#🔌│tech
1 messages · Page 40 of 1
or I guess would neeed a lot more that just 128 cores for gpu, maybe like 64 chiplets of the tiny c cores lol
i will buy a 7990xt in a nanosecond if they do multiple compute dies
waiting for the 6090 to be longer than my ultrawide screen
dont think they will this gen
since that means next gen is almost guaranteed to be more than one GCD for amd, wonder if nvidia will have a jump to that aswell or have a single GCD design as well at first
tbh i doubt we can call amds design chiplet
the design inherently has no scalable compute
mcm ✅
chiplet ❌
go intel route and call it tile based approach since they cant call it chiplets after they trashed on the words some years back lol
they didnt
again they were talking about how zeppelin was a glue
which it was
each chip had all the functions
intels tile approach is not chiplet
considering it uses a substrate
chiplet > substrate imo
anyone know how I can make stable diffusion not make every image blue
why is that thing obsessed with blue
and by substrate i mean active substrate
I mean yeah the glue part wasnt false and I think the biggest joke was intel calling it "repurposed desktop product" when it was very clearly the opposite
nah
amd was bragging about that part...
they only needed to add 20% area to their desktop part to make it useable as part of their server stack
and amd was absolutely right about every part of that
the only problem is the part wasnt that fast
yeah it's been a while, was sure they said they made it scale down instead of up but dont remember any specific quotes either way
it specifically doesnt scale down tho
but while we are on the memes part
Keeping the dream alive
The cool thing is they could use that 1 base unit to make all the products
Msdt 1 chip, hedt 2 chip
Server 2 or 4 chip
well with scale down I mean they have mcm in servers and can scale that down to work with 1 chip on client
Considering they had more client volume than server that gen it is in fact scale up
And considering each chip always contained every part needed to function as the base unit
from a volume point absolutely, but was that the design point?
hmm, pop os does not use grub
or was it they made both and neither was a first focus and they just made both the focus from the start, which seems more likely to me
Well considering chiplet is the scale down design and glue is the scale up design it's also that way from the design standpoint
Obviously tho this is a purely academic or linguistic discussion tho
It is able to be small or big and do well either way
Chiplet is just better at that
oh yeah absolutely, that's why I also have the same opinion as dr cuttress on intels approach for meteor lake and co
they just have tiles that dont scale
well.. that'd be amazing if it was real
I mean if you're willing to put on VR goggles it isn't far off right now
quest pro video
hand and eye tracking would be so nice
if it was real youd have 500 more ads in your view
if only I didn't have a small ass apartment then I could actually use vr properly
I can barely get enough space to do room scale and that is by moving tons of stuff
just dont have a bed and use that space
I already don't have a bed
it does actually on laptops
On laptops, yes
I have one myself which is why I like that aspect ratio so much
It feels great
But on desktop it's mainly just older office displays
my desktop second screen is a 1920x1200 screen
Based on a company that do selling of stuff directly from major warehouses - suppliers
That's how many 16:10s there are on the market rn that are available in EU according to them
fewer
Mansen
I will strange you with a ring made out of bread crust
The soft part I will eat in front of you
Buy EIZO EV2730Q 26.5" 1:1 IPS Monitor featuring In-Plane Switching (IPS) Technology, DVI and DisplayPort Inputs, 1920 x 1920 Resolution, 1,000:1 Contrast Ratio, 300 cd/m² Brightness, 178°/178° Viewing Angles, 5 ms Response Time, 2-Port USB 2.0 Hub, Auto EcoView/BrightRegulator, EcoView Sense. Review EIZO EV2730Q
1:1 would be great for CAD work
Ah that apparently falls under 16:18
Why is this a thing, though 
Just combined two 16:9 displays into a single continuous panel
Why not just get 16:9s on a arm mount?
I already said that
Sign up for Private Internet Access VPN at https://geni.us/inEpD
Companies have tried to make monitor bezels as small as possible, but it still leaves a gap. ASUS might have a solution to make them completely gone, but does it work?
Buy ASUS ROG Bezel-Free Kit
On Amazon (Paid Link): https://geni.us/aBoYh6
On NewEgg (Paid Link): https://geni.us...
remember these?
But it's just uncomfortable
how
What if you want to do side by side instead
Or higher refresh rate
Adaptive sync
Those "office" panels don't have those features
They sure are colour accurate, but not much else
Nah that one is for Bitwit
at least Jay is capable of building a PC successfully on the first go
I take breaking something that can happen to anyone over insisting being smarter than you are and spreading false info
Fair...
I would take the same. Though who's insisting on being smarter than they are spreding false info?
