off-topic-tech

[night girder]

I will look into it. I am not a big fan of "It just works" implementations without knowing the ins and outs of how it works. Especially when it comes to security.

[stray badger]

Tailscale is based on wireguard

[night girder]

And I can remember UPNP has risks. Was mentioned earlier on how Zerotier works.

[twin dew]

So needs to be VDSL modem that implements specific VDSL standards up to specific revision levels.
And VDSL2 Profiles up to 17a.
And doesn't do something stupid against the spec in actual use.

[night girder]

Indeed.

[stray badger]

tailscale doesn't use UPNP unless you want it to

[twin dew]

G.993.2 Annex B, Annex K 3, Annex N
G.993.5 Vectoring
G.994.1
G.997.1
G.998.4 G.INP

[wooden vessel]

what risks would that be? it's a twenty year old technology. hey, router, send this port on the outside to this ip/port on the inside. plskthx. if you don't trust the inner service then don't do that then?

[twin dew]

Need to check what those K 3 and N are.

Because for example Asus only says A and B for 993.2, and doesn't mention G.997.1.

[night girder]

www.howtogeek.com/122487/htg-explains-is-upnp-a-security-risk/

[twin dew]

UPnP is IMHO very bad idea,

[night girder]

I just remember vaguely, that UPnP has drawbacks.

[twin dew]

The whole point is to allow devices and software behind NAT make dynamic port forwardings.

[stray badger]

its security risk

[stray badger]

not good if you get malware infection

which is why tailscale doesn't use upnp

[wooden vessel]

if you are owned, upnp makes it marginally easier to exploit the ownership
this is absurd

[twin dew]

Lot of normal software try to make UPnP requests to allow outside access to them, even when most don't need that functionality.

[wooden vessel]

yeah, like games

[twin dew]

Ah, That G.993.2 Annex K.3 just means being able to operate in PTM TPS-TC mode (vs K.2 being ATM)

[twin dew]

So basically from this list only that G.993.2 Annex B, G993.5 and G998.4 are relevant.
Everything supports the rest if it support those three.
And basically everything has supported all of those three for a long time already.

So basically this says:
Any modern VDSL modem that isn't broken and actually follows the specs.

And then that out that if someone connects something that isn't following spec, ISP can go to that clause 4 to kick the user out or force them to change the device.

Modern, as in designed or updated after 2016 or so.

[night girder]

But start finding a router based on the specs

[twin dew]

And you expect ISP to list all the worlds VDSL modem models?
Because basically anything will officially fill it, and you cannot know really without testing if some model is broken or not.

[night girder]

world? No, but atleast a few models.

They do support Fritz!Box. Probably because of a contract/deal they made.

There is no justifying this one. They do this with bad intentions, 99% sure.

To comply, while not really complying, but still complying with the law.

The fact is, ISP want us to use their overpriced shitty routers. ISP don't want users to have more control or privacy.

[twin dew]

Where would they put the limit?
And if they listed some models they didn't themselves already provide, the ones producing stuff that is not tested and listed could start lawsuits or other trouble.

[night girder]

pff

We got two ISPs.

Telenet and Proximus. Tell me Baldur, why can Telenet do it?

www2.telenet.be/residential/nl/klantenservice/internet/modem-en-andere-toestellen/eigen-modem.html

If it's so insanely hard to list all the worlds routers ... but yet, one ISP did it. And the other didn't.

[twin dew]

Those are all OEM modems, not retail ones.

And seems they just checked spec sheets and didn't actually even test:

[night girder]

This is a coda 57, does seem useable for home.

[night girder]

It's still better than what Proximus does. Which is not helping at all.

[twin dew]

Point was availability for consumer to buy.

[night girder]

There is a difference in a company trying to help it's costumers, and saying "we might be wrong".

Or throwing a bunch of specs to them and say "find out yourself, we cba"

[night girder]

No, point of the law was, that costumers can buy their own routers **__easily __**and don't have to be forced to pay for shit ISP router.

[twin dew]

No, that the list of the second ISP was pretty useless as the products on it aren't usually available in retail channels.

[night girder]

By throwing up these barries, people will just keep buying ISP routers so nothing changed.

[twin dew]

Only to ISPs.

[night girder]

Let me check.

AVM is just fritzboxes.

[twin dew]

Forgot those have started to have actual retail availability.

[night girder]

tweakers.net/pricewatch/1396562/avm-fritzbox-6660-cable.html

[twin dew]

Didn't have for long time.

[night girder]

and that one is available.

[twin dew]

Tried to look for one about 10 years ago to buy IIRC.

[night girder]

I have a hunch.

That ISP hopes, router companies will start labeling their stuff with "Proximus/Telenet compliant" ...

And the reason is, because there is a contact e-mail at the botom of the page for them 😄

Costumers: go to our community blablabla.

Companies: contact us through e-mail.

[edgy hazel]

9950x3D being nuts

[night girder]

And some charts about how great the new CPU is.

But, no clue what source is.

[twin dew]

Would seem like AMDs own slide.

[night girder]

Black Myth Wukong 🙊 1%.

[twin dew]

Because all the difference in that chart is probably coming from the higher frequency.

[night girder]

maybe I am just reading the chart wrong.

But this doesn't blow me away?

58% increase on counter strike 2 right?

But for the FPS* hungry games, like CP and Wukan etc. Not much increase.

[twin dew]

Like most 7000 to 9000 game comparisons?

Because mostly the parts that benefit games didn't get IPC improvements from Zen4 to Zen5.

And IOD is the same.

Zen 5 was major AVX512 improvements for server space, and setting stage for Zen 6.

[night girder]

So it's little incremental improvements.

[twin dew]

Very large front-end changes.
But the backend and RAM are holding that back.

IIRC

[night girder]

no, I mean the next generation of CPU's from AMD.

[willow pike]

Dell??????

fuckin my god intel are cooked

strix halo in dell to complete with apple

who knew

[twin dew]

There is some AMD keynote in CES ongoing?

[willow pike]

the 9800X3D showed performance gains in the likes of cinebench, while the 7800X3D did not; this tells us zen5 is memory bound

[willow pike]

it's live

[twin dew]

Yeah, but not interested in watching.
Just expected that to be the source.

[willow pike]

i have it on in the background

it's mostly been shareholder bukkake

AI every 5 seconds

[night girder]

Oh, we get more AI?

[willow pike]

strix halo has the XDNA NPU on it and can have 96GB RAM, that's AI crack cocaine

[twin dew]

That Strix Halo 385 (AMD Ryzen AI Max 385) seems interesting laptop CPU.

[willow pike]

i want it in a steam machine

[twin dew]

8 cores, 32 CU, 45-120W

https://videocardz.com/newz/amd-ryzen-ai-max-300-strix-halo-apus-bring-up-to-16-zen5-cores-and-40-rdna3-5-compute-units

[willow pike]

the top end is 16 cores 40CU

[twin dew]

Yes, which will raise the price a lot.

[night girder]

temps?

[willow pike]

had my eyes on strix halo for like a year

[twin dew]

How would we know yet?
Only articles with the slide deck out.

[willow pike]

lmfao they didn't mention radeon at all

[edgy hazel]

The new AMD Ai Ai chip with Ai Machine Learning Ai

[willow pike]

god

[night girder]

you said laptops. I think temp.

[twin dew]

And that will depend on the laptop.
Just that my brother is planning to get new gaming laptop in fall.

In 1500e range if possible.

[night girder]

Damn.

[willow pike]

mfw no radeon

[edgy hazel]

tbf radeon is usually end of the year

and they also said that

[twin dew]

That 9000 series should be coming in next few months.
But this was too early compared to the rumors.

End of this month or some time in february for the announcement.

[willow pike]

official slide

https://tenor.com/view/infinity-war-avengers-death-cjb1967-gif-18666706

[night girder]

So, they split up the 9070 into two types? XT and non XT version?

while 7900 was only XT variant?

[willow pike]

top end 9000 ~ 7900 XT ~ 4070 Ti

[night girder]

Ah ok, I understand slide more now.

[twin dew]

That there are some 9070 and some 9060 models coming out.
And at least 9070 and 9070 XT.

