[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: da.png (289 KB, 668x617)
289 KB
289 KB PNG
Where have you been when NVidia was kill?

>NVidia RTX 6000
3nm

>AMD UDNA
2nm
>>
>>106959819
>nm
marketing nonsense that no consumer actually benefits from
>>
>>106959819
>2nm

so 2 being lower than 3 means its a flop. Nvidia wins again.
>>
>>106959820
I know bro, I can open Wikipedia and read the line that says it's a marketing term that doesn't refer to any specific length of anything.

Nonetheless, the smaller the better.
>>
>>106959819
Do they have an answer to nvidia's domination with CUDA yet? If not then it doesn't matter for the larger gpu market.
>>
amd gpus suck ass lmao
>>
I don't know what that means.
>>
>>106959825
you just dont get it, do you
>>
>>106959819
in english, shill-kun
>>
File: 1740029273928374.jpg (17 KB, 268x284)
17 KB
17 KB JPG
>>106959819
>it's another "AMD's architecture is so dogshit that Nvidia can sandbag onto an older node and still match or win on perf/W" episode
>>
both have gone full ai slop and don't make gpus anymore
you'll either upscale from 600p or 500p
make your choice
>>
In english doc!
>>
>>106959819
*where were you
>>
>>106959825
rubin = ai shit (token coding and generative shit)
mi450 = ai shit (oracle's 50k supercluster)
in other words: ai shit for corps burning through billions (soon to be trillions)
>>
File: 1332735103272.gif (519 KB, 1006x575)
519 KB
519 KB GIF
>>106959820
all i know is that when nm number goes down, my temps and power usage also go down
funny that
>>
>>106959819
>indian twitter thread
really bros
>>
>>106959819
>amd last ditch effort to be relevant in AI, overpays for 2nm advantage
>amd poojets thinking their gamer scraps will also be 2nm
kek
>>
>>106959819
you can even see the beard on this tranny lmao, he's like the clone of the nvidia chink
>>
>>106959819
oh boy, i can't wait for
>amd card that's 85% as good as competing nvidia card for $100 less!
this strat can bring amd to 3% of the gpu share from 6%!
>>
>>106959833
wrong
power usage has gone up
>>
>>106959833
You're a fucking retard, how do you people function outside of the internet?
>>
As soon as they make the superior product ill buy it
>>
>>106959835
I don't know what's more pathetic. Falling for another AMD lie or falling for another 12 GB overpriced garbage "mid-range" card from nvidia.
>>
>>106959819
your nm don't matter.
prices matter.
>>
>>106959821
quarter pounder vs 1/3 pounder all over again
>>
>>106959819
>no cuda
in to the trash it goes
>>
>>106959844
>muh RedditUDA
What are you homos doing with it?
>>
>>106959819
>nvidia was kill
I do remember them selling somewhere like 20x more RTX3000 series than AMD sold RDNA2 cards.
AMD was on real TSMC 7nm
nVidia was on Samsung 8nm that was really 10nm.

I think they'll do just fine.
>>
>>106959845
Making AI touhou GFs.
>>
>>106959845
generating lewd images
>>
>>106959845
>reddit
Loves AMD...
>>
just give me 48gb vram and ill be happy
>>
>>106959850
You can already get that.
>bbbut I don't want to pay extra for pro cards
Too bad.
>>
File: 1742688813277104.png (1.6 MB, 1320x1320)
1.6 MB
1.6 MB PNG
>>106959851
im not spending 10000 dollars om 48gb of anything
>>
Isn't nvidia now backed by the zogverse? There's no reason to put any effort into such childish hobbies like gayming.
>>
>>106959852
The rest of us don't want to pay $5000 for gaming GPUs that will be scalped by china AI fags.
>>
File: 1732183592423348.jpg (10 KB, 259x194)
10 KB
10 KB JPG
I'm gaming
>>
File: 1751342154690283.png (359 KB, 651x523)
359 KB
359 KB PNG
just buy a switch 2 and feel like you're 12 again
>>
>>106959819
This happened (even worse) when it was Ampere and RDNA3 yet Nvidia still BTFO out of Radeon in sales
>>
>>106959847
>>106959848
Thought so.
>>
>>106959846
>>106959857
Oh, really? Sorry then.
NVidia bros, we're so back.
>>
>>106959819
The first generation yield on this node is unlikely to compare favorably to more mature nodes you find under the banner of 3N. It will also likely not have the numbers for larger consumer audience.

It is better suited for halo products like MI450. As well as chiplet architecture. Nvidia still seems committed to monolithic.

I do hope AMD doubles down on matrix multiplication and ray acceleration though. This will be necessary to be competitive against Nvidia.
>>
>>106959856
nogames and completely different function.
>>
>>106959861
>nogames
it already has 2 contenders of game of the year lmao, and it's getting another one soon with the return of Metroid
>>
>>106959860
AMD ditched chiplets on GPU.
They aren't coming back.
>>
>>106959862
The baby kart game? Or do you mean neo palworld? I wouldn't call those GOTY material
>>
>>106959819
And nvidia will still win because it will be more efficient and just as, if not more powerful
>>
File: 1733326360556674.png (176 KB, 480x480)
176 KB
176 KB PNG
>>106959864
mario kart world is fucking revolutionary and donkey kong bananza is one of the best platformers of all time. it's clear you're never even played them.
>>
>>106959856
>and feel like you're 12 again
because you're buying a fisher price tablet?
>>
>>106959866
LOL
>>
>>106959863
Chiplets might come back, if AMD can somehow solve the bandwidth issues. It would give them a huge manufacturing advantage, after all.
>>
>>106959863
>AMD ditched chiplets on GPU.
I will take it on your word that this has happened. I believe chiplet are still used for MI450. Correct me if I am wrong.

However I disagree with this:
>They aren't coming back.
Chiplet will be mandatory in the era of HighNA EUV as reticle size will shrink and large monolithic dies will not be possible.

