#programming
1 messages Β· Page 153 of 1
shouldnt that just be an automated task?
you can make it one yes
You can set it as such, it just doesn't by default
2 in fact 
I mean i think it would be reasonable to even just ask if you want the GC to run once a week or smthing
.
theres probably lots of people with some 100gb of garbage just because they are unaware of the gc, not even knowing of the option
i really doubt this user actually exists
Like if you're just watching this chat
anyone using nix is gonna know that nix has a thing called a "store"
if it's too large, you are gonna search for ways to make it smaller on the documentation, which is freely and easily available
lurkers aren't supposed to speak
you're not a lurker now
if you dont, then really you have just decided to ignore the problem
Exactly
Lurkers should triangle
πΊπ»πΊπ»πΊπ»
curious
gamemode doesn't work π
I don't have the command so clearly I have goofed somewhere
but I do have the other gamemode commands
okay so i need to make a dll shim on 9x to replace some syscalls
oop
according to microsoft system32 always loads first
but to everywhere else apparently it says that placing a dll smack dab in the middle of the directory should work
@sage crag not much better
70k
try with the igpu though?
every part of my soul screams at me that this is a horrible idea
but I wanna see what happens, so go for it
i wanna try it with wine first to test so i dont go through the trouble of sending it into a vm if it doesnt work
but i guess ill just
do it
fair
man, you guys are making my 30k fps score look just bad
I thought about it and I still think that there are plenty of people who dont know about it. one anecdote is that lots of people are running their 144hz (or higher) monitor at 60hz because they are unaware of the option to set it to 144hz and just expect it to be that way automatically. same with people not overclocking ram / enabling XMP, just running at base speed
8.9k here
lots of people just explect the default behavior to be good and only need modification if their usecase is different from the expected, where as the options expect the user to tinker with them regardless of the usecase, so there is a mismatch in expectations
its completely non-analagous though. buying a pc with components, and buying a monitor and plugging it in, is a trivial action. it doesnt inform you that there is a manual either. nor does one need any level of expertise to set it up.
people who use nix configure their system manually in a programming language. there is no way around doing that, that is literally the point.
@true hemlock We love llvmpipe
try setting LP_NUM_THREADS wayyyy lower
3 seems to be optimal for me
who would've guessed that 32 threads are bad at rendering a small triangle very fast
you can enable autohardlinking
old wiki
new wiki
old wiki rejects vpn ips 

its so sad that our team is rank decaying
Wait, reducing resolution has a benefit here
@mkrek please 
fake fan
@amber fractal are you still doing llvmpipe shenanigans?
It's funny
how many threads are you using?
1
fair
Anything else was slower
somehow 3 was fastest for me

Fair
src/common_structs.hpp is where that info lives
ok I count modifying the source code as not part of the challenge π
It was explicitly allowed by sam 

We had memes of 2048 frames in flight
Quack tuned everything he could for his spot
I personally have best luck with 6 instead of 7

Road map
Reached (7%)
What

rust fans rn: 
!triangleupdatelinux 1,70854
Updated entry "owobred" (69767 -> 70854)
@umbral thorn can you please change the bot pfp π π₯ π
i should probably try my triangle on wayland
this is actually petty cool
its a shame how inaccurate it is
it takes a minute to write it in a way it recognises the word
changing the type or something should be fine as long as it shows the accurate fps numbers (you do not change the multiplier.)
i think we need a way to validate this case so people don't cheat with modifying that lol
true
Just make sure its acurate
Integer devision does not count as accurate I think
you have -ffast-math anyways
nothing is accurate
i guess
it is accurate
It just needs to be accurate to a certain degree
it rounds toward 0
I guess its fine then
π
I reduced 3 cpu clocks(in main thread) 
probably more without the lib calling idk
IMO needs to be at least accurate till the tens place. The ones place doesn't matter too much
Huge tbh
For optimising I'd like to be in the hundreds but otherwise I would personally be fine with only thousands
Regardless, rules are rules
Having the tens not be acurate would be fine if we didnt have people with only 300fps

can I set rasterizerDiscardEnable = true so it doesn't draw anything 
it still process the stuff, it's just not outputting stuff into the buffer
meh that lose the fun
No

My personal edit is removing the atomic factor of the frame count by simply never resetting it.
Taking one from rust, only one mutable reference.

