#off-topic-tech
1 messages · Page 137 of 1
It's almost a non problem really
It doesn't matter what kind of cable it is somebody gigging wiill have at least one spare anyhow.
01:30, and I should have headed bed around 21:00...
plus to pull things back into perspective a quality guitar cable is likely going to just work anyhow
Was about that interference part at this point.
it's the cheaper ones and unless i hear otherwise any of the big guitar company branded ones because they probably outsource to who knows where or who
Even the cheaper one't don't just constantly fail they're just way more prone to it
To be able to better shield that cable to connector part.
Metal outer body plugs do help a lot on that.
oh yea those are uterly useless msot the time
they just break after the "kink protection"
Even with 6.3mm plugs, just that cannot be fully done with them.
The failures i see are always busted solder joints and those bendy things that are supposed to protect things... ehhh i don't see how they're doing much if anything
the cable still bends, its still got the same ability to stress what's inside
I did say that most of that kind of things are just fake.
On consumer/audio cables.
and again a problem plug it should be locked down tight anyhow
i have cables with nothing past the metal and they are not an issue
The real deal kink protection springs do work, but need to be tuned to the specific cable, and be different materials than the fake stuff use.
but i've had them open to make sure they were locked in, if i put that much stress on it i deserve iron time lol
im soeaking stictly guitar cable
i'm sure they sometimes do that but... they're a big too tough for that usually
big enough that the stress ends up inside the connector usually
The point between the RF connectors, and metal outer shell 6.3mm plugs is that the insulation portion will still be a hole in the shielding.
And at that I was talking about the shielding quality by changing to new connector type.
Not about the rest.
But that metal shell plug already provides MOST of the interference protection.
Well the cable can pick up noise but that's if it just doesn't have any shild or one that's just bad
the pickups are the real noise beast
and some amps 😕
could be
but mostly, like you said, pickups.
90% of the time I'm using a bridge humbucker so not too bad, i feel bad for peop[le who only play singles
i actually need to look up the specs on the hot rail i have, not sure if it's just hot or a double in a single package
I'll have to ask my neighbor exactly what it is... it can't be a hot rail because that's seymour
def not a double, but also no labeling anywhere even under the cover. I could see through some holes with the cover off that it is def a single coild and not a double
I think I remember now though, he replaced the neck pickup in one of his strats with a hot rail is, that's what it was. This is just a stock fender american standard neck pickup
Quality apple product
new meme
Just remember. Sega does what Nintendon't
i present to you
NVIDIA Flame Generation
https://www.youtube.com/watch?v=1z7aco-y4RE
Introducing the All New Nvidia Flame Generation Technology, available on all of the 7 50 series RTX cards we have sold so far.
The more you buy, the more you save
- Daddy Jensen
3090 burning render from:
https://www.youtube.com/watch?v=VbV7xWplOSI
honestly we need to switch to 48V rather than 12V if we're gonna be shoving so many watts down wires
would probably need a new connector design to prevent arcing but you could get away with thin gauge wire that wouldn't immediately melt
power drills started moving to higher voltage years ago for similar reasons, makita went from 12V to 40V
the cables just need to be regulated
like they were on the 3090 TIE
it's also more efficient to send higher voltages
Molex Mini-fit Jr. series is specced up to 600V.
That is what ATX, EPS and PCIe power connectors use.
yeah well none of these shitheads will self regulate
Der8auer's video: https://www.youtube.com/watch?v=Ndmoi1s0ZaY
Patreon: https://www.patreon.com/buildzoid
Twitch(mainly gaming streams): https://www.twitch.tv/buildzoid
AHOC Shirts: https://actually-hardcore-overclocking.creator-spring.com/
Bandcamp: https://machineforscreams.bandcamp.com/music
My FPS game: https://buildzoid.itch.io/hv-shooter
B...
fascinating
so 12V really is just a compatibility thing then
nop
its planned obscolence
and another feature NVIDIA can advertise
by this I mean, a 48V standard would not be compatible with 12V, so you have the chicken and egg problem
Introducing the All New Nvidia Flame Generation Technology, available on all of the 7 50 series RTX cards we have sold so far.
The more you buy, the more you save
- Daddy Jensen
3090 burning render from:
https://www.youtube.com/watch?v=VbV7xWplOSI

Yeah, everyone would need new PSUs to go with the new GPUs at minimum.
And would be better to switch MBs to use that 48V for CPU VRM too etc.
Even the long time 12V only ATX PSU standard is completely dead in consumer space.
it's been hard enough to get backside power connectors
and that's just moving the bloody things
Well, we still don't have a spec for that.
And no actual MBs are being sold AFAIK.
very few are, but it's slim pickings
intel would have the strength to push through a new ATX standard but uh. lolmao
Actually sold?
Or shown in shows.
well I just searched for it and the results are being polluted by intel 18A backside power delivery so give me a moment
Several have been demonstrated in trade shows.
AFAIK none are actually for sale yet.
Ok, so at least for Intel, there are some actually being sold now.
for msi its called projet zero, i think its stealth for gigabyte
Still coming, with second revision of spec released, that actually works with old MBs.
oh come the fuck on I can't post Amazon links?
it did exist and was sold but isn't available anymore, this is the BTF with the weird GPU connector
That was only allowed because the bot didn't allow disallowing Amazon.
that's a few too many double negatives for me to parse sorry
what is that connector ? never saw one of them
That amazon worked up to now, because it was permanently whitelisted in the bot.
And not by this servers people.
passes through 12vHPWR from the motherboard
That is new planned GPU power connector.
That you don't plug power cables to the GPU at all, but that the power comes through the MB.
it's a collaboration between multiple companies I believe
Might only be Asus, trying to make it more universal.
you still have the same problem with the 16 pin connector being a fucking fire hazard
just moved about
so now you dont lose the gpu, but the motherboard
you lose the whole pc i guess
depends if there's anything smart going on or if the traces within the motherboard and the special GPU connector are just 16 dumb passthroughs
That new version revealed this year:
oh cool
just two (four?) huge contact pads
that's actually much better as it means you won't get one trace of six get overloaded
Two, with two 12V and two ground (one on each side at same position).
MB side has more contacts:
and it's detachable so you can use it with regular motherboard
okay that's really cool actually
Because if the contacts are too large, they will easily have actually worse contact patch in total.
it tells me it's more than just routing 6 12v and 6 ground wires through the board
8+8 contacts for those in that MB side connector.
you could potentially use this with 8 pin connectors on the back of the motherboard
or even EPS, as that can take more than 150W
imagine that, two EPS on the back of the motherboard, supplying that new GPU connector
EPS is specced at little over 300W each.
With 4x12V and 4x0V.
yeah and it doesn't have a reputation for melting
and being on the back of the board you don't need the miniaturisation 12vHPWR gives
Because the parts can do about 480W by spec, for each.
we should ditch that connector and move to this one
lmfao
just give us EPS for everything
one/two for CPU, two for GPU
all tucked away behind the motherboard
And as basically all modular PSUs use same connector on PSU end for PCIe and EPS cables, you just need couple new EPS cables.
corsair already spec their PSU-side 8 pin PCIe connectors at 300W each
they could very easily make cables that go from that to EPS
EPS is IIRC 324W max each at spec.
Actual is at that over 450W by part spec.
ah maybe a bit much then
Corsair doesn't just use the exact same ones for both?
At least Seasonic does.