Also I believe AMD is still forward looking to chiplet. Their last MI300 chip featured fan-out interconnect, as oppose to traditional SerDes for the sake of better chiplet interconnect.
>>
>>106959819
who gives a fuck if all games are shit and you can already run them at max settings
technology advances far more than dev skills, modern devs cannot even take full advantage of 15+ year old cards without having to inject a shit """modern""" engine into their games and let it do the work for them
>>
>>106959866
Not that anon and have played both.
Bananza is a genuinely amazing game, but it's not GOTY material. Especially against the media's E33 darling. (KCD2 is my GOTY, but it won't win either)
MKW sucks though. I wanted to like it, but it's just boring straight roads between actual courses you're not allowed to drive more than a lap of: the game.
Knockout mode is the best part of it but it's not enough to save the rest of the mediocrity. I played MK8DX 3 days ago, it's been months since I've booted up MKW - most disappointing MK since the shit phone game.
>>
Remember the bit wars? I remember.
>>
>>106959819
>using a stock symbol to refer to companies
yeah, it's a stock market grifter
ignore and move on
>>
File: 1757307928053448.gif (776 KB, 853x480)
776 KB
776 KB GIF
>>106959819
nvidia-bros how do we cope
>>
>>106959869
As mentioned here:
>>106959870
Chiplet will be mandatory with HighNA, as it will shrink reticle size to 429mm^2, making monolithic dies like the one in the rtx50xx series infeasible.
>>
>>106959870
On Instinct they aren't going anywhere.
On Radeon it didn't work. RDNA4 chiplets were cancelled apparently due to AMD's ancient enemy from years past - microstutter.

Wild Speculation: I think they'll probably do 'chiplets' on UDNA as the Display Controller+Media Engine and other common things outside the memory controller and shader engines, will get shuttled to their own chiplet, with the rest on single big dies. Like the AVIVO/Radeon Theatre days.
>>
Ah yes. AMD definitely has the GPU know-how to capitalize on this situation. Just like they did back at Pascal vs Vega. Vega wasn't a power hungry, hot, whole hog of a card which had everything going for it (smaller node, newer architecture, came out after) compared to the competition and yet, still lost in every respect to its rival cards in the end.

What I'm saying is that, even though AMD is in the exact same situation now, it'll definitely make a card that's better and make full use of the smaller node. This won't be another wash for Nvidia. AMD have this in the bag!
>>
>>106959876
I don't know what nVidia will do.
They've apparently been looking into Intel fab, so maybe nV will do FOVEROS?
AMD don't need chiplets for their GPUs though.
Officially Instinct are 'accelerators' now, since all the G part got torn out with CDNA.
>>
LOL He doesn't know what nm is!!!
THIS
IS
BIG
HUGE
MASSIVE
WOW
LOVE
FROM
INDIA
>>
>>106959877
I think that is all fair, and your thought process about separating dies by function is the standard approach to chiplet. This is suppose to be one of the selling points, because you can acquire different dies on different litho from different suppliers to meet your specific need.

However the industry is spending a lot of resources on attacking the interconnect latency and power over head issues. This is because the writing is very clearly on the wall for monolithic dies.

We will be forced to move to chiplet, and we will have to get better at interconnect. AMD is part of this with fan-out interconnect design.
>>
>>106959878
They didn't really have a node advantage there.
TSMC 16nm was miles better than GF 14nm.
Even nVidia's own parts told that story.

They did have a distinct node advantage with RDNA2 though.
>>
File: cat hoodie.jpg (433 KB, 828x798)
433 KB
433 KB JPG
Nvidia is better at capitalism. If they don't already have a better or equal tech to release tee'd up behind the scenes I assume their corporate espionage division will be right on it, and break that barrier shortly.

Therefore I will wait for the matching nvidia product, as it will definitely have better drivers.
>>
>>106959879
I don't know what Nvidia will do in the consumer market either. However, in the AI industry they are putting their money on CUPE and Photonic interconnect, which I see as the real end game.

In the future every data line should be photonic. This includes data lines in the NoC.
>>
>>106959820
weird that this line became so prevalent just as nvidia fell behind.
>>
>AMD has worse drivers
Goddamn Nvidia shilling meme needs to die already.
>>
>>106959819
>UDNA
"oo dna" mean at the bottom in russian
>>
>>106959883
>le drivers

