pads though

[hardy idol]

It's not polished tho

[hardy idol]

My man

I live pretty much next to the westeschelde

💀💀

[old marlin]

Otherwise hdpe is prob the safest bet

[hardy idol]

It'd be a matte finish though

[old marlin]

Due to how soft and maleable ptfe is you don't really get any dust etc

[hardy idol]

I can get it polished
But that's not as consistent as if i have the surface treated for smoothness

[halcyon moss]

goddamn fucking chatbots, i hate them

as a support tool

ive yet to find a good one

[mild dome]

The production of this stuff produces a lot of waste that then gets into everything. Lowering PFAS usuage overall would probably be for the best
Though a single set of mouse skates ain't gonna make much of a difference
Though you can wind up with a reverse tinkerbell effect.

[wheat plume]

Pixel 8 Pro leak!

Temperature sensor .... Hhmmm

[hardy idol]

The benefits though

PTFE is a really interesting material.
It's incredibly useful (I work in food industry machinery production)

[mild dome]

There's no low friction alternative that's as good is there ?

[hardy idol]

For some purposes it's really the only option

[mild dome]

Ideally it's usage should be lowered as much as possible. I mean teflon pans should probably be eliminated entirely

[hardy idol]

Chemical resistant

Non sticky

Can handle heat

[mild dome]

Actually it outgasses toxins when exposed to heat

[hardy idol]

Really useful stuff in stuff like ovens

[mild dome]

💀 bad idea

[hardy idol]

Depends how much heat

[mild dome]

It breaks down at a surprisingly low temperature

[hardy idol]

I mean we don't use it directly inside

Because that's mainly metal and we have a different plastic plating for that

Idk the exact material it's black but that's really no indicator

We can opt for PTFE when there's stuff like wipers, shutes etc
Depending on product and it's environment.

[echo brook]

hey guys im new to this wooting thing

[mild dome]

Although the average person hasn't received enough exposure to PFAS to cause problems, we are seeing it accumulate in the environment, it's almost literally a "forever chemical". It just doesn't decay and that's alarming, especially since we see this stuff does cause damage to those who involved in its manufacturing.

[hardy idol]

But tbh it's been a while since I've seen a full PTFE sheet here

[echo brook]

and it's actually genius

do you guys know where i can buy a wooting keyboard?

[hardy idol]

Our goto plastic is HDPE

[echo brook]

one that ships/delivers internationally

[mild dome]

Wooting's store

[hardy idol]

I read immediately my bad.

But apart from the PFAS stuff. ptfe is very interesting in its use cases

[mild dome]

https://wooting.io/

[hardy idol]

But as I've spoken with Aeon yesterday (i think)

Purposeful use of material is super cool because it allows the use of cardboard for those drones xd

[mild dome]

I know PTFE is really good at what it does, but I believe there should be some alternatives that do the job almost as good, but might cost more.
Like the issue as I said before is we don't have a cost effective way to remove PFAS from the environment, it just keeps increasing

[hardy idol]

That's the thing, purposeful use of material

Like of course you could have a house made of ptfe

Of course

But that's not purposeful

[mild dome]

What is, and isn't purposeful use of material?
Where do you draw the line, especially if alternatives can do the same job most the time?

Like we need to cut back on this shit, just like we gotta cut back on carbon emissions. This PFAS stuff isn't causing mass damage YET, but it keeps increasing

[hardy idol]

Criteria are:
Environmental impact
Cost effectiveness
How well it does for the application

I think?

[mild dome]

I don't want to give our kids a world where we need to inform them "Sorry Timmy, but because we wanted to use Teflon™️ pans and cheaper waterproofing on our cardboard we filled the environment with carcinogens we can't remove. So you'll be enjoying a substantially lower quality of life then I did, hope you enjoy half your friend getting cancer by 20!"