And that 9070 models will fill up from RTX4070TI/RX7900XT down to 4070 Super and RX7800XT
And then 9060:s will be below that.

[night girder]

And the 9060 is competitor of the 4060TI?

[twin dew]

Yes.
And 7800XT to 7600XT

[willow pike]

they gotta be cheap

[night girder]

So no competitors for 4080/4090 as asual?

[willow pike]

not just affordable, cheap

[twin dew]

Not even to 7900XTX.

[willow pike]

I want this shit bundled in cereal boxes

[twin dew]

AMD did say they weren't going for high-end in this gen.
Officially just because they didn't want to.
In reality probably because the third design was chiplet based with need for advanced packaging.

[night girder]

Fair. It was no critique to AMD.

[glossy glacier]

i thought they said they'd do it to gain more market share?

[twin dew]

And once we got close enough to today, AMD saw that the highest end design couldn't be produced at realistic price in current situation.

And didn't have time to do alternative one anymore.

[willow pike]

we're pretty certain that RDNA4 is a true chiplet design

but the underlying tech is the same as what the MI300 uses

and they'd much rather sell those

so instead we just get chip, singular

[twin dew]

Dropping one silicon variant from same architecture wouldn't save that much design costs.
And AMD has been doing three variants each generation for long time.
Now there are only two.
And the rumor was that AMD had designed the top one to use much more advanced interposer than what 7000-series used.

[willow pike]

the 9080 XTX would have been two 9070 XT dies

[twin dew]

And had to drop that one because of packaging at suitable price just wasn't available, because of both TSMC development problems and because of that highest end packaging being all taken over by enterprise AI accelerators.

[willow pike]

even if it was only 70% faster than 9070 XT, that's very high end territory

[twin dew]

Will probably take years until we hear what really happened.

[willow pike]

my money is on AI sucking in all the funding like a black hole

we're in a gold rush and MI300 is the shovel

[twin dew]

Not funding, but fab capacity for the interposers and packaging.

[willow pike]

ironically the 9080 XTX chiplet design would have been done to get costs down

now we have to pay whatever jensen wants

[twin dew]

When that mid-point TSMC interposer stuff failed because of technical issues.

[willow pike]

yeah they didn't plan on AI demand being so high

they order years in advance

also means the 9080 XTX order existed and was rescinded

"yeah we paid for your factories but we wanna use them all for this instead"

[twin dew]

AFAIK there was supposed to be new lower cost, lower performance interposer tech that TSMC was developing, that just didn't work in the end with enough reliability.
Between the one used in 7000-series and MI300

[verbal raft]

so

guys

we have the FSR 4 DLL

or at least im being told

[verbal willow]

Alright you PC experts. I got a problem that is stumping me and the Internet lol. My computer doesn't like my ram. I tried to boot up my PC and the Q-LED check lights freak out and continue through a repeating sequence, yellow(dram), red(cpu), white(VGA). Reseat, try one stick at a time. No luck. So swap them with my gfs sticks(different brand and older) and both PCs boot just fine with the swap lol. But my PC won't boot with my original sticks. The Internet is giving me no help with this strange issue

[verbal raft]

or at least it seems like it

[twin dew]

Have you reset the BIOS/UEFI to defaults?
Have you updated it to latest version?

[languid gulch]

ok, steve from GN said that for the 9950X3D, one of the CCDs will have the L3 vcache, but the other cores would be higher frequency. does that sound right?

[verbal willow]

Updating to latest version is my next step now that I can actually get into anything. With my sticks I couldn't even get into bios

[verbal raft]

yeah pretty much the same as the 7950X3D

[languid gulch]

would that be to distribute heat, or is the interlink & schedule good enough now that it wouldn't matter where the cache is

[stray badger]

i think its possibly a cost thing too

the vcache is probably not cheap

[willow pike]

👀

[verbal raft]

well what i'm being told is the FSR4 DLL😂 🤣

[languid gulch]

just kinda wondering why it wouldn't either all be stuffed onto 1 CCD, or evenly distributed across both CCDs

[verbal raft]

(on the optiscaler server)

🤔

would it work if i just you know

[languid gulch]

i can't wait for AI FSR where it still can't figure out writing

[willow pike]

those madmen

[twin dew]

Nowhere near good enough to use the other CCDs cache even in 9000-series.
And that V-cache cannot easily be less than that 64MB.

[willow pike]

going between CCDs makes performance suck CCDs nuts

[twin dew]

So either you have one CCD with it, or both.
And the V-cache CCD needs lot of extra steps to make, and second full processed die.

[verbal raft]

OK this has a 99.9999999999% chance of not working

[willow pike]

or it's a virus

[languid gulch]

hang on. so with the 9800X3D, you have 1 CCD, with all the L3 vcache piled onto it, right?

[verbal raft]

that

[twin dew]

Yes.

[verbal raft]

to me these seem legit enough

[willow pike]

9000 same as this

[languid gulch]

and with the 9950X3D, you now have 8 cores at higher frequency, but not the L3 vcache, and the other 8 cores at a lower frequency, but the fat L3 vcache. am i right?

[willow pike]

ye

the frequency difference will be very small

it was much bigger on 7000 where the vcache hurt thermals

[twin dew]

Well, those per core L3 cache numbers are crap, as they average between CCDs, would need to be two numbers for the dual-CCD ones.

[twin dew]

Still does, as even with VCache below the CCD, there is still structural silicon add-on on top of the CCD die.

[willow pike]

yeah it's dishonest saying it's got 144MB

[verbal raft]

well so much for that LOL

[willow pike]

nowhere near as bad, but still there a little

[languid gulch]

144MB, but not all of it is available evenly across all cores

[willow pike]

let the optiscaler boys work

[twin dew]

9000-series:

[verbal raft]

also it seems like it wasnt leaked at all

[languid gulch]

now i'm wondering if they could do a L4 fat stack of cache where the 2nd CCD would be 🤣

[verbal raft]

https://download.amd.com/dir/bin/amdxcffx64.dll/

[willow pike]

oh lmao

[twin dew]

AMD could already do more Vcache layers.

[languid gulch]

well yes

[stray badger]

price

[languid gulch]

but man that empty spot feels kinda lonely

[verbal raft]

there

[languid gulch]

oh i know

heat pipe!

[willow pike]

no

[languid gulch]

window in the IHS & stick and RGB LED there?

[willow pike]

the structural silicon is there to stop it cracking

[twin dew]

It isn't actually empty, but has dummy silicon in it for structural purposes AFAIK.

[languid gulch]

oh i know, i just want it to do something

[stray badger]

price

[willow pike]

you should see MI300

[stray badger]

EPYC probably uses that empty space

[languid gulch]

engrave the name of one of the workers on it

[stray badger]

you gonna pay EPYC price?

[willow pike]

same tech as 3D Vcache

GPU on top of IO die with built in cache

[languid gulch]

does ryzen use HBM as its cache?

[willow pike]

it does not, too expensive

this is what they killed the 9080 XTX for

[languid gulch]

ah

[willow pike]

it's not for graphics, it's for AI

[languid gulch]

can't wait for the RX9090XTXTi Super LC OC Strix

[willow pike]

chatGPT uses it

[verbal raft]

9090 XTX

instantly better

[willow pike]

CCDeez nuts

[verbal raft]

https://tenor.com/view/cat-cats-what-wut-crazy-gif-20900490

[languid gulch]

🤣

[twin dew]

By XTX

No, what was the X<something>X GPU maker?

[languid gulch]

ngl whenever i see "liquid cooled overclocked" abbreviated to LC OC, i can only read it in a french accent as "le cock"

[twin dew]

XFX!

[languid gulch]

SWFT

sooft

so i assume the 9950X3D is supposed to beat the 9800X3D

[willow pike]

in games, nah

[twin dew]

Probably again +100MHz from 9800X3D to 9900X3D cache-CCD, and then again +100MHz from that to 9950X3D cache-CCD

[soft bloom]

watching random YT suggestions can be blessing

[willow pike]

maybe some will use the extra cores, slightly higher clocks, but mostly the same

some games will cross the CCD barrier and performance will tank

[twin dew]

It did, when right CCD was used.
But the extra price and extra work to make it work right made it not weren't worth it for pure gaming build.