Intel drivers are still shit for the Arc GPUs ;-;
Sponsor: HYTE Y60 computer case on Amazon - https://geni.us/3ngWP
We’re reviewing Intel’s Arc A770 Limited Edition GPU and benchmarking it vs. AMD and NVIDIA GPUs, including the RTX 3060, RX 6600, RX 6600 XT, 6700 XT, and more. Benchmarks include 1080p, 1440p, and 4K gaming tests for FPS and frametime consistency, plus tests of cooler flatness, ...
jay
and yes we already talked about intel drivers twice now
Shame, cause id be tempted to get one otherwise
although maybe not, id need a new CPU also
Will AMD also be madly raising GPU prices like Nvidia did for 40 series?
We don't know.
makes sense
if i had to choose between them i think id go intel, but im here with a 3080
so am good
If Intel figures out the massive issues with the drivers, then maybe id get one
idk how they are like on Linux yet
Useless pre Kernal 6
Makes sense
wait, is it like their iGPUs where the drivers are built into the kernel like AMD?
Or do you need to install them?
Makes sense tbf
Would be sad if Intel killed of their GPUs
More competition is better
Also i bought a Steam Deck which also started off poo 
Still has many issues
but a lot better, its just how it is with gen 1 tec h
I love my Steam Deck <3 <3
i love mine too
I do have a teeny tiny scratch on the screen ;-;
I love my AirPods Pro gen 2 
Should have got a protector, i didn't use mine much till screen protector.
Was to scared
That feeling when you buy earbuds the same price as a steam deck.
Steam deck for 250£?
I know that if I messed up installing it, misaligned slightly or trapped air bubbles id just take it off and not use it ever again
same happened with my phones
I somehow did mine perfect first try
Anyone wanna buy my Steam Deck? 
Has no scratches.
And surprisingly AirPods don’t have much less transistors either
Why?
LOL
Yes whatever you do don't get AMD GPU amirite
even others would say it just in a different way since a product less is better for them than a product that has massive issues for many cases
AirPods Pro actually may have similar amount of transistors as steamdeck 
no?
Steam deck has 2.4 bilion
H2 has 1
Question is if there isn’t one H2 in each airpod
Well there is
So if apple meant 1 for each
That’s 2 bilion
- the transistors in case
With the U1 chip
And also 1 bilion is probably not the exact number
ok and if I take an entire wafer I have many multiples of some billion
wtf is that comparison
What?
comparing one chip vs 3 doesnt make much sense
you dont pay for transistors you pay for die size
if you put 5 or 5 billion on them doesnt really matter
Well the node will also be more expensive
Anyways it’s just weird that AirPods combined have that many transistors to me
Especially putting it into comparison next to steamdeck
Doesn’t change the fact I find it surprising
if you dont have high performance libraries you can cram a lot more into the same space
I mean I'm just saying why it's nothing special
I am not saying it’s special
Just didn’t expect it
Or well I expected steam deck has more like 10 bilion so the 2 bilion AirPods have would seem as less in comparison
I mean sure, it's just a thing that doesnt make sense to be compared as just an amount so it might be surprising because that leaves out a ton of factors
True I suppose
But it’s still weird M2 has 20 bilion transistors while steam deck 2.4 to me. But I suppose you can’t directly compare them at all
why not?
m2 has a ton of accelerators built in that are actually pretty solid
thats why it can do so well in so much stuff
it also has a huuuuuuuuuge gpu in comparison to the deck
thats also leaving off the wireless controller (s) and display control ics the deck has
rf uses alot of transistors
also the node helps a lot, after all amd managed to put like 50% more into zen4 than into one zen3 chip
also of note, the h1 has integrated memory
presumeably h2 too
8gb of dram is 8 billion transistors too
and thats just for the storage itself not including the addressing
yea the h2 has all of its storage and memory onboard
Which is all because of Apple/M2 kol
what
no, the only reason 5nm exists is because of asml
uhh sure I guess but if not apple someone else would've funded it, they say like 50% of a chip is DTCO at 5nm and that part is going up so basically a custom node variant already
Who funds that
german government
they are dutch
we have zeiss, so part of it but not all of it
Lol
USA also pumps money in it, so does intel
TSMC is 34% if ASML's revenue
ibm is shockingly large for a company who has fallen so far from their peak
Grab a GN Tear-Down Toolkit! https://store.gamersnexus.net/products/gamersnexus-tear-down-toolkit or a GN Large Modmat! https://store.gamersnexus.net/products/modmat-volt-large (both in stock & shipping at time of posting)
In this video, we're tearing apart the brand new Intel Arc A770 16GB Limited Edition video card (it's not actually limited)....