Sure the frame timer will explode, but it'll take awhile
idk how you'd make that work then tho
unless you keep a copy in the main thread thats one loop behind and you comapre the diffrence
That's the attempt I went with
The code is not liking me trying to switch out of floating math so I still need to apply that
I swear, if only I could guarantee that the fps got called every second

Letting that run to see if it is off ever
frametime does always seem to be 1000 unless you are literally quitting the app.
nvm
Some cases do exist
Highest I've seen is 1004 out of 1000
That one attempt was lucky
Shame as any amount off is over the limit. So that's one that we'll have to eat.
the program cannot be faster now
it's hardware capped

there's no way you need that many resources to render 6x7 pixels
2 pixels and still capped 28k

it's cpu capped
rdtgyuhjiopk
yep
evenat 1080p is also pretty much cpu capped
but you just remove a microsecond of the cpu waiting at smaller res

Could you simply not be late?
by 1ms
Granted it is less than it was before
If this stops being late then no need to divide and mutiply
That's an ugly fix but it technically works
basing off of this, moving the SDL_Delay into the brackets and doing 999ms instead of 1000ms
Essentially busy waiting until 999ms becomes 1000ms

@olive sable Jank code improvement

I could check == instead of >= but CBA
im not sure if its actually faster since the outside of the loop executes a lot more now, but it might be
At the very least it combines the other changes
I guess weighing a multiply and divide over busy waiting
I've gotten better results personally
This chat is so smart Iβm trying to understand coding/programming but itβs been way to hard for me to understanding
The first step is to understand the language and concepts your working with. Programming is very different to reality no matter what level your at. (I'm looking at you JS)
Right now it is punishing being on time, Ideally it'd not be punished at all and just make it.
lets do the math
if we're at 999ms or 1001ms, we get an error margine of how much?
0.999000999 till 1.001001001 % accurate
before lowering, I saw only 1001 after implementing fast math
ye
so that case was either 1000 or 1001
before fast math it was 1000 - 1004
fast math being caibi's changes
70K with that error margin gives us 69930.06993 till 70070.07007
which is fine
i think
is the current threads being hyper-threaded
probably, idk
Depends on the user, I have hyperthreading disabled by default and quack said it increased theirs
it should, but idk if you can change that in the code, you'd have to turn of hyperthreading in the bios probably
yea it's in bios

Thing I'm wondering is the more optimized the fps calculator gets. The less needed the 999u over 1000u is needed
if google is correect hyperthreadign on gives about 80-90% the performance for single-threaded programs as hyperthreading off does
So eventually we won't even need any checks
and set 999u back to 1000u
i think just do delay 1000U and we'll accept the error margin
not printing it should help too
maybe 999u, since the fps stuff does take time but im not sure
That's my reasoning for including it
If 1000u exists, the if check also pretty much is redundant
actually, since we're jsut gonna be delayign by the same amount each time we wont have to do get ticks
which was one of the more expensive functions
i think jsut do 1000u
Begone if statement

By removing it we are making it more accurate
Classic accuracy check making it less accurate
wat is 1000u if u don't mind me asking?
At your point, it should change it as long as u have as little background tasks running as possible I belive
The windows install has crashed 3x due to how much was removed from the OS
How does your software renderer beat my dGPU???
Speed of light being slow
takes time to transfer mem. from CPU to GPU.
If u do CPU rendering less distance the electrons need to travel
insert pepecrying emote here

thank u
Idk electracity is hard, I tried to understand it multiple times, and never did. It is magic to me to this day
5+ times, actually

Alright let me push this to a repo @midnight sigil is it alright if I upstream your patchset?
Mine pretty much relies on yours existing
Still need to ask, otherwise the fps counter would take too long and mine wouldn't be accurate
πΊ
Well granted the only thing left is setting fps to a uint
ye
tysm
I'm back from overclocking
hope my cpu won't fry