multiple times this year nvidia has released bugged drivers. At this point I don't bother updating for months, since its usually more trouble than its worth.
>>
>>106959887
Cool, but yesterday some Russian nigga teabagged me in BF6 after I failed to knife him.
>>
>>106959886
My AMD card had water flying in the sky for months in Hogwarts Legacy (2023), go fuck yourself
>>
>>106959890
thats literally part of the game world go fuck yourself
avadacadavra
>>
>>106959819
hopefully theres finally meaningful performance jump AND good VRAM next generation from both companies
>>
>>106959838
power usage has gone up because the RTX5000 series is using basically the same process node as the 4000 series, but with larger dies. The transistor density actually went DOWN from the 4000 series. It's long been the case that for the same die size, a node shrink will allow for higher transistor density and better power efficiency.
>>
When should I sell my AMD bags? They're up 100% (my Intelaviv ones too).
>>
When will they become affordable?....
>>
>>106959856
>can't afford more than 5 games
I rather not.
>>
>>106959894
If it's good enough to screenshot (or tell us about it), it's good enough to sell.
>>
File: 170346794602.jpg (111 KB, 1440x810)
111 KB
111 KB JPG
AMD cards and drivers are fine. They are definitely fixed and better than they were and not hideously broken in the most popular games of the last 5 years.
You will not regret purchasing AMD.
>>
>>106959898
this guys on like a 6600xt or something with dated studio drivers
>>
File: 1742671623602260.png (962 KB, 1353x900)
962 KB
962 KB PNG
>>106959896
you dont need more games than 5 because they're so good. pic related
>>
>>106959819
I wish mommy Lisa dropped her bombshell on my faec
>>
>>106959898
>>106959890
>its muh amd
>The "Hogwarts Legacy" NVIDIA water bug can often be fixed by adjusting settings like the shader cache size in the NVIDIA Control Panel, disabling fullscreen optimizations for the game's executable, or checking the game's files for corruption. Other potential fixes include lowering graphical settings, disabling ray tracing, or ensuring your graphics drivers are up to date.
>>
File: 170346923569351.jpg (154 KB, 1440x810)
154 KB
154 KB JPG
>>106959899
Of course. It's always the user's fault somehow. AMD drivers are perfect and would never cause issues across multiple cards for 6+ months after release even with updated drivers.
>>106959902
>google The "Hogwarts Legacy" NVIDIA water bug
>all the results are AMD
How odd. Why are google colluding with NVIDIA against AMD? This is a conspiracy.
>>
>>106959888
/v/ is techlet central. They're still pushing a decade old meme about AMD and emulators.
>>
>>106959820
>former José is twice the size of the current one
we could get a fuckhuge perf increase because they can go back to brute force by shoving twice the current transistor count. Of course it won't happen because they'll milk this 2nm process for the next decade so we'll get only marginal improvements for double the cost as usual.
>>
>>106959898
>>106959903
>playing harry potter games
Legitimately why?
But also; my 7900xt ate everything I threw at it, including the absolute mess that was Oblivion Remastered without issue.
>>
>>106959903
this guy is crying about a single game lol
remember when nvidia cards burst into flames
remember when nvidia cards catch fire because of the cable connector they now use
remember when nvidia gpus caused monitors to blackscreen randomly for 3-6 months when the 50 series came out
>>
>>106959903
As recent developments have shown, AMD have no interest in supporting anything but their latest generation of GPUs.
There's no point buying AMD in case you're a richfag who can update constantly, in which case, why aren't you just buying NVIDIA instead?
>>
>>106959907
It has absolutely nothing to do with the cable, but the port itself on the GPU. ideally 12pin connectors should have three power delivery rails. They did on Ampere for 3090/3090ti. On Ada they had 2. On Blackedwell they have 1.
That's why the fires keep happening.
>>
>>106959819
AMD will find a way to snatch defeat from the jaws of victory, don't forget Susi and jacketman are cousins and seem to have zero intention of actually going into a market war for the best product.
>>
>>106959839
you dumb cunt
>>
>>106959910
Sucks how Lisa Su saved their CPU division from the brink, but managed to nosedive the GPU division into utter irrelevancy. They exist just so nvidia doesn't get hit with an anti-trust suit by the US government.
>>
File: images.png (9 KB, 367x461)
9 KB
9 KB PNG
When do we get Nvidia GPU handheld? AMD fucking sucks. FSR is a flop. I need DLSS.
>>
>>106959903
>>106959898
>see bugged graphics
>instead of updating drivers spend rest of the day taking screenshots of glitches to prove a non existing point
>>
>>106959819
Per Watt the Switch 2 on 8nm (offshoot of 10nm) blows the absolute fucking shit out of the Steam Deck which is 6nm (offshoot of 7nm) these both use GPU architectures that launched the same year, the same is true for the GPU's from the same generation, Nvidia 30 series are Samsung 8nm, and RDNA 2 are TSMC 6nm and TSMC are fuckloads better than Samsung, yet the RTX 30 series blew the absolute fuck out of whatever the fuck Radeon generation it was, who really cares nobody uses Radeon except consoles.
This matters shit all until Nvidia end up on the same fab process in which case they will just have an even bigger advantage.
>>
>>106959913
Nvidia doesn't make gamer products anymore. Why would they? Even though they could release a Shield 2 amidst this explosion of chinkheld retro pockets, they won't. Because it's not worth it.
>>
>>106959912
I really REALLY want a AMD win because their drivers are so smooth on Linux, but I know it won't happen, best we can hope is for the AI bubble burst to crash GPU prices for a little bit (until another minable shitcoin gets popular and we're back to square one).
>>
>>106959819
I don't care nigger. 2GB cards are more than enough. Just make good games again, instead of le bigger number video cards to enable worthless AA and AAA studios.
>>
>>106959819
>30 and 20 angstrom processes
we can't go much smaller than this
>>
File: 8348348.png (478 KB, 1284x242)
478 KB
478 KB PNG
>>106959856
They're not getting my money before I get a special edition for a game I like in nice colors. I can wait until the next smash game. Mystery nigga threads get me excited.
>>
>>106959917
The drivers are fantastic, but the Adrenalin suite is like 50-80% of your GPU's features now, and that's only for Windows. I will NOT install Windows 11. I'm just not gonna do it hahahaha.
>>
>Njudea shill defense force is in full swing already
kekaroo kill selves
>>
>>106959920
Didnt you say you bought one like 2 weeks ago while posting your mario dance webm
>>
>>106959915
You're so absolutely wrong it's not even funny. All Ampere had over RNDA2 was ray tracing. In regular rasterization, RNDA2 trounced Ampere at every single performance tier.
>>
>>106959920
you should buy one now and then just buy a second one whenever they release a special edition. your doing a good deed supporting a company that actually cares about video games, and since it's portable its extremely nice to have more than one console.
>>
>>106959924
Cool, but the age of Redditerization is over.
>>
>>106959819
First America invents AI then they're gonna invent the first chips measured in picometers and finally, FINALLY we'll enter an unironic cyberpunk age
>>
>>106959925
>portable
>dies in 1h30
>>
File: 170439623035.png (73 KB, 1090x394)
73 KB
73 KB PNG
>>106959914
>just update your drivers bro
>>
>>106959913
XeSS XMX > fsr4 > DLSS4
>>
This isn't news you fucking nigger, amd achieved 2nm lithology much earlier this year.
>>
>>106959819
She looks like a fucking man
>>
>>106959926
Ironic shitposting is still shitposting, faggot.
>>
>>106959924
Switch 2 shits all over the Steam deck for a fraction of the power, this really matters for nothing when the truth is that Nvidia's GPU architecture is just outright better, per cycle it does more work, so miniaturization just doesn't matter for shit.
>>
>>106959819
wow. amd goes from 2% market share to 3%
>>
>>106959930
In your dreams.
>>
File: 1695050968609148.png (39 KB, 184x184)
39 KB
39 KB PNG
>>106959929
>muh decade old card wont play the potterslop
>>
>>106959819
Doesn't matter, all the AI stuff works on nvidia and that's 90% of the industry now
>>
>>106959929
>rx400/500
lol hes literally indian
>>
File: 1738267221998934.png (41 KB, 252x267)
41 KB
41 KB PNG
>>106959934
>switch 2
>dedicated gpu
>deck
>shitty 4core apu with 4 compute units
the switch is stronger wow really you don't say
>>
>>106959935
More realistically 2.2%
>>
>>106959928
>for a fraction of the power
the power: >>106959934
>>
>>106959825
chip makers made up marketing terms like 4nm, and AMD's marketing made up number is lower than Nvidia's made up number so that means it's betterer.
>>
it was supposed to be this order sorry