revenue isnt investment
It's a public for-profit company.
also only because intel ordered after tsmc and asml cant keep up lel
yes and they get money from investors to spend on developing the tech
well, the big purchases came later, obv they already have some euv machines for intel too
the revenue goes into paying back the investors or lenders for purchasing the raw materials
ahaha I love how GN just takes the 4090 unbox embargo and just shows it in a 770 video and no further comment because unbox embargo are dumb
TSMC spent 5.7 billion at ASML in 2021. Outside investments and stock selloffs accounted for 1.1 billion.
how much did intel spend and how many orders of intel are still backed up at asml
lol yeah, as soon as I saw that the videos were unboxing only, I just X out
revenue is after the thing exists
investment is before the thing exists
Find me the earnings report that shows it theb
how much did tsmc give asml that wasnt in return for hardware
If you know it to be true
investments arent earnings 
You do know that a publicly traded company has to show outside investments on their earnings reports
sort of
its complicated
the important part is that tsmc gives asml money after the tech exists
It's only complicated because you are talking out of your ass with nothing to back it up
speak for yourself
Intel is the largest investor in ASML and did it solely for EUV
And only invested for 4 billion.
https://patents.google.com/patent/US20080107994A1/en
Intel also was actually developing their tech on their own before they started working with asml
Embodiments of the present invention provide EUV (extreme ultraviolet) photoresists comprising photonic crystals, as well as other components. Photonic crystals in general provide the ability not only to block light transmission, but also to create resonant pockets in which light can propagate. The photonic crystals are based on bio-relate...
TSMC is only in for 1 billion
omg the arc backplate doesnt even have a thermal pad as connection, so it makes the heat worse to look fancy lmao
And again, no one is in as deep as TSMC lol
Also be aware that invested in asml is not the same as "is responsible for a node made on asml machines"
And no one spends more on bleeding edge silicon than apple
They literally have the best production silicon available, every year
Year after year
Except ofc for ibm
they have the best high volume silicon, aka leading edge, not bleeding edge for some science wafers at ibm or so

Apple is insanely safe
As safe as they can possibly be
They do not do anything that hasn't been done before
They just do it well
And give up if it starts going poorly
No sunk cost bs for apple
And what dose IBM have production at even 5nm?
Quantum computers are so far beyond 5nm that it's not funny
And yes they are in mass production of quantum computers today
Lol
oof, february 22 is when the baseplate for the card was made that GN tore down, that seems a bit long ago for a launch now
first time seeing thermal pads that looks fabric
You know that Eagle is built in 7nm fab right?
Sure
Osprey and Condor are set to be 2nm
But no announcements yet
So IBM dosent have the best production silicon
Because all the fancy quantum happens with superconductors on 7nm silicon
There are no chips that have even remotely any match to Condor or z16
Like it's not even funny how far they are ahead of the competition
That has nothing to do with silicon chips
They only use the fab as a base
To built the multiplayer chips
Z16 is purely normal compute silicon
Yet it has a worldwide monopoly on financial transactions
the insanely over engineered LED strips

custom pcb strip for the glowy rings lmfao
guess we found why the limited editions costs 20$ more
And it manages to do that on a very stable node
Z16 is still 7nm
Like there's not a single market where 1 silicon vendor has over 90% share
Except for ibm mainframes
Technically by installed mainframes they are under 90%
But by transactions performed they are way over
Intel loves adding needlessly wastefull pcb's for their rgb strips
I mean who doesn't.
circular ones around the fans are especially bad in that regard.
Most of the time the rgb strips are made quite efficiently, not just a solid pcb in a cricle or u shape
then again if they design it in such a way that all the parts come from the same board it may not be that bad.
Samsung already has "4" NM in production
Though it's not really much denser than 5nm
modularising manufacturing is good for servicing and repairability
just a shame you'll never find parts for these types of products 
Pretty sure Samsung's 4nm is less dense than TSMC 5nm
Sapphire has the best fan replacement mechanism
In wish we could by GPUs with just 120/98mm fans
or with just a heatsink that can mount 120mm fans
we can, if its an AIO card with a 120 rad 
Yet it's actually running circles around it's competition
What competition
Z15 lmao
I want a gpu with this installed stock
MORPHEUS II CORE EDITION, whole heatsink black coated, a superior high-end VGA cooler, designed to meet cooling capacity up to 360 Watts of TDP. Morpheus II core edition comes with 12 copper heat-pipes, 129 fins, mirror copper base, and the option to install 2×120mm fans; those outstanding combin...
add your own fans
That's what I have
And it's only actual competition thanks to how dominant ibm is in that market
Fanless in my torrent
oh nice
torrent have great fan support
lol interesting approach, passive on core component, but case full of fans
Big fan with no restrictions move air quietly
It's honestly silent
Even with my HVAC off I can't hear it
does the torrent have a solid panel option?
