[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / 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]


>AIEEEE XI-SAMA I WAS WRONG TO BAN THE CHIPS
>PLEASE STOP TRAINING YOUR NEW MODELS ON HUAWEI
>I'LL GIVE YOU A 25% DISCOUNT
The top 3 model open source model, deepseek v4, glm 5.1, kimi are all trained on huawei. It's over.
>>
art of the deal
>>
>>108908188
lmao china doesn't have any money to buy nvidia chips even with discount, another big Trump W
>>
>>108908188
>Top 3 models
Nah
>>
File: 1761037610209.jpg (125 KB, 800x1000)
125 KB JPG
>>108908188
>0% marketshare
Even jenson knows the days of milking China by selling his is over thanks to the retard in chief. no dependency on cuda = no money. It's over.
>>
>>108908235
He said open source, retard
>>
>>108908258
Still incorrect
>>
5d chess move right there
>>
>>108908188
They don't call him the Fat Orange Retard for nothing.
>>
File: 1701550480824980.gif (1.43 MB, 640x360)
1.43 MB GIF
Nvidia is highly profitable and almost all of the better node production is fully booked for AI. Somehow it is a problem that the demand isn't even greater and consumer goods even more expensive?
Shut the fuck up retard go back to /biz/ this isn't NVDA.
>>
Are Americans really OK with the fact that their main export is GPUs now?
>>
>>108908376
americans main export is intellectual property lmao
>>
>>108908359
Nvidia is sitting on an inventory of unsellable GPUs and it's costing them millions when they're only worth trillions.
>>
16 months of constant failure, just 32 months to go KWAB
>>
>>108908188
>The top 3 model open source model, deepseek v4, glm 5.1, kimi are all trained on huawei.
fake news.
none of those were trained on huawei.

https://github.com/deepseek-ai/DeepGEMM
>DeepGEMM is a unified, high-performance tensor core kernel library that brings together the key computation primitives of modern large language models — GEMMs (FP8, FP4, BF16), fused MoE with overlapped communication (Mega MoE), MQA scoring for the lightning indexer, HyperConnection (HC), and more — into a single, cohesive CUDA codebase.
>>
kek
>>
gimi
>>
>>108908254
>tanking american company sales in China
>Jensen HUANG
Hmmm
>>
>>108908188
The worst part is that his "tarifs" for China were not even real. They had exceptions for everything meaningful and were only there to distract the american goy.
>>
>>108908188
America doesn't even have open source models anymore. The last company, meta, went full closed.
>>
>>108908188
I wonder if people know how simple those AI chips are.
I mean, consumer NVIDIA GPU is astronomically complex and packed with tech, both new and decades old. But if it's just a matrix multiplication accelerator? It's dead simple. It will work flawlessly even if your engineers are idiots and even if you don't have latest 1nm tech or whatever, it will work just fine using teh old one. China has plenty of energy and they literally use sea cooling to take care of the increased heat (older tech dumps more energy as heat).
I'd be surprised if China hasn't figured this out. In fact it is weird that it took em so long. Some random Arabs did exactly that. Funded those ASICs for AI, they work flawlessly for inference, NVIDIA was eating so much dust they aquired some of those startups and probably are going to kill them off silently.
>>
>>108908254
You don't even NEED CUDA with AI anymore. The main thing CUDA has is momentum, because of how much FOSS CUDA shit nvidia maintains, and infrastructure that was already built on CUDA. Current models like gpt 5.5 and claude opus 4.7 can write vulkan compute damn near perfectly. Add a few debugging rounds and fixes and they'll get it right. The main thing holding models back right now is the small context, because large codebases are too big for the effective context window of current models. I think if models could correctly reasonable about like 10M context, programming might actually be solved to the point that a single person could compete with major corporations. Of course that's kinda unrealistic right now, unless you take those SubQ guys serious, which you shouldn't.
>>
File: 20260526194902002396.jpg (215 KB, 2100x1500)
215 KB JPG
>>108908291
>>
>>108911734
>I think if models could correctly reasonable about like 10M context, programming might actually be solved
Anon, I really don't want to call you stupid, but you are asking for it.
>>
>>108908188
Let the jew make the AI in the real world so the chinese can destill it.
>>
>>108908519
>>108908376
Number 1 export is gold
>>
>>108908291
Which are the top 3?
Pony, WAN and Qwen?
>>
>>108911792
Every day massive quantities of gold is shipped from USA to London where it gets molten into ingots.
The next day those ingots are shipped to Switzerland where they get molten into different sized ingots before getting shipped back to the USA.
Questioning why this happens in antisemitic btw.
>>
>>108908218
Buddy, they have plenty of dollars to spend
>>
>>108911792
>>108911834
It's obvious, no one wants the dorrar, it's monkey money meant for the tax peasants in afrimerica

There is no goodwill left, once the gold runs out over there it's over lmao

Now this is some 3D chess stuff
>>
art of the deal
MIGA
israel first and only!!!!!
>>
>>108908188
Yet another Trump policy that backfires spectacularly.
>>
>>108908313
Donlads worked extremely hard for that moniker, he earns it every day anew.
>>
>>108913018
pls understand, he spends all of his time finding the most vile criminal jews and pardons them
>>
>>108908535
then maybe they shouldnt have abadoned their consumers to please dick sucking shareholders.



[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.