[hardy idol]

The poster story is of course the CFK one

Because that shit was kinda too serious

But yeah we shouldn't repeat this

We cannot outright ban PTFE production because it has it's uses
But we can limit the uses by exploring into substitute materials.
Most of the substitutes aren't yet mass mass produced and therefore numerous times more expensive
PTFE is an older material, kind of maxed out I guess. So it'll be cheap to produce, unless environmental impact tax

In some cases there just isn't a replacement (yet)
And in some cases the impact of ptfe is bigger than the alternative, but gives the product a much longer lifetime, so much that it makes it worth it.

It's a complicated matter, but I do think environment but also durability should be taken into account in more ways

[mild dome]

An environmental impact tax would probably be the best solution to making the market follow through on fixing a lot of issues. Force them to use crap plastics and PFAS only when necessary

[hardy idol]

Product lifetime is important to weighing out alternatives.
I can buy one more expensive charging cable that lasts me 1.5 years
And I can buy a new aliexpress cable every 3 weeks.
In that case, it's going to be better to pay more initially.

Same with environment
If the alternative simply isn't good. It can be that the use of PTFE is more or less required.
But in a lot of cases, there will be alternatives that require more than 10 seconds of thinking

[hardy idol]

Ptfe has become a quick answer to a lot of questions, and while it's not a bad answer in two of these aspects, the third suffers.

[hardy idol]

Define unnecessary
I think it's a matter to discuss per application, and it should be weighed with alternatives based on the 3 criteria

[mild dome]

Via taxation you can make it easier for a system to operate
If a company believes its necessary then they utilize material with the extra tax applied and pass on the costs. This means the market will look for alternatives, and apply them if they exist

[hardy idol]

It's not an or situation.
Environmental impact tax should be part of the thinf

The issue you got with that is that your phone is going to cost you an incredible amount

A simple household item like a fucking fridge is going to be more expensive than a good secondhand car

And a car

Well

A car would be more expensive than the entire spaceshuttle program

(jk but

You can't just tax more is one of the problems

In the Netherlands theres an idiot that says oh well tax the rich more to increase minimum wage and then people can afford their groceries again

No, of course not
If you the bakery staff gets paid more per hour, the bakery products....
More expensive

(short version it's more detailed ofc)

I saw someone summarise it very effectively, socialism is only fun until someone else runs out of money

But that's the tax part. The material part is that people use some materials just because they've grown to default to them. Which is wrong, but cheap.

[old marlin]

not just low friction

also for chermical resistance etc

basically impossible to find a replacement

[mild dome]

IDK, some League of Legends players seem pretty resistant to corossive environments, did you try using them?

[hardy idol]

Lol

Look

In the case of mouse feet

It could be worse

They could just give you a ptfe desk and cloth skates

I want a metal or glass mousepad yea

I am going to experiment with my own stainless steel mousepads

Lmao

Glass is dirt cheap

Mousepads is finished glass though

And that takes time and effort

Also textured

[hardy idol]

*as a resource

Yknow what's stupid

They're all HUGE

I told (idk anymore) yesterday

I use like 13cm of mousepad effectively

Square

Don't think my sensor ever goes out of its 7x7cm box tbh

Mhm I'll try

I want the ROG Moonstone because it'll look good with my new mouse

This is basically sex

It is interesting

In addition to that pad I will experiment with my own stainless steel pads

It's not as bad as some other tbh

Msi gigabyte etc

I do like their design but that software on windows bothers me

BUT

I'm on linux

I have no issues at all with it tbh

[wheat plume]

when TF you swap to Linux?

welcome though!

[halcyon willow]

Lana...

[wheat plume]

?

[hardy idol]

Long ago

[wheat plume]

oo

https://youtu.be/Bup5QREzqNI

Using ChatGPT as the virtual assistant, interesting

[reef patrol]

I see no way this could break immediately at a whim.

