#🔌│tech
1 messages · Page 1023 of 1
if you do that, most of the issues besides power draw can be avoided
and as long as you install the chipset drivers it will run fast
but like
im really particular to the feel of windows desktop
and uh
am4 just aint it
its somewhat fast on my mega stripped windows live stick
but it still feels slower than my old system with the same conditions
i am literally in the process of doing a fresh install now tho
so we shall see if that helps
hmm
i did only get the cpu a week ago
built a new pc with a 5950x yesterday, feels the exact same in windows as my 5600x and the same as my 8600k
there isn't any difference in just windows
guess i am more perceptive
or more likely something just is wrong, if there is a difference
but yea if you disable the power savings features, the performance drops
but if you leave them enabled, its not as responsive
thing is
most people dont really optimize their systems
and sure feel free to not do that
but uh
if im gonna be paying a grand, id hope to get better performance
im also not happy about how amd reports ecc errors in software instead of hardware (hwinfo cant see the ecc support until it reports an error)
where on intel theres always an ecc errors line (it just says 0 when there arent any errors yet)
you really want to use ECC and miss out on xmp on something below threadripper 
ecc bdie is cheaper than regular bdie 😛
yeah but ecc only works at jedec timings
my ram was perfectly happy throwing ecc errors at 3200 c16
which very much isnt jdec
also
what
I don't really know much about ram. All I know is i've got 3000 C15
why would ecc not work at custom timings
oh god
uh
5000 or 10700k you should probably get new ram
i know a few folks running 3800 sub 70ns on ryzen ecc
and just using the ecc as a way to get stable slightly past what the hardware can actually do
an ecc error an hour about
and everyone i know with a 3175x is running overclocked ecc memory
except one dude who was running it not overclocked
i was lucky my 5900x runs 2000flck with 64GB of bie 3600 running at 4000cl14
stable?
linpack doesnt stress fclk much
use occt
occt large is solid for fclk
oh also
gib aida64 latency
aida64 no work on linux
i mean
AMD WHY
linux would of booted in your issue
nope
linux doesnt boot amd raid without significant trouble
but yea my board wont do over 1900 fclk
it just dies
oh yea you can run intel mlc
it wont give a good idea of the true latency but it can be compared to my mlc numbers
There is literally no difference for feeling between cpu's at desktop. For GPU's maybe yeah
Nvidia and AMD draw the cursor a bit differently iirc
my last windows+amd setup was a 290x
and i had tons of issues with the cursor lol
Both have the same amount buffering (DWM) on desktop, unless you globally disable vsync on Nvidia's (Feb 2021) MPO'd driver
AMD lacks a MPO driver atm but it's probabiy in the pipeline somewhere..
I have really hard time believing someone could spot differences of 2 digit microseconds between CPU's at Windows Desktop
the fix for the cursor was to flash the bios with a faster 2d clock
or turn on cursor trailing
im talking about moving the mouse
intel has a boost ramp time of 50ms which is terrible, so a set clock is able to feel alot more responsive
amd is supposed to have a boost ramp time of sub 1ms on 5000 series
but it doesnt feel that way
it still feels like my mouse is a solid few frames behind
is your driver from after feb 2021 btw?
what driver?
nvidia
yes
i think my mouse weighs as much as those 3
try forcing vsync off globally, see if you can feel a difference
considering ive stripped the mm710 too, probably
maybe your settings were reseted.
i need utility than speed though
it feels better now that i have a fresh install
if you force vsync off globally, you have 2 frames of input lag less at desktop, moving the mouse
but if a fresh install feels like a 5 year old install did on the 9900k
what is it gonna be like in a year or two?
uh no
yes..
i can assure you that that was not the case because i didnt have 2 frames of lag
it was only around 1
and changing it didnt change anything either
so 
I know you did not read, but desktop (w10) is double buffered by default, MPO'd driver with GLOBAL VSYNCH OFF has 0 buffering
good for it
yet somehow
so you're not feeling 16ms of input lag?
absolutely not
Why would you feel between intel and ryzen then... 😄
to hell
god damn this is so dumb
why did i want am4 again?