POST is taking long mhm
it boots at least 
Small performance improvement soon mhm
booted into my ultra mini os as well
While you were away:
- FPS calculator math improvement
- My 4070Ti got beaten by a software renderer
there is a command to put your cpu at max power in linux, but that made my poor 5950x do 5.1GHz at idle, and still only 4.6GHz when runnign the triangle and the same fps
so i stopped doing that
Probably can't sustain full speed under multicore load
I just started disabling cores 
I actually disabled p cores and left e cores
power limited moment
More like some of each
@olive sable sent
okay
done
no way

caibi added and removed a singular line of code 
oh ye that was the leaderboard update
And
account is just a bot
yep
I can pr mine later
I'll say it was the one line that I had left from their patchset
R8 colorspace moment
color has to be a vec3 it seems, so I can't disable write for GB
we should just do R1. its either red or it isnt

I wonder how much it would take for the triangle to run entirely in the CPU's L1 cache
I removed more lines than I added so at least more of it is there now
dlls are too big to fit
the very fact we have to use a graphics api means you're already filling L1, L2 and L3
vulkan takes like half a gig
Just wait until someone with one of those CPUs with 1,1GB of cache comes and beats us all
computer go brr
I swear there was a site for colorspaces, maybe that was opengl tho
L1 is probably too small
L2 maybe
I think it should already fit in L3
It's fine, we're only considering the hot path, not the whole stack
I see an R8, no packed versions tho
VK_FORMAT_R8_UINT might work
9-15 are R8
yep nothing, if we need packing. One could just pack R8
wait how do you do software rendering again
LLVMPIPE
You may have to change the thread count
.
I had luck with 1, bred had luck with 3
I'd switch to lu and maybe switch the frame counter to the 64 bit fast version, if there is a fast one?
Not my original code
!trianglewindw
Windows Leaderboard:
- 32024 by Quack on Tiny 10, i9-13900 (ES Q0L4) + RTX 4070 Super
- 29250 by Caibi on Windows 10(19045.2965; AtlasOS 0.3.2), i7-14700F + RTX 4060
- 22800 by iwbtvb on Windows 11, i9-12900KF + RTX 3090
- 17118 by Fridge on Windows 10, Ryzen 7 9700X + RTX 3070 TI
- 13531 by Vanyilla on Windows 11, Ryzen 9 9950X + RTX 4090
- 11000 by Sam on Windows ??, Ryzen 9 5950X + RTX 3090
- 10032 by owobred on Windows 11, Ryzen 9 9950X + RX 9070 XT
- 0 by Adesi on Windows ??, Ryzen 9 5950X + RTX 4080
- 0 by Quack on Windows 10, Threadripper 3990X 64C + TITAN V
33k is huge on this board
im changing the event handler code too
who said I only got 33k 
sam how's your fast shit going
well 69k
i cant get llvmpipe to work, its jsut not on my system
Got to install it
have you installed Mesa yet
install Mesa
I believe arch has a separate package for it
I want to try and figure out how to do the opposite now (obviously crashing is considered cheating)
as litle fps as possible?
yes
Gotta lose to samsung smart fridge
the real solution is to jsut not render anything
crashing obviously cannot count
if you want it to technically be a working renderer, but its jsut slow
just run the most inefficient checks possible
We are cpu bound anyways, so 1mhz vm could also work
then make your framebuffers 32K x 32K, put some really expensive shaders in there, render millions of objects, and limit your pc's power and amount of cores
true
tbh as litle fps as possible is easy, gettign a high amount is hard
We are only making your life harder as we continue to improve the codebase
where's the repo btw
!triangleaddwindows caibi,34447,Windows 10(19044.3324; Tiny10-Modified),i7-14700F,RTX 4060,2x16GB DDR5-6600 @ 6600MT/s C32
Added entry "s caibi | 34447 | Windows 10(19044.3324; Tiny10-Modified) | i7-14700F | RTX 4060 | 2x16GB DDR5-6600 @ 6600MT/s C32 | "