>>106959934
>for a fraction of the power
the power: >>106959928

and no docked doesnt count just buy a ps5 at that point
>>
File: Screenshot_20251021.png (406 KB, 2125x1236)
406 KB
406 KB PNG
>>106959929
>mod says AMD already fixed it
lol.
https://www.nexusmods.com/hogwartslegacy/mods/995?tab=description
>>
>>106959819
>2nm
>3nm
That means absolutely nothing. It has no bearing on any real world numbers. It's literally just "mine's smaller than yours" talk.
>>
>>106959924
Ok and now factor in DLSS
>>
>>106959939
Perhaps, but you sound like an Amerimutt.
>>
>>106959923
I've never posted that, so no.
>>106959925
lol no
>>
>>106959929
dumb jeet
>>
>>106959949
the mario dance webm poster was always saying he was waiting for a special edition in a color he likes before buying though so its you
>>
>>106959889
my condolences, git gud
>>
>>106959819
The problem with AMD is in its drivers. I want to buy an AMD card but production on their hardware sucks.
>>
>>106959940
They are both APUs.
It has 8 CUs, the Switch 2 has 12 SMs, but the Switch 2 is also way more power efficiency, but the Steam Deck has 2.5x the battery capacity.

>>106959944
It's a point of comparison that the Switch 2 at a higher fab process is better than the Steam Deck, and nice that you mention the PS5, since that showed us that higher clocks beats more CUs/SMs, with some games ending up with better performance than some games on Xbox Series X, but Switch 2 at lower clocks and more SMs beats the Steam Deck, 1.6ghz vs handheld Switch 2 at 561mhz while in handheld.
>>
>>106959954
why does performance matter if you cant play for more than an hour 30
>>
>>106959819
Wasn't amd 6000 series on 7nm TSMC node or something, whilst the Nvidia Ampere 3000 or whatever was Samsung 8nm? How the fuck did AMD lose so badly to Nvidia's 3000 series? Fucking joke AMD makes TERRIBLE GPUs AND their CPUs run TOO FUCKING HOT
>>
What would happen to Jewvidia when China cracks processor production?
Imagine all those fake monopoly dollars that would magically vanish when the stock market acks itself
>>
>>106959954
>"they're both APUs"
>one apu is actually stronger than the other
>"the switch is le better!"
okay.jpg
>>
>>106959825
Smartest nvidiachad
>>
>>106959937
>>106959939
>it's normal for 4 year old cards that are beyond the requirements to have shitty drivers with water flying everywhere
Maybe for AMD
>>
>>106959951
I don't feel like it's a unique stance man
>>
File: 1740776220421507.png (326 KB, 774x720)
326 KB
326 KB PNG
>>106959960
Yes, Sukdeep, it's normal for a decade old card to struggle with UE4/5slop
>>
>>106959960
>4 year old
the rx480 came out in 2016
thats 9 years
>>
Amazing, what games are worth the upgrade? Not one? Shame
>>
>>106959963
SAAR
SAAR PLEASE UNDERSTAND
THE RX480 DID NOT REDEEM IN MUMBAI UNTIL 2021 SAAAR
>>
Unfortunately power efficiency doesn't matter outside of mobile devices.
AMD needs to start releasing shit with giant dies that draw 500W if they want to compete.
>>
>>106959955
You'd be able to play for 5 hours if it had the Steam Decks battery, that's just the gap in performance/efficiency, which is of course funny because it's a generation behind in terms of miniaturization, Switch 2 on 6nm, would have made the Steam Deck look like a complete fucking joke, and most likely we will see this come to pass for the Switch 2 refresh model, it's going to be like comparing Mario Kart World sales to Sonic Cross worlds sales.

>>106959958
They are both APUs, absolutely, and yes undoubtedly one would be better than the other, it's what happens when you compare 2 things where there's an objective way to measure superiority between the 2, absolutely, and yes the Switch 2 has shown to be better.
>>
File: pepe-angry-1304654900.gif (73 KB, 498x380)
73 KB
73 KB GIF
>>106959819
>Implying AMD is viable ever
The CEOs are related, they're both in on it. Lisa Su has massive shares in Nvidia. They're effectively running a controlled monopoly all of this is smokescreen tactics. Just look at how hard they're both fucking over Intel for trying to launch ARC.
>>
>>106959967
Yes the higher tier GPU is stronger than a weaker tier one.
>what do you mean RNDA2 is better than Ampere? My 3080 is more powerful than a 6400!
>>
>>106959964
Modern games are so unoptimized that you actually need to start upgrading your CPU more. I was looking to go from AM4 to 9950x3D to get better framerate on Palworld in bigger worlds (since my fps would dip to below 20fps when pals went to sleep), but when when I saw it would cost 2300-3000€ to upgrade mobo, cpu, memory, psu and stuff, that's when it hit me that maybe modern games are just not worth it anymore. So now I game on Switch 2 instead. If I miss a game on switch, it's because the game was unoptimized. If even Cyberpunk is on switch and runs ok, then there's no excuse for the other game to not be. So any games not on switch is probably not worth playing because developer couldn't optimize it enough.
>>
>>106959850
I usually upgrade my card when I can double the VRAM for the same price.