"oh wow amd is so great now they fixed all the issues"
-_-
try to check if dragging a icon on desktop lags behind the cursor, if it lags on upper part of the screen you still have 2 frames of input lag on
if it stays on top of the cursor its 0 frames
on bottom of the screen - the icon should be faster than the cursor
when moving slowly its perfectly synced, but move too fast and it desyncs
which way it desyncs?
it trails behind the cursor
by a pixel or two
its the same at the top or bottom of the screen
rip, unless nvidia changed when cursor is drawn on the screen
maybe you need to boot PC after having turned global vsync "off"
thats not a thing nvidia controls
it is
no it really isnt...
its not doing anything strange when im dragging windows around
Iirc the cursor is a GPU sprite update (overlay over framebuffer), but everythng else is double buffered if DWM is not skipped, so this is expected behaviour
If you can disable DWM the behaviour of dragging changes
What else could be causing said difference of feeling
if the cursor itself is the part thats taking longer to start moving...
idk, I can feel no difference compared to intel system. Maybe its your raid setup interfering with USB
i had a raid set up on intel too...
whats your DPI btw?
1600
oh ok that should be fine
yup
but also dpi shouldnt affect the hero sensor
only the other two
the thing is
on this system my mice and keyboard are directly attached to the cpu
where on my intel it was sharing bandwidth with all my drives
it should be significantly faster on this amd system
it affects how fast movement is started since lower distance is required to cross over counts, but it's only like 1-3 ms at 1 m/s speeds
Odd, maybe some pcie 4.0/chipset/agesa issues or interference
Hmm
fuck amd nvme raid is so sloow
intels raid just slaughters it its not even funny by how much
Take the keyboard out, test alone with the mouse. Intel has 50µs USB interrupt moderation on W10, but it was like 1 ms on W7. Who knows if AMD has longer.
Maybe the Lekker is delaying the mouse
its not like i can go back to the intel system to compare at this point
i dont have my lekker yet
im still using my boring old gmmk
actually
no
im getting a he
which isnt called a lekker any more?
I just use 1 USB device (mouse) in the cpu port. Too paranoid about interrupt moderation 😄
yea but i care about keyboard latency too...
and this keyboard doesnt send data when its not being pressed anyways so its not a problem
maybe, but worth a shot
i do have a keypad that does but its not attached rn
didnt change anything
still feels alright
no faster than my previous system
Odd, my system feels much more responsive than my old Sandybridge, but that was lagging really hard. I get like 125 ms avg on humanbenchmark, so I'm not a complete potato yet.
sandy bridge
im talking 9900k to 5950x
sandy bridge is capable of running modern titles
the 9900k is actually competent
shrug, if anything the 9900k should feel sluggier due to patched vulnerabilities slowing down I/O. Or actually idk how much they patched in 9th gen hardware side
there werent any patches that affected io
and either way
thats not the problem
the 9900k is fast
but i know it slowed down over time
why is the 5950x the same speed as the 9900k is now
it should be faster
I don't trust my own perceptions regarding minuscule latency differences. Maybe you've aged and become either faster or slower yourself?
the problem is i have an ice lake laptop here that is faster than both of these systems to compare to
and the ice lake laptop has always been the fastest
but it used to be less faster if you understand what i mean
Uhhh, maybe the laptop is using some intel gpu magic
yes
which is why it should be a solid comparison
especially since i havent changed it at all in that time
time to start clicking humanbenchmark or that 9k fps program, and logging differences between pc's
problem is it isnt click latency that i care about
its mouse movement latency
and whats that 9k fps program?
either way too late to run it on the 9900k
Idk what else there could be that explains it, aside from what we discussed
maybe the mm710 is scuffed cuz its coolermaster lol
the 5950x is signnificantly slower single thread
i main the superlight tho
the mm710 was just something i had on hand to compare with
there's a 9k fps client version of humanbenchmark.. but it makes me feel weird
so I can't use it
🤷♂️ im well over 10ms of inconsistency in humanbenchmark
well firefox/chrome are double buffered and capped to hz, unless you disable it from the cmd
making firefox run at 3k fps heats my pc tho lol
🤷♂️ im talking about moving my cursor to things before clicking on them
not about how long anything takes to open
warning: if you're epileptic
but what really pisses me off the most is the crazy high idle power usage
its so dumb
if amd is including chipset power usage in the core power usage that would explain it
but uh
that doesnt make any sense
gonna try 1.2.0.3B tomorrow if it solves anything
oh yea on 1.2.0.3 it still kills my usb hdd if i try to use it
it would be great if they fixed the usb drive issue first tho
evga says it requires a 750w psu for their 3070ti
requires or recommends? They all recommend much more than the cards pull to cover themselves
a 3090 only pulls like 350W
They're worried about spikes.