!trianglewindw
Windows Leaderboard:
- 32024 by Quack on Tiny 10, i9-13900 (ES Q0L4) + RTX 4070 Super
- 29250 by Caibi on Windows 10(19045.2965; AtlasOS 0.3.2), i7-14700F + RTX 4060
- 22800 by iwbtvb on Windows 11, i9-12900KF + RTX 3090
- 17118 by Fridge on Windows 10, Ryzen 7 9700X + RTX 3070 TI
- 13531 by Vanyilla on Windows 11, Ryzen 9 9950X + RTX 4090
- 11000 by Sam on Windows ??, Ryzen 9 5950X + RTX 3090
- 10032 by owobred on Windows 11, Ryzen 9 9950X + RX 9070 XT
- 0 by Adesi on Windows ??, Ryzen 9 5950X + RTX 4080
- 0 by Quack on Windows 10, Threadripper 3990X 64C + TITAN V
buh?
er
erm
erm

!triangleaddwindws caibi,34447,Windows 10(19044.3324; Tiny10-Modified),i7-14700F,RTX 4060,2x16GB DDR5-6600 @ 6600MT/s C32
No access.
bwaa
nice one
Thought I'd check because some commands have a typo
!trianglewindw
Windows Leaderboard:
- 32024 by Quack on Tiny 10, i9-13900 (ES Q0L4) + RTX 4070 Super
- 29250 by Caibi on Windows 10(19045.2965; AtlasOS 0.3.2), i7-14700F + RTX 4060
- 22800 by iwbtvb on Windows 11, i9-12900KF + RTX 3090
- 17118 by Fridge on Windows 10, Ryzen 7 9700X + RTX 3070 TI
- 13531 by Vanyilla on Windows 11, Ryzen 9 9950X + RTX 4090
- 11000 by Sam on Windows ??, Ryzen 9 5950X + RTX 3090
- 10032 by owobred on Windows 11, Ryzen 9 9950X + RX 9070 XT
- 0 by Adesi on Windows ??, Ryzen 9 5950X + RTX 4080
- 0 by Quack on Windows 10, Threadripper 3990X 64C + TITAN V
I'm on the linux leaderboard now
oh lmao
!trianglelinux
Linux Leaderboard:
- 70854 by owobred on NixOS, Ryzen 9 9950X + RX 9070 XT
- 55272 by Quack on NixOS, i9-13900 (ES Q0L4) + RTX 4070 Super
- 36036 by Iggly on NixOS, i7-1260P + Xe Graphics (integrated)
- 35600 by superbox on Mint, i7-12700K + RTX 4070 Ti
- 34447 by s caibi on Windows 10(19044.3324; Tiny10-Modified), i7-14700F + RTX 4060
- 32100 by Semicolon on Arch, Ryzen 9 5950x + RX 6800 XT
- 30573 by Iggly on CachyOS, Z1 Extreme + AMD Phoenix
- 27158 by Shuni on NixOS, Ryzen 9 7950X + AMD Raphael (integrated)
- 26364 by konii on NixOS, Ryzen 5 7600 + AMD Raphael (integrated)
- 22809 by Prism on Custom, HP Prime G2 RevD + (Graphing Calculator)
- 18230 by Sam on CachyOS, Ryzen 9 5950x + RTX 3090
- 17000 by toast on NixOS, Ryzen 9 5900x + RTX 5070 Ti
- 13768 by Sqbika on NixOS, Ryzen 5 3600 + RX 9060 XT
- 13337 by Sqbika on NixOS, i7-9750H + GTX 1650M
- 13037 by Assist on CachyOS, i3-8100U + UHD 620
- 9458 by unkomp on Arch, +
- 672 by Quack on TizenOS, Quad core ARM cortex? (Samsung Smart Fridge) + Mali GPU
- 0 by Elvyn on NixOS, Ryzen 9 7950X ES + RTX 3080
LAMO

so is this what you all do in your free time
!trianglerem 5
No access.