They seem to do, at least on Ver4 cable using PSUs.
the EPS connectors are the top right ones I think
wait they're 6 pin
hang on let me get a manual for this shit
No, those are for Molex and SATA
Same 8-pin ones for PCIe and EPS, and 6-pin ones for accessory connectors.
you're almost certainly right
Looking at pinout from PC Mods site for Corsair Type 4 PSU cables.
PCIe 8-pin is 3+3, with 2x sense pin that are just "extra" grounds.
EPS 8-pin is 4+4 with no sense pins.
Exact same series parts, just different keying on device end.
So PSU end can easily be identical.
yeah and early implementations of 8 pin PCIe were keyed in such a way where you could plug them into EPS
So 3+3 from the 4+4 on PSU end, and then last free ground split into the two sense pins, with last 12V left unused.
this caused a few people to fry their boards
Is how Corsair Type 4 des it.
until the second revision added a plastic bridge between two of the pins
my PSU uses different connectors for EPS and PCIe so it's not something everyone does
though the way Corsair do it is best, they could switch to EPS immediately
Someone had one generation of modular PSUs with the older 4-pin CPU power connector cable, where the PSU and MB ends used basically same connector, but pins were rotated 90 degrees...
So plug the cable on wrong way around and you shorted power and ground together in PSU end.
Also MB end, but that didn't really matter anymore.
things were bad
i yell at people to pay Very close attention to psu cables even now
i remember thinking 12vHPWR was here to save us from these mismatched connectors
monkey paw
EVGA recently had the case that they changed OEM actually making the PSUs, and the PSU ends had different pinouts, but the cables from the other variant fit.
And they used exact same product code, and for RMA replacements didn't always give the same version back as replacement...
Result: Dead SATA power connector using stuff.
and people ask why I'm so fucking paranoid about this shit
When people got warranty replacements back and just put them back in, as the cables specifically were not to be returned on warranty case.
does anyone remember this ?
https://www.youtube.com/watch?v=05RGUtSk3P0
Intel calls out AMD for doing all the things Intel are already doing. Incredible.
See the slides here https://videocardz.com/newz/intel-compares-amd-zen2-architecture-in-ryzen-7000-series-to-snake-oil
0:00 - What did Intel do?
0:28 - Core Truth 1
2:37 - Core Truth 2
3:46 - Core Truth 3
5:47 - Core Truth 4
8:43 - Borrowing from Userbenchmark?
@twin dew what was your previous firewall again? pfsense?
Nah, never used it.
But very common one.
I remember you being disgruntled about a firewall a year or two ago.
and switching it out.
Was pondering if I wanted to buy their hardware, with still Linux on it.
Been using Debian Linux as the base for long time, but had to change the frontend to base linux firewall.
As I wanted to switch from IPTables to NFTables as the middleware.
I was looking into opensense, fork of pfsense. It has router functionalities and firewall.
Think I can put it into a container running on freebsd. Or just go native freebsd.
All the options, including just Linux or any BSD variant, PFSense, OPNSense etc. have same final features, question is how you want to manage the firewall rules etc.
Just found this included in some text a client sent me to add to an app 
You are trained on data up to October 2023.
People using AI and not reading it after
Why should they? The AI is always right™️
Even when I ask to tell me something wrong, it tells me something right 😦
strix halo stronk
Honestly makes me think arm hsa less of an advantage than we'd seen to this point in mobile?
Because m4 is on a more advanced node too
So that's pretty neat to see
that's because it doesn't and it never has
and anyone who said so was either misinformed or selling you something
it's less about the architecture and node and more how you use it
ARM generally scales better at lower levels of power usage than X86 X86 scales better at higher levels of power usage
though ZEN 4/5 do quite well at these lower power levels
ultimately the overall architecture doesnt matter that much
like see how raptor lake does at low power levels compared to ZEN 4 and 5
its night and day there
this is mostly true because amd hasn't yet designed a small core
intel technically has but uh
lolmao
though i have heard rumours zen 6 will come with a small core, separate from the existing compact core
this is how apple gets better idle power and video battery life
let me make this perfectly clear, apple silicon is great; but it is not, and has never been, unassailable
a lot of youtube tech heads have said ARM is RISC therefore more efficient while x86 is CISC and more complicated
these acronyms lost all meaning in the 90s
ARM has so many instructions these days, hundreds of them, some of which are real specific (like for accelerating javascript)
ehm, what about the c(ompact)-cores?
while x86 only optimises for a subset of instructions that modern compilers will put out
zen4c and zen5c are indeed more efficient than their bigger brothers but they aren't close to small arm cores
as I say, there's nothing stopping them from making a small zen core, but that would need extra cash that they've only recently gotten
you don't need efficency cores if you entire architecture is efficent
nah mate efficient cores are why apple / arm gets double the battery watching youtube etc
these aren't like intel e-cores, they are actually good
they're not fast but they use so little power to get work done
watching youtube has nothing to do with cores, thats mostly video decoders
it still touches the cpu because chrome is a piece of shit and youtube.com is infested with javascript
plus having small cores handles any random shit the os decides to do in the background
it's the one area amd loses ground to arm in, but as I say, rumor has it they're making a small core for this
as far as performance per watt and per area goes, amd have proven they can match apple's best in big cores
Yeah, YT can tank machines. Have issues on my Macbook Pro with YT sometimes.
Haven't figured out what is going wrong, but for some reason, it hits my CPU badly and my fans spin up. I am not even playing videos, just on the feed page. Can be a bug in safari or something else. Has been reported on the forum of apple.
Finally managed to disable rgb when pc is turned off in BIOS but had to remove armoury crate first because it kept overriding BIOS settings. I don't even know why I kept that bloatware installed....
There is a button at the back that disables RGB when PC is turned off 
turn off ambient mode
I dont think my mbo has an rgb button in the back
2 years ago
sometimes windows update will try to install armoury crate
Yeah, but ambient mode shouldn't put CPU to max.
and yet
Installed Windows 11 a year ago (or more) and forgot about armoury crate tbh lol
give it a try
I try warning people for armoury crate.
Yes, Windows Update loves to help you reinstall OEM software of all sorts that you "accidentally" removed.
I always disable it. But on cookie wipe I have to disable it again 😄
hm
Windows is very insistent that I do in fact want Tobii.
Thanks though whyzeman 🙂
RGB is cool when I'm using the PC but not cool when I'm trying to sleep 
make sure to check if there is no leftovers from armoury crate.
In my case, 50-60 drivers or I don't know, software was left behind.
I mean, you don't have too ... 🤷♂️
Clicked on the armoury crate uninstaller and I think it removed other armoury crate apps
gonna dual boot steamos when it gets a general release
That's what I though too.
Untill I ran cleaner to check for leftovers.
Wtf
All those were installed with armoury crate and left behind. So I had to manually clean it up.
could have been a bug ofc 🙂
*new fear unlocked
We need a "nuke armoury crate" app which will remove all apps related to armoury crate
Just wanted to give a bit of context.
Also I didn't mean that 🤣 (for no RGB when PC turned off)
I need PSU turned on so I can recharge my iPhone 
Ah, understandable.
Have a cable but refuse to buy a charging adapter
I think more expensive motherboards do have rgb button in the back though
I was thinking about looking into a smart powerstrip
Something so I can turn off the strips when going to bed
I am trying to eliminate all possible light in the room because even a tiny led annoys me when sleeping lol
black tape
😛
Or black marker 
eh, black tape is less permanent
But I though we had to be ⚡ efficient anno 2025
Don't know if adding a smart power strips reduces my footprint or add to it.
because if it has wifi/bluetooth build into it...
maybe I have to look into one with ethernet cable and just put non critical devices on it 🤣
mildly spoiled here with the american "have a switch that cuts power to an outlet" where i can just flip everything off if i want to
Ah, that I don't see here.