I got 12GB for $300 two years ago. I'm willing to spend up to $400 for 24GB.
>>
>>106959819
Who cares? That's just going to increase costs more while they squeeze 10% more actual performance out of a card that uses 15% more power. Then slap some AI shittery on there to pretend it's actually 6x faster.
>5070 is le 4090 performance for $549!* ** ***
>*in select titles with maximum dlss/framegen versus no dlss/framegen
>**in supported titles only
>***don't play for longer than like 30 minutes or else
>>
File: 173062962350.png (68 KB, 1044x185)
68 KB
68 KB PNG
>>106959963
Even the RX590 had the same issue and that was 4 years. Definitely not a "decade".
All because AMD don't know how to make water work over 45 FPS lol.
>>
>>106959968
Yeah the FTC needs to do something about the duopoly shit. Intel used to be fucking based I know some their recent CPUs have had problems, but I really want intel to be back because for very long they were the best way to game and I have nothing but fond memories from Intel. My games started running lower FPS the moment I switched from 5820k to 5950x and I don't know wtf happened
>>
>>106959969
They are the same tier, Switch 2 is just better designed, at least on Nvidia's side, although Steam Deck is kind of disappointing hardware wise, no idea why they went with RDNA 2, they should have waited for RDNA 3, RDNA 2 was complete shit.
>>
File: 1746433049384202.png (597 KB, 700x700)
597 KB
597 KB PNG
>>106959971
>I'm willing to spend up to $400 for 24GB.
>>
File: 1753656947279668.png (538 KB, 912x681)
538 KB
538 KB PNG
>>106959973
>2016 card
>"4 years old"
>>
>>106959946
>>106959943

the smaller your process the more transistors you can fit in the same physical space. the more transistors you have the more powerful the chip

so yeah, even a single nanometre is a huge fucking deal here
>>
>>106959970
>but when when I saw it would cost 2300-3000€ to upgrade mobo, cpu, memory, psu
It would cost $1000 at most. On Shitch you're locked to $80 30 fps 720p ports. Don't try to equate the two you dumb fuck.
>>
>>106959957
It's already happened.
Huawei is within 10% of H100 performance at half the power and double the ram.
China is banning new purchases of nVidia and forcing the larger companies to move to Huawei.

However, now since Trump says US needs to beat china on AI, anything not sold to China will just be sold to US for 4x the price. Leatherjacketman still won.
>>
>>106959975
>t-they're th-the s-same t-t-tier!!!
lmfao no
>>
>>106959981
Yes, they are handheld hardware tier.
>>
File: F-ZyjJgWEAASKsC.jpg (27 KB, 616x598)
27 KB
27 KB JPG
>>106959982
>there's only one tier of handheld performance
I thought you were trolling, but now I'm sure you're just stupid.
>>
>>106959979
>It would cost $1000 at most
If you wanna do an experiment, go on parts picker, and choose top quality parts worthy of a 9950x3D system, without overpaying. 64gb-128gb memory kit because browsers keep eating up memory (but 250-350 for memory), 1200-1300watt PSU, mobo is 350-500 or something (probably MSI or other brand with quick release button for GPU). I did this and in my country it was 2300, and this time it wasn't with 700euro asus darkhero mobo cause I learned that you need to have quick switch release for GPU on your mobo
>>
>>106959983
And of which they are of the same tier, it's not like it's an Rog Ally Xbox, come on ally.
>>
>>106959856
when i was 12 the vidya industry still had soul.
>>
>>106959985
>And of which they are of the same tier,
source: your curry-stained ass
>>
>>106959978
Which is why they're using it as marketing talk but it has ZERO bearing on the actual manufacturing process. The number is meaningless. It's just an arbitrary number. There's no telling if AMD will be faster than Jewvidia because they use entirely different manufacturing and metrics and they're just figuratively racing to the bottom with their numbering schemes.
>>
>>106959987
My source is the comparable performance of the two devices.
>>
File: 1742136972560955.png (244 KB, 626x650)
244 KB
244 KB PNG
>>106959989
And where is your methodology, you barehand ass wiper.
>>
https://www.nvidia.com/en-us/products/workstations/professional-desktop-gpus/rtx-pro-5000/

>72 GB GDDR7 with ECC

NVIDIA WON
>>
>>106959984
>9950x3D system, without overpaying. 64gb-128gb memory kit because browsers keep eating up memory (but 250-350 for memory), 1200-1300watt PSU,
Ok so you're retarded good to know. Stick with your tablet.
>>
>>106959992
9950x3d + mobo alone is more than 1000€ . Anon you failed, why must everything be ragebait on this website now?
>>
>>106959991
>can run 2077 with path tracing at 24fps before upscaling/framegen
waow!!!!
>>
>>106959984
>I learned that you need to have quick switch release for GPU on your mobo
confirmed for moron.
>>
>>106959990
comparable flops at least when handheld is considered as well as comparable performance between games, they are similar, but like I stated the Switch 2 does that for significantly less power draw, because the hardware is just better designed, just not on Nintendo's end.
>>
>>106959995
>jayztwocents is a moron for recommending consumers to get a mobo that doesn't run risk of breaking when trying to release a massive GPU that covers the regular PCI-E release switch
LLM tier bait
>>
File: 1679681196505689.png (170 KB, 329x290)
170 KB
170 KB PNG
>>106959996
More numbers you pulled from your unwashed backside?
>>
>>106959997
>jayztwocents is a moron
Correct.
>>
>>106959999
>thinks he knows better because he has small dual fan 4gb gpu
that's cute but we're not in 2012 anymore
>>
>>106959974
But that would be communism. Or at least socialism. Or at least a step against your beloved free market.
>>
File: Screenshot_jpg.jpg (787 KB, 2098x1719)
787 KB
787 KB JPG
>>106960000
Solved 20 years ago without needing stupid fad/meme shit.
>>
>>106959998
Well now you're just trolling me, stop @ing me you retard.
>>
>>106960002
Which part of
>large GPUs of today covers the regular swtich from being accessible
failed to register your brain? You need to use a tool to remove large GPU from a PCI-E slot that doesn't have quick release switch, like MSI and some other brand has
>>
File: IMG_5740.jpg (58 KB, 976x850)
58 KB
58 KB JPG
So what happens when they get to 1nm?
>>
>>106960000
>4gb gpu
>2012
What? my 2011 560ti had 1GB of Ram and my 2015 960 had 2GB of Ram.
>>
File: 1716018709612720.png (796 KB, 2000x2000)
796 KB
796 KB PNG
>>106960003
I accept your concession, Ranjeep.
>>
>>106960004
No you don't.