!triangleremlinux 5
Removed entry "s caibi | 34447 | Windows 10(19044.3324; Tiny10-Modified) | i7-14700F | RTX 4060 | 2x16GB DDR5-6600 @ 6600MT/s C32 | "
!trianglepush
!triangleaddwindw caibi,34447,Windows 10(19044.3324; Tiny10-Modified),i7-14700F,RTX 4060,2x16GB DDR5-6600 @ 6600MT/s C32
Added entry "caibi | 34447 | Windows 10(19044.3324; Tiny10-Modified) | i7-14700F | RTX 4060 | 2x16GB DDR5-6600 @ 6600MT/s C32 | "
!trianglewindw
Windows Leaderboard:
- 34447 by caibi on Windows 10(19044.3324; Tiny10-Modified), i7-14700F + RTX 4060
- 32024 by Quack on Tiny 10, i9-13900 (ES Q0L4) + RTX 4070 Super
- 29250 by Caibi on Windows 10(19045.2965; AtlasOS 0.3.2), i7-14700F + RTX 4060
- 22800 by iwbtvb on Windows 11, i9-12900KF + RTX 3090
- 17118 by Fridge on Windows 10, Ryzen 7 9700X + RTX 3070 TI
- 13531 by Vanyilla on Windows 11, Ryzen 9 9950X + RTX 4090
- 11000 by Sam on Windows ??, Ryzen 9 5950X + RTX 3090
- 10032 by owobred on Windows 11, Ryzen 9 9950X + RX 9070 XT
- 0 by Adesi on Windows ??, Ryzen 9 5950X + RTX 4080
- 0 by Quack on Windows 10, Threadripper 3990X 64C + TITAN V
git is broken as usual
the mods are gonan hate us if we keep filling chat with this

They already don't like it
i need to do push everytime it changes and we want it in git
!trianglepush
did they say so?
Took a bit to update nvm
we can move to #bot-channel np
I know vani made a comment on self botting, not sure if exactly that tho
ye thats fair
mods ping me or quack if we need to shut it down
actually ping both probably
#programming behavior
annoying neurocord mods since 2023 
Quack lost the windows leaderboard 

im going to retake both tonight

well, i installed mesa now so time to reboot
programming has moved from a mostly ai focussed chat to le triangle
I already ignored it 
self-bots are annoying
fuck it, im going to use debian
T R I A N G L E
also i figured out how to unlock my 13900
Wdym? Can you block an accounts messages?
Yep
Interesting
i will retake what's rightful to me
im not gonna ignore it, but i hope people jsut wont spam it
There are no sections missing (real)

should I publish my code for validation
im going to add cooldown

i just got my new monitor
imma lock in

Imagine losing your monitors to triangle
i shoudl add easyeffects to my startup programs
otherwise my ausio is always set 6Db too loud
deepfried_vineboom.sfx
i'll get obs on my debian to also prove i didn't change the fps counter
i actually put the audio -6dB so i have more bass lmao
just incase if i got insane numbers
obs probably obliterates your fps tho
Was about to say
oh
just take a picture or something i dont care
bye bye nixos, hello debian
just to avoid suspicion of me cheating 
you should use capture cards
okay so
ok i managed to get llvmpipe, lemme see the fps
single core maxed out only pull like 40-50W
thats a lot tho
and my cooler can handle 200+ W
imma do extreme overclocking
hope my CPU survives

try not to resize
just run and hover to taskbar to see fps
managed to get llvmpipe working, only 6K fps tho. yall said there was some magic comamnd with threads?
lemme also switch GPU
maybe to my 4090 idk
won't fit in the briefcase
but i have a riser
4090 is gonna obliterate us
Resizing the window is banned as it deletes the Vulkan context and stops rendering anything
cursed briefcase moment
huh, the magic command does work, i get 16K fps on llvmpipe with LP_NUM_THREADS=3 make run -j ARGS="--use-igpu"