Yeah that's what I have
hmmm tempted to migrate
Black on black
yeah I'm not a fan of tempered glass
I have a Silverstone 1000w platinum
Which is fanless most of the time
And no mech drives
I'm super tempted to go 2x180+2x140 rad, but air is working great
what are your temps with the X3D
like 80s with ciniebench
does it throttle?
ahhh
155-160 watts
i mean more for zen 4
oh
hmmm, I'm wondering if the IHS will be thinner because of the v-cache
at least thinner in a pocket
same heigt
zen3 it was same height they just made the parts thinner where the cache went so it stayed same height
likely going to be the same here
you dont want an air gap over your IO die lol
the ccx are already thicker than they need to be to match up with the iodie
so the 3d stacked cache isnt a issue
wonder how temps will be with liquid metal and if that is worth it with the ihs still
we know it did a bit with direct die but I dont think we had delid tests with just LM yet
with ryzen 1/2 io die was much taller
no? even intel puts it against a 3060
yeah its horrible vs amd
Ryzen 1/2 didn't have an iodie
then was it 3000 that had the difference hieghts?
one did
and it was a pain for direct die
3000 and 5000 have the same iodie
Dunno height differences
But the silicon for the iodie is the same between the two
hmm wonder if derbauer is going to make a custom frame like for ADL but not for pressure but to use a smaller IHS so the cooling can be better lol, could be usefull when x3d comes
since I doubt they already improved the cooling of 3d stacking for this generation by anything significant
finally they have fixed the fucking charger bugs!
if there wont be something better for cooling let me know when you will direct die your v cache so I'll know if its worth it 
lol
i thought it was a power limit not thermal
well it's speculation on my part I just assume the vcache would run into thermal limit, assuming they improved it so the clocks aren't going down this generation on it
it only like 4320 vs 4600mhz stock.
5800x3d vs 5800x
but 110 vs 130 w stock
yeah so my assumption is they might have improved it so the clocks are closer/the same at the cost of higher temps
at least it appears likely to me
well they said that it was limitations due to it being the first implementation, we just dont know how much they improved in the relatively short time till zen4v
yeah
it could be an internal thermal issue
or just plain voltage limit
- can get a bit more out of it
99.999% of the time those two are the same
my mobo will do 103 bclk
at least based on these results I assume x3d will do well in my main game so I'm planning to potentially get it even though my 5600x isn't having any big issues
7950x3d hype
now that I'll likely have money to spare because I managed to avoid having to pay VAT on our solar installation I'm getting very tempted at least 
i really wish someone would benchmark vr performance in games
oh man, looks good. Shame I prefer a lower profile back hump. More of a mid hump person
thats literally a gpw with a further back hump
Yea people enjoy that shape
its as much lighter than my main mouse as the entirety of the weight of my secondary mouse lel
no im ragging on the cloning
DAV3 is probably my upgrade
and yes i palm a m2k
Everything is a Zowie clone so who cares
eh
g305/pro wired
and id say everything is more an intellimouse clone than a zowie clone lel
I just can't get over you knocking the shape but you use a m2k
thats not what i said lel
I just want the G Pro shape with good side buttons why I am interested in the Lamzu
just because im making snide comments about how its a clone of the best shape doesnt make the shape pad
in fact quite the opposite
@gusty kiln mmmmm just waiting for him to talk about the side buttons
I need G pro shape with god tier side buttons and that will be my endgame
whats wrong with the superlight side buttons?
They are trash
Could be just my copy that broke, but seems like a common complaint by others also
i looked in my garbage bin for the superlight side buttons, and they arent there
they are on my mouse
is logitech not replacing them?
Yea they are replacing mine, just waiting for White to come in stock
They said 2-3 weeks
could just be hitting the .1% afr
They offered me the other colors but I just want my white one back
if only they can jam the side buttons from the G603/G703/G403 into the super light
they sell enough of those that there should be thousands failing for random failure reasons
a lot of people just don't like how they feel
🤷 doesnt make the buttons trash
it works fine, but just not subjectively to their liking
@gusty kiln THISSS!!! The 703 side buttons are god tier, but I hate ergo mice
i still think the next step is passive mice with active sensor mousepads
is Pulsar still using dot switches for the side buttons on the X2/X2 mini?