You're a braindead idiot who needs instructions from a video but can't see with their own eyes that the switch is ABOVE the socket, so it won't be covered by the GPU.
You're a c o m p l e t e brainlet.
>>
>>106960005
The universe resets and we go back to using single core everything as god intended
>>
>>106960008
>No you don't.
Then you don't have a large GPU, which why I said, your dual fan is cute, but it's not 2012 anymore, and don't try to pretend you know. You're arguing out of malevolence and ignorance
>>
>>106960007
There is no concession, the performance of the graphics parts of these APU's is well known, they are comparable due to their configurations at least when comparing handheld Switch 2 to the Steam Deck, we know this, we know this from gameplay captures measuring performance, we know this from calculations comparing raw power capabilities and we know this from architectural comparisons. You're the retard claiming that I'm just making this shit up so really you are at this point completely full of shit.
>ranjeep
Painful, there's nothing worse than someone claiming you're an indian, the lowest of humans on this earth.
Every post you've made so far has been you spouting nonce sense from a place of ignorance, seriously either make a real point or fuck off.
>>
It wont mean lower prices and AAA games arent worth playing so the power is wasted.
Unless you're a slop prompter then fair enough.
>>
>>106960006
not that poster, but you could buy a GT710 with 4GB of ram.
Why? Who knows, but they made them.
I remember because I laughed at them a decade ago.
https://www.amazon.com.au/GeForce-GT-710-Graphics-Consumption/dp/B0BVVFPDMH
>>
>>106960010
Your GPU doesn't go above the socket moron.
>bbut you don't have a large GPU
I have a larger GPU than you.
>>
>>106959819
can't wait for udna! amd gonna kill nvidia with their $50 discount, let's go team green!
>>
not vidya related. get this AI shit over at /g/
>>
No, not /g/ now someone can point out the flaw in my argument, I'm done for, at least nobody will reply to this thread anymore.
>>
>>106960016
I accept your concession.
>>
File: file.png (68 KB, 1340x460)
68 KB
68 KB PNG
>>106959819
>2nm
Won't matter if Nvidia's chip has twice the die size...
>>
>>106960030
3090 got mogged by 520mm2 though.
>>
>>106960005
we switch to picometers.
>>
>>106959970
I just got a 5700x3d from china for $150 earlier this year. Hasn't struggled with anything, bf6 is pretty cpu heavy and it's pushing 150fps in 64p multiplayer.
>>
>>106959849
I'm not buying cards based on what you think reddit does, dimwit. Go finish licking the windows.
>>
File: 1717955302261699.jpg (91 KB, 336x442)
91 KB
91 KB JPG
>>106960011
If you had any kind of facts to back up your bullshit, you'd post it. But you don't so you wont, Sukdeep.
>>
>>106960100
Thanks for confirming that you're just a retarded faggot, but you know what
https://www.techpowerup.com/gpu-specs/switch-2-gpu.c4313
https://www.techpowerup.com/gpu-specs/steam-deck-gpu.c3897
ops per cycle x CUs/SMs, x frequency, this gives us around 1.7 Tflops for the Switch 2 in handheld and around 1.6 Gflops for the Steam deck, which doesn't have a handheld/docked mode, this gives us comparable at a base hardware
https://www.youtube.com/watch?v=CLJajeFkmhQ
You can see similar performance within these benchmarks, but obviously the Switch 2 in handheld does this on a significantly smaller battery.
Switch 2 is definitely better when it comes to RT and in reality Switch 2 will end up better in some games, depending on optimization, hence titles like Star Wars whatever the fuck that new Ubisoft one was called having RT features while running significantly better on Switch compared to Steam Deck.
https://www.youtube.com/watch?v=fngLoH69jZc
Star Wars is closer to what a game is like when it's properly optimized for Switch 2, the kind of advantage you get out of using known hardware and avoiding bottlenecks in that hardware, which can't be replicated just by changing options.
Fuck You
>>
>>106959968
AMD makes 10x more money from nvidia shares than they do from their own GPUs.
Retards who defend AMD Radeon garbage are just shills, those fuckers make it seem like /g/ has 95% AMD GPU marketshare.
>>
hasn't the "nm" in lithography nodes been marketing for like 10 years now
>>
File: a.png (2.07 MB, 1920x1080)
2.07 MB
2.07 MB PNG
>>106959906
i recently replayed HP1, it was comfy. looks quite nice in 1080p i think
lol i found the screenshots and apparently it was a year and a half ago. recent enough.
>>
File: 1575089917078.jpg (16 KB, 450x450)
16 KB
16 KB JPG
there'll be zero stock with all these dogshit "AI" deals no matter what fancy process
>>
>>106959844
anyone told u how stupid u sound today?
>>
File: file.png (310 KB, 1384x1606)
310 KB
310 KB PNG
>>
>>106959819
Oh yeah well intel has 1.8nm now so I'm gonna sell my 9070xt for a C580 when it drops