debian + 13900 ES overclocked
then if im not satisfied im switching to a 4090
you know what fuck it burn that 13900 i need more numbers i can always buy another
now that im rendering on cpu, changing the resolution would surely matter a lot more
make sure it fits avx vector size
idk what the sizes for that are
what's your cpu again
but at 10x10 pixels im getting 20K fps
5950x
20K still
4x4 because it might be running off 64 bit(?)
ok so while 4x4 does increase the fps to 21K, it breaks the window
does that count for the score
from my experience on my install of hyprland, if a window is 5 pixels wide or less in any direction it jsut breaks and becomes 720p
imo no
the triangle needs to be visible
and im just getting a black window
6x7 was the right call imo
6x6 is the min amount of pixels i can render and seems to be the fastest
im getting 21600 FPS
πΊ still visible?
kinda
imma consider that valid ngl
i dragged the window next to the counter, its the glowing thingy
Triangle 2 
le trianlge
smol

let me try on a laptop with igpu
aight my cpu can do 6.7GHz sustained
Yep, there sure is a triangle there
I can go up to 9 ghz in my Intel processor
LN2?
Yep
22187 
First i will coat Waterproofing
1.42v
My FX runs at 5.5GHz 
For the lols I was searching for how much a tsmc wafer was. $20k USD for high end 3nm.
If it stays at that price that is.
28nm wafers seem interesting tho, I've seen 2-5k depending on the manufacturer. Don't ask me how this affects our triangle endeavors but decided to look it up.
That seems a bit high
Don't burn your ES CPU
im not joking when i tell you that they actually handpicked just really tiny volume of samples when making the es
which means that they really picked out the best ones
so its no surprise that ES can sometimes outperform the actual retail products
engineering sample cpu as your main daily driver is seriously underrated.
I think vedal is training from discord
Only issue is to be within the rules, one would have to make their custom chip vulkan compatiable
And what's the evidence?

i have a 28nm wafer
MY THOUGHTS ARE TELL ME
Vibes, noted
22875 
Doubt they happen to support vulkan
blue one is a samsung GDDR5 28nm process node
let alone everything else needed
LET ME USE IP GRABBED IMAGE
the other ones a random FPGA/HPC on 12nm process node

I mean if yall want to make custom hardware for the sake of rendering a single red triangle 
im tempted yeah

I like your hardware farm
If you're at the point of making custom hardware, just hardwire the display out to a red triangle
I'm changing bitys of png
Needs to be vulkan tho
so can't hardwire
in theory you could just make 60+ nm ASICs that pulls thousands of watts but would outperform any scores
Triangle oven 
the traingle accelerator (tm)

6.9GHz sustained
nice
as long as its under 1.9v should be fine
2v is instakill
1.9v it may still run for few minutes
1.85v, run it for like few minutes and you took months of its life away
Can't wait for
tomorrow to be like "can't get Elvyn PC, have to get another new CPU"
Seems like the world record is 9.0GHz. So we got a bit of room
I'm 100% that is latest hardware tho
American source moment
So that is the goal to triangle
I should look for deals here, if I had budget maybe I can emulate Quack's luck
Any one has better motherboard board that can Handel (15 v )
Your CPU will explode if you put 15V on it
Test nitrogen colling
I hope that doesn't burn core
you can try that on dummy test chip made for thermal testing afaik
at least on 100+ nm process nodes
I9 to dead
lower than that it'll short out and burn
aight i figured that the IMC is also unlocked now
Guys i made a linux version of the compile shaders bat file, and i updated the shaders to be faster
time to go for 4400MHz
Test file
I pushed them
bent pins category goes crazy
dram voltage to 1.55v surely nothing will go wrong, i'll 1dpc it though
Gotta love me some pin 62, nice even power distrobution 
Let me research via grok ai or chatgpt
get out
From what I know, 1.4 or 1.5 was the max suggested continuous voltage for DDR4 overclocking long term
You're probably fine
I personally never increase voltage
I only increase the clock speed without increasing voltage
So many 5600 for less than $50, too bad it is a gamba
Still, my RAM runs just fine at 1.35V (stock) with 3800 MT/s and whenever I get the chance, CL18
Prove for #bot-channel message
hol up, 2 bent pins for 35 
They are even on the outside
max I can pump is ( 2.0 v)
bid, nvm
Thats too much imo
IF the CPU CAN Handle (15 v ) but
I want to try if available (45v)
Cpu only Handel is 2.0 even u using nitrogen colling
(One time only )
I just don't want my hardware fried because I can't afford more of it
Same
Still planing mode
But i want to try it
why is the triangle smol
Faster
We're doing cpu rendering now
Gpu doesnt care much about amount of pixels, cpu does
this guy be like "i overclocked my CPU to 15 volts and undervolted my monitor to get negative response time, i also overclocked my 4 PSUs running in unison"
Cpu also likes big amount of frames in flight
(btw 2 volts will instantly kill your CPU. that's obvious)
0.01A with 130V
also
I do kinda wonder what kind of silicon I've got in my CPU and how fast it could actually go
Maybe so
its also pulling shit tons of current so i had to switch to my AIO
- its single core, heat density is fucking it up
wdym what type of silicon
do you mean isotopes

