This Nvidia Challenger Says Its AI Chip Is 10x Faster Than A GPU

6nm!!! lol!

NVDA down 10% in 2 weeks. Every hyperscaler is building chips to replace them. Jensen's "installed base" moat is backwards — Apple/Google/MS own the real one. CUDA covers
training only. Inference is hardware-agnostic. $4.85T at 30x P/E while moat erodes. Intel 2000 all over again.

TBH, it doesn't look like their current hiring demand equals any sort of large production and after sales support that you'd expect to see with a large scale-up.

Probsbly a lie

On Trump's to-ban list for sale to China.

these guys are smoking shish, 10x faster? 3 times cheaper? try get passed that CUDA moats where others trilions dollar company can only dream of.

🧐

Sounds scammy.

tsmc guides all designs it makes, accessing its patents and proprietary technology free to customers

if it's made in India, it's made from curry

Just check the founder's background. It is mostly f*ing marketing bg. Will you trust someone with marketing bg or engineering bg?

The guy is from india🎉

are indian scammers evolving???

Assuming their chip really is faster, how will they compete against CUDA? Nvidia has virtually all the best AI talent trained on their proprietary software.

10x is not that impressive. Try to get an interview with the Q.ant CEO he will tell you about REAL speed up by using photonics.

doing is harder then saying m

nice bussom

all these BS! everyone trying to get a piece of that Ponzi AI bubble.

All these chips are built on digital logic, using 50 year old logic synthesis tools. That is where the energy inefficiency is.🤔. No one has a solution(incl. NVdia!, since TSMCs digital tech is also dependent on it). sram writes are not energy efficient compared to dram, though has faster read access speeds, benefits from same digital process node, that also compensates on density due to smaller geometry. Seems like a stop gap solution(just for inference, since training involves lots of writes), rather than permanent one🤔. NVidia will have to adapt soon🔮, when low power training chips using non-boolean logic(while still using digital cmos tech🤯) hits the market🤔. Just some 💭

12 Codes of Collapse feels like someone took all the AI fears people avoid saying out loud and put them in one book.

weirdo launch needed ???

un cpu curve on connait !
mais de là faire un socket comme un tank ! c'est n'importe quoi

AI is now in era like bitcoin miming was first with fpga and suddenly with asics, so this is kind if asic not a gpgpu for AI, that should happen years ago if someone would looknjusta decade back, nut not even that, just took a look from nvidias parrern, double the transistors, double the power plus(all this on more advanced node that should lower the power plus generational ipc uplift and still hardware itself was barely double, most of the time much less than double stronger ) rhere were advancements in some accelerators in this gpgpus but not general gpu performance, it was the algorithms that got more and more refined and optimized and able to utilize that better and better

Why is Blackrock selling 1,000 copies of the same Bitcoin under their “derivative” scheme?

Random companies saying they have AI breakthrough to rocket up stock price

Just go run a Kwik-E-Mart.
What a joke. This company is garbage.

Substance-free "journalism."

when i see the founder is an Indian i lost interest

It smells like 6x more bull** than competitors

We'll see

How this scrap could be a competitor!

Well, not yet. D-Matrix has a long way to go before beating Nvidia. Its only edge right now is that it isn’t shackled by U.S. export controls, unlike Nvidia. But when it comes to raw innovation, Nvidia is still miles ahead.

Major biotech company acquiring pro kidney buy now

Blah, blah, blah. It's the same mindless AI slop and PR spin every day, making easy money from our "eyes on the screen". If AI is so good why aren’t the AI “experts” and influencers on their super yachts ? LOL. We know the truth. AI companies are shonks and grifters. AI can’t “do” maths, it doesn’t store knowledge, it has no verification system, it can’t be used in law, it’s not always right, and it can hallucinate. How stupid is that ?! Soon the AI fraudsters, charlatans and their clueless fanboys will disappear. Hooray

it is just raw power no real use just benchmark stuff in real world use cases unoptimized hardware doesnot matter after all these are asics not flexible enough to get real power unless tpus like years of optimizations and robust software stack

Even if it really better than Nvidia's China will still want it 🤣

Definitely no surprise to see new dedicated AI chips being faster than GPU, surprised to not see Nvidia doing something like that

Raw performance is only one piece of the equation. The real challenge is building an ecosystem that developers want to use and customers can deploy at scale. Competition is healthy, but the AI race will be won by companies that combine great hardware with software, networking, manufacturing, and access to global markets.

bs

you come for the king you better not miss