If they add some type of texture or grip to that LA-1 may have to check it out
That or the Lamzu
gpw side buttons aren't the nicest feeling, kinda mushy and undefined feedback but haven't had problems with their actual function
and i find most mice side buttons to be over expressive
my favorite side buttons were on a microsoft mouse, but i forget what model
i liked the gpw side button swapping situation but the gpws makes more sense
time to glue a m2k on the side of an m2k for sidebuttions 
side buttons is tbh the reason I dont daily the m2k
i need to be able to alt tab and ctrl tab on my mouse
we need more gaming foot pedal that's not for racing sims
hmmm
thats a price wow
isn't that just on/off tho
Little pricey tho and its not really for gaming, more just a 3 button stream deck
i would be shocked if there werent accessibility tools that would work for this too
they should just add midi cc support to games and render a ton of overpriced gaming accessories obsolete 
@astral haven anything in particular? Somebody might have already made a plugin
voicemeter apparently includes midi to macro adapters
no like i mean hardware
xbox adaptive controller + trigger footpedals would definetly work ofc
Have a feeling those pedals feel like shit tho
probably terrible pre and post travel
Watch a couple streamers who use it, but they only have Discord mute set to it
shame you can't just slap usb on a $25 expression pedal and call it a day
xbox adaptive controller
yeah that thing is really neat, still a $100 though before you've even gotten to what pedal you want to connect
i mean you can use ardiuno and xinput
yup but its not pnp
I mean you could, but then you have monkes like me who can't write a single line of code to save their life
and don't give me that "just copy someone elses and put on there" I can't even figure that part out 
you can buy premade and programed ones
with screw terminals
linkeroo
I suppose, still a bit involved for the plebian that just wants the most braindead consumer friendly way of hooking up a pedal
if its premade and programmed it is pnp
there are a ton of them
thats just first result
That's not 3.5mm jacks tho
i said screw terminals
🤷 still wont work with a pedal without doing a bit of snipping
what foot switch is 3.5mm?
you can also get $10 hid usb foot switches
or
no snip
yeah
heck you can use the two buttons on the adaptive as is
but cheaper
barely
you can get usb hid foot switches for 10bucks from china though
yes
i just use my flight pedals
got usaed to that in arma for analog lean
and i can set them to press keys at points in the axis
😭
thats still p good for the price tho
useless for me tho
i paid like sub 100 used
that led me to the "riser card matrix"
and they have 1 active gen 4 riser
and its rather strange looking
RSC-S-68G5
I think it's using oculink for the connection to the motherboard
yea ikr
gives you 8 total x16 slots
also for 8 grand i want gen 5
dont game on your workmachine
i dont
you dont need 100g to game
i want to have my games stored on my nas
Do it - Cause then you can save VAT on business expenses that happen to be gaming related 😄
wont comprmise gaming
also my workstation is fully provided by work and is basically a super expensive thin client since it remotes into my main machine at work
wont store games locally
in iops over networking with windows doubt it
guess
nothing over the network is faster than the same device local
actually thats not true
these ssds specifically require a core and take 2gb of ram per tb of storage
it really bogs down my main system to use them
WTF are you using sheshhh
i let my server do all the raid nonsense with them, and roce takes care of the rest for minimum overhead
oh and these ssds also have onboard dumb plx switches
its really fkn annoying bc they are gen 3 switches, but the ssds themselves run at gen 2 speeds, and it refuses to boot on an x4 slot
its only 2.0x8 of bw
but i need an x8 slot to use it
id love to see you actual prove its faster
get anywhere near a million iops over network and id be impressed
and actual proof
cache disabled
not saying it works
im not at a million on x8 3.0
in fact thats not physically possible
wait no it just barely is
you can do a million iops on 10g
not at 4k
1m 4k iops = 4GB/s of purely data
40 gig can do a million iops
but I cant with just 1 server
I actually dont have results at 100g oh noes
heres what I was doing on 40g tho
both of those are uncached numbers
to get anywhere near that you need either cache or optane
also yes that 3 way optane raid is slower than single drive on everything but seq read and write
i mean thatrs slower than a single p41
its all non-volitile
its using rdma fam
lol
its fully atomic
what dose the m in rdma stand for
memory
and what memory dose it refer too?
it using ram for cahce one way or another
nope
this is in fact my field of expertise
its fully atomic
rdma allows you to send data directly to devices on a remote system with the cpu just acting as a controller
if you disable the caching on the drive itself and on the client side, it doesnt cache at all
its literally memory to memory
memory in the data science world doesnt mean what you think it means
the m in nvme stands for memory too
yes as in flash memory6
rdma isnt direct to nvme yet
it still though host memory
try it with a single stick of ram in the server
they literally sell ethernet nvme ssds now that do that
and check the speed again.
signle channel?
ddr4?
ya
thats rather fast for server ddr3 ram
here it is
roce nvmeof ssd
yeah and the ethnernet adapter for the ssd totally wouldnt cache at all
caching slows things down at the high end...
so they just slap ram on there for looks?