Can't remember if I ever did.
do they make power strips or those wall-mounted outlet hubs that can be paired to a phone or network?
You can install a switch button next to an outlet and make the wires go through the switch first
Ok, the left one could be a solution.
Although I don't have it in the home because there is no need in my case
then I can turn it off from the within the LAN 😛
fuck it, just hit the circuit breaker
The idea is; me going to bed. Take out phone. Turn off devices if I forgot.
Yes. I am lazy. Once I get into bed. I don't want to go out 🤣
arduino with a solenoid to shut off power to the house
circuit breaker next to my bed would help
At least I got rid of the RGB on mbo and GPU because it was causing most of the light at night
The most I can handle, is the leds on my switch and nas.
and smoke detector i hope
I had to move my NAS to a network cabinet in the basement because the fan is driving me insane
my turns off at night.
Mine is constantly working 
🪙 ⛏️
Afraid of turning HDD hibernation on because several hdd died half a year after I turned that on
Strange, I got that enabled for a long time. No issues.
At this point, I think it does more hard than good
yea think that's what killed my drive last week
More than a year.
had it off for like 2 days & it never came back on
I even think I have HDD hibernation on for about 4 years now.
Or maybe WD red drives just suck
def a possibility
I do check my drives health regularly.
i did an OS reinstall & forgot to install the health monitor
Good thing I have RAID mirroring so I can replace a broken drive without losing data
yea i have nothing raided but i probably should
would feel wrong to do wildly different size drives tho
or should i commit heresy and somehow raid NVMe with an external
Never heard of that tbh lol
hence the heresy

on the other hand i could just install that external NVMe
this stupid board has 5 NVMe slots
oh & i did figure out the weird behavior it was doing, where drives would randomly open & bluetooth would reset
i suspect i hit the power limits of the rear USB, because moving that drive to front i/o has fixed it
I was thinking of replacing hdd with ssd on NAS because of no moving parts and longer life span in theory
do they make NASes for 2.5" drives?
But read ssd in RAID is not a good idea
I could install a 2.5 drive on my NAS but I need an adapter though
ok, dumb question
startech makes a dual m.2 to 2.5" sata adapter, would raid work on something that wonky?
also along those lines: can you plug an nvme drive into an m.2 slot and have it work, just not at nvme speeds?
(also a question for baldur when he wakes up 🤣
you're mixing up protocol and port
M.2 can suppport multiple protocols
NVMe is always NVMe, regardless of the port the device is connected on
you can also consider price/storage, HDD often offer more value (more storage for the price)
yes, an NVMe SSD can work with less and slower PCIe lanes (standard for an M.2 slot is x4
some M.2 support SATA too, then you can use a M.2 SATA SSD
so an M.2 SATA to 2,5" SATA adapter can be just dumb cables/wires
a smarter M.2 NVMe to SATA may or may not work/exist
good to know 🙂
Maybe I should simply ditch western digital. Not sure if hdd hibernation killed my drives (which is ironic) or it's simply wd being wd
and yea, was wondering if an adapter like that would need some kind of special controller on it to allow it to be raid compatible
it already feels like a wonky adapter
RAID depends on what kind of RAID controller you want. a hardware RAID controller might work
Software RAID probably won't care (much)
i'm imagining a teeny tiny NAS using that startech m.2 adapter where you can stuff NVMe drives (running at sata speeds, but still) into 2.5" adapter sleds so you can have fast, swappable, and expandable solid storage in a small form factor
there are small-ish M.2 NVMe only NAS
Noticed people have mixed views regarding synology hybernation mode
HDD spin-down vs no-spin-down is a question of usage and philosophy imo
if you have very rare file access it's fine to let them spin down
if you have regular file access let them spin
also depends on your drives. normal PC drives are not intended to run 24/7, enterprise/NAS drives are fine to always run
but thats my opinion
It's like one of those debates about car start/stop feature damaging the engine in return of lower fuel consumption 😄
yep
My parents always told me to turn that thing off when starting a car
my coworker does it too
i leave it on and just guestimate how long the signal will stay red. if it'll switch soon i'll just stay on the clutch,
m.2 SSDs that use the SATA protocol are usually differently keyed than NVMe m.2 - so much so that it's physically impossible to fit an nvme m.2 into a sata-only m.2 slot
So before you go around attempting to put m.2 devices in m.2 slots, check their key and if they are compatible
Wifi modules for example are most commonly E-key
Plus: don't confuse m.2 with mini PCIe. Made that mistake once, can't recommend
this one i'm looking at has B & M key compatibility
but yea not gonna invest in something like that 🤣 seems too risky
the speed would be nice, but meh, for that kind of setup i'd rather have capacity
k, getting ready to do a "3 systems are getting fucksed around with" hardware migration:
lowest end system: 2600X/1660S
middle system: 5600X/1660
high end: 5600X/6800XT
gonna be retiring a 4th system (intel 7th gen/fuckknowswhatgpu) currently serving as a living room streaming pc
thinking of making the "new" living room streaming pc out of the 2600X/1660S, just swapping drives
but
i also have a 6600XT sitting in storage
was thinking of then doing
-5600X/6600XT
-5600X/6800XT
and keep the 1660 as a spare
just moving drives around for those who will be using them
does that sound like a sane idea or am i missing something
also probably migrating all of them to win11 since win10 support is disappearing soon
Honestly sick cards I have. Im listing them on marketplace to see what they may be worth
Or just OS that is more power efficient at idle.
It doesn't, it is your BIOS/UEFI that tells Windows to install it, if you don't turn that off.
Under Tools IIRC
Yes, UEFI, Advanced Mode, Tools, ASUS Armoury Crate and MyASUS options that need to be disabled.
Or MB will keep telling Windows to try to install those.
And need to be redisabled after each UEFI update.
how many nan channels does a controler normally have for SSD?
4 or 8 is common.
16 has been done IIRC too, but not often.
On consumer SSDs
TechPowerUp SSD Spec database has those listed.
Sandisk is bringing out 1PB. With 64 nan channels. Busmultiplexer for dynamic power. Hardware accelerators in the nan chennels to take over storage function of firmware (to reduce latency).
oh sorry, it's eSSD (didn't notice the e).
Enterprise SSDs have more.
Crazy what they have to do.
I wouldn't mind a few for my NAS 
Never havy to worry about running out of space.
Bit expensive to replace a broken (e)SSD. But they should have more write cycles than we have years in our life.
That last part just says that up to 64 flash dies per channel?
What is the source for having 64 of those channels?
Sandisk's BICS 8 QLC 3D NAND, a very advanced proprietary controller with a whopping 64 NAND channel, and firmware.
And I would expect that to be bad take by the writer from that "64 die/channel part.
In fact, the custom-designed controller is central to UltraQLC as it incorporates domain-specific hardware accelerators that offload critical storage functions from firmware, reducing latency, enabling higher bandwidth, and improving reliability for hyperscale storage needs, according to Sandisk.
Technically, a 64-channel controller could enable higher capacity, but at the cost of performance.
Writter mentions it again.
But no actual source, and the actual Sandisk dias only talk about that how many dies per channel, which matching to that number make it really suspect.