[languid gulch]

are there games where the 7800X3D/9800X3D bog down because they don't have enough cores?

[twin dew]

Sorry, +400MHz for the V-cache CCD from 7800X3D to 7950X3D?

[twin dew]

Just from the extra frequency on cache-CCD.

[verbal raft]

well the experiment failed but it was worth a shot

[willow pike]

report back if they can't get it running

[verbal raft]

i guess its called driver-based for a reason

[verbal raft]

it runs fp8 which as far i know only ADA supports

[twin dew]

Or the APIs in DLL changed too much.

[willow pike]

weird

[twin dew]

To just get it to work by replacing a DLL with new major version.

[verbal raft]

or the current optiscaler mod is interfiring

couldnt even type what i was going to before it crashed LOL

[mental oriole]

Nullptr 🙃

[twin dew]

FSR3 plugin called function that didn't exist.

Basically part of the FSR3 support is in the main exe, and tried to call FSR3 dll, but you had provided FSR4 DLL that didn't have same function.

[languid gulch]

lemme know when FSR & AI stuff is good enough to overlay upscaled textured on older games in real time

[twin dew]

So that FSR4 dll cannot be used to replace FSR3 dll.

[languid gulch]

waiting for my system to be able to just fake upscale to 4K textures because there's 0 chance the game's getting it natively

[languid gulch]

i had this one recently. face caved in

[gilded helm]

SO what's the scheduling situation with Zen 5 X3D dual CCD?? Did they make it automagic?

[verbal raft]

so then that's why they were boasting about some driver/software feature that can update FSR

[twin dew]

And from now on, you should be able to update the DLL for any game that supports any version of that FSR4 with DLL swaps, like you can with Nvidia DLSS.

Which you couldn't do before.

With FSR1/2/3.

[warm blade]

Yo so i need a GPU upgrade from a GTX 1050 TI to a Intel B580 (I have a AMD Ryzen 5 1600 Six-Core CPU) and the budget is about 500 euros, im in Serbia so some sites are like Gigatron or Technomania, idk what else i need to specify really

[twin dew]

Does your motherboard have ReBAR support?

Because not all old AM4 boards got it even with latest BIOS/UEFI versions without modding the image.

[verbal raft]

Wouldn't recommend the B580 with a CPU that slow

The driver overhead will murder your fps

With that CPU

[warm blade]

yeah but i can upgrade the CPU too

[verbal raft]

Ok then as @twin dew said check if your mobo has REBAR

[warm blade]

not really into PC parts and building, wdym mobo and rebar

[twin dew]

Motherboards BIOS/UEFI Firmware needs to have been updated to allow enabling feature named "ReBar".
If the system doesn't have that, Intels dGPUs don't really work properly, as the drivers have been coded expecting that to be available.

[verbal raft]

Oh,sorry Mobo=motherboard and REBAR is required for ARC GPUs

[twin dew]

And it was intruduced after Zen 1 in actual usage.

Lot of motherboards got those firmware updates to allow enabling it after.

[warm blade]

god wtf did yall just say 😭
im so tired bruh

[twin dew]

After AMD intruduced it as AMD SmartAccess Memory, which was just rebranding of that optional PCIe spec feature of Resizeable Base Address Register

[pure karma]

i think im in need for this one

[warm blade]

well imma check in tommorow or something, but tommorows christmas (im a orthodox christian) so im not sure

[pure karma]

long story short you cant get a intel arc GPU which would likely be best for your budget unless your motherboard supports a feature called REBAR if it doesn't you would need to go with something else

[verbal raft]

Or upgrade your motherboard and CPU

but at that point you'd be better of with something else IMO

[gilded helm]

Mostly I'm concerned because AMD do not appear to have sent the 9800 X3D to mobile, only the 12 and 16 core versions.

[twin dew]

GPU-Z allows to see if ReBAR is enabled in BIOS/UEFI, but No doesn't mean it isn't supported, just that either there is no setting, or setting is set to disabled:

[verbal raft]

From what I distilled from what GN said it should work

[gilded helm]

If the stars align and I can get an X3D laptop with 15% Rakuten, student or senior discount, and credit card statement credits, I'll be very tempted.

[jagged snow]

Original DOW was my first ever pc game

Got that right with cim city 3k

[willow pike]

even with ReBAR, this is a terrible combination. buy literally anything other than an intel GPU

[winged valley]

Do you even know anything about arc battlemage

The B580 is far more than sufficient as an upgrade from a 1050ti, especially at its price point

Also he's coming in from a discussion we had in dms. When he said he has a 1600 I told him he needs to upgrade cpu first since his main game is satis

[wooden vessel]

i didn't see a mention of motherboard chipset

[tough owl]

The NVIDIA GeForce RTX 5090 is set to retail for $1,999 USD while the GeForce RTX 5080 will come in at $999 USD, the GeForce RTX 5070 Ti at $749 USD, and the GeForce RTX 5070 at $549 USD. The GeForce RTX 5070 is rated for around GeForce RTX 4090 level performance.

[languid gulch]

can't wait for the independent reviewers to get their hands on the 5070 & discover that the "4090 performance" comparison is in power consumption only 🤣

[tough owl]

They did the same thing with the 3070 and 2080 TI

[languid gulch]

wtf could they have discovered that'd let them get on par performance with half the bus width and 38% of the cores

and half the vram

[twin dew]

Now that I had some time to passively think about this, this probably only means that Recall is installed.
Not if it is working or not.
Just if it is ticked as installed or not in this screen:

[twin dew]

It says that 5000-series is with framegen in 4x mode.
So 1 real frame, 3 generated frames.

[languid gulch]

god i hate frame gen

[twin dew]

And 4000 series with just current 2x mode

so 1 real, 1 generated.

So to get like for like, the 5000-series numbers need to be halved.

And to compare to anyone else, halved again for both.

[languid gulch]

well yea, a lot of monitors have solved the ghosting problem, time to bring it back at a software level

can't wait for them to wreck a monitor's color accuracy with some kind of new thingy

i hope that reviewers choose to stop using upscaling & frame gen in their testing

[edgy hazel]

Next up: ColorGen

[languid gulch]

true, but some seem to include it as some kind of metric

"this gpu is crap, and even worse for the value, but if you just lower your resolution down to something from 2008, it's fine!"

[dire igloo]

Still not acceptable imo

it just means that it doesn't invalidate the channel as a whole

[languid gulch]

can kinda "get away with it" because a laptop screen is small

[twin dew]

Where if you use FG on anything less than 60-100 FPS, the result is just a mess...

[languid gulch]

to me frame gen & upscaling means "just duplicate 1 rendered pixel into 4 on the monitor, but change the lighting & color a little bit"

[twin dew]

So it really cannot be used to get playable FPS from unplayable, as the image after FG from unplayable FPS is so bad that you cannot play.

[languid gulch]

what really sucks is that reviews on youtube don't/can't really show what frame gen stuff actually looks like to the user, you have to have the gpu & monitor in front of you to see it

if there was a way to kinda have it to apply to only those 1%/0.1% lows, and then go away under normal gaming, that'd be mildly acceptable

dude is a gender neutral term as well 😛

mine likes to occasionally lose signal & just spam the last keystroke it registered

super fun when it's a movement control

HA

oh that's just dumb

sounds like spyware to me

"you must install this software that's directly attached to every keystroke you make"

[twin dew]

Friends don't let friends buy Razer...

[languid gulch]

i almost bought their mmo mouse until i discovered the redragon one for like 1/4 the price, figured it was worth the shot

i do get a kick out of anyone who's not a billionaire thinking that any of this AI stuff is good

snakeskin leather so it counts

look at where it goes small to bigger, long straight lines. topside/underside of snake

unless they were using baby crocs for it

[edgy hazel]

He went full snakeoil merchant

[languid gulch]

i'm not seeing any of those bumps in the middle of scales that crocs have, so i'm thinking it's snakeskin

these things

unless it's completely fake & just for looks, which rich people love to do

[languid gulch]

yea XPS got nuked

interesting. PCB in the middle so both fans are fully flow thru

[twin dew]

And where are display connectors?
Seem to be missing from that pic.