[wheat plume]

LOL

I like the design but not that color

ooooo it has a purple version

I must see it

damn it, cant make art

[final flame]

creating wallpapers other than nature seems iffy at best
the amount of trash I have gotten out of even a quality model is a lot more than the average consumer is willing to handle

very rarely do you get something usable right out of the ai

For example these 3

[hardy idol]

Damnn

I prefer real art though

It costs a fortune but it's nice

And I've befriended a great student artist

[final flame]

Yea real art will always be better

I do like the idea of people creating ai art for commissions tho, works good when trying to describe certain things to a artist for commission

[hardy idol]

This is my latest commission

I mean it's not wallpaper type art. She's yet to master making backgrounds

[supple torrent]

Hello guys do you think its worth having double 24.5 inch monitors if i want to have one horizontal for chat and discord? Or is it better to buy 27 inch for that? Let me guys know… My main monitor for gaming is 24.5 inch

I mean something like this and i was thinking whether its worth buying 24.5 inch for horizontal or 27

[final flame]

depends on how often you use discord

it can be nice but at the same time as someone who has that same setup, it also is just sometimes easier to leave discord on your main monitor

if you do plan to do it, make sure the 2nd monitor is at least ips tho

[supple torrent]

Quite often to be fair xD Atm i have it set up the same way but the monitor is old 60 hz with huge bezels and takes up a lot of space so i am thinking of replacement purely because of the bezels. And was thinking just for cheap alternative 140hz kooruii

[final flame]

get a good and cheap ips monitor

[supple torrent]

As you see the bezels are huge since the 2 monitor is really old.

Deal: KOORUI 24 Inch Computer Monitor -FHD 1080P Gaming Monitor 165Hz VA 1ms, AdaptiveSync Technology, LED Monitors with Ultra-Thin, HDMI X2 /DP, VESA Compatible, Tilt Adjustable, Eye Care 24E4 https://amzn.eu/d/dRLPmk8

What about this one?

Its VA

[final flame]

Va's are better, but ips really gives you the best viewing angle

you don't want to have to move your secondary anytime you move back in your chair

[supple torrent]

iiyama G-Master G2470HSU-B1 24 Inch Fast (FLC) IPS LCD, 165 Hz, 0.8 ms, FreeSync Premium, Full HD 1920 x 1080, 1 x HDMI, 1 x DisplayPort, 2 x USB, 2 x 2 W Speakers https://amzn.eu/d/61YPo2q

So this one you reckon will be a better choice)

I am looking for cheapest options since secondary monitor i think doesn’t have to be the most premium xD

And the real question is if 24 or 27 inch xD for horizontal view paired with my main 24 alienware

[latent mural]

VA isn’t that bad after having used one

Wouldn’t get it as a second monitor though, Glarxses is right

Get two 24+ inch IPS monitors and call it a day

[supple torrent]

Yeah i will go for IPS i only cant decide what size to buy. Since i just want it for horizontal view xD my main is 24.5 alienware for gaming

[ashen spindle]

why though

[supple torrent]

I am not buying the 2 monitor for gaming only purely for horizontal angle and discord and chat and so on xD so idc

[ashen spindle]

yeah second screen IPS is great, I had a TN once and man the viewing angles were a pain

[supple torrent]

Yeah i have got th old 60 hz secondary

And its not good

Ur right xd

[ashen spindle]

again though its for a second screen youre not watching HDR movies in portrait mode

[fair gull]

easy fix make it so the panel conforms to some radius

just put the monitor through this

ezpz

[halcyon willow]

i think its something like 5000:1

great for blacks

in cinema context

no screenshots please

[ashen spindle]

https://twitter.com/aschilling/status/1699408468472385617

Andreas Schilling 🇺🇦 (@aschilling)

#Navi32 @amdradeon

Twitter · Sep 6, 2023, 1:05 PM

[final flame]

the only reason to get a va is if you're using it for content
smearing on va's is a issue, not really worth the extra contrast

[crimson wigeon]

BEST VALUE YET

[hardy idol]

Inb4 I insert an Apex Pro

[pale sigil]

It's just a 6800xt but a little better?

