Thread 511989277

>China's First all domestic made GPU, the Lisuan g100 has reportedly outperformed both the RTX4080 and is slightly behind the RTX5060 in new benchmarks

In less than 3 years, Lisuan graphics have improved the performance of the g100 from in 2023, lagging behind the GTX660TI to now beating out flagships like the RTX4080

The west is fucked. By choking out China and banning them from buying chips we've forced them to innovate and now they're nearing TSMC chip making capabilities.

GPU's are the new oil, Nvidia made $224 billion last year.

By 2026 China will surpass Nvidia, mark my words.

>>511989277 (OP)
This is what a free market looks like

>>511989277 (OP)
I'll wait for the benchmarks

Don't believe it. Will never buy it.

I'm sure Microsoft and the other US/Israeli/vassal state software and hardware manufacturers won't just vendor lock the Chinese competition out by introducing as many compatibility issues as they possibly can with their products.

*sets your house on fire*

>>511989277 (OP)
inb4 the drivers are somehow even worse than amd and nvidia

>>511989839

That's a feature...

>>511989277 (OP)
Lets see the chinese market uptake.

>>511989277 (OP)
it outperforms 4060 not 4080

>>511989277 (OP)
do not buy chinese hardware, jewvidia is better for you goyim, and at affordable prices, isn't the free market wonderful?

>>511989277 (OP)
I too believe everything the CCP says.

>>511989277 (OP)
it's so over

When patents don't actually mean anything, like in China, it must be easier to do that. In the USA, making a GPU costs a fortune in various licenses.

>>511989277 (OP)
Total Niggervidia Death

>>511989277 (OP)
It’s going to mine crypto and spy on you in the background without your consent. You’ll be trying to get to sleep wondering why your PC sounds like a ramjet engine 24/7. Bet

>>511989277 (OP)
if it has more than 32gb of vram without costing $2000 i will buy one.
unreal how bad all 3 manufactures failed to read the room.

>>511989277 (OP)
So China supposedly managed to replicate years of R&D and technological feats by Nvidia/AMD/Intel in the GPU space at breakneck speeds and we’re to believe it was all thanks to their sheer determination? Why don’t I believe it. They either stole a lot of shit or there are some real hardware limitations they’re not letting on. And no thanks to the rootkit.

>>511992837
What good is vram if it’s computationally shit at drawing polygons depending on the variables at play.

>>511993568
because even a 2014 processor with 64gb of vram would demolish a 5090 in AI inference if the model can't fit in vram and needs to be swapped out constantly.
it's why people have been modding 3090s with bigger memory chips.

>>511989277 (OP)
>Gayming video card
Who gives a shit?

Completely fake most likely.
The dicklets post these lies every month, by the time the last lie has amounted to nothing - as it ALWAYS does - they are telling the next lie.
Chinks are worse than Poos.

>>511989277 (OP)
OK. cool.

>>511993925
So the shitty card is still shitty.

>>511992837
32gb for games or running models? if the latter you can just buy 2-3 separate cheap cards

>>511989277 (OP)
I really don't need the latest specs, if China actually pumps out shit that's one gen behind but less than half the price I'm never buying niggerdia or AMD again

>reportedly outperformed both the RTX4080 and is slightly behind the RTX5060
this doesn’t even make sense in the slightest

>>511989277 (OP)
The U.S. government banned it from sale in the united states like they banned Huawei phones and Kei trucks.

>>511994091
And shills here still carry water for every single thing they try and tout, while not realizing all their credibility vanished years ago.

>>511989277 (OP)
Sure it does

>>511989277 (OP)
>Better than 4080
>Behind 5060

Anon.

>>511989277 (OP)
>beat 4080
>behind 5060
uuuuh ???

>>511990290
oh ok, make more sens.

>>511990290
Not a GPU expert.
Isn't the 4060 an older card, and the lowest quality version of it?
My understanding was you needed the 90 for anything AI / local models, and the modern nvidia cards were the 5000's.

>>511995714
amd released first good video card 9060 so jews in nvidia had to step it up to kill competition.

as a linux fag I do not care I will have to use amd because of better drivers support.