[languid gulch]

yea noticed that

could they do a ribbon cable to a PCB attached to the rear mount?

[twin dew]

Very high bandwidth/frequency signals, so no.

ribbon cable just doesn't cut it.

[languid gulch]

maybe a 90° vertically tilted PCB coming off the PCIe side, then slotted into another PCB with the outputs on it

[twin dew]

Will see once someone gets actual product in hand.

[languid gulch]

something like that maybe

16 pin connecting the daughter board, then another 16 pin connecting to the output PCB

time for weird: this means that for the first time, you could have bottom intake, full GPU passthru, CPU air cooler in updraft, and top exhaust fans, all in line

i think the only other time i've seen that is when LTT stuck that noctua cooler on the GPU

[willow pike]

multi frame generation

https://tenor.com/view/team-america-vomit-gif-3459509

[twin dew]

Bigger Number More Better!

[willow pike]

so 30fps becomes 120fps

fucking gross

[languid gulch]

oh no doubt, most valuable company in the world doesn't want to be reliant on anyone for sales

[twin dew]

Been squeezing their "partners" out for long multiple generations already.

Reason EVGA stopped making GPUs.

[languid gulch]

EVGA probably had the slimmest margins, which means that everyone else has been charging you extra for being not-EVGA 😛

[twin dew]

Nvidia is bundling GPU chip and VRAM chips in bundles you have to buy from them, for very high price.
You cannot use other sources of VRAM.
And all designs have to have approval from Nvidia.
And then all possible ways to differentiate outside of coolers was toned down over time.

Where the price difference between that bundle of GPU and VRAM, and FE card retail price was small and getting smaller.

Basically started to be too small to actually make a GPU and sell it for profit.

Forget about warranties etc.

[willow pike]

this PCB is hideously expensive

nvidia can eat the cost as they make more money than god

partners cannot

EVGA saw the writing on the wall

[languid gulch]

100%

[twin dew]

They stopped when they saw that they would either need to end up with loss on the GPU division as whole or compromise on production quality and warranty in next gen.

The designs were ready, but when Nvidia told initial pricing, they saw they couldn't make it work.

When Nvidia was taking so large cut for that basic bundle.

Basically if Nvidia had to themselves pay that price, the FE cards would each be sold at loss.

Just for production costs.

[night girder]

But are you 100% sure about that? Do you trust MS?

The problem Chris Titus had, is when they removed Recall as a feature. It also bricked File Explorer.

I know, they made update video.

People need to learn to read.

"had", is pass tense no?

But my whole point is, if we take recall as a feature, we can atleast say it's ambigious to know what depends on it. What breaks. How it works etc.

[twin dew]

But just checked the documentation for those commands, it just tells if something is installed or not.

[night girder]

Yes, make sense since it DISM.

But then again, do you trust it?

As in, do you want to rather have recall on your system. And trusting MS it's not doing anything.

Or do you rather not have it on your system at all.

They could push it under here:

And it's not listed under there I believe.

But still shows up installed under DISM.

This is just messy.

[twin dew]

And it wasn't even installed in new install of Windows 11 Pro N 24H2.
Probably because of that N version stripped to fit EU requirements.

That didn't install Store shit either.

[night girder]

Fuck me, I whished they let us chery pick Recall tbh. Or atleast AI features.

Just push out 24H2 with a few optional AI features. But I guess co-pilot and co, are baked deep into the system.

And the only sane solution, anyone ever offers, is "Linux".

[twin dew]

And I was just able to uninstall whole Copilot in that Windows 11 N?

[night girder]

Who know what you broke 😉

[twin dew]

Just with normal Windows Apps control uninstall.

MS own uninstall process, not some weird stuff.

[night girder]

Oh, yeah. Don't trust that.

[night girder]

I don't even have co-pilot listed under my apps 🤨

Either I already uninstalled it (but there are references in my OS to co-pilot still) or somethings is off between versions (I am on older version).

[twin dew]

Not completely completely, the WebView part will be still left.
Just the browser UI bits will be removed.

[twin dew]

IIRC it is only separate app in 24H2 and newer.
Not in 23H2.

[night girder]

So it's not completely removed. Point remains; it's rooted into the system.

Depends on who you ask. Some people really don't want anything to do with co-pilot, Recall etc. Some don't mind.

That's not a good thing, but that's another discussion 😄 (chrome browser)

That's good.

[twin dew]

And RAM

With centralized updates for security issues etc.

[night girder]

I haven't seen thjis one.

Proof of human presence with Hello with Secure Bio? 🤔

https://www.youtube.com/watch?v=2SQQWJG3ULg

[twin dew]

That to use it, you need to use either Hello face detection or Hello fingerprint to login.

[night girder]

So many videos of it being hacked already rofl.

Nah, I am just making fun of Microsoft slide.

How they want to make it sound so secure.

[twin dew]

Yes, it can be hacked, but that you cannot even enable Recall without having TPM and either of those login types set up.

[night girder]

But when you just google the windows hello for example, you see it can get hacked.

So this one goes out of the window.

I mean, there is proof probably there is a human presence, but can be someone malicious 🤣

[night girder]

False sense of security?

[twin dew]

And if someone has enough access to enable it, they can already do the same stuff via other ways with malware implants?

[night girder]

1. Windows announces Recall feature.
2. Backlash from privacy and security community. Which bled into business community.
3. Windows adding false sense of security while not addressing privacy issues.
   (4. Since MS is unable to make the feature, while having privacy in mind, they gave option to just outright delete it.)

Job well done.

Privacy issues comes mainly from companies. And if you think, from a company perspective, yeah ... no.

The security and privacy community, are mostly people who use linux already. It's a minority, but they can be vocal. This got picked up by enterprise/bussinesses and they got worried. And then MS changed course.

Yup, but that's all after (major?) backlash.

And really, I wonder how much productivity the Recall will add to certain people.

If it's really worth all this back and forth between MS and community. And all the extra development power.

arstechnica.com/gadgets/2024/06/microsoft-makes-recall-feature-off-by-default-after-security-and-privacy-backlash/

it had a plaintext DB even

and a large plaintext database tracking everything that users do on their PCs

And in the end it didn't work. All the time rush I mean. And development shortcuts.

[willow pike]

strix halo workstation

looks cool as fuck

[night girder]

is that an aux port?

[willow pike]

wat

there's no price for it yet

yeah workstation tax I guess

some are saying the CCD is different from desktop zen5, which makes sense given it's not using basic MCM packaging

PS5 Pro is explicitly RDNA2, this is RDNA3.5

they didn't change the architecture for PS5 pro on purpose so they wouldn't need to recompile every shader

that's directly from mark cerny

you appear to be misinformed about the PS5 pro

I don't blame you

everyone misreported it

https://www.youtube.com/watch?v=lXMwXJsMfIQ go to 5 minutes in

this is the lead architect of playstation

there is no such thing as a more authoritative source

PS5 pro is RDNA2 with some additions, it is far from RDNA3

he also dunks on the teraflop numbers people reported

and this video only has like 300k views and didn't get big headlines

so people still think PS5 pro is RDNA3

truth is a thing of the past

n- no

[twin dew]

RDNA 3.5 is specific IGP gen, used in many of the new AMD APUs

[willow pike]

that's exactly what he said it isn't

[twin dew]

Upgraded RDNA3

But unless some actual data comes out, I would probably count that PS5 Pro as RDNA2.5?

RDNA2 upgraded with RT improvements from RDNA 3.0/3.5

Much easier to just replace the RT subunits than try to keep binary compatibility with full change.

Until Sony or AMD officially says something about the exact tech, or someone does deep level silicon comparison, cannot know.

Between old PS5 chips, RDNA2, RDNA3, RDNA3.5, PS5 Pro chip

RDNA3.5 is official AMD arch.

With separate instruction set documentation etc.

Not used on any dGPU products, only in APUs currently.

Used in both Strix Point and Strix Halo

Point was, RDNA3.5 is RDNA version used in AMDs own Zen5 mobile CPUs IGPs.