[ashen spindle]

and cheaper

[pale sigil]

Well in some cases ye

[crimson wigeon]

YEAH HBM!

[hardy idol]

I guess it all depends on the fan configuration and thus airflow in the cases.

[ashen spindle]

simple, rdna3 is not memory starved at all

N31 was supposed to have a SKU with 3d stacked memory

not happening with the cores being gutted and they cant even saturate the current memory

so making a bigger bus wouldn't really do anything if you already can fully feed the cores

[rough osprey]

no one uses higher then 1080p 😀

[vagrant pagoda]

https://youtu.be/2s-OKMFJqlU?si=aDdPVxUbB9-LjxnZ

[rough osprey]

lol

[vagrant pagoda]

You do how the make smaller gpus right?

They design the GPU in units that they can stack together to make the gpu

Memory interface is part of that

So when you remove units you lose memory width.

[ashen spindle]

well in this case its a different layout with 4 instead of 6 MCDs as well so things are a bit different from monolithic dies

[vagrant pagoda]

Also 384 it has been the standard for high end GPU for ages

And has scaled perfectly will the new generations of gddr

I think only the 1080ti was the only flagship with less

Had to go up 2 titans to get 384

[astral haven]

Also gotta remember that 1080p is half the pixels of 1440p which is half the pixels of 4k

It is actually quite fantastic scaling to have around half the performance at 4k vs 1080p

[vagrant pagoda]

You need the CU to talk to the MCD

You can't slap 8 MCD to a 7800

[astral haven]

That's not guaranteed to be an architectural issue on rdna3 or whatever it is now

It was definetly an issue on maxwell

[vagrant pagoda]

I mean literally every other GPU has been like thst

[astral haven]

But it's theoretically possible to make a GPU that doesn't have that bottleneck

[astral haven]

Not a100

[vagrant pagoda]

What is a cut down a100

[astral haven]

And rdna2 had huuuuge caches

[astral haven]

A100
There were a bunch of different a100s
Some were 5 stacks but a few were 6 stacks hbm
I think there's also the a80?

[vagrant pagoda]

There isn't a GPU with a completely separate complete memory controller

[vagrant pagoda]

A80?

[astral haven]

A800 oops

[vagrant pagoda]

That's just 80gb a100

Or is it different?

[astral haven]

That doesn't matter
What you are talking about is cus wired directly into specific busses of the memory controller

[vagrant pagoda]

It isn't the cu wired to the memory controller

It's the io for memory being part of a block of cus

[ashen spindle]

given that n31 could have had double the MCDs I think the N32 wgps are perfectly capable of having a bigger bus to use if that would have been needed

[vagrant pagoda]

Which is the same on rdna3 as 2 it's just cache and controller are chiplet

[astral haven]

Yes so if that system is designed to be fed from let's sty an aggregate of 1.6TB/s of l3, you could actually just increase your memory bandwidth until you match that

[vagrant pagoda]

We aren't talking speed

We are talking bus width in mid range

They way GPUs are designed

[hasty thunder]

The fuck did AMD think when they launched the 7800XT?

[vagrant pagoda]

When you remove cus

[hasty thunder]

[vagrant pagoda]

You loose memory interfaces

[astral haven]

Yes which doesn't matter if you are passing through the imc and that imc has the l3 controller

[vagrant pagoda]

What gpus has a monolithic memory controller?

None

They have many controllers

Wether they are chiplets or ondie

[hardy idol]

No no no
With your question you're implying they thought in the first place.

[ashen spindle]

N21 is phased out and N32 worked well enough is the thought there

[astral haven]

Yes but they also aren't inside of the cus
The imcs are around the edges of the chips, and the cus need to have the ports on the cus sufficient for the cache bandwidth

[vagrant pagoda]

I'm not saying they're part of the CU you're missing that

[hasty thunder]

Well, it takes less power, but not significantly less and yet performs worse than the 6800XT

[vagrant pagoda]

I'm saying they designed in blocks

[hasty thunder]

WHY?