>>511995714
The 50 series have multi frame generation, so even though they have lower clocks, they perform better on paper (though it's fake frames and they are lying to your face, your lag will still be insane).

>>511989277 (OP)
>era of dirt cheap GPU's is here
I kneel glorious chairman Xi
I wasn't familiar with your game

>t. Ching Chang Chong University
Credible

>>511989277 (OP)
im so fucking glad china wasnt a thing years ago Im glad im still alive and didnt get instantly turned to ashes playing diablo 2 with some cheap chink piece of radioactive shit

>>511990105
If they're smart they will just fully open source the drivers.

>>511993389
>So China supposedly managed to replicate years of R&D and technological feats by Nvidia/AMD/Intel in the GPU

Both NVIDIA and AMD are run by ethnic Chinese (Jensen Huang, Lisa Su).
The majority of their technical staff is Chinese or Asian, even if they're American citizens.

The chips themselves are manufactured by Chinese people in Taiwan.

All our chips are already Chinese.

>>511998258
The 60 series are trash. I have a 3090 for work and ai gooning, and 3090 beats the 5060. They purposefully make the lower end cards terrible to upsell you to a more expensive one.

>>511998258
>My understanding was you needed the 90 for anything AI / local models
depends on what you're running but for image generation SDXL can be run on lower end hardware, you'll just have slower generation times. for large language models though you typically do want flagship models since those are very vram hungry. and as far as the 4060 vs 5060 goes, the uplift for each generation is total dogshit and the 5060 barely beats cards or loses to them from even the 2000 series when it comes to things like pure rasterization. nvidia often scams consumers by claiming a huge leap in performance using fake AI frames, but those are like putting a new coat of paint on a piece of shit car with 3 wheels. it still runs like shit even if it looks smoother, you still have massive input delay since the base game is still running at like 25-35 fps before fake frames are inserted. and multi frame generation also tanks native frame rates making it worse. its okay if you have a decent card and want to bump up frame rates for high refresh monitors, but its a pure marketing scam when they apply that same logic to lower end cards.

This is what a free market looks like

>>511989277 (OP)
It’s so over for the west

>>511998471
The writer is just a retard and it was about 4060 and not 4080.

>>511989277 (OP)
>Better than RTX 4080
>Behind RTX 5060
The fuck? What kind of benchmark is this?

>>511989349
lol, our government will probably not let us buy them, or slap on a huge tariff.

>>511989277 (OP)
Can it run Clair 77 in ultra settings at 60+ fps (hard limit)?

I say this as someone who works in AI.

The majority of talented people in this industry are Chinese. I meet with them and visit conferences full of them. I’ve written research papers with a few of them. Their work is of excellent quality and they do so while hindered by the fact all AI research is conducted in English rather than Mandarin.

The whole “Chinese can’t innovate, only copy!” is a complete and utter boomer cope, their rate of learning and advancement in tech is unparalleled, plus they have all the manufacturing capabilities, educated population, quality infrastructure and cheap energy to back it up.

I’m convinced the west is fucked and it’s mainly due to our hubris. Expect China to drop a GPU superior to Nvidia within the next 18 months and stock market chaos to follow.

>>511989277 (OP)
I'll stick with my current card thanks

>>511998950
The majority of the people who do the research and engineering are in the US. The US puts a CEO in charge of the companies to appeal to Asia like they do with jeets, otherwise people would complain about colonization or what have you. AMD/NVIDIA is a duopoly like coke/pepsi the people who develop and own the licenses are the same ones who develop the software.

https://developer.nvidia.com/gpugems/gpugems3/contributors

Most of these aren't Asian, there's a surprising amount of South Americans however. If China wants to stop being cheap labor and finally become a proper 'developed' nation then that's great, but for now it's just the same shit jeets do talk about how they're durgasoft tech geniuses for getting tech licensed to them. Can't have your cake and eat it too, even Brazil has more integrity than this.

>>511999640
I agree with this as far as bugmen can hive mind and hyperfocus a massive group of workers on a single task. I still have doubts on their innovation ability though. And yes, the west fooked itself in the 2000s when we shifted to a service economy and thought we could expolit labor forever.

>>511989277 (OP)
Built by a Chinky Polack named
Sum Dum Fuk