?
you mean the dram that this dual port controller uses to convert a single port drive to a dual port drive?
again
roce is quite capable of being fully atomic
ive provided plenty of reading
what part of "no data copying into system buffers" is confusing you?
Here's while I spent days banging my head against a wall figuring out how to get infiniband working
this is with caching on a 2.0x4 drive
so probably pretty easy to beat
Aw man digging up old pics is fun
This was 40g
gig thunderbolt card?
Ya
i had the x399 designare
Yea thunderbolt works fine on amd
As long as you don't plug in anything while the system is powered on 🤣
Hype for USB 4 boards
nah that wasnt even the issues
Wha?
yeah
hot plug is fine if the board supports thunderbolt
the asrock and gig cards are cross compatible too
no idea about am4 stuff
the boards just mostly dont have the headers
nope that still works
im gonna need a citation
cause pcie hotplug is what those are for
yea on old fw the giga card will unsafely hotplug the pcie interface
some boards work with that
like
i had the header and used it
hot plug was never the issue
it just was flakey
speed wise
in what situations?
random
lol
plug in nvme device
transfers at 10mb/s
stop transfer
restart
transfers at 2GB/s
i used it for hotswap storage
then i got a usb 3.2 nvme dock
and didnt use tb anymore lol
lel ikr
dose 2GB fine
neato
i have one port with a m.2 to u.2 adapter
and it works without power for most drives
the big boi intels need more power though
shocker
just hotplug the raw pcie adapter wcgw
😭 I really was wishing for that 4.0 switching riser
Yea but then my plx card doesn't get x16 still
And I suspect for VR every bit of bandwidth is nescessary
Also the upgrade would cost too much
It's GPU or CPU for this year and I think I'm going to pick getting a new bike unless again multiple gcd 7990xt
bruh the intel gpus look so clean but made with tape
like wtf
just use screws
why
u dont stick ur m.2 into the pcie slot directly?
weird
@astral haven recommendation for good affordable ups?
id go either apc or eaton
good joke for my 27m^2 apartment in the middle of berlin
who knows we might get power outages in the winter
surely that will keep my pc powered for long enough to be shut down
yeah just run it in 10W mode 
powering your hopes and dreams with that one
this one good?
eaton and apc in general are good
the pricier products just have some more features usually
like network management and usb power off stuff
if you plan to not have automatic shutdowns the cheaper units are solid
Is this the lowest I should go or is there a slightly cheaper option?
ngl for the cheaper ones id say browse ther catalog for your VA needs
i usually sell bigger rackmount ones
just make sure backup time is enough
I just want something to be able to safely shut down my pc in case of power outages
the one u linked has 9mins on half load
ups
are they good tho
I'd say so?
cheap easpecially in ups world isnt always good or long lasting
but ya in many small to medium sized companies ud see APC or eaton
be it rackmount or under desk
7 votes and 27 comments so far on Reddit
How does XMP remove the ability to put the system to sleep
Apparently it might be limited to only a few vendors or ASRock only
Other X670s work fine
https://noctua.at/en/na-stpg1 man people really are too lazy to use tape or smth I guess
The NA-TPG1 is a simple yet highly effective means to preclude thermal paste build-up in the cut-outs of the heatspreader of AMD AM5 CPUs. When the mounting pressure of the cooling solution is applied, excess thermal paste will be squeezed outwards. With AM5 CPUs, this excess paste tends to accumulate in the cut-outs at the sides of the heatspre...
This one looks good right?
If Apple swap to only wireless charging instead of using USB C in 2024, then will they try to make it just as fast?