Those Strix Point and Strix Halo are the chip names for those AMD Ryzen AI (Max) 300-series chips.

[willow pike]

you need to recompile shaders on a driver update, let alone a whole new GPU. if you can run the same shader on PS5 and PS5 Pro, they are fundamentally the same arch

[verbal raft]

i think the RT is from RDNA 4 not 3.5 IDK though

[verbal raft]

it would be cool if AMD gave more details on amethyst

on the 21st

[tribal kraken]

www.gigabyte.com/Graphics-Card/GV-N5090AORUSX-WB-32GD#kf "Patented leak detection"

Clock speeds TBD and no buy button 😢

Actually the buy button links to 2 Finnish webshops 😄

[willow pike]

https://youtu.be/JV83G744yrI

did he really say the 4090 was a great investment before announcing the 5090

[twin dew]

And basically that comparison was that 5070 will be half the performance of 4090.

[night girder]

Oh, lets take snoop dog word as value.

[twin dew]

Nvidias own slides.
But it was with 4090 with 2x FG and 5070 with 4x MFG

[pure karma]

I got a phone but i literally dont need one so its basically just another burden i have to deal with…

i guess its still a free phone so thats one less expense later on

[night girder]

Not what I am on about.

Do you know who snoop dog is?

[twin dew]

Ah, not going to even open that video.

Yes, and large part why I'm not going to open that video.

[night girder]

Right… that was my point.

Regardless of what slides snoop is looking at.

Its snoop dog.

Btw havent watched this one. But that thumbnail makes me curious.

https://m.youtube.com/watch?v=_YXbkGuw3O8&embeds_referring_euri=https%3A%2F%2Ftweakers.net%2F&source_ve_path=Mjg2NjY

28 fps to 242 fps

[twin dew]

Because DLSS4 brings ability do more than 1 real, 1 fake.
To different amounts of fake frames between real ones.

At least to 1 real, 3 fake (4x)
And possibly to 1 real, 7 fake (8x)?

Maybe up to 1 real, 9 fake for 10x?

[tribal kraken]

I blame console players 😄

[twin dew]

Ok, that was DLSS completely off vs. DLSS4 on on unknown settings.

Ah, it was in the description:

Captured on GeForce RTX 5090, 4K, RT Overdrive, Max Settings, DLSS with Multi Frame Generation (4X Mode), Ray Reconstruction, Super Resolution (Performance Mode)

So very shitty actual render resolution from that Performance mode DLSS, and then 3 fake frames per real frame.

[tribal kraken]

DLSS never really looked good for me, tested in few games on different profiles

F1 2023 was horrible exprience, even for weekend arcade racing with friends

Waterblocks for FE cards going to be interesting, but most likely will not wait for them

[willow pike]

it's not actually snoop dogg looking at anything lmao it's niktek meme

[willow pike]

eIGHT??????????

SEVEN AI FRAMES?

[twin dew]

Was just 4x in that video.
The other 4x improvement was from lower resolution from Performance DLSS upscaling.

Just looked at the FPS difference and it was almost 9x

But combination of much lower render resolution and then 4x frame gen explained that together.

IIRC DLSS Performance was 25%?

[willow pike]

yeah I can't find anything about more than 3 generated frames

I was about to go vomit

[tough owl]

https://tenor.com/view/getting-a-dell-gaming-pc-dude-gif-14395733393982085552

It’s back. The meme is back

[night girder]

So, is the future is just more and more fake frames?

[twin dew]

To Nvidia it is.

Bigger Number More Betterer!

[night girder]

Fml.

This is what we get for being so focused on FPS

[willow pike]

some poor schmuck is gonna buy a 5070 12GB, turn on all the DLSS, get 90FPS after 4x frame gen, see the frame rate is high and feel good, but then wonder why it feels so shitass

and start to notice everything looks smeared to shit

[night girder]

You forgot a part.

[willow pike]

and then a game comes out that 12GB will not do for and it'll tank

[night girder]

That poor schmuck will go to forums to complain graphics look off.

[willow pike]

most won't but some will

90fps with dlss4 is actually 22.5fps

I don't give a shit what reflex is doing

that's not enough

[jagged snow]

And blame the devs

IMO the move to dlss and framegen feels like the move away from disks to streaming

Worse quality and you're paying more for the chance to get steamrolled by a greedy corpo

[verbal raft]

also : reflex 2 wont help latency with DL-FG

[willow pike]

not anymore than it already is

[verbal raft]

anyway i think we are gonna se a few people upgrading from the 4090 to the 5070 BC they think that the 5070 performs like a 4090 without MFG

[willow pike]

poor bastards

[verbal raft]

yeah NVIIIIIIIIIIIIIIIIIIIIIIDIA marketing is gonna get a few people

satisfactory is on there

[night girder]

Yeah, I can see that.

Funny thing before streaming, they tried to push blurays instead of dvds.

[jagged snow]

I liked bluray

[night girder]

So we went more quality on bluray, to fuxk it all streaming

[jagged snow]

Yep

Bluray is still the best quality you can get in-home

[night girder]

You can also dee YT quality go backwards.

Unless you pay I guess.

[jagged snow]

That too

[night girder]

Sorry for typos. Am on phone.

[jagged snow]

All good, me too

[night girder]

Quality vs disk space will always be interesting balance. Flac vs MP3 (128, 256, 320 kbps I think?) and variable bitrate.

And then apple had their own stuff for music. Forgot extension. I think .mov for video.

[jagged snow]

aac for audio?

Or was Apple wav

[night girder]

Yes! And same for pictures btw.

All media files are either optimizrd for disk space or quality.

Raw image vs jpg.

[night girder]

I think they changed. Who pushed mp4? Wasnt that also Apple?

[twin dew]

MP4 is just container format, not video or audio codec.

[night girder]

Quicktime makes mp4

[twin dew]

MPEG consortium

[night girder]

I think thats my association

[twin dew]

Lot of stuff uses MP4 containers for video or audio.

But then the actual thing is what video and/or audio codecs are inside that container.

AVI, MP4, MKV etc. are container formats.

[night girder]

Oh, well I get movs, mp4, aac, wav, mp3 all from apple devices

[twin dew]

That don't specify what codecs are used for actual compression.

[night girder]

I like mkv.

[twin dew]

Very versatile, but with just one video and one audio stream doesn't differ from MP4 in any meaninful way.

The bonus is able to include many, and subtitle streams etc.

[night girder]

On certain websites, you will find a lot of mkv movied

I havent seen mp4 used widely 🏴‍☠️

[twin dew]

Yeah, but point was that you need to differentiate between container formats, and codecs.

[night girder]

Ok. But we were just talking quality vs disk space.

[twin dew]

And container doesn't come into that in any way.

Only the codecs

[night girder]

Not the finer details of codec vs containers.

[twin dew]

So for example MP4 doesn't matter to talk about disk space vs. quality.

[night girder]

No matter how you look at it, quality = more data.

E.g. 100 pixel picture vs 1000000 pixel picture

[twin dew]

As it is just a container format.
The codecs used for the actual video and audio data are what matters.

[night girder]

Its true for gaming, and all sorts of media.

I can understand that streaming companies do it.

Less data to stream etc.

But games… i dont know. Maybe its possitive for people who have low end hardware?

[twin dew]

And it is actually triangle, encoding time, size, quality.
With limits in each direction.

[night girder]

So they can still play games with 1 real frame and 3 fake ones.

[twin dew]

Slower encoding allows for better quality at same size, up to certain point etc.

[night girder]

Well encoding is not what Ibdo.

[twin dew]

Was about video or audio compression, not the original talk about DLSS etc.

[night girder]

I watch movies. I listen to games. I look ar photos. And I play game. But I dont encode anything afaik.

I listen to music

[twin dew]

But when that movie or audio release is made, there is encoding step at some point.
Unless it is completely uncompressed like basic WAV files.

[night girder]

Agree

My point is. That is not a concern for everyone. Its niche community

Way more people just use media and dont provide it 🙃

[verbal raft]

UHH

unexpected

[night girder]

Also isnt encoding going faster and faster with each gen of hardware?