Linus did a video about...was it 3090s in SLI and showed just how bad spikes could get even at stock and why nVidia recommended overkill wattage for a reason. (Ultimately their own design flaw but here we are)
yea if you raise the power limit it uses alot more power than youd think
which is a problem for the boards with really high default power limits
but yea
a 5800x or 5900x will draw near 200w at peak
a 3070ti draws near 300 at peak
and thats fully stock settings
you are hammering that 5800x like a motherfucker. Never even been close to 200
Doesn't seem to matter to evga though. They say 750w for anything above a 3060 ti regardless of whether it's a 3070 or 3090
not talking about amd reported nonsense
im talking about vcore in
and the 3090 asks for 850w if it has a power limit over 350w
ok nvm they must have changed it
thats very odd
evga doesn't have any of those. Maybe the OC premium versions out there
there are plenty of seasonic 750ws that cant handle the 200-500w swings the 3090 produces
the 3090 ftw3 does 450 with power limit maxed
the kingpin does 540 stock
wait no 512
yay my computer doesnt freeze whenever i open my usb hdd 
On my 5950X, mouse movement is stable and snappy enough I can tell the difference between my different monitors, how much pixel response time each one has
good for you
Doubt it's the CPU that's the issue
its not what anyone would call slow
its just not as fast as either the previous cpu i had in here or a fucking tablet
Then you have an issue in the pipeline somewhere, and I doubt it's the CPU
well what changed was uh
the cpu
and i removed some extraneous drives
so uh
it has been given every advantage
oh yea ofc the motherboard is different
but its a direct successor
Aight bois, with the Etherium difficulty bomb inbound you should all be keeping an eye on ebay for mining cards later this month.
Around the 23rd we'll be seeing an exponential increase in Eth difficulty until the London hard fork gets integrated into the mainnet, if you wanted an RTX 3000 GPU for gaming, soon will be the time.
I'd expect maybe a little above MSRP.
TL;DR: More than 100TH/s of GPUs need to move somewhere, some of them will doubtlessly end up on ebay.
Nvidia: "Our cards don't go to miners. We swear."
rolls eyes 100TH of second hand cards and years of purchasing.
Most of the farms I'm seeing are RX400 and 500 series with the occasional GTX 1000 and RTX 2000/3000 series.
But hooo boy there's a LOT of RX580s in farms.
Y tho wouldn't you want to keep your hash rate even with reduced difficulty to mine more eth?
Difficulty bomb increases the network difficulty.
And currently Ergo is more profitable.
@young pecan did you end up making profit with the cards you bought?
good luck hope you can recover from it either mining or selling
from the debt I meant
Ah yeah.
The math adds up, I'm just gonna keep mining.
Hell, yesterday I found a solo Ergo block, so that's pretty pog.
How much taxes do you need to pay on income from selling crypto?
@zenith sonnet 30% here if it goes above 1k in profit
Besides much like with Bitcoin doesn't difficulty correlate with value?
Technically none if I withdraw nothing.
As long as it stays crypto you technically own nothing
but whats the point, one day you want the profit
One day you might wanna turn it into money
Sure, but so long as it's not money I don't have to pay tax. :)
So I just...
I trade purely in crypto.
:)
As far as the law sees the graphics cards arriving at home are gifts.
Or if they legalise it as currency, you'll have to pay taxes on it
In which case I pay taxes until I somehow become a resident in a country with no taxes. :)
Corporation fuckery, bois!
ehehheh true I guess
I'll happily just move somewhere with no taxes so I can just...
Not pay for asshats to waste my money.
If my government actually did a good job I'd be more than happy to pay taxes so they could do things.
But we spend WAY less than enough on basic shit for a good society and entirely too much on stupid BS.
If I could pay taxes and guarantee them all toward education I'd happily accept 40% tax.
But... apparently not. :L
:(
People shit on the Catholic Church a lot, but despite all that's wrong with them, they're actually the world's biggest spender on social welfare and education
At this point I'd rather the church take my money than the state spend it on shit like an airport in Berlin, a train station in Stuttgard or give it to political parties to fuck around with and blow it on stupid meetings
I'd say the child indoctrination thing shouldn't be ignored there.
All parties have their deals with the devil and the Catholics have some dark ones in their time.
Idk, I try not to trust groups that don't care about me. That just seems like basic self preservation.
Funny enough, the Church owns a lot of good hospitals in Germany. Many kindergardens. Senior residences. Care services...
Thank god
old news..
And he did step down today, so that's pretty good news
oo
This doesnt give the right format despite me having the right formatting options according to the documentation .-.
and as usual we can't read your head and know what the expected format is...