👀
honestly think you should use an adblocker setting that also disables youtube comments
it'd benefit you
yes the stickers are nothing new, it's been a thing for ages on various boards, nobody else besides that person had a sticker peel that way, for many others it was a piece of plastic
and for the most part it was showing installation and that it would take a long time on first boot so people wont think something was broken
ASRock (and others) have had stickers on DRAM slots and PCIe slots for years. Why is this now only an issue for some people? Why this generation, rather than ANY OTHER? Why are you even ripping it off rather than doing the slowly slowly approach
147
I have literally never seen an internal sticker like that
And it isn't an excuse not to use super light glue 
anyone else who got a board with a sticker had either no issue removing it, had a plastic piece there or got a newer revision that doesnt have any at all any more
so its literally an article about an issue one (1) person out of many had
either way my main point was don't put any weight into youtube comments because 95% are talking out of their ass
That's definitely a different matter then. But I don't see how a faulty sticker can ever be anyone but a seller/manufacturers issue 😏
a one off wrongly applied sticker is hardly the same as "asrock sabotaging its own production"
💯
4* also it's mostly been lowered with bios updates so it's not taking as long anymore even for first boot
thats why newer boards with newer bios revisions also dont have the stickers anymore
I dont even remember if my asrock board had a sticker but it was ddr4 so prolly not
Interesting
im upset
my deck screen already has a tiny scratch
and my switch doesnt yet I never use the switch in a case
;-;
No protector?
I dont use a protector on any device
Because I know if I mess up applying it, I will not use it
Probably something for OE or SI
You just peel it off and try again lol
doubt it, they wont use LM as OEM and then excess paste doesnt matter
It's just for paste. And OEMs and SIs already use paste templates.
ik but people dont want it there so I think its for that market of people who dont want it there even if it doesnt matter
idk if oems want to ad cost for no benefit really
Have you never seen the tools used on an assembly line?
I saw some brand made steam deck glass protectors and they came with a template you placed on the Deck and it perfectly lined the glass protector onto the Deck
I forgot what brand it was :/
They spend tons of money to just make it less likely to mess up
I know sapphire screen protectors come with them, but don't think they do the deck
unless its LM I dont see this mess something up though, and idk if for those places noctua would make it a public product and not just advertise it on the B2B channels
This is an tool that has existed for lots of processors. Noctua just decided to make a end user model for this gen.
Also someone sold a oem style hex template a while back.

Thicc
Sponsor: Get 10% off Squarespace purchases (https://geni.us/BqEpf)
We met with NVIDIA's thermal engineering team to cut an RTX 4090 Founders Edition cooler (not GPU or PCB) in half with a water jet. This was for science, as it allows us to see inside the vapor chamber, the heatpipes, and the cooling structure to understand how coolers are intern...
Wasn’t it TT?
Yep
Is it safe to use a computer if I lay it sideways?
Like the not glass side being down
Yeah it'll disturb the flow of the electrons
LOL
hard drives dont like moving while on
my PC is fully ssd
but modern ones dont care about actual orentation
the only thing that hates movement is the DVD drive
When you start plugging the motherboard into the GPU instead.
@wheat plume replying here cuz im hit with slowmode in #archived_wootstations
| Why is your About Me section a spec list?
HAHA it was something that people from the FinalMouse server did and I picked it up from there
BOY THAT AINT THICC THATS A BOWL OF OATMEAL
He had a bit of a mild meltdown two days ago? and plainly said he needed a break and left. And then came back a couple of hours later and posted more and... not sure since.
ah :P

See even a newcomer agrees that Fleimi is a good lad 
i was yessing to this actually. never interacted with fleimi
What 
maybe he left before i came
Are you insinuating that fleimi is not a good lad?
lmao
although, i was greeted with an arse when i first came here
literally my first message which was a question stirred someone up
won't say who
I mean not everyone is a fan of gourds
those who know, know

ah yes you were there
he did not return
I just happen to post this frequently
Could have sworn I saw a message later.
Ah well. He's entitled to a break. 🙂
slanted alice southpaw when?
inb4 "we put an xbox in a 4090" video
power consumption so high it'll shut down the entire german grid
pretty sure thats a 4090
isnt the connector on the 30 series rotated by 90 degrees?
yea the font change was unessesary, I guess it was done to make the 40 series more distinct from the 30 series, other than size difference
I thought the cards were almost the same size
Just bigger fans
Think 4090 is a almost 1 slot thicker
That's why on the 4090 the fan is into the trim area
Same 2.5
Like almost 3 but not quiet
and here I thought my 2080 was a large card when I first got it
Chonk, what happened to high end 2 slot sized cards feels bad man
3090 ti*
thought the same about my 1070ti
The ti is thicker than the non-ti
2xxx and 1xxx cards were the same size. Lots use the same coolers
then i got a 3080
Well non ti to non ti, it's thicker
And you know they still have a 4090ti in the back
Yea all I can say is Eaton
I haven't found anyone actually reviewing ups'
Lol that's where i work at
Wouldn't call Eaton affordable
I should get one at some point
Yes but I also can't afford to replace my PC if it gets zapped to death
Apc and cyberpowet have warrenties too.