I think I read nvidia just added more encoders to their gpu

2 encoders and 1 decoder if not mistaken

[glossy glacier]

"I watch (encoded) movies. I listen to (encoded) music. I look at (encoded) pictures. And i play game(s with all three)"
Hows encoding niche again? You may not encode yourself, but you enjoy the benefits in everything you do

[night girder]

🤦‍♂️

My dad never encodes.

My mom never encodes
Neither does my sister

My niece never encdes.

How is this practice mainstream?

[twin dew]

They do every time they use any kind of digital camera for example.

[mental oriole]

That's 💩

[twin dew]

Just with settings fixed by the equipment or software manufacturer in most cases.

[night girder]

They don’t do it conscience, with knowing they are doing it.

[glossy glacier]

Any picture send online. Any voice message, Any Video uploaded

[night girder]

Jeez. Nitpickers. Fine. Encoding is super important.

[twin dew]

But mainly the point was that it isn't just size vs. quality.
But that there is third dimension, on how much processing power is being thrown at the encoding process.

[night girder]

More important than quality and diskspace. You win.

Its super important. More than Ai.

[glossy glacier]

everything is more important than ai

[twin dew]

You can get large size files with crap quality if you aren't willing to use the processing power.
And after certain point more processing will give very limited gains.

[night girder]

And I am adding ???? to that extra dimension in importance.

Critical thinking 🤔

Quality vs Diskspace vs Encoding vs Decoding… how far you want to go.

[twin dew]

It is very important for YouTube etc.
Where is the balance between processing power needs vs. low enough size, vs. still usable quality of the result.

[glossy glacier]

How long it takes your Smartphone to encode a picture and how much battery i uses is not important?

[night girder]

In the end most users just see 2 things. Quality and space (how long download takes).

[twin dew]

Same with cameras and video cameras, those usually go for less processing with ok quality with large file sizes.

Which is why recompressing on computer can give still same quality with 1/4th file size.

[night girder]

Not ad important ad qualiry and diskspace.

[glossy glacier]

So 2 minute encoding before you can send it is fine?

[twin dew]

And that isn't what the manufacturers are actually doing.
Most are optimizing much less for diskspace than fast encode time.

[night girder]

Ok, but you do know the difference between the average joe and tech nerds?

Go out on the street. Ask 100 people to order in importance: quality, diskspace and encoding. How important it is to them. My feeling. Encoding will be 3rd.

[stray badger]

lumen can use hardware RT so i guess their arent wrong then. also dlss

[glossy glacier]

The Joe wonders why sending a picture takes ages.
The nerd nows why.
Both are anoyed

[stray badger]

because nobody knows what that means?

even among the film nerds i know, me and only one other person know what the different codecs are

[twin dew]

And encoding happens always.
That third optimization dimension was the processing power needed for it.

[night girder]

I actually never had issues with encoding of my phone pictures. But I do have when I need to send 100mb 🙈

[twin dew]

So on fixed hardware: Time vs. Quality vs. Space

[night girder]

Indeed.

I wonder wtf you all are encoding

That you all get so triggered 😂

[glossy glacier]

Time vs power vs quality vs space?

[stray badger]

blackmagic RAW -> DNxHR -> h.264

[night girder]

i never had issues with it.

[stray badger]

i would rather use AV1 but literally nobody knows how to use AV1 properly

[twin dew]

I did say on fixed hardware.
Power comes when you can trade time for more powerful hardware etc.

[glossy glacier]

Ah, fixed (amount) of HW

[night girder]

How long does it take?

For you.

[stray badger]

the encode to h.264 takes about a minute and a half

for a 5 minute video

DNxHR idk, resolve is also rendering so it takes a lot longer

[night girder]

Ok, movies are different story for sure.

[stray badger]

RAW decodes quite fast tho which is nice

[night girder]

But when I did photography, I never ran into encoding issues.

[glossy glacier]

Is RAW even encoded?

[night girder]

It was all fast.

[stray badger]

photos dont take as long

[night girder]

Its raw.

[twin dew]

Usually lossless compression.

[stray badger]

well, yes, any format is still encoding, but raw is not compressed or anything

[night girder]

And I had giant projects of 1Gb for one picture.

[stray badger]

yeah, RAW

[night girder]

Nah photoshop shit.

100 layers etc + raw.

[stray badger]

most of your speed at that point is taken by rendering. encoding a single PNG doesnt take very long

[night girder]

I think thats where my surprised reaction comes from.

[stray badger]

alls ima say, is that i wish my phone had a better h.265 encoder. Its not the greatest quality compression

[night girder]

But its fast

[stray badger]

which is why im happy to have switched to an actual camera

[stray badger]

not really, thats the problem

[night girder]

Rofl 🤣

[stray badger]

its lowering quality in order to meet performance standards

better encodes take longer

[night girder]

Not when you got better hardware v

[stray badger]

if i crank the quality in handbrake, it will take longer to encode

[night girder]

Shiiiit

[stray badger]

that is my point is that the hardware should be better

so it can encode at the same speed, but better quality

[night girder]

Want my old canon 5D? 😂

[stray badger]

cinema pocket 4k seems to be just fine

[mental oriole]

Translated to "frames"
~50.97ms -> ~19.7 fps (FrameGen x2)
~55.50ms -> ~18.0 fps (FrameGen x3)
~57.30ms -> ~17.5 fps (FrameGen x4)

It's still pretty "unusable" 😄
Do they have comparison to previous generations?

[stray badger]

oooooooof

[verbal raft]

the ms is latency

[mental oriole]

I know

I'd rather call it frame time, but ye same thing.

[tough owl]

Who let nvidia make the 5090 575W

[verbal raft]

are those the fps amounts before FG?

[twin dew]

Latency from click to response?

[verbal raft]

no they are PC latency

[twin dew]

Complete measurement which is that click to photon one?

[verbal raft]

its software

[twin dew]

And incomplete then leaves most of input latency out, and output latency out.

[verbal raft]

reporting

[mental oriole]

I just simulated a frame time of those numbers in my engine by setting a forced update time to that from the average.
If they showing ms measurements, they need to compare to the old generation.
Like if old generation hw is 150ms for frame gen x2. Then sure it's good. But what is the comparison?
This only compares the different frame gens on the same hw, which doesnt really say much.

It only says x4 is marginally slower compared to x2

[verbal raft]

got it thanks

[mental oriole]

Are those numbers from nvidia directly?
Because if so, their data set is misleading.

To be fair it's full RT at 4K so it's not too bad, but I wouldnt call it playable fps 😄

[twin dew]

Yes, Nvidias own YouTube, with best case video, very dark and slow moving without any major sideways/turning components.

[mental oriole]

Thanks ❤️

[pure karma]

Themselfs

[verbal raft]

so true LOL

[pure karma]

On another note Im at a standstill for what i want to upgrade next

I want a cpu upgrade but getting a 14900K would be dumb and a swap to am5 wouldent be worth it so im at a standstill

[visual tree]

I don't use Facebook but this must be a joke:

So, Instead of using fact-checkers to check if 2+2=4, they will rely on the community to do fact-checking lol

Good luck with that 🤣

I mean, fact-checkers can be biased but leaving fact-checking to the public is much worse

[jagged snow]

It's worked semi-well on x, but that's a wildly different culture

[verbal raft]

5090 samples are out in the wild now

[night girder]

"just fine" is very subjective.

[night girder]

Found source:

RTX 5090: 3x encoders, 2x decoders
RTX 5080: 2x encoders, 2x decoders
RTX 5070 Ti: 2x encoders, 1x decoder
RTX 5070: 1x encoder, 1x decoder

Dan voor de laptops-GPUs:

RTX 5090: 3x encoders, 2x decoders
RTX 5080: 2x encoders, 2x decoders
RTX 5070 Ti en 5070: 1x encoder, 1x decoder

[willow pike]

what the fuck

[verbal raft]

also...

LOL

[twin dew]

It was DLSS-FG (2x) vs DLSS-MFG (4x).
So 5070 is half of 4090 or so.