Im hetting 23954 fps with 512 frames in flight. Cpu really like big frames in flight
I mean how good it is
Weird, it was worse on my end
my fps goes down if frames in flight goes above 32
I'm also rendering in 7x6 with some changes to the shaders 
In my experience from testing just now, around 64 works the best
You know what would be even slower than any computer? Hand-executing the Vulkan draw callls

Aight i got 24044fps 
Im happy with that
caibi ur dead
Just draw triangle on paper by following the Vulkan draw calls, swap the paper, do it again
Finally at a decent fps 
I assume too many volts
5GHz -> 7.6GHz sustained, 7.9GHz short period
3200MT/s to 4800MT/s CL16
oh caibi ur so fuckin dead because im trying windows first
the program is extremely session-based, the same executable can lead to stable 28k fps or stable 33k fps
Crazy how my RAM instantly crashes if I try to run it any higher than 3800 MT/s
And
is here at low DDR5 speeds
my ram is 6600MT/s C32 

With DDR4 level CAS
In the words of Elvyn, nothing ever happens I'm going all in
Cpu dont have lungs, the 13900k can dmoke if he wants
i switched sticks to the one i got from my dealer. s8b binned, its rated for 4000MT/s CL18 on jedec
Oh, that explains
quacks dealer
its so expensive though because s8b bins
grabbed some actually good RAM and then overclocked it and I assume increased voltage to reach +800 MT/s
1.45v yeah
my 3600MT/s sticks can only do 4400 CL16 max at 1.55v
wait i think i can do 4800 CL14 on the s8b chip
surely
im running at 24044fps max now, which is 0.04159041756779238 ms per frame. with 64 frames in flight that gives us 2.66178672434ms latency
thats fine
I can overclock my ram to 7800MT/s CL32, actually

im tempted to get a ddr5 mobo and get a 2x8GB ddr5-7600
I'm happy to be in 3rd on linux if it means that I am the highest mobile chip
get better hardware
oh btw the reason i don't use this s8b ram sticls for daily driving is because its 2x8GB, lmao
but quack is going after my spot as well
i need a sponsor
anyone willing to sponsor me
???
EWWWWWWW off center
eeew
dYmanic isLand
ew
dynamic my ass
hold up i'll go to bios again
-SanVanTriangle, 2025
I know being off center is bad, but may I have you look at #3 compared to the others
okay
it's not on top
and uh
yea kinda ugly

Under windows
Linux is about to go insane

YESSSSSSs
caibi ur so fucking dead
congrats
Imagine not having context on what we've been doing for the last 4 days

All of this is for a high fps triangle
I don't really chat that often here
but still more active then #artist-alley and #gaming
I've basically never checked #artist-alley, haven't done anything I want to post there