Its this, but it isnt outputting correctly
I have the same options that the document shows
Unless they changed in the new python version :/
😄
Does order matter?!
i dont see how though
It shouldnt but its python, id not be surprised
the use the same options here and it works.... WTF?!!
how can I make the color of the bar be different based on its height?
if (height > 40) { col = red; } else { col = green; }
Anyone got any ideas on getting a Skypad top for a powerplay mousepad?
The powerplay has a dimension of 34cm by 28cm, Skypads are 34cm by 30cm, doesn't fit.
Anyone know of a custom measure for that?
I have to get the largest number from the second column for each hour on the first column..... how do I do?
the first hour has 10
the second has 16 as you see
How I get these numbers? 10, 16 etc?
In Python
count how large that cell is? or count how many rows till the next one starts
Can that be done in Python?
Nah gonna have to do that in assembly
Anything can be done in python
Die
I don't use python, but if python can't do subtractions, I'd be very surprised
I can just find tutorials on how to set cell height, as if you are writing data with Python to spreadsheet
not get it
then don't look for that and just count how many rows till its the next hour
how tho
literally just counting how many rows there are, probably some way to make it easier but this works without knowing any pyhton
I need to get the max number in the second column based on the height of each cell in the first column :/
if row 2 has hour7 and that value changes in row 12, then you have 10 rows of data
thats all
Found easy way to only select certain columns for the dataframe.... this is great
O.o this is dramatically different

Oh... This would be handy to have
just be aware that it won't work for HDDs
there are ones with an external power brick that do
They need more power
True
Will work with 2.5 hdds
not sure about that one, even if ETH isnt that profitable anymore, why not just switch the coin? 🤔
Oh everyone that knows anything will.
But there's always the few scared ones that'll liquidate.
hmm well that might be the case but i dont think it will be that much. But it will definitely not hurt someone to keep an eye on it.
BUT what does prevent other miners to just buy the cards from the scared miners? 
Nothing, that's what I'm doing. :P
We'll see who beats who to the cards. :D
But hey, better that everyone here keeps an eye on them, y'know?
Same here 😛 Atm running 6 rx6700 as an experiment since i just love experimenting with tech lol. Was luckly to get them for nearly MSRP
But if i can find another 30XX card for MSRP holy why should i not buy it ^^
btw. speaking of switching coin... have u seen this? Its quit much speculation in there but if hes right, hes a pretty rich men 😛 https://www.youtube.com/watch?v=L0m55Qr__tM
What is the best budget GPU for mining Ravencoin? In this video we're having an in depth look at Nvidia graphics cards to see which is the most economical for mining Ravencoin, taking both cost of the GPU and electricity into account. This all came about after watching Son of a Tech's video (https://youtu.be/jIxrW2ePRf0) where he hypothesised th...
Yeah, I've seen it.
The theory is sound, but holy shit lmao.
Like what really?
Farms use tens of thousands of watts easily.
Yo, anyone know where to get a good GC carrying case?
The old ones aren't made anymore and I'm looking for ways to store old stuff.
Yeah, but idt those are made anymore
i didnt know you could date format in a f-string. can you actually? yes you can
I got it working, turns out it doesnt display it totally the right way in XML but when you open the XML in Excel then it does show the right format
Can scatter plots on Matplotlib have various size dots depending on a variable?
So, in a bee related incident, I managed to spill water on my rig, which soaked the gpu and shorted the PSU. I was only able to get a Smaller factor PSU on short notice, so the cables had to be jostled and it doesn’t fit the case as well, but happily the GPU works! Was not looking forward to buying a new one.
For context, the bee was trying to get out of the window above, the glass of water was on the edge of my desk (the brown one) and the Pc was between the desks…
If you got demi/distilled water into electronics it not that bad. The minerals in water fuck things up
Tap water, but we immediately started drying the GPU out and it works fine now.
The water dripped through the top vent pretty directly onto the GPU
The PSU shorted within 10 seconds of the spill, so we were worried the damage would be more severe
Cut power immidiately and remove the water. its all you can do. If there is no currents running anymore it should do no further damage. Still need to remove the tap water as it tends to leave mineral deposits when drying up, which are still conductive
PSU is most likely to still die with their big output capacitors. GPU doesnt need them so runs out of power much sooner
MI-200 is crazy
kinda surprised we still dont know the announcement for it
makes me think it's gonna be late in Q4
It's kinda odd that they only have 750 accelerators on 200 000 CPU cores
So, 1 accelerator and 4x64 cores per system?
Blade*
you havent seen the press release
its blades
The way this system is likely set up is not homogenous especially since epyc can't do 3 links
and hp/cray did el and it was blades.