[verbal raft]

yeah

[twin dew]

On Full-HD, not 4K.

[verbal raft]

that is the only way it makes sense

[twin dew]

It was in the actual slide.

DLSS Performance and otherwise same settings, but with FG on 4090 and MFG (4x) on 5070.

So 1+1 on 4090, 1+3 on 5070.

So base rendering on 1920x1080, and then doubling of frame rate for 4090, and quadrupling of frame rate for 5070.

[jagged snow]

That's some pretty sketchy marketing

Curious to see if they get blasted for it

[verbal raft]

oh

[verbal raft]

no they arent

at least not in the discords i've been looking

(other than this one ofc)

[twin dew]

Because almost no-one looks at that small print that hides the actual details on what the graphs show.

[verbal raft]

HUB did call out their BS

[verbal raft]

yeah

[jagged snow]

Moreso thinking once perf reviews drop

[willow pike]

https://youtu.be/ORXoOKND1Tk

[verbal raft]

yeah im curious on what the non AI fluff perf will be

[willow pike]

Sup GAYMERS it's ya boi ryzenMAAAAN comin' atcha LIVE with a Graphics Card Recommendation, pay attention, this is which one you buy, geforce or radeon bros:
||don't||

[verbal raft]

"ONLY GAYMERS KNOW THAT JOKE"

😂🤣

[willow pike]

i retire from pc gaming

i am not buying a new pc

i am done

[verbal raft]

😂🤣😀😄🙂😕🙁☹️😣😖😫😩😭😭😭😭😭😭

[jagged snow]

I was originally planning to upgrade this cycle

I don't think it's happening

There's no reason for me to

[willow pike]

"what graphics card should I buy" get a new hobby

go get a playstation

or steam deck

turn your settings down

[verbal raft]

🤣

well linus is NOT calling them out for their BS

https://www.youtube.com/watch?v=3a8dScJg6O0&t

[twin dew]

Because the money Nvidia pays is more important than integrity to him.

[night girder]

https://tenor.com/view/tyler-james-williams-wow-whoa-shocked-shocked-face-gif-20698709

[twin dew]

And Nvidia would absolutely stop any ads or partnerships with LTT if he said anything.

[verbal raft]

nothing new

[night girder]

and LTT needs to pay a whole company. (his words once)

[verbal raft]

https://www.youtube.com/watch?v=FGW3TPytTjc

[night girder]

he has a responsibility over other families. So money will alway win, if you apply that logic.

[twin dew]

So doing lot of bad decisions and then expanding too much too quickly should be accepted as reason?

[night girder]

And it's not my logic. It's what he told us in one of his videos once.

[twin dew]

But the whole thing is entertainment, not reporting.

[night girder]

Exactly.

[twin dew]

Just that lot of people cannot tell the difference 😩

[night girder]

I had some reaction about "I left adobe behind..." clickbait video posted here a while ago.

It was entertaining, but had little value to me.

The video itself is top notch editing skills. And fun to watch.

But the fact they make it every x year that they leave adobe 🤷‍♂️

[verbal raft]

is it just me or would it be quite funny if that 16-pin started burning behind him?

[night girder]

I want a GPU to fall off while someone is recording.

Just because of the massive weight some of them have without a bracket 🤣

[visual tree]

Linus drop tips

[night girder]

Speaking of support brackets.

Anyone experience with these?

[verbal raft]

hhhmmm

[night girder]

Because my current support bracket really sucks.

[visual tree]

I was thinking of 3D-printing a superhero figure as a support bracket

[night girder]

It's not very precise.

[night girder]

I need something more precise.

[verbal raft]

is that snutt?

[night girder]

This is the one I have, so I only have the holes to pick from.

[visual tree]

Like this

[willow pike]

I got an LTT screwdriver and now I wanna de-brand it

[night girder]

So my GPU is still sagging a bit.

and if I put it a hole higher, then my GPU is bending upwards 😭

[visual tree]

I guess you could say he is carrying the weight of the GPU on his shoulders

[night girder]

Also, look at the tiny little clip that hold the GPU.

It's all resting on tiny piece of metal.

While I think something like this, you could put in the middle of your GPU more or less.

[visual tree]

I had sagging issues with gtx 970 but not anymore with rtx 3060 ti

[night girder]

Then all I need is this thing:

I wait for Baldur to wake up. He will probably tell me if it's a good idea or not.

Because with the fans on the bottom, there is not so much choice I have to put sag bracket.

[visual tree]

No turning on your computer untill the civil engineer confirms the GPU is perfecly leveled

[verbal raft]

guys

the cinebench accelerators got beat

[night girder]

I will probably discover that my floor isn't level. So the table on which my PC stands isn't level. So my GPU can't be level.

Or it can be level, but then all the rest will be off 🤣

top of the cherry will be that I will break my GPU while leveling it out.

wait... cherry on the cake will be 🤣

[visual tree]

I don't think anything in my house is leveled thanks to the civil supervisors 40 years ago who didn't care about anything lol

[night girder]

same here, but I am not complaining. Have a roof above my head. Food in my belly.

[visual tree]

I have issues with walls and door frames that are not straight, not to mention the idiot who tought it was a good idea to cover phone cable with cement instead of installing a tube so I can pull an ethernet cable through it

And if there is tube inside the wall, it's not going straight and looks like a snake lol

which made me damage the fiber connector on one end of the cable when preparing the house for fiber in advance by pulling fiber cable through conduit

[night girder]

ugh that sucks

[visual tree]

Not that today's handymen are any better. The good ones have moved to Germany and the only ones who are left here are self-trained ones who would probably do worse job than myself

[verbal raft]

🤔

[visual tree]

Gone are the days when you could buy a top-tier GPU for €600 ☹️

[soft bloom]

Or just get a ball and test if it rolls

Also should work with some cylinder to test in 2 directions

Theoretically even water on repelling surface

[tough owl]

TIL you can run llms in a web browser

https://github.com/mlc-ai/web-llm-chat

[night girder]

That's cool, what's the benefit over running in a docker container?

WebGPUNotAvailableError: WebGPU is not supported in your current environment, but it is necessary to run the WebLLM engine. Please make sure that your browser supports WebGPU and that it is enabled in your browser settings. You can also consult your browser's compatibility chart to see if it supports WebGPU. For more information about WebGPU support in your browser, visit

https://tenor.com/view/trombone-pusheen-musician-instrument-gif-11434220432919976776

[tough owl]

Oh yea you need webgpu

[night girder]

😇

[tough owl]

Enable it

[night girder]

NotSupportedError: WebGPU is not yet available in Release or Beta builds.

It's not supported for firefox 😦

updating firefox.

Didn't help.

I read it should work on firefox nightly.

[pure karma]

this is a problem i wouldent have expected to run into

my gpu is starting to be unstable under low loads because it gets TOO COLD

[verbal raft]

https://tenor.com/view/cat-cats-what-wut-crazy-gif-20900490

[river mountain]

https://tenor.com/view/doubt-press-x-la-noire-meme-x-button-gif-19259237

whatever the problem is i don't think it's that

[pure karma]

either that or its another incoming vram failure

because last time i experienced this was weeks before the 1660 blew

[verbal raft]

just the avarage activity on the optiscaler discord

[night girder]

Thinking the same.

[verbal raft]

this is true though

[stray badger]

its actually a pretty good game tho

[pure karma]

yea but also it highlights every flaw with FG and DLSS extremely well

so its a nice double edged sword

[verbal raft]

i havent played it😭 ....just that its like the one game where every NVIDIA feature is at its peak and every AMD feature is at its worst

[stray badger]

FSR in cp2077 is not that bad

also my 7900 GRE runs it pretty well (no RT)

[pure karma]

FSR and XESS outperform DLSS by a good amount in CP2077 but they look like trash so it depends what you value more

[tough owl]

I swear you need a prebuilt from Dell with as many warranties as they can force feed you

[pure karma]

isent that exactly what i had

[winged valley]

I know the whole dont ask to ask thing but im not trying to start an ark thread in satisfactory discord

If anyone in here hosts/has hosted an Ark Survival ASCENDED dedicated server, preferably on windows, please dm me bc I need help figuring out why it wont connect to the server list

[safe trench]

saying "nvidia is lying about 5070 performance for using dlss and mfg!!"

its the same thing as saying

you cant use features you paied full price for

and i am SURE they use it to protect amd

[pure karma]

well yea just i think the main argument is its not fair grounds because the comparisons arent under the same conditions but yea i absolutely agree the argument is dumb

[gilded helm]

Satis was in one of Nvidia's slides. That gives me a little hope we can use DLSS and power lines won't look like garbage.