[vagrant pagoda]

With cu cache etc and memory

[ashen spindle]

it performs about the same and is cheaper

[astral haven]

Yes and that means that however much memory controller they put on is a design decision

[vagrant pagoda]

So when you remove those blocks for a smaller die

You loose memory

[hasty thunder]

I won't upgrade this generation then i guess

[vagrant pagoda]

They aren't going to do a major redesign for a mid range card

[astral haven]

Ion just look at the dice they're very obviously not designed all tied together in One singular unit

[ashen spindle]

I mean the thing is it acts like a 7700xt, the name being different doesnt make the product worse

[astral haven]

it's not Maxwell

[vagrant pagoda]

Idk what you're trying to argue

[ashen spindle]

if it was a 7700xt, performance of a 6800xt then it would have been normal product release

[vagrant pagoda]

They design them in blocks and stick blocks together

And the blocks are made of blocks

[astral haven]

Think of the memory system like tiers
L1 and L2 on Radeon are attached to a cu
But above L2 there's a GPU bus before you go to l3

[vagrant pagoda]

Yes but none of that is a singular unit

There is no way to take the blocks and build a low-end GPU with a full memory bus width

[astral haven]

Exactly!!!

[vagrant pagoda]

You're arguing the same thing that I am that's what I don't understand

[astral haven]

How could that possibly lead to that conclusion???

[vagrant pagoda]

Because that's how they design gpus

They don't clean slate every single model

They take their building blocks and just use less

[ashen spindle]

uhhh

[astral haven]

Be more specific
You are making a huge leap in logic

[vagrant pagoda]

Which is why memory with scales exactly with core count

And has on every generation of GPU

It's also why you end up with a 3.5 GB GPU lol

[astral haven]

Again making huge leap in logic

[ashen spindle]

basically its launched at 150-200 lower msrp, so normal product upgrade cycle really

[astral haven]

Maxwell and earlier did this therefore it must be the only way

[vagrant pagoda]

??? Look at any modern die shot

You have multiple memory controller complexes with crossbars

[astral haven]

Exactly!!!

[vagrant pagoda]

You are missing the point or dumb I can't tell

[astral haven]

Ion
What is a crossbar used for logically

[vagrant pagoda]

You can't gut cus to srink the die and leave the memory controller with out a redesign

If you don't redesign you end up with a 970

[astral haven]

A redesign to what?
Specifically

[vagrant pagoda]

Where part of your memory is slow as fuck because of crossbar

[vagrant pagoda]

A different core/cache/memory controller layout

[astral haven]

That is just plain wrong

[vagrant pagoda]

No it iant

Show me 2 GPUs in the same arch with different core memory layout?

If you cut cus and not memory, the die doesn't srink

[astral haven]

6950xtx and whatever fuckoff mobile chip happened to use the same architecture

Entirely different memory system using the exact same cus

[vagrant pagoda]

Mobile vs desktop are entirely different chips

[astral haven]

You are correct that some individual part in here needs to be changed, I'm just making you say what exactly it is

[vagrant pagoda]

Different die size and transistor count

[astral haven]

Desktop vs desktop is also entirely different chips most of the time...
Your point?

[vagrant pagoda]

No

Look at refresh 3090, 3090ti and 3080ti

Same die

Same l3

Same memory width

Same cost to manufacturer

[astral haven]

And not a single one of those products shares a die with more than 30% of the Nvidia product stack

[vagrant pagoda]

Exactly

To srink the die they can't just remove cus

They have to remove an entire design block

Also the 3080 is on the same die

With a block lasered off

[astral haven]

Holy shit dude
You are the Usain Bolt of leaping to conclusions

[vagrant pagoda]

How

It's common knowledge

That's how chips are designed

That way most of the design cost is shared

They only have to do the auxiliaries when they shrink

[astral haven]

And what exactly are the part(s) that they need to redesign

[vagrant pagoda]

Notice any patterns?