Blades as in the technical definition or blades as in the compact server variety?
lol
they are blades
the are nodes that lack standard io and power
the connect to a backplane that provides both
infact has cray ever made non blade super computers?
Yea so they aren't traditional blades where you just have CPU and ram on the blade, with pcie going out to attach to other blades
they are still blades
These use ethernet
Yes they are still blades but they aren't classic blades where you can have 2 dual CPU blades then a gpu blade with no cpu
They are using separate gpu and CPU nodes
Lmao
Amd outsold Intel in the new supercomputers list 1.5 cores to 1 yet not a single new supercomputer used instinct gpus
bit hard when they aren't out yet/mostly still custom made for the needs of the DoE and such
M100...
and currently available instinct cards are like first step of zen, a start but not amazing
Again mi100
..yes? apply previous sentence
no cuda, rocM is still far aways from the competition
pretty much all of the effort that went into it so far was to make it useable by DoE as they are by far the main customer
Mi100 is 3rd gen that supports rocm...
There's a big mining shift happening over the next week
No miners are buying cards this week, and next week at least some will sell cards
and based on amd's own slides rocm 4 or 5 not sure, is the first one to cover all areas, previously it wouldn't, so aside from the hyperscalers nobody used it. not adoption in lower tiers like for cuda means less people capable of using it, or knowing how to use it better than cuda etc.
Yes I'm aware instinct is garbage, but usually the supercomputer folks are the ones to try anything if it can give them slightly faster results
And the fact that zero were deployed in a serious supercomputer is kinda yikes
Especially when most of the CPU compute deployed was AMD...
I mean they announced frontier and el capitan quite a while ago and everyone expecyed those to be the first to really use what is new at that time
el capitane is 4th gen instinct tho
im talking about now
and frontier is the same story
yeah for that one frontier is the first one to probably get post-ES mi200
and why would you want to risk something that usually takes years to finalise on paper on a software stack that is only halfway there and might at that time even under best circumstances only be on par with the safe option
https://www.servethehome.com/amd-radeon-instinct-mi100-32gb-cdna-gpu-launched/ if you compare the example they show vs what nvidia shows on a slide for their hpc solutions, the fields are much smaller and limited. Same as to why Rome was a much bigger hit on datacenter than Napels
i mean amd was yacking up a storm about how fast their accelerators are 
I kinda see mi100 in the same spot as napels
show it works and all and on the next one people might jump onto it
with the difference, nvidia was not facing issues like intel
so even harder to convince HPC buyers
i mean
intel sold more cpus, just fewer cores
but thing is
the mi100 is pretty unilaterally faster than the v100 in real world compute
and well
there were new v100 supercomputers deployed
v100 was an option that was solid and had a track record and came earlier, and these contracts have often a long lead time
again
I dont think the "slightly faster" was really how it looked when it was time to decide who to get
- I wouldn't be surprised at all if nvidia was greasing the gears quite a lot in cases people were debating on getting instinct
ah yes the conspiracy angle
we know intel did it, nvidia does/did their gpp bullshit, this would be a lot lower on the list in terms of being unethical so I dont see why not

also what if in those cases they did test it and for either efficiency or power reasons or whatever nvidia was the better option?
opteron still got deployed...
so 30% or whatever number of them are idiots and should've bought instinct or what is your point?
ah yes definetly
that was a question because I don't know what your point is...
im just saying it takes skill to mess up that badly
where you have none of your attached accelerators deployed in a period where you were winning on total cpu compute
so is your point amds marketing was not good enough to get sales and the gpu was good enough?
Ok so apparently setting up infiniband p2p is a huge pain so I'm gonna try Ethernet I guess :/
ehhhhh
not that bad
tbh
if you are lucky equipment is dirt cheap as well
Can you help me get it set up? Because it's sure as hell not working for me?
I have a pair of connectx3 pro dual port cards, 1 in my server, one in my desktop, attached by a proper fdr cable
What gen are you using?
server on the bottom desktop on top :/
I know it thinks it's working...
Fast charge is such a life saver
In the hospital and got to 10%
Now back at 60% in no time
At the moment I'm stuck on 1g :/
i mean at least it isnt wifi
Seen the specs for Samsung's gen5 ssds, that will be outdone on single drives soon lol
It's already outdone by a single gen 4 ssd
But a single gen 4 ssd costs more than my server did lmao
custom built with almost entirely parts lying around
i mean sure i could sell all the parts of this server for around 2k but shh
only parts i bought to turn it into a server were the nics, the cable, and a dummy hdmi
i could swap the hardware into the server i actually got for free if you want lol
besides the only ssds as durable as this guy of the same speed would be crazy expensive, not quite p5800x expensive but still very much up there
unfortunately i have requirements
minimum requirements
Hmmmm.