[soft bloom]

Condition is that you use product at its max capabilities.

That is a totally fair way of comparing

[languid gulch]

we'll just have to see. yanking 75% of its frames out of nowhere had better be incredible

[twin dew]

Or just check the distance between PCB and case floor in two positions with ruler or tape measure?
Or just by eye?

[languid gulch]

personally i prefer the "like-for-like" direct comparisons as a primary judge, with the "other goodies" of newer gen capabilities done as a supplement

[languid gulch]

are there any games that people with 4090s play just because they have the hardware to max it out, or is it more of a "it just runs it with less effort" kind of GPU for games

[dire igloo]

It's a mislabeled Titan, the only reasons people get a 4090 are "I need it for XYZ productivity thing" and "I had the money for it"

[languid gulch]

doing some gen to gen comparisons, and it's kinda interesting

at least on TPU's list

4090 is 26% ahead of the 4080 non-Super

Titan RTX is 22% ahead of the 2080

3090Ti is 32% ahead of the 3080

it's definitely a jump up, but that's only because it's current gen. i bet that Radeon in 2 gens will match or beat it

but i do agree 100% that it's a mislabeled titan

[twin dew]

So anyways, 5070 (without architecture improvements), should have about 37% raw performance of 4090.

And 5090 should have about 127% of 4090.

Actual differences will be smaller, as that didn't take into account inefficient usage of CUDA cores, power limits, architecture changes, VRAM bandwidth etc. effect.

[verbal raft]

R.I.P.

[bronze jasper]

They're drama bait YouTubers that talk a lot without much to show. I've never been impressed and figure it's all marketing.

[willow pike]

is that the channel the yt algorithm is trying to shove down my throat

yes it is

[bronze jasper]

He says junk like "we're the only studio trying to do optimizing" and shits on every other dev constantly. Yet they've never published a game. It's like the spoiled kids on the playground that think they're the best because momma always told em so

[willow pike]

oh wait they're developers????

[bronze jasper]

He has slivers of truth seeded in his videos but it's not enough

[willow pike]

that's heinously unprofessional

[bronze jasper]

It's all algorithm bait and ego boosting in my opinion

[willow pike]

game engine grifters

[bronze jasper]

I wouldn't doubt he bots his views and sub counts

[willow pike]

was only a matter of time I guess

[bronze jasper]

It's just one guy and he acts like he runs a full studio

[verbal raft]

it is a small indie studio

[verbal raft]

so when did he ever lie?

[bronze jasper]

Have never published a game

[verbal raft]

that is a non-answer

[bronze jasper]

Talking like he's the only guy taking optimisation seriously is a big giant lie.

Don't simp for him. He's a snake oil salesmen

[twin dew]

Most I have seen have been on large part intentional misrepresentations and intentionally leaving all nuance out etc.

[bronze jasper]

The game Dev streaming community does not want grifters like him around

[twin dew]

He has his own absolute view and then he fits anything he makes to that view, and anything that isn't what his view is is absolute bad in every way.

[soft bloom]

note to future self: don't group dates by day as category

[twin dew]

Basically the videos aren't objective.
They are to reinforce his pre-existing view.

[soft bloom]

are you talking about Jonathan Blow?

[twin dew]

That Threat Interactive stuff

[soft bloom]

oh

[twin dew]

Like the "Nanite is horrible for performance" video etc.

[soft bloom]

yeah YT gave me some videos and I am kinda sceptic about if it will be used in practice, but demos are promising

[twin dew]

Yes, badly done Nanite is horrible for performance.
But well done Nanite only has minor impact, and lowers the dev workload a lot.

[bronze jasper]

It's a lot of toxic hate bait yellow journalism tabloid style dialogue

[twin dew]

And that dev workload is also very important thing to get anything out.

Making small optimized demo is in the end pretty easy.
Scaling that to whole game is completely different ball game.

[bronze jasper]

Yeh he talks a big game but has Nada to show

And to use a death from a loved one as a shield now .....

That's a huge lack of class

[willow pike]

lmao

[bronze jasper]

Go morn. Stop posting toxic shit on Twitter

[verbal raft]

[willow pike]

https://tenor.com/view/nvidia-jensen-huang-the-more-you-buy-the-more-you-save-keynote-2018-gif-12315008507302833354

[night girder]

you all reinventing the wheel

[twin dew]

I just didnt understand what the point was between water level and GPU PCB droop bracing install.

[night girder]

most gotten lost in the chat then

[night girder]

Point I was making, this thing is highly inaccurate. And kinda a pain to work with.

Either my gpu is sloping down, or upwards. You can't tweak it. So I was thinking of getting another one.

[twin dew]

Also supports from wrong side (slot side)

[night girder]

Something like that looks more precise.

Just wondering if there was anything to keep in mind when using the one I just posted above.

[twin dew]

It has infinite adjustment, just that it cannot be adjusted while GPU is in place, and doing small ones is real pain.

[night girder]

Which one?

The ones with the holes?

[twin dew]

The MB hole mounting one you posted originally.

[night girder]

the MB hole mounting one can't be infinite adjusted.

[twin dew]

I have plans to get one as extra to far side support already in place.
But only as extra.

[night girder]

You pick a hole and that's it 🤣 It's mounted on the motherboard.

[twin dew]

It can, the holder side screw hole is long so you can move it then tighten the screw down

Where that adjustment range is large enough that it covers whole range when you combine with the screwholes in main part.

[night girder]

Trust me, I spend a lot of time with it.

You picka hole, put in GPU and pray it's good.

You pick another hole, put in GPU and pray it's good.

My GPU is sloping slightly down. If I pick other hole, it slopes upwards.

[twin dew]

That can be moved upwards from that "natural" position by that amount, then tightened.

To get infinite adjustment.

[night girder]

Mine doesn't have that 😉

I don't even know where I got it rom. I though it was with my MB. But it's not listed in manual.

[twin dew]

Then you got even worse copy than the one in the pic you attached.

[night girder]

Trying to find exact picture.

[twin dew]

Original is by Lian Li

For that model.

[night girder]

This one!

[twin dew]

https://lian-li.com/product/gb-001/

[night girder]

Ok, that's a very small one. But look at the holes. You can move the other piece.

[twin dew]

Yes, that still has that slit on the L-part to get more adjustment range.

[night girder]

But it's still a pain in the arse to adjust.

[twin dew]

Yes.

[night girder]

and once the GPU rests on it.

It lowers a little bit too 🤣

nah, I am getting a better one. Fuck this 😄

[twin dew]

Yeah, like I said, that is on wrong side of the GPU to matter much.

As it is right next to the slot.

Which already provides support.

[night girder]

Agree, only put it in because it was in the box 😄

"Better than nothing".

And it does help.

But not efficiently.

omg, this even comes with a water marker.

🤮

[twin dew]

Yeah, lot of ones with RGB.

And that actually has limited adjustment without extra pads or similar.

[twin dew]

And if that happens, then the screw wasn't tight enough and it isn't giving any support, until the L-pieces slots top hits the screw.

And it might not be even possible to tighten the screw enough without stripping threads.

[night girder]

No.

The rubber padding.

[twin dew]

Ah.

[night girder]

So, this is what I got from Lian Li. Tiny pieces.

Not easy to work with. It was with the case, but still, pretty much sucks.

I readjusted my GPU 3-4 times before I gave up.

[tribal kraken]

Dont let anyone see your GPU, problem solved

Motorsport and eSport should be the same, equipment malfunction is a lose, and dont show what you have under the hood 😄

[tribal kraken]

Must be the same spirit level that you get with TV wall mounts. That style support was supllied with MSI 3090 Suprim