It's almost like it's built from Lego blocks...

[astral haven]

Be specific or I will state examples that obviously disprove your point

[ashen spindle]

I got lost what’s the point that you are trying to make

[vagrant pagoda]

Have you literally never seen a BLOCK diagram for a GPU?

There's a reason they call it a block diagram

Please go read on of Nvidia technical related on GPU design.

The design blocks

Nvidia calls them graphic processing cluster

All of the skus use the same gpc

[ashen spindle]

Like that’s all true but if the point is still about bus width, that can still be changed it’s not set in stone without having the option to change it at points in the stack

[vagrant pagoda]

You'd run out of l1 to l2 width then

[astral haven]

AMD can and does mix and match to use different parts fed with different bandwidth

[ashen spindle]

That can happen but it’s not guaranteed to happen, depends on the design and what widths you are comparing

[vagrant pagoda]

3080 has 6 gpc with 200gb of l1 per gpc

It has 760gb of memory bandwidth

So that's max of 1200 l1-l2 and l2-memeory at 760

That isn't a lot of wiggle room to srink

Sure it would help with latency

But making something like a 3070 with 4 gpc and still having full size l2/memory isn't going to make it magically faster at 4k

Mainly because you'd eat the cost with the extra die size and memory modules so the price would go up

And unlike a ti card in the middle, you wouldn't have more cores

[astral haven]

Assuming this was true, radeons terabytes/s of l3 proves the limit for their current cu count is at least 3x their current memory BW

[vagrant pagoda]

It has more cache bandwidth not more memory bandwidth though

[astral haven]

What is preventing amd from adding more memory bandwidth if they wanted to?

be specific

[vagrant pagoda]

The fact that they probably know it doesn't scale well with the cost

More memory width is more memory chips

And we are talking mid range not the high end halo

[astral haven]

While that's true, if only that is true, it proves your entire argument up to this point is B's...
So what is your point that would make you say anything you've recently said

[vagrant pagoda]

The halo cards work

And don't need more

They asked why MID range cards have weak memory widths

[astral haven]

If the flagships do, the midranges do as well, and scales it quite linearly

[vagrant pagoda]

I didn't say they needed more

I said why they don't have more

The block layout works

And everything is built around that

The reason that mid range don't get more memory width is the build in blocks. And the blocks can't/don't need more width

[ashen spindle]

I mean in rdna3s case its more like it's already overbuilt with memory due to fuckups in shader xD

[astral haven]

This is untrue

[astral haven]

This doesn't make it true

[astral haven]

Rdna3

[vagrant pagoda]

Proof?

[astral haven]

Wrong

[astral haven]

7900xtx and 7800 use the same cus with different memory bandwidth

[vagrant pagoda]

Did you fail math?

7800 is2/3 the die size

It doesn't have the same cu count at all

[astral haven]

and?

even if you assume that theres a l2 to l1 bandwidth limit

the l3 bandwidth proves that it has plenty of room for a larger memory bus

[ashen spindle]

I mean there were plans for a double stack N32 just like there were 0-1-2 hi N31 chips so they could have had 8 MCDs on n32

[astral haven]

amd chosing to balance the memory bandwidth by going hard on the l3 bandwidth doesnt mean that they couldnt go bonanza on the memory bandwidth if they wanted to

[vagrant pagoda]

Look at the xtx vs the 7800s core to width ration. It's almost the the same. The 7800xt actually has a better memory width to core ration

And still dose poorly

And because l3 scales with width on these

The 7800xt has MORE l3 and memory bandwidth per cu than the 7900xtx

[astral haven]

exactly!!!

[vagrant pagoda]

So it scales worse

Despite have more memory and l3

[astral haven]

yep!