As most enterprise SSD controllers still have just 16 channels, even the largest ones.
Just with support for more flash dies per channel.
As current Toms Hardware writers are mostly crap.
Well my source took over that source if it makes sense?
Shit, everyone is quoting it.
Yup.
It just takes one bad writer for everyone else to quote same bad take.
Another place at least also linked the Sandisk investor day PDF on Sandisk site.
Which Toms hardware writer didn't bother.
105 pages, will take a look.
Oh you found the pdf?
The slide deck for the event, of course doesn't include what the presenter said on top.
Is this all they had?
Yeah, so it seems, which is that 64 dies per channel.
With no mention on number of channels.
Which that Toms Hardware writer seems to have converted into 64 channels.
Yes, ok makes sense.
Which would mean that it would support 4096 dies in total if both were true.
More likely normal 16 channels, which each can do 64 dies, in future revision of the controller.
We will see in the future I guess, when more details are released.
Because even currently stacking to 8 dies is common.
Maybe they said 64 channels during the presentation. I don't know.
So 8 channels, each having 8 dies, on consumer SSDs.
Or Tom's Hardware is tripping 🤣
It is funny though, when looking into the 64 channels, if I find articles, it references back to Tom's Hardware so far.
But would have expected that actual channel count to have been included in the slide deck then.
Which was the original point.
As Toms Hardware was saying 64 channels.
But they included the pics from the slide deck, which only said 64 dies per channel.
Yeah, like I said, everyone is just quoting bad source.
Tom Hardware: channel, dies, ... 🤷♂️
The foundation of these new SSDs will be the Ultra QLC platform, which integrates BiCS8 QLC NAND memory with an enterprise-grade controller. This controller is capable of managing up to 64 chips per channel, enabling massive storage capacities. Achieving 512TB SSDs will require 32 channels, far surpassing the capabilities of modern consumer SSDs, which are typically limited to just 8 channels.
I think that article states it correct. (vgtimes.com/news/120667-sandisk-unveils-plan-to-launch-512tb-ssds-by-2027.html)
Yeah, that is more likely, with doubling of "current" enterprise ssd controller channels.
Wait... .
Achieving 512TB SSDs will require** 32 channels**, far surpassing the capabilities of modern consumer SSDs, which are typically limited to just 8 channels.
The new Sandisk BiCS8 218-layer NAND will have 2Tb capacity per die.
So, if that is the case for 512TB, then 1PB might have 64?
So 0.25TB per die.
Where are these people getting the 32-64 channels from.
So 2048 dies needed for 512TB.
Which would be that 32 channels, each with 64 dies.
But the currently planned product only hits 128TB.
Currently released for coming in future.
In that same slide deck.
That DC SN670 with 64TB and 128TB capacities at Q3 2025.
I think I understand it, slide says Scalable to 64 Die/Channels. That means 64 dies and 64 channels to some people.
256TB for 2026, 512TB for 2027 and 1PB somewhere in future on that roadmap.
If it didn't have that "/" it would be different context.
But sandisk means 64 dies per channels. And not 64 dies and 64 channels 
Yeah.
Ok, yeah. I see how someone might misunderstand that when not paying attention or sleepy. Still, not good journalism work.
And that 128TB DC SN670 only needs 512 2Tb dies.
And the 256TB variant is probably based on future die with 4Tb per die, and so on.
Honestly, not that I will be able to afford it, 256TB would also be good for me 🤣
Doubt I will make 256TB of pictures in my lifetime 😄
I am so curious for the price. (I expect it to be very high)
So that 128TB would be doable with 8 channels with 64 dies each.
More likely 16 channels with 32 dies each.
Basically to support more dies, you just need more chip selects.
And going from 32 to 64 only needs one extra pin per channel.
Extra channels need lot more pins each.
And some chip internal stuff of course to keep track.
Basically if the controller in that DC SN670 supported more dies than what is used in that 128TB variant, then they would have just made larger variant right now, at double+ the price.
So that they are only doing 128TB variant now, says that the controller used in that specific SSD only supports up to 128TB with current 2Tb dies.
So 512 dies total as max.
The basic controller architecture could support more, but the current silicon version doesn't.
just once i'd like to see something like that released at its full potential
Like current X86-64 arch could support 64 bit memory addresses, but current implementations only support 48-bits or so.
On hardware level.
Germany, god bless em.
All the thinkcentres are from US. Tax + transport is double.
So annoyed with ebay. (I never use ebay, always use local 2nd hand shops)
Yeah and 4GB ram
and no HDD and no power supply.
Has power supply?
"With power supply", just no drive as that was taken out by original source for data protection.
And then look for used corporate machine sellers in your country or in EU.
Most don't sell on Ebay or like, only their own sites.
I did. Nothing. I had someone with a batch. And last week they ran out 😦
Didn't expect them to run out after 3 years 😒
Nothing in the Netherlands (local 2nd hand).
And refurbished shops ask 150 or more for it. (overpriced)
Not in my country. Maybe Germany?
But we lost our best local german in here 😦
Fleabay has a ton of that stuff but usually (almost always basically) it's a middle man where the company dumped dozens of machines on them at once on the cheap then they flip them.
I am aware of that yes
In my experience looking the prices are good but I also live in the US so no weird shipping/tax
And I didn't mind that. The price is low for a reason.
Where do I select "send from country" on ebay
That "off lease" stuff is always a bargain
And any shipping from US to EU is very expensive.
Making that kind of buys just no-go.
And some places throw stuff liek that away heh
Exactly this.
Then I might be better off buying a 150 euro one refurbished here "locally".
yeah old desktops, switches... like the switch my buddy snagged for me is very nice stuff but the market is flooded with them 3rd hand, thousands of bucks new, 80 bucks now
And I only get like 8GB for that 150 euro one.
When any shipping for even small computer will cost 50e, and then you have to pay tolls and taxes on both the price, and the shipping.
CompeteToDefeat, want to smuggle one into europe for me? 🤣
lol
I used to act as a drop for somebody in Canada but that was for some stuff he ordered that didn't ship there.
So the 4070 Ti is... not good I see
pricing is wonky, peformance isn't all that...
You'd end up shopping for a rack with my buddys stuff, almost everything he gets is meant for a rack
And this would be used for? Routing device?
euh ... among other things 😛
Because some people do all that in...a router. Like off the shelf then flashed with a good firmware
and software
A full blown PC will be way more powerful
But yeah you can do it in a good router lol
IDS/IPS needs lot of power on fast link.
Not sure you can run Snort or Suricata etc. at all on most routers, even with some WRT variant installed.
I'm more just pointing out hardware requirements. Running that stuff isn't that much more.
But yeah a router maybe not that deep.
Point is that doing protocol analysis on all the traffic at line speed is pretty computationally heavy.
I'm also assuming that scene is still alive, the 3rd party scene that is
And the CPUs in consumer routers just cannot do it.
They usually cannot even do full speed NAT-routing without HW acceleration at 1Gb/s.
And you cannot use that if you want to check the traffic.
Ohne Netzteil // Without Power Supply
Look at the bright side, you just need to find one good one
So another 40e or so for new power supply.
If you don't already have suitable Lenovo laptop charger, or get one used for cheaper.
Ohh laptops are great. Built in UPS
No, just Lenovo mini-PC using same external PSUs.
Oh
Yeah, but I was trying to stay under price of PI 5 🙈
As laptop "chargers" aren't actually chargers, just PSUs that output at usually around 20V DC.