Ok, ok.
Hear me out.
Free raspberri pi with a free SSD attached.
Impossible to get better price per gigabyte. 
holy fuck im such a moron
jesus christ it takes effort to be this dumb
qsfp cables click when they go in
i didnt check if they had clicked
thats 6 hours of my life i wont get back
Material scientists have developed a fast method for producing epsilon iron oxide and demonstrated its promise for next-generation communications devices. Its outstanding magnetic properties make it one of the most coveted materials, such as for the upcoming 6G generation of communication devices and for durable magnetic recording. The work was ...
Yeah no. That’s a big claim and it comes with artist concepts of what it looks like.
Scientific papers or bust!
Being able to make epsilon iron oxide 30 times faster whilst also making it purer and for cheaper?
Something there smells off.
https://pubs.rsc.org/en/content/articlelanding/2021/TC/D1TC01242H#!divAbstract
you can just scroll to the bottom of the page and click the scientific paper
In this study, we demonstrate the synthesis and the investigation of magnetostatic and magnetodynamic properties of epsilon iron oxide (ε-Fe2O3) nanoparticles of various sizes. Silica gels were obtained by rapid hydrolysis of tetraethoxysilane in an aqueous–alcoholic solution of iron(iii) nitrate and then dr
How does one install U.2 on AM4 for example?
m.2/pcie to u.2 adapter
they are pretty cheap
since all of them are just pcie in disguise
OR you get one of these two boards
since those have one natively
Ugh... Google still haven't rolled back their terrible UI changes. Guess I'm sticking with an "out of date" version.
of.. the play store?
Chrome.
@old marlin damn.. you just plug an m.2 <-> U.2 adapter to the m.2 slot?
Time to do a dist upgrade..
Just thinking about buying U.2 Optane drive
but how hard that would be to slot on the AM4 cpu m.2 slot
probably hard..
Yes
IIRC there were some orthognal aligned ones as well
doesn't look like they are still a thing
time to switch to firefox 😉
Not old enough.
And a temporary workaround at most.
But yeah - Might have to migrate to Firefox at some point, if only to stop the pestering from Chrome about being too old.
Once I find a smooth way to migrate... literally everything about Google.
Such a boomer
Some of us like having everything in one place 😛
Yes dear
and the "shield up mode" or whatever it is called
seems to be very proud about everything it blocks
Ye lol
yea uhh
I don't want that
the point of an adblocker is to not see distracting stuff
so having a windows open up to distract from content
seems counter productive?
ah should mention
am on ios now
and safari has 0 ad protection
at least that is what it seemed like to me
just wanted to say
will sure not work for me
I couldn't screenshot
ah
not qutie what I meant
maybe I am just doing so weird gesture
but everytime it blocks something
it opens a window that covers half the page
and tells me that it blocked x amount
I'm sure sure tbh
Everytime I open a window irs just a picture how
@old marlin
Minus the top sites which I chose
the special character placements
are just bad
even gboard does not have a number row over the regular keyboard
so on the second layer the special characters are moved to random positions
@astral haven you had some optane drives? what size and form factor? 🙂
swiftkey all the way
I use Openboard :p
Like I at least need the comma on the first layer
I'm so tempted by optane drives 😩
find a way to borrow one for a week to make sure you'd actually want it for your usecase
Not possible in here ;/
No Amazon to exploit? 
I assume whenever an app wants you to scan your fingerprint, that it calls an Android API to get the OS to scan it for send the result back? or does it access your fingerprint data directly?
@ashen spindle Hm, do you know if optanes (e.g 900p) can use directstorage?
dont see why not
that's a pcie thing not something the flash itself has any interaction with
no idea lol
wait shit, I think optane requires Intel nvme driver to be recognized
or atleast the ones in the pci-e slot do, idk about the ones that can be connected via M.2
hm maybe not
Mad
In Ocado's grocery warehouses, thousands of mechanical boxes move on the Hive. Are they all individual robots? Or is this one giant hive mind? • Thanks to Ocado: https://www.ocadogroup.com/technology/technology-pioneers (this video is not sponsored, and they had no editorial control).
Reference: Strandwitz P. (2018). Neurotransmitter modulation...
You can also get x4 to u2 and even x16 to quad u2
Im getting my 12yo cousin to install Kubuntu on my old laptop
Its so painfully slow....
Took 3 minutes to boot up
but now we in the installer xd
The only question he asked so far is "Will I use entire disk?"