Quack's score to beat:
1. caibi 34447
Overclocking competitions got nothing on this
Unstable system
nvm
it might just be instability
this monitor apparently pulls way more power and even my molex to usb adapter can't supply it
i used its own power brick
and it works
using molex to usb to power a monitor
Classic, my portable ones just shut down instantly
24245
I really want to tear one of those open because the panel can do a lot more than 4k60 and the usb interface is getting in the way
Oh wait a damn minute
Yall said lm num threads 3 or 5 was good, so i tested and landed on 4
But 1 gives me 31K
I landed on 1
imma boot to tiny 10
31609 
and triangle there with all hthe overclocking
also i can get 4933MT/s CL11 now
@olive sable give me the new src
The what?
My stuff is merged, don't know if anyone else needs to upstream
My computer is hell
wait have you pushed the new one
The github is up to date
We're not i dont think
I know none of the regulars would
because we all know this channel is useless to scrape
40k
34K
kind of lower than expected though
Blame windows for all issues
41554 seems like it
Are we printing the fps now?
i used caibi's zip
oh hold on
@true hemlock sam changed the frames in flight in a leaderboard commit, better set that back
I have 6 because igpu is still faster
Im switching to printing now so its comparable to what quack is doing
Probably less expensive as well, considering the one botch fix in the fps calculations
True
Igou is faster than software for you?
I cant compare
I'm probably to blame
either that or somehow i accidentally triggered the buffer bug again
?
can't
I havent tried on windows lately so 
you entirely skipped 40k
Nah thats good

thank god dynamic voltage is also a thing so my cpu doesn't run constantly at 1.7v
That means linux will be even higher
So you got a chance to beat that leaderboard as well
quack what motherboard you have
@trim valve im coming for you
just a random B760m with like 8 pin eps
bred apperanty did no modifications, so it better be by a bit
i just disabled ALL my cores except for 2
is it the DDR4 one
yeah
aye got ya
so only 2 cores pulling all the power because my mobo isn't 8+4 or 8+8 EPS
elvyn's lucky though, her motherboard has 8+8 EPS
fuck it
i need to make my windows score beat my linux one first
55.3k is to beat
nvm then
definitely can't
UNLESS
IF I GO FOR 1.75V
nah
enough torturing my poor cpu
time for linux
we need the cpu to live for linux
nixos or debian chat
32232 is my pb on linux rn

Whichever is less of a pain to compile on
6+2+1 
Previously when using gpu rendering there was always a pinned core, now with llvmpipe there isnt
sam what's your best score
I got 32232 now
I have an 8+8 eps for cpu power, and 16+1+2 stages/phases
Still linux cachyos
that sounds very desirable 
32276 
need to have a better cpu before trying to kill my current one though
what's your gpu? the 4070S one right?
Im pulling 62W on occasion according to zenpower
just for a single triangle
100W / core is just frying amounts
Its not just a triangle
Its the fastest goddamn triangle on earth
I just recalled this
why did I fucking buy an 8+8
and a non-k series cpu
????
This should become a nerdy yt documentary, it would be funny
We need a video essay
I'm mad wasting money on my motherboard ngl
I checked for ES and the only ones I could find are far too old
Guys im in the bios, how do i speed?
Engineering sample
ESs are scary
no ES are fucking peak
you never know what have happened to it
Game boost?
CNY$600 for a 14th gen cpu sounds like a scam fr
Surely the big button with gameboost would be faster
which one
you also need to look for the revision version and label
mine is rev 2 (Q0L4)
if you wanted 13900, get Q0L4, avoid Q0D8 even though its way cheaper
Hmm
also 
What do i do?
advanced dram config
LMFAO
leave it if you don't know what you're doing
you only touch max 4 parameters if you're not doing some extreme stuff
While i know the basics, im out of my depth here
Real 100k here we come
60% efficiency you meant?
Welp, bold ram oc from 3600 to 3800
Not the first time, but second time did
The normal timmings
means it could be instable lol
if it's not consistent, you might face some stability issues
Ye its probably unstable, lemme lower it
try 3733