[vagrant pagoda]

Also the 7800xt is lasered cu

[astral haven]

thats why amd doesnt go ham on that spec despite their clear ability to

[vagrant pagoda]

As the block for rdna3 desktop is 32

Or 16

But still 60cu midrange vs 96 top says snip snip with the laser

[ashen spindle]

its not snipped its a different die, not a cutdown

[astral haven]

far as we can tell 7800xt is full fat navi32

[vagrant pagoda]

Nah

[astral haven]

https://www.techpowerup.com/gpu-specs/amd-navi-32.g1000

[vagrant pagoda]

The block is either 32cu to 2 mdc or 16 tocu to one

Think 7900xtx is 96cu 96 l3

Why would it be 60 to 64

[astral haven]

because those are independent variables lmao

4 mcds is 64mb

[vagrant pagoda]

I'll put $100 on there being four more cues on the die when we get a good die shot

[astral haven]

and ill bet $1000 they arent enabled and dont need to be

[vagrant pagoda]

I never said that they were unable

I said that the 7800 got lasered

Because they build gpus and blocks

And our DNA 3 for desktop is built with 16 core blocks

[astral haven]

if your statement is true, then they must be enabled

[vagrant pagoda]

The rest of the block is enabled

I'm saying you can't go from a high-end GPU to a low-end GPU keeping the same memory bandwidth and shrink the dye

That's the reason that memory scales with die size

They build them in blocks

[astral haven]

be specific as to what is preventing that

[vagrant pagoda]

Because they build in blocks

[astral haven]

blocks that contain what EXACTLY

[vagrant pagoda]

It is not a coincidence that dies scales with width, minus lasered off ones

[vagrant pagoda]

Everything

Literally everything on modern silicon is blocks

[astral haven]

state one thing they contain and i will prove that wrong with an example

[vagrant pagoda]

Scaled out

[astral haven]

and i mean specifically that are contained in a block and are not available outside of that block

[vagrant pagoda]

On rdna3 the block is everything from cu to l1

[astral haven]

anything else?

[vagrant pagoda]

Everything between those

[cursive summit]

What's wrong with this chart

3070 same performance as 4070?

[astral haven]

exactly!!!
l2 and l3 and memory controller are not part of that

tho we dont know about l2

[vagrant pagoda]

The cross bar is though

[astral haven]

do different gpu cores have the same crossbar?

[vagrant pagoda]

The crossbar goes between them all

[astral haven]

cores as in n31 vs n32 vs n33

[astral haven]

are they the same crossbar physically on these different chips?

[vagrant pagoda]

I haven't looked at AMDs core layout in a while

[astral haven]

oh so you dont even know if youve been bssing this entire time

holy smokes man you are a neutron star of density

[vagrant pagoda]

And you haven't provided any proof either

[astral haven]

proof of what?

different chips are different?

if you need proof that n31 isnt the same chip as n33 im sorry i cant help you

[vagrant pagoda]

No that they are made of blocks that are the same core:cache: memory

[astral haven]

i didnt make that claim

[vagrant pagoda]

That's what in claiming

And you're saying I'm wrong

[astral haven]

https://www.techpowerup.com/gpu-specs/radeon-780m.c4020

this rdna3 gpu has the same cores and l1 and l2, yet has zero available l3

oops my bad it has more l2

[vagrant pagoda]

That isn't desktop and is a igpu

The argument is why midrange cards don't have more memory width

[astral haven]

yet it still contains the exact same rdna3 cores

[vagrant pagoda]

And the fact is desktop GPUs are made in blocks

They do not redesign the block for a mid-range GPU

An entirely different class of product likely has some redesign

[astral haven]

true statement

[astral haven]

true statement

[vagrant pagoda]

But when they go from a high-end GPU to a mid-range GPU they don't redesign the block

[astral haven]

possible statement

[vagrant pagoda]

Without redesigning the block you're not going to get any benefit from more width

[astral haven]

true statement

[astral haven]

statement that requires proof

[vagrant pagoda]

Cost/pref wise

[astral haven]

entirely different claim

[vagrant pagoda]

More cores is always better than more memory for mid-range

So why not do both