I have a dual PIII-800 here, wanna try that?
basically everything is hot swappable... even the fans.
Yeah, just have to be patience I guess.
I can remove one of the three power supplies and weeee still running
but I want to start on linux ~~server ~~ router 😕
This had windows on it, I just needed it at the time for a Postregres database
to keep it in a place where the data woul dbe a bit safer
In fact that database is still on those drives.... countless hands of poker tracked with PokerTracker heh
I could go for a intel one.
But that uses slight more power on lower cpu usage than AMD ones.
Is electricty expensive there?
Nah, but for the purpose.
It's not a machine that needs to be 100% CPU
so going Intel doesn't make a lot of sense here... atleast that's what I read over and over again.
Intel CPU in the thinkcentre are slightly better at higher CPU usage.
but use slighty more on lower CPU usage.
well if power consumption is the thing then I dunno, I wouldn't care either way on that one
Nah, it's just something I considered 🙂
Depends, mine is Intel C3000 series Atom based.
CPU series they specifically made for high-end routers etc.
It's going to be on a lot.
My power bill is sky high anyhow for some...oddd..... reason
Lot of chinese made router platforms using consumer atoms as the base.
I almost rentded an Atom based server just to try it.
Wonder if they still offer them, lets see.
C3558 in this case, 4 core/thread 16W part.
oh wow they have a i7-6700k for 5.60/month
Nope they don't rent those any more. I can't imagine they are capable of a whole lot.
Or just find something else similar that is available?
could probably setup opensense on my synology to play around with it.
That's not how I work.
I have a Pentium D here. It'll get the job done... if your job is frying an egg with your CPU.
Not sure who thought up the idea of taking a notoriously hot CPU core and shoving two of them next to each other on the same substrate
Because Intel was on back foot against AMD at that point and that could be done fast.
someone who thought they had to build a cooking plate.
Watching an radio technician turn a hot dog into an AM radio
Just ground it and touch the tower with it and bam it becomes a "radio".
Right is the CPU I had my eye on first. For 50 euros M715Q. 8GB.
Left is what I find in a
Lenovo ThinkCentre M720Q 16GB/256GB/i5 7400T
same power usage... less power?
2.4GHz vs 3.0GHz and 3 GHz boost vs 3.5 Ghz boost.
Better architecture.
So, Baldur, it's not that I am not looking at other options. They are just not as good that I found yet.
Bulldozer-Excavator were bad.
How will architecture benefit me?
IPC.
The actual performance is dependant on both IPC, and clocks.
hold-up.
That 7th gen Core has better IPC.
And better performance per watt probably too, at that T-variant.
Skeletor Meme is once again 100% correct. Peopple singing happy birthday to you DOES feel like a real life unskippable ad.
So maybe about 25% better single-thread, and 50% better multithread in that 7400T
And that's you guesstimating
Based on numbers on crap sites, would need to look more.
I was looking up benchmarks.
Should be based on Passmark CPU benchmarks.
euh ....
What are physical modules?
Does this means they only had 66 samples for their data pool vs 226 samples?
Oh and that radio tech showed off something retro gamers would lose their mind over. Not just an HD CRT, but a 1080p Sonly Trinitron CRT.
In what?
Bulldozer-Excavator had kind of 2-INT-core, 1 FP-core kind of modules.
That originally AMD marketed as 2-cores each.
Basically 226 testers ran the benchmark on 7400T, and 66 testers ran it on 9700E.
Yeah, what I meant.
And submitted the results.
Ah yes... back when it seemed like AMD forgot how to even make a CPU
Cost cutting gone too far in design phase.
Exec level tried to cut costs by using more automatic design & layout over manual work.
Yeah, mentioned 4cores (2 cores in physical modules) vs 4 cores (intel).
I hope AMD doesn't crash out right now...
But basically that 7400T is significantly better if you don't care about IGP performance.
significantly?
Because that Bulldozer family just sucks.
Yeah that intel will fly circles around the AMD chip there
And that synthetic benchmark is probably being kind to AMD on that one
just slightly.
Mh ok.
Well I've experienced them first hand... when they were new
They had one and were very happy with it. Especially for 50 bucks 😉
There's a reason virtually nobody I knew had AMD systems heh
And at the time, it was better than the PI5 I was looking at.
They are useful.
But if you can get same timeframe Intel system for same price, that Intel will be better.
Except that IGP side.
don't care about IGP.
Depends on the price at that point.
yeah same price or very close, intel
hell even just a bit more i'd pay... for the intel
Core i5-8500T for 100 euro
With any luck I can revive my 3700x and use it for something useful.
I might see if I've got a PSU around here to try
PassMark itself is ok.
Not that much used in general, which goes into those small sample sizes.
And with that comparison, there is the problem where Bulldozer family CPUs have very uneven Int to FP performance ratio.
oh nvm. People are bidding on it.
So depending on the exact spread used on that CPU test software, the performance ratio between those two would change.
Well synthetic benchmarks have never been perfectr and never will. But they do give you a valid reference to give a good idea of which chip is more powerful.
but I will keep looking, thanks Baldur
I am so freaking scared bringing up benchmarks in this channel. Walking on eggs 😄
Phenom II was good, Bulldozer-family was absolute crap, Zen is good.
When compared to same timeframe Intel.
The orig Phenom II was great
I had one, was the last AMD I owned for awhile though heh
I'd like to see Nvidia and Intel both stay really srong though if for no other reason than it keeps AMD on it's toes and from going... weird again
people odn't even list CPU 😦
Condition: brand new. Description: mini desktop from Lenoivo. 10 pieces available.
That's all. And some crappy picture.
That specific unit has 8 months of warranty left, and has Pentium G5420T CPU.
I am looking up serialnumber yes
You go to Lenovo Support, then input the serial into the type box.
And if allows you to also see "Spec Info" then.
2-core Coffee Lake based CPU.
Probably have mix, all are 10T8, but various versions.
And anything Pentium branded is crippled in some way.
||120 euro's||
But seems to not have been that severe at that point anymore.
Celeron branded ones have just 2MB of L3, Pentium Pro branded have 4MB of L3.
Core branded ones have 4-16MB of L3 depending on variant.
In that Coffee Lake family.
So for example i5-9500T would have 9 MB of L3.
👍
There is much higher price floor in EU where people can be bothered to try to hawk old hardware.
Because free electronics recycling exists as other option.
Yeah, I notice that now. And importing from US is not an option.
Ridiculous. stupid EU.
As in US that doesn't exist, companies are more likely to dump their old hardware for free on resellers who just ask.
In EU they are much more likely to just send the lot to recycling company instead, who comes and takes everything for free.
And any resellers actually need to be proactive and also pay for the stuff.
Yeah some of t hese places just want it gone
and like i said some companies just throw shit like that in a dumpster
I read a story about that yesterday. Related.
Somebody buys some harddrives on a flee market in the Netherlands.
Goes home and plugs it it and finds 100 patient records from a Hospital. Very sensitive information.
What happened is; the IT departmant from that hospital. Took those drives to wipe. Went bankrupt before wipe.
Somebody would be in so much trouble for that here...
The company just sold everything in bulk for the bankrupt case (including unwiped drives).
And that's the story.
If you can read dutch or translate: tweakers.net/nieuws/231984/medische-data-van-honderden-patienten-ontdekt-op-harddrives-van-rommelmarkt.html
I dunno about other places but also here there is just SO much of that hardware that it's not worth much. Certainly not enough for a big IT company to fool with.