>>
>>106959973
the rx590 came out 7 years ago my negro
>>
>>106960005
same thing that happened when we went past 1µm
>>
>>106959849
Nvidia sub has over 100k more users than the AMD one in Reddit. Nvidia is Reddit.
>>
>>106959865
No one gives a shit about efficiency outside of third worlders who can barely afford power. If you can afford an Nvidia GPU, your power bill should be at the back of your mind. Or maybe it shouldn’t be because the card will melt anyways lol.
>>
>>106959882
>RDNA 2
And lo and behold, it was their most competitive generation in years even during a time where Nvidia had an even larger featureset advantage.
>>
both are still 5 real nms at best
>>
File: energy-efficiency.png (57 KB, 500x1170)
57 KB
57 KB PNG
>>106959828
When was the last time that happened in your pea-sized brain? Because the actual last time Nvidia were on a worse node, AMD raped them in terms of efficiency.
>>
>>106959819
didn't work for intcel. Won't work for ayymd
>>
>>106961146
Once you compare marketshare, AMD has more redditors per capita by far.
>>
>>106959841
The winning move is not playing at all
>>
The worst 2nm tsmc process node has a density of 320mtr/mm2 while the best 3nm has a density of 220
>>
>>106959863
No they didn't
AMD wanted to ditch rdna since rdna2 lost on compute gimmicks so they're not fully committing to rdna4 hence why they're monolithic
It's all hands on deck for udna and a skeleton crew for rdna4
>>
>>106959823
There's over 200 posts in this thread and everyone ignored this post which is the only thing that matters.
>>
>>106959824
nvidia drivers are worse in 2025
>>
>>106961180
the funny thing is that the semiconductor density wasn't even that much higher while the difference between 2nm and 3nm is huge
>>
>>106959846
>>106959915
>>106959956
You guys are on crack. Ampere lost so hard. a 6900xt is worth more than a 3080ti today.
>>
>>106959967
They are SoCs, you sperg. APU is AMD's marketing term. Nvidia doesn't make APUs, just like AMD doesn't sell Quadro cards or Xeon server CPUs. Fucking /v/tard.
>>
>>106959823
See >>106959845
>>
>>106961416
Using professional software.
>>
>>106961381
an apu is any soc with more than a rudimentary amount of graphics cores
>>
>>106961430
APUs usually have less than SOCs
>>
>>106960257
Hey man, which model is the one you use on your wallpaper? You know, the one with the blue swimsuit.
>>
>>106961456
no, soc is the term that encompasses the weaker ones like phones and all the smart internet of things shit
if anything consoles are textbook apus, they just have sony/microsoft branding and they could easily let you install linux/windows on them
strix halo has more gpu cores than the base ps5 yet it's still an apu
>>
>>106961500
No, SOC is a single system on one chip, it would be more than an APU as it typically has wifi etc. on the same board not weaker. An APU is a type of SOC but a SOC isn't necessarily an APU.
>>
>amd fags buying amd because of nm, energy efficiency, chip shape
>but not because they're actually good at running games
ha ha
>>
>>106961564
>nvidia fags buying nvidia because of marketing
i'm glad you admitted that you don't understand silicon at least because process node is the single biggest determinant of performance
>>
>>106961430
Only in your /v/tard sperg brain. Its AMD's marketing term they coined back with Brazos and Llano. An intel Sandry Bridge chip is not an APU.
It is one narrow and specific marketing term for AMD only. Again just as Quadro and Cuda are for Nvidia or Xeon for intel. You fucking 50 IQ retard.
>>
>>106961630
my gpu is faster than anything amd make so I'm not sure what point you're trying to make?
>>
>>106962003
that you're retarded as evidenced by you dismissing the process node as a measure of performance
>>
>>106959819
Let's be real, It's AMD.
They'll find another way to screw up like they did with the rx6900xt
>>
>>106962019
I'm dismissing AMD based on them having a long history of being shit and all AMD fags can go on about it energy efficiency while ignoring how the perform [badly] in GPU tasks
>>
>>106962041
so you came to /g/ to reinforce the status quo despite your lack of understanding of the matter?
imagine admitting to being a fucking npc
>>
>>106962060
Please answer why you bought AMD?
It wasn't because they run games better
It wasn't because they have more features
So what was it? It cost less or some energy efficiency bullshit ?
>>
>>106962072
please explain why you came to this thread to not actually discuss things and spew the status quo without being willing to discuss performance indicators?
>>
>>106962086
I read the thread, my comment was after reading why nobody was actually discussing performance. It was all about muh efficiency, which is all you have and is evident by not answering the simple question of why you buy AMD.
>>
>>106959819
>Where have you been when NVidia was kill?