Whoo. Green screen of death achieved
All all it took was Windows 11 thinking it was a grand old idea to do an nVidia driver update while playing.
🥳 🤣
😬
Oh great. I think it might have gone into a delayed GSOD loop
Just died. Again.
Aaand sure enough.

Time to see if it will let me rollback to 10 
No one wanted to use Lumberyard, so now it's relicensed and fully open source or something https://o3de.org/ 
Watch out UE5!!! And Unity!!!

I hope it's actually good and gains traction
The world needs more prominent open-source software
I guess we'll see in 4-6 years
time to wait for someone with no hobby to spend a week to compare it vs cry engine and figure out if it's even worth touching it
guess the license might be more attractive in some cases?
Yes
I'm using 3 add in cards in raid on AMD, they must be using somewhat standard nvme
They are a 750gb p4800x, a 280gb 900p, and a 480gb 900p
Lmao nice
Screw china
Actually, everyone wins, defi gets more de and china isn't wasting as much power
I mean there is also a massive chip shortage
👏
lets bring that hash rate to reasonable levels, across all cryptos
lower power consumption, lower hardware prices for all
a PC gaming Renaissance
@astral haven ahh nice, so yours are not using the intel nvme driver? As in you didn't download it from the intel website during OS installation
I was looking at the P4800x and the 905p
I guess you're just using a U.2 to M.2 cable?
Back when I was just booting the 900p I didn't either
All pcie aics
I had to fukn install the AMD raid drivers which was 2 nightmares and a half
Ahh
😄
I guess Optane with directstorage should be ok, if you can just use the default windows driver
I'll try to snatch one.
Especially because I assumed I could just set the raid up, then clone my drive over and have it boot with my current windows install
Biiiiiig mistake
pci-e slot ssd's seem to have a lot of issues
I'm just planning one optane as a boot/game drive
raid seems too painful
Nah they are identical to the u.2 ones
Only difference is the physical interface
Electrical, protocol and driver side are the same
both my miners are making more bank 😄
I had a pcie ssd back in the day when SSD's first started appearing and it would BSOD my system often, OCZ something.
had to use their driver otherwise it wasnt recognized
@astral haven The new P1600X seems interesting. Wonder if it's gonna be cheaper than ebay 280gb 900p's
1600x?
Oh dang
That's neat
Replacements for the awful 800p
Still would want a few of them in raid :/
Most of those were sas ssds on a raid controller
7µs read latency vs 10µs on the 900p's
Meh
The 900p specs were very very heavy underestimates
And besides it's usually CPU limited without very well optimized systems
Yup
if only the senate would've been directly involved in the decision 
Hah
UwU
huh the 10B contract gone and microsoft stock isn't down even a bit? wonder if that comes tomorrow or why it's not affecting the price
weird situation for sure
I agree the contract was probably given to MS due to not kosher intervention
yet I still think it was the best outcome for anyone aside aws
https://twitter.com/Underfox3/status/1412502324644298756 damn thats a lot of AMD patents that seem interresting, gotta look at that thread tomorrow
no
keep up the pressure
force the competition to innovate and provide better products at better prices
Quantum computers, NVDIMMs, automatic core and memory overlocking, massive CPU improvements, NVDIMM improvements, and more.
Dude, seriously calm down for a sec there.
Intel needs some time to play catch up.
Yes, but that's a problem.
Aeon, I'm sure they're gonna trickle them out as intel catches up
If Intel can't catch up we'll just end in AMD stopping improvements sometime down the track until someone comes along and pulls a Ryzen.
doesn't make sense to do that again
people will just end up using 10 year old CPUs again
:P
better to constantly improve CPUs so that you can sell new ones every few years
They're also doing AI and multi-GPU things again.
Cxl will make all of this easier
I thought that was deader than disco
Apparently not.
now I'm interested...and clearly can't do SFF again
can't fit mulitple GPUs in a case under 20 litres
Nah
But then you have to fit the rads somewhere.
Gtx 690 was 3 chips technically and was standard length
For high tier cards that's a real hard ask.
most can fit a solid 240 or even 280mm rad
it's enough with an undervolt
I suspect AMD will probably do both since they're loving chiplets so much.
Folks, mcm is possible on one package
It's definitely not gonna be 2080ti sli style mgpu
Why not implement a better Crossfire and have chiplet designed GPUs?
Not like they can't with the available technology at no particular effort.
Because integrating those two premises is alot better
makes chips cheaper to make, and allows for way higher bins and core counts
what's not to love?