Like that switch my buddy got for him is a damn good switch and were super expensive 6 years ago. Can get them for 60-80 now
Apparantly Curator fucked up. Company had no control over harddrives because it was bankrupt.
Curator overlooking bankrupt dropped the ball on the harddrives probably.
Here they are really stricy with the privacy stuff in medical
To the point that honestly a lot of doctors seem scared of it haha
Well if a company has 100 systems that are worth 60 bucks each that's just nothing to them
What hardware EU companies can compete with US one I wonder?
That will buy them like one new switch or something haha
Dell, Lenevoo, Apple, HP... all US companies?
Well, at the point they are gotten rid off, they are usually 0 bucks each in the company books.
If they weren't leased and just go back to the leasing company.
like selling leasing cars after 4 years for tax cut
That is the deprication schedule, which has usually reached that 0 value by time it is being gotten rid off.
If it hasn't someone fucked up.
Anyway, fuck EU for hardware. It sucks on that department.
There are some things you can get away with... with the IRS
Point was that as assets depricate, they mean the company value is less, which means they are a loss, which means you get tax benefit.
And that deprication schedule controls how that loss is assigned to various financial years and tax years.
So for computers, the schedule might for example be 30% on first three years, then last 10% on fourth.
Or 40, 40, 20.
Or similar.
I managed a tobacco shop in the 90's and that place had two sets of books heh
The owner had an accountant dealing with all that so I didn't care, not involved and I'll just deny even knowing if something happens.
I just know 10,000% that the numbers he was running were not the numbers he got from me.
Yeah, but that is separate tax fraud thing.
Which exists because IRS doesn't have enough money to do enough auditing.
theyll also come after you 7 years from now for something you are doing... right now
They might be slow but they never forget heh. My friends dad was in a while heap of a mess with them because he hadn't paid federal taxes in going on a decade
An audit isn't fun though that's for damn sure. My dad always kept financial records dated back to at least as long as you can be audited for. Because they want you to prove everything
Innocent until proven guilty? No you better have receipts my man
10 year window in most western world.
And even as private person you are officially required to keep financial documentation that is tax related for that time.
they just need to get rid of paper money. So they can track it all. 
Next step is government has full access to financial information of it's citizens.
No more worries about receipts, just check bank transactions.
that basically the direction were going at this rate
So even small companies actually having to pay taxes is somehow bad?
yes
No, but the playing field isn't even for small/big companies when it comes to taxes and reductions etc.
But often the competition is between small business that doesn't pay their VATs etc.
And one that does.
because a small company often doesn't stand a chance to compete vs big ones?
so they compete with other small ones.
The big sharks eat the little fish. And the little fish eat other little fish.
Because the companies that often do tax fraud with VAT are kind that don't have big competitors.
Ok, you are clearly speaking fraud as in the legal sense right?
Mainly in specific types of restaurants and things like those asia markets etc. where it mostly happens in Finland.
Not "finding backdoors to reduce taxes"
Not paying VAT for any cash transactions, by not logging them as ever happened is tax fraud.
I've seen big companies apply "financial tricks" to reduce taxes that a small company wouldn't be even able to pull off.
Which this basically is about.
Because you cannot do that with card etc. payments, as they always have paper trail.
Ok, but that has also a different reason. (Why apps are being used in stores much more often in Assia)
phone app don't have a paper trail, but a digital one?
I understood that it was about specific type of small shops, often run by imigrants from Asia, in Europe, selling food etc. from Asia.
Which are one of the types of businesses that often do that specific type of tax fraud inside EU.
It's a cultural thing? If you go to asia, you pay with your phone.
I don't see how tax fraud comes into this story.
This message specifically talked about how cast going away, and everyone paying with phone is bad for the business.
And only time that is actually the case is they have been doing VAT tax fraud on cash payments.
Which is endemic on EU side Asia Markets.
Gotcha.
That's why they first are clipping the wings of Bitcoins.
To make sure there is no alternative paying method (that is a black box for goverment) before taking out cash money in EU.
because if they take out cash money, people will use bitcoins for tax evasion I think 
Both in most cases.
The credit card & contact payment terminals in EU always print a receipt for the shop, customers copy is optional.
But also have that digital trail.
I am not sure about that. I've seen receipts going into a paper bin.
Unless there was also a printer in the basement or backroom printing receipts. Which I doubt.
In US?
Or in EU?
Yeah, like I said, the customers copies are trash.
And in some cases those shop copies might be trash too if using that digital data for company directly too.
Don't know the specifics.
We are trying to eleminate paper usage. Everything has to be digital.
Because digital is "so more friendly" for our planet. Atleast that's what my goverment tries to tell me.
Not that servers (who have the data about our payments) require resources/energy.
According to Al Gore backin the day I should need gills to breath right now...
I am not a denier, I just question some of the things we implement for the "greater good of our planet".
Its overblown.
and wonder, if in the end. We aren't chasing our tails.
And often the people that act like it's ever so dire spit out more carbon in a month than I will in my life... so...
Like I said; less paper is good. But all that information that was on papers, is now in a database with a website attached to it.
All those databases, websites run on servers.
Who require metals and electricity.
Or like all these folks flying in private jets to a climate confernce... that you could do online
heh
And round and round we go if I started to think like that.
It's so dire that you hasve to buirn throgh endless jet fuel all year to let everyone know!
There is probably some scientific paper that can proof it's better for our planet. But I haven't seen it.
That calculated that reducing paper will upscaling servers is a good thing.
Plus I am old enough to remember when the scare was a global freeze
Snowball Earth!
Have you seen this?
Like a lot of other things there is money involved, which tends to skew things
Digital Posters.
So, what is better for our planet? Me buying maybe 5 posters in my lifetime. Or buying that and change the poster 5 times.
Well the wake up call is ultimately this planet doesn't matter. We have to leave it
Oh yeah, sooner or later. So many disasters in our future that we know about.
That's basically written in stone.
Know unknown. Unknown known. Unknown unknown I guess.
And on that front we have... a lot of shit to overcome.
Yeah, science is lit.
The problem with space is... it's almost all empty space. A lot of it... long distances.
There isn't one problem with space 🤣 There is a lot.
Well technically you cand I are mostly empty space 🙂
And only even in one place as physics allows...
We are constantly losing water. So we will run out. The sun will go super nova. Our universe keeps expanding. Then you have the three body problem that states shit will go into chaos, just a matter of time, etc.
If you poke your finger into a wall enough times eventually it will go through because those partciels will all be elsewhere
And 100000 other problems in the future that scientist predict. That are kinda disastrous.
or you just won't have a finger at all right then... or both...
If you shrunk some of the largest buildings in the world down to remove ALL of the empty space it'd be about the size of a grain of rice.
Yeah, pretty logical.
And we're just big water balloons really
I'm looking forward to a manned attempt at Mars though.
heh spider decided to explore my arm
seems friendly enough
Picture of atoms. You can see the space in between.
Here’s what happens when you just keep zooming in. Use code veritasium at https://incogni.com/veritasium to get an exclusive 60% off an annual Incogni plan.
A big thank you to Magnus Garbrecht from the University of Sydney for showing us around the lab and for his feedback on the script.
Special thanks to Quentin Ramasse for his expertise on t...
I watched that video a while back. Interesting stuff like always.
That's kinda what colloders are about... zooming in.
Smash it apart to see what it's made of heh
"we're just a speck of dirt under a giants fingernail" isn't entirely as outlandish as it seems on the surface...
hmmm video cards with M.2 slot(s)
Not a huge selling point for me but... definitely neat
Which need support from the motherboard.