wonder if the 4090 coolers all come form nvidia having planned for higher tdps earlier on and now the designs are already made and done
funny to think 95% of that thiccness is still just the cooling
yeah I know I saw that but that makes no sense
They aren't that much bigger than 3090ti coolers
there is no way nvidia went from samsung to tsmc as late as february
Cyber power is known to not honor them
And apc is stepped sine up to the price of the eatons
Well since aib don't even know the price of cards before Jensen announce it, they may not know the final tdp as well and they just design for the worst case scenario?
yes that's what I speculated, just the timeline igor showed doesn't make sense
a 600W 8nm samsung chip would've been above the reticle limit
Nah
with the same performance yea
Stepped sine is fine for modern psus
But any ups is better than no ups
I don't think AD102 would've physically been possible on SS8
It isnt the terrible square wave that dwells at 0v, it will be fine on a modern switch mode PC psu
It still absolutely hammers them and is known to significantly affect output stability
Easily
Remember it's solidly possible to make the chip, it just may have required an uncoolable amount of power
yeah but that ignores one of the variables, 600W AD102 with same performance as now, that just doesn't fit
apparently you can't access two wooting boards plugged in at once via wootility yet
Why not?
Run the second one on a vm?
Samsung 8 isn't that far behind?
I suppose you could just program one of them first, save to board then unplug it
then plug it in again once both are programmed
can access a flaretech and a Lekker board tho
yeah fully separate softwares for those
The cyberpower one claims to be "true sinewave" which is why I linked it
urgh forgot wikichip put articles behind a timed paywall now, can't find the numbers to compare the nodes
You can if one is Lekker and on isn't
But you have to use pre Lekker wootility and the Lekker beta.
But it works
ok here we are, AD102 is at 125 Mtr/mm² so almost twice of A100, while GA102 only managed 45, I dont see how that is something they could've gotten on the same node within limits
with N4 also clocking higher as well
yeah, am not too keen on grabbing an old optical one though
and since settings are stored on-board you can still workaround it
the reticle limit is something around 850mm² so with 628 the GA102 die is only 30% or so away from that
so yes, samsung isnt that far behind, but thats compared to N7 not N5, nor 4N which is their custom variant for it
N5 is only 80% as dense as TSMC 5
yeah but in the alleged comparison they would've stayed on 8LPU
ADA LL is N4 and 7xxx is N5 rite?
it's 4N because the N there is for nVidia, both are just using variations of tsmc5 high power nodes
so as much of the same node as it gets nowadays with DTCO
its not the first time they did that
Good that I asked, I nearly went for NV this go since I read on reddit nvidia using denser node
intel arc has denser node than ampere, still doesnt beat it lmao
thats not a relevant metric for the last decade or so
yeah I just care about efficiency atm
see also zen2->zen3 was all done on the same node
Density of the node is barely relevant to efficiency now
literally in some points its too dense so they need to relax it or it would cause thermal issues lol
Still reading online that ADA is using tsmc N4 and RDNA 3 will use N5
Vs stock settings, most GPUs can halve power usage and get 90% of the performance
30 is a bit odd where you could cut 30% and lose actually zero perf at all
dno what to beleive
Both are right
its accurate
but its just node names for variants of the same base
Modified how?
Apparently Nvidias version is particularly dense
Well each company chose their own modifications
about 50% on 7nm or lower is co-desigend from fab and customer to get the product right
so take that for big volume customers like amd/nvidia 50% is the same and 50% (or more) is made together and likely unique
at leat now they keep the names similar, unlike 12n that was just a variant of 16 lmao
or was it 22 to 16N, one of those
Tldr ignore node just look at the reviews
yup
Fkin HDR lags on OW2
damn thats one nice collection https://twitter.com/bl4ckdot/status/1578015645882789888/photo/1
too bad galax is unlikely to fill the KP card hole
How so
They have always been #1 or 2 for XOC cards
Just much much more limited
yeah from what GN and co mentioned it was mostly evga/KP who pushed for more stuff they would be allowed to do on the high end cards so that is lacking now
Limited in terms of availability not in terms of capability
Yeah
also that, much lower presence in western market
As far as the us folks know
HOF cards have always pushed the highest limits
Hof ln2
And generally leak the XOC bios
The retail versions are often quite different
also a good chance at least parts of what HOF cards could do would come after evga pushed for it, not entirely obviously, but as war as we can tell it certainly seems likely
Luckily my 1080tis have the XOC bios
EVGA pushed it to be public. Galaxy was always fine with it being behind a wall
I mean doesnt really do much if it exists and nobody can touch it lol