In theory crossfire and sli are terrible
Modern tech, tho.
In practice they work better than they should
I think there's something there with the new technologies of the past few years.
But multi die single gpu is waaaay better
driver level fuckery is a powerful thing
Yes, it is.
Thanks nvidia
isn't that a pain in the ass to do thermally?
Nah.
Gimmie 4 slot cards with like 6 GPUs in it.
agreed, way too hard to support, and pretty much dead in terms of support from even the manufacturers
I'd be happy with a single 3 slot card that had 2 or 3 GPU dies on them, but read as a single unified GPU
Chiplets
Water cooling. 👌
...
2 slots 3 chiplets
thermosiphon

I want a thermosiphon GPU card so bad
watercooled, no pump noise, no BS attached to that
Mmmm.
let gravity do the work
yeah, this is a 5-10 years down the line thing
I mean it will be better for software to support the equivalent of multiple numa nodes
but for gpus
i mean
i mean the software that dose support it dose it well
you just have to start out with it in mind
it isnt terible
Receive a $25 credit for Ting today when you sign up at https://linus.ting.com/
Use code LINUS and get 25% off GlassWire at https://lmg.gg/glasswire
A year ago the prototype of the IceGiant ProSiphon impressed us with it's cooling performance, but does the much smaller retail unit keep up?
Check out the IceGiant ProSiphon Elite: https://www.i...
here's the latest on the one commercially available one I know about
seriously impressive
they dont noise normalize though
i could strap some 40mm thick deltas on a hyper 212 and beat that
but it would be 100+db
then you could do the same with the thermosiphon and well...dominate
yeah what im saying is the temps at a fixed noise is more useful for the average user than max temps
at max fan
yeah
(whenever it's audible)
No the thermosiphon is limited in cooling capacity where normal jeatpipes can handle alot more
until the fluid boils in the wicks
yeah
but with a heat piple
they will scale as long as you keep the finstack cool]
you get a 1+inh2o fan pushing 100+cfm THOUGH the fin stack
and it will cool hundereds of watts
yes
so, makes sense for server farms, not really for future consumer products
which is why a max temp at max fan chart is moot, because who ever is loudest wins
having a fixed noise and thermal load
then comparing them
at multiple different loads and temps
you have to control something
2ghz is hard at stock volts for some
I'd try undervolting my 3070, but it sits at a comfortable 66 degrees at full load...somehow
no value in dropping that temp further, 'specially since I've never done anything of the sort before
it already boosts from 1.9 to 2.1 GHz in games
can I really squeeze much more than that from it?
not much
we will never have good gpu overclocks from nvidia again
they killed it
and amd has to bin so hard to compete with nvidia there isnt headroom
i rememeber my 290x's
pushing 400+w each
i went from a 1000w to 1200w because when gtav came out it would trip ocp
You can usually drop volts hard without losing any performance
Manual vf curve is pog
Nah there's headroom it's just factory locked 
Hmm.
That won't last long, not with the sheer amount of ASICs that run SHA256 out there.
I give it 2 weeks.
^ Alot of them are being rehomed rn
Aye, since all of the China people have to move and setup the infrastructure.
Probably a good gambit, buying some Bitcoin.
just wondering, what are the clicky switches called that come with the wooting one?
the wooting 1 clickys are blues
if by clicky we mean tactile bump
orange is the manufacturers name though
Wooting went with the red/blue convention people were used to
Same
What mouse
game mouse
My 3070ti will be here tomorrow
Nice!
M2K?
Ofc
300 euros holy shit
lol
I guess it serves a purpose for somebody out there but I can't immediately think of any

to have a mouse as light as possible with a wheel etc
although
same weight as the M1K without a wheel
so to exist is the purpose lol
I'd be interested to know how many it benefits with a change that is measureable
Wait, people still play quake?
you could look at acceleration etc
ye
even quake lives still gets to 400-500 players daily
dunno statistics of champions
I'm curious how it plays in osu
It's also on paper very close to the correct size for me despite the odd shape
Also driverless 8khz is pog
Can never go wrong with zaunkoening, but I realized I prefer the stability from palm support so I think I'll main Hydrogen

they're pretty close in features
which ones specifically
is it about EM shielding?
cause even european mouse companies don't give a fuck about that
WEEE compliance, I'm not sure I should talk about it :I
ah
He was interested how Wooting deals with the EU regulations, WEEE, he seemed quite pissed
Do most companies just cheat on those?
ik the chinese ones do
hehe
g-wolves put recycling and "don't throw away" logos on the same mouse
regulations



murica