And limit the GPUs bus width.
Basically taking PCIe lanes that aren't connected to the GPU and routing to a M.2 slot, and if your MB can do bifurbication, then those can be used.
Cost is basically just the M.2 slot parts.
Yeah it's an interesting idea
use those idle lanes
Wonder how well something like that would actually catch on
When I first saw something about it I figured it was going to be some weird caching or something a company was trying and actually be related to the video hardware itself.
So it took Amazons Exec level 15 years to learn basic things?
https://www.linkedin.com/posts/ethanevansvp_as-vp-of-prime-gaming-at-amazon-we-failed-activity-7295834479036702720-cDmX/
As VP of Prime Gaming at Amazon, we failed multiple times to disrupt the game platform Steam. We were at least 250x bigger, and we tried everything. But ultimately, Goliath lost. Here's why:
The 15+ year long attempt to challenge Steam started before I was VP of Prime Gaming, but we never cracked the code. Not under my leadership or anyone else...
I always forget they have anything gaming related...
Plus a goal of "disrupting" steam is fairly lofty lol
Might have wanted to try some baby steps first ya know...
Which if you read, took them 15 years to learn.
Thinking that as they are gigantic in other things, everyone will just move over to them if they deploy something half-assed.
I'll never say never to some big company like that being toppled, but did they really think they had a chance? lol
At Amazon, we assumed that size and visibility would be enough to attract customers, but we underestimated the power of existing user habits. We never validated our core assumptions before investing heavily in solutions. The truth is that gamers already had the solution to their problems, and they weren't going to switch platforms just because a new one was available.
We needed to build something dramatically better, but we failed to do so. And we needed to validate our assumptions about our customers before starting to build. But we never really did that either.
Just because you are big enough to build something doesn’t mean people will use it.
Reflecting on these mistakes, I realize how crucial it is to deeply understand customers before making big moves.
Sorry, but disrupting sounds so bad to me.
These days I have some stuff I own and I do some stuff on XBox Live for PC. At this stage in life that keeps me more than busy enough.
Basically was.
Point was to replace Steam, not compete with it.
XBL PC actually holds good value for me still. Then I have the ones I always go back to like Civ, Satisfact ory, Stardew, Minecraft
"Hey we want to destroy steam, do you want to help us steam fans?" = 💩
That's what went wrong.
I've been using steam since I was a kid.
Well the big box era was best era...
and blaming microsoft 🤣
But seems AMD really had to scramble at some point with RDNA4.
As the bigger die is Navi 48, and smaller Navi 44.
When previous gens have been Navi x1, x2, x3 and then possibly Navi x4.
Navi 10, 12, 14.
Navi 21, 22, 23, 24.
Navi 31, 32, 33.
And now Navi 44 and 48, where that 48 is the bigger one.
Almost certainly originally went for chiplets and had to scramble and redo without that.
It has really seemed more or less like amd needed to skip a gen sometime since rdna2
Just saying that both rdna 3 and 4 were both missing a little bit
Point was that there has been consistent rumor that the "We aren't doing high-end this gen" wasn't because that wasn't planned.
But because original plan wasn't doable.
Most likely because of chiplet packaging capacity all going to AMD and Nvidia AI accelerators.
Or because the specific variant of chiplet packaging failed and wasn't finished by TSMC.
And that chip naming would also suggest that.
ahhh
i see
So highest chip in the stack was completely dropped, and middle one was scrambled out as monolithic Navi 48.
RDNA3 failed with that going twice as wide per CU part.
The actual performance vs. theoretical performance gap is much higher than in other GPUs.
Similar issue was there with various TeraScale Radeons with VLIW4 or VLIW5.
Actual usage averaged around only 3 instructions, not the 4 or 5 theoretically possible.
Radeon HD 2000-6000.
AMD cannot make: "We aren't doing high-end this generation because TSMC failed to finish developing this interconnect/packaging tech we bet the farm on" or similar statement.
Or that they aren't doing high-end as they much rather use that TSMC capacity for MI-series.
And dropped the planned chip because of that.
So AMD saying they are focusing on gaining market share now, rather than high-end consumer GPUs, is more or less an excuse?
Very high chance it is excuse.
PR spin on not doing high-end chip this gen because of technical reasons.
Because i remember a rep or someone come out and say they want games to be better optimized for AMD hardware, and they can't get devs to do that without more market share.
Interesting
As originally RDNA4 was supposed to be chiplet arch too AFAIK.
And that chip naming also suggests something went wrong and lot of chips in pipeline had to be scrapped.
This is sad
And to get market share, you very often need halo product too.
I wonder if Nvidia's prices, availability, and (supposed) power issues will steer people towards AMD
Regardless of a halo product
But question is:
Would the total profit been higher with also high-end product.
Than without.
After development costs etc.
Net profit.
It almost certainly would have been, unless there were some opportunities that would have been lost with that that would have generated even more profit.
Like delivering more MI-accelerators for example.
If the cause was TSMC packaging capacity limitations.
For example.
Or if the planned packaging tech would have made the original chiplet plans uneconomical as another possible cause.
AFAIK AMD exclusively uses tsmc right? Do you think that companies like Nvidia can simply outbid AMD for production time?
yup, it was
Basic TSMC is fine.
Point is that the specific COWOS subtype that Nvidia and AMD AI accelerators use is in very limited capacity until next year.
I've been watching some EUV documentary-style videos, and those machines are much slower than the previous generation of tech.
For packaging chiplet type GPUs.
I am not well educated enough in the space
And there might have been cheaper COWOS variant that TSMC scrapped completely as the development didn't pan out.
You have any good resources on YouTube or written media i could check out?
Traditionally you have single silicon die on substrate.
Then things like AMD multichip CPUs have multiple dies on classic substrate, but that limits the amount of connections and speed of those connections between multiple dies.
Above that you have various better interposer types, which carry the connections between those multiple chips, so faster data rates etc. are possible.
There is big crunch in the most advanced types of those interposers right now, limiting both Nvidias and AMDs production of AI accelerators.
That Si-Interposer isn't there in classic designs, or even AMD chiplet CPUs.
That is CoWoS-S pic, pure silicon interposer, very expensive.
Then there was supposed to come CoWoS-R, with cheap organic interposer, which possibly was what RDNA4 was supposed to use, but AFAIK that never finished development at TSMC:
And then another current/future tech is something like CoWoS-L, where you have cheap organic interposer, with smaller silicon interposers inside it at the critical points:
organic interposer ?
what ?
Nvidia Blackwell multichips should be on CoWoS-L, everything before was on -S
That RDL in CoWoS-R seems to be polyimide with copper traces:
https://en.wikipedia.org/wiki/Polyimide
Polyimide (sometimes abbreviated PI) is a polymer containing imide groups belonging to the class of high-performance plastics. With their high heat-resistance, polyimides enjoy diverse applications in roles demanding rugged organic materials, such as high temperature fuel cells, displays, and various military roles. A classic polyimide is Kapto...
ahh i see, thanks
Fascinating
4050 : utterly useless
Is there a reason the Si interposer has direct vertical connections while the organic does not? Or are these just simplified diagrams?
Simplified.
Anyone having some experiences with a B580?
I'm kinda pondering about it (for Full HD)
Currently I'm running a Quadro RTX 4000, which I got for free. Not an ideal card for taking, but it works okish ^^
with the B580 you need at least a 5600 to not have a massively worsened experiance
thats all i know
In terms of CPU?