they got killed last week already
its going to be a hard time to find a buyer for all those cards
>>
>>106962105
you clearly don't understand what a process node is
go on, explain to me why a 6950xt competed with a 3090ti on performance despite the latter having a die that was 100mm2 bigger?
and the 6950xt's silicon was only 15% denser
imagine what happens when it's almost 50% denser like with 3nm vs 2nm?
>>
>>106962126
You keep dodging my question. Why can't you answer it?
>>
>>106962126
Density is the least important parametric here.
N3 is FinFET while N2 is multi-channel GAA ribbonFET. For HP parts there is a 34% power difference between them in ISO comparison.
>>
>>106962134
why can't you discuss things? because you know nothing about the matter
>>
>>106959819
>here's how AMD can still win
>>
>>106962157
discuss what? that AMD have made a claim 2 that this time in 2 years they're going to make a super duper GPU which will beat Nvidia. I see this every AMD release, followed by the same old excuses by the dumb retards who bought them.
Now answer my question. Why did you buy AMD?
>>
>>106962171
you're so desperate for a win you're trying to talk about people rather than concepts and ideas
sad really
>>
>>106962182
I'm asking why you bought AMD.
Should I run it through an English to Indian translator for you?
>>
>>106962182
here. is this better?

आपने AMD GPU क्यों खरीदा?
>>
>>106959819
1. It's priced in
2. Good luck producing 2nm at scale
3. Cuda
>>
>>106962194
you should code switch yourself away from being an insecure normalfag
your mindset is that of a post privacy zoomer that needs every aspect of his existence translated to the internet
>>
>>106962239
>you're an insecure normalfag zoomer
I've already won. You own AMD. I don't. You're so ashamed of your retarded buying decision that you can't even answer the simple question of why you spent money on it and why you're here defending the company.
What a loser lol
>>
>>106959940
Plus the Shitcher 2 released after the 'Eck. Like, no shit: Newer hardware!?
>>
>>106962265
you'll grow out of this online retardation when you're older
>>
>>106959913
go buy your nintendo switch nvidiot
>>
>>106959913
never lmao
no one wants to deal with nvidia's shit especially when it comes to such a low profit margin device
>>
>>106962277
>shamed AMD owner calling other people retarded
All you do is throw out insults. You said you wanted a discussion but you can't answer simple questions without acting like a sperg.
>>
>>106961354
It's about component complexity pretending to be density.
At some point we will start to measure ICs by height.
>>
>>106959823
apparently every new gpu gets rocm support but thats probably still not enough. No idea if this is still valid https://newsletter.semianalysis.com/p/mi300x-vs-h100-vs-h200-benchmark-part-1-training
>>
>>106962355
you really don't fit in here and it shows
i invite you to go to reddit, they love letting pompous fools like you farm karma
they also have personal banners that let you show off your specs so you can have your gay parasocial shit there instead
that website is made for you
>>
>>106962426
>more insults , no discussion
Why did you buy AMD?
>>
>>106961477
same one >>106962493
>>
>>106962485
maybe you shouldn't have tried to drag it down to a personal level?
not my fault you chose this avenue and you have to cry over being called a redditor despite acting like one
>>
>>106962527
>still can't answer a simple question
Are you a moron? Do you buy things without researching first? Do you not have a reason why you purchase AMD GPUs?
>>
>>106962543
do you have a reason why you're trying to talk about people in a thread about technology?
do you watch kardashians too?
>>
File: wisdom.png (1.99 MB, 1396x1435)
1.99 MB
1.99 MB PNG
>>106962553
>ask simple question that someone shilling AMD should be able to answer
>20 mins of insults, cope and spaz-outs later still no actual response
I know why you buy AMD. You know why you bought AMD. We both know that you'll never admit it though.
>>
>>106962576
we both know you know absolutely nothing about silicon and for some reason you think you can throw your weight around like it's reddit
last (You) you're getting, you're genuinely not worth responding to, I could extract more wisdom out of a brazillian that has just build his first poverty pc
>>
>>106962597
>bought AMD GPU
>"y-y-you just can't comprehend my vast knowledge on this subject"
So explain why you bought AMD then?
>>
entire thread is one bot retard nigger spamming why did u bought aymd
holy shit absolute state bring back cunny posters now
>>
>>106962682
>this entire thread is someone asking me a simple question, which I'm refusing to answer for some unknown reason
answer the question.
>>
Reason to buy AMD:-
>A - it cost less
>B - It is more energy efficient
>C - I'm a huge slobbering spastic
>D - other (please explain)

I can't think of any more. I've made it multiple choice , you can choose more than one.
>>
>>106962796
AMD GPUs are less efficient than their direct nvidia counterparts.
>>
>>106959819
Great, can't wait for amd to make an underpowered midrange card out of it. Stock shelves with maybe up to tens of them, and then never make one again.
Can't get TOO competitive, after all.
>>
>>106959819
None of this matters because Nvidia will continue to bribe companies into making their shit run worse on Radeon products like they've done for 20 years.
>>
>>106962941
So you're saying that Nvidia GPUs can do things much better than AMDs?
>>
>>106959819
AMD stock went up 50% this month the fuck you mean nobody has caught it.
>>
>>106963051
>I saved $10 per year by buying a worse performing gpu
>>
>>106963051
>9070 in efficiency terms
5070 and 5080
9070 will casually chug like 300w to match a 5070 that's pulling 200w
>>
>>106962796
linux drivers should also be considered
>>
>>106963092
>E - I trooned out
>>
>>106963099
pride flag taskbar on windows lmao
at least linux lets you choose what you install while windows forces every line of trooncode through
>>
>>106963039
the real play is long term holding stock from both
>>
>>
>>106963098
No I do, the claim that the 9070 is efficient compared to the rest of the current gen gpus is basically just made up and only applies in cherry picked scenarios.
>>
>>106959820
fpbp
>>
>>106963121
no you literally do not
nobody wants to conversate further with you because you keep repeating shit like a monkey without any knowledge in the field
enjoy your dead thread
>>
>>106963113
>I know I've cut my cock off but windows had a pride flag for a month 3 years ago
>>
>>106963136
windows hires hundreds of trannies and you have no choice but to accept their code
imagine defending windows after 7, actual zoomoid
>>
Why is this thread just one retard telling everyone else they don't know what they're talking about?
>>
>>106963157
>windows hires hundreds of trannies
did you apply for a job?
>>
>>106963134
People assume efficient to mean how much power to hit whatever level of performance, on this metric the 9070 isn't necessarily great.
>>
>LE REDDIT BROS WHY DO THEY KEEP CALLING ME OUT REDDIT BROS
holy shit this board is filled with literal retards
>>
https://youtu.be/4b3DrLXYt2g
>>
>>106963120

>precedes to google tdp of completely different gpu to prove his point
HAHAHAHAHAHAH HOLY SHIT
>>
>>106963170
have you considered being worried about trannies that code for microsoft that the public can't review?
or are you selectively mindbroken because you're an online activist that serves corporate interests in the end?
>>
>>106959819
>muh nm muhfugga
Nvidia still softwaremogs AMD softly.
>>
>>106963208
What makes you think you make me worried?
>>
>>106963223
so you don't? you actually like trannies that code for microsoft that are partly responsible for windows becoming shittier because it plays into your tribalism?
glad to hear that you don't apply your standards equally
>>
>>106963237
You're weird as fuck.
>>
>>106963238
i'm not the one that's mindbroken about trannies but not willing to actually commit to it
you choose what happens with your linux
with windows trannies always win
>>
>>106963276
Do you not realise how much of an overly dramatic faggot you are?
>>
>>106963312
>trannies make windows and i can't do anything about it but it's okay because i'm fighting a strawman where only linux has tranny involvement
either commit to your tranny hatred and stop using computers altogether or drop your retarded position
>>
>>106963330
jesus wept.
>>
>>106959968
I never understood how the entire media, consumers, institutions and userbase seem to be okay with AMD and Nvidia CEOs being related.
In any other industry this would've been a scandal and one of them would've been forced to step down very long ago.
>>
>>106960013
Yeah back in the late 2000s and early 2010s VRAM was plentyful, and manufacturers were releasing entry-level low end cards with insane amounts of VRAM for the time, that they didn't even have enough fillrate to take advantage of, mostly for marketing purposes.
Think something like a RX 6400 or GTX 1650 with 16GB VRAM being released today, that's what they were doing around that time.
>>
>>106964206
So you bought an Intel GPU, right?
>>
>>106964478
>So you bought an Intel GPU, right?
My i7 12700K has an Intel UHD Graphics 770 iGPU, does that count?



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.