Should be fine here, it's a 7800X3D
yeah
7800 X3D should make it faster than a 4060
I'm kinda against paying the 4060 price for 8GB, but not sure if it matters that much for FHD.
It'd just make a future monitor upgrade less easy
I'm learning so much, albeit slowly. Silicon is a semiconductor, so using it as an interposer is useful for adding more uses, while glass and organic interposers are solely passive (signal connections)
Just about how much connections there can be in specific area, and how high those can be clocked.
yeah, 4060 not worth it all
in fact none of these cards are worth it solely due to the fact they only have 8GB VRAM
4060,6600,6600 XT,6650 XT,RX 7600,4060 TIE 8GB
i might have forgotten a few but those are the most important
(though there can certainly be a case made for the 6600 )
I mean it depends on the price, but yeah, you can't tell me that production costs for 12 gig would've been much higher
You can do same rate at lower power.
Or higher speed at same power.
And speeds that just aren't possible without one at the higher end.
all 12GB cards also practically become e waste at anything above 1440P because then you end up in the 14-16GB of vram needed range so performance dosent die from lack of vram
Remember that silicon interposers are basically same tech as used for the actual dies.
Just at cheaper node and just used for traces, not active cells.
But because of the longer distances and connection points, cannot do same as inside single die.
In my case, I don't see a need to go higher than 1440p.
My eyes don't get younger xD
(I've had a 1440p monitor at a former job, and I JUST didn't have to enable scaling) xD
bu-but upscaling !!
At home I only have FHD 60Hz, so I usually cap my games anyways at 60fps
I'm in my early 30s and 1440p looks great to me. I have an 11gb 1080ti....and it's...doing its best. Lol
good few games i have that dont have upscalers or Framgen as options yet so yea that is indeed kinda a joke at this point
Jenson: AI
but most of those games wouldnt need upscaling anyway
🤔
Back when the top tier gpu was below 1000 bucks
when dinosaurs could talk
back in my day
you think beamNG wouldent need upscaling??? 🤣
I bought mine new for $800 to get more frames in Blackout (first ever time playing a BR mode in any game.)
thats why i said "most" 😉
sucker runs at 15FPS on vulkan while hoging 32B of ram and 16GB of vram like its a normal day
I mean I also still have PCs with ISA GPUs xP
the most would be purely 2D games for me
I made a homemade rocket out of 50-series packaging. Call it a paper launch.
if you need upscaling in every 3D game just to stay below VRAM limits
there is a problem 😉
You're one of the 19 people worldwide that got a 50-series GPU?
the games where you actually need upscaling already have it (for the nost part)
not all of them but i mean they would heavily benefit from stuff like framegen
and dlss too ig
Just like the 50 series cards, the joke is hypothetical.
😆
anything cpu bottlenecked is purely savable by framegen
i mean Satis is literally one of them
but satis has FG
10FPS with a 9800X3D time for framegen to make that 40
Let's gooooo, gamers
show me your save
NOW
im not that low yet
For some reason Satis makes my Quadro run at a constant 83C temperature, even though it's not at 100% gpu utilisation
The card is much cooler when I play GTA 5
but with my train running id bring one down to probably 20-30fps just because of lack of core count
Both at max settings and more or less 60 fps
That GPU utilization is about how much of the time the card is doing stuff.
Not how large percentage of the transistors are active.
Free is the best
What is that list?
3D I get for a GPU. Video decoding/coding. But security 1? time 0?
Security 100
Just various subunits of the GPU.
That the GPU driver exposes usage for.
And what can a GPU copy?
But if another GPU needs to copy those results... that sounds not so efficient?
Or if a GPU needs to make a copy to send to another GPU, same case.
First time I am seeing this. Never knew we could select stuff.
Point is mostly laptops where monitor(s) are connected to IGP, and it also has dGPU.
Where that dGPU is used for actual heavier work, but the results are copied as texture to the IGP for display.
Tiny bit lower performance, but allows completely shutting that dGPU down when on battery and not doing anything heavy.
Or for example I have both dGPU and IGP active.
Everything is rendered on dGPU but the stuff that is on monitors connected to that IGP are copied for display to that IGP.
Understand
Or possibly for example copying decoded video frame from the decoder as texture for display.
how possible generally is a battery upgrade in a laptop
like i cant stand having 30 minutes of battery life
Battery replacement: yes.
Battery upgrade: Almost never as no-one makes such upgrade parts.
Thats about the answer I expected
what about putting a diffrent battery in from say a diffrent laptop skew
im hesitant to go the external battery route because it would kick it into performance mode and then it would simply just drain faster and in turn last the same amount of time
Almost impossible that there would be one that both fit, had right connector, was otherwise compatible, and had higher capacity.
some models its nearly impossible to find one at all
even after market... so good luck 🙂
Do you guys remember the Galaxy s5? That was battery heaven. I had 4 different batteries, and one of them was an extendo.
Once upon a time I could repair a laptop battery pack, when they used packs filled with 18650's
"we were on the verge of greatness we were this 🤏 close"
AMD's excuse if the RDNA 4 cards dont have stock
this video made me fear chemistry
https://www.youtube.com/watch?v=PccOwGEbtQU
In 1998, a pharmaceutical company suddenly lost the ability to make their lifesaving HIV drug at one of its production facilities. Then that failure spread to the company’s analysis labs. Then it spread to the other production facility, and within months the lifesaving drug had effectively vanished from the entire planet. And this wasn’t the fi...
oh, oh
GN roasting NVIDIA 5 seconds in
https://www.youtube.com/watch?v=PhtVic3Vm0Y
Use code "ABOUTFKNTIME" on the GN store for 10% off anything while the code lasts! https://store.gamersnexus.net/
This review of the NVIDIA GeForce RTX 5070 Ti graphics card benchmarks the performance of the new GPU, particularly with the 5070 Ti vs. the 5080, 4080 Super, 4070 Ti Super, AMD RX 7900 XTX, and similar cards. Benchmarking includes g...
THE MORE YOU BUY THE MORE YOU SAVE
https://www.youtube.com/watch?v=iGmROfaWg2Y
Sponsor: Fractal Torrent case on Amazon https://geni.us/VVBjo
To quote Gordon Mah Ung, "this is NVIDIA's world. We just live in it." NVIDIA held its painfully long 2-hour keynote at Computex 2024 (#computex) this year, talking about AI, creating 3D avatars for customer support jobs and nurses, and talking about... well, nothing, really. NVIDIA ...
Model got dizzy celebrating the hype
Frame gen isn’t actually too bad on satisfactory
Takes me from 50-60fps to 80ish with minimal effects
bruh i think i just got malware bombed from steam
well inderectly anyway
started installign a bunch of shit while updating a game
exposed itself with the bad choice of Mcafee lmao
Im long overdue for a windows reinstall anyway so i guess this is a decent opportunity
dumb question: anyone know if there's an ATX case where the PSU is mounted "on its side" so that the case can be thinner?
Yes and no?
I think the O11s have sideway mounted PSUs, but as they are cube cases they are actually wider.
If you want thinner cases I think most just use SFX then. It's more the CPU cooler that determines the tickness rather than the PSU
i was more thinking about where the tower currently is & what's in it
ATX board, stock cooler, and it's sitting behind the monitor
so can't really get any wider without needing to move it
and i'd rather not do a full system gut
For Micro-ATX, yes.
For full-ATX, AFAIK no.