Support NeoGAF

[winjer]

Step Aside Intel, TSMC Chairman Says NVIDIA Will Be World's Biggest Chip Company

Santa Clara, California-based semiconductor designer NVIDIA Corporation will be the biggest chip company in the world in 2023, according to the Taiwan Semiconductor Manufacturing Company's (TSMC) chairman, Dr. Mark Liu. NVIDIA was the darling of Wall Street in 2023 as investors rushed to pile in...

wccftech.com

Santa Clara, California-based semiconductor designer NVIDIA Corporation will be the biggest chip company in the world in 2023, according to the Taiwan Semiconductor Manufacturing Company's (TSMC) chairman, Dr. Mark Liu. NVIDIA was the darling of Wall Street in 2023 as investors rushed to pile in the shares after the success of OpenAI's ChatGPT in late 2022. NVIDIA's shares provided the investors some respite from the stock market devastation due to high interest rates, and the firm kept up with the hype as its latest quarterly earnings report reported 3x annual revenue growth.

The firm's latest report for the third quarter of its fiscal year 2024 saw NVIDIA's revenue grow by 206% and its profit grow by 588% annually in a remarkable set of results for a firm of NVIDIA's size. This pushed NVIDIA's revenue to sit at a cool $18.12 billion for the period. At the heart of this explosive revenue growth was NVIDIA's bread-and-butter Data Center business division. Data Center revenue during Q3 was $14.5 billion, a sizeable growth over the $3.5 billion during the year-ago quarter.

On the other hand, AMD and Intel's latest quarterly revenues sit at $5.8 billion and $14.16 billion, respectively. Additionally, NVIDIA's latest market capitalization is currently $1.19 trillion, while the market cap of AMD and Intel is $198 billion and $185 billion respectively.

NVidia selling AI chips now, is liken to selling shovels during the gold rush.

 

[Bashtee]

Yeah, the gold rush comparison is very on point. They earned it, though. NVIDIA pushed CUDA support hard almost 15 years ago, and AMD is still playing catch-up. HPC might be a bit different with ROCm, but anything AI is usually easier to launch with a NVIDIA GPU.

 

[crackajack]

I would have thought they already are. Intel and AMD are fighting in the CPU space, Athlon, Core2, Zen were always rather close with some edge to one or the other, but with thin margins in the mass market products, while Nvidia is basically without proper competition since the 9800Pro days and turned themselves into some luxury brand like Porsche, still selling some Boxsters to everyone, but focusing on 911 and Cayenne and leaving the Honda Civic, VW Golf let alone Skoda Rapid or Dacia Sandero type of prices behind.

 

[Hudo]

Get ready for $5000 consumer cards.

 

[E-Cat]

Get ready for $5000 consumer cards.

Consumer cards are so yesterday.

 

[Wildebeest]

It seems like a lot of the hardest computing problems are more practically solved on GPU than CPU, and Quantum Computers are still a dream. But a large part of that is the software development environment nvidia have built, which is not so easy for Intel or AMD to just replace with equivalent hardware made using the same chip fabs.

 

[Gen.Grievous]

Consumer cards are so yesterday.

They have plenty of left over silicon they can put in consumer cards to sell. It's still a sizeable portion of their revenue.

 

[E-Cat]

They have plenty of left over silicon they can put in consumer cards to sell. It's still a sizeable portion of their revenue.

For now. The crossover to getting most of their revenue from datacenters has only just occurred.

 

[Griffon]

They have plenty of left over silicon they can put in consumer cards to sell. It's still a sizeable portion of their revenue.

I remember a time when we didn't just get the "left overs".

 

[Buggy Loop]

They have plenty of left over silicon they can put in consumer cards to sell. It's still a sizeable portion of their revenue.

It’s peanuts and low margin compared to AI cards though

Legit, I think desktop GPU will clash during peak AI supply.

They’ll either raise up prices so high that you’ll be connecting to GeForce Now to pay to play via cloud, or they’ll distance the releases quite far apart and with maybe only a select few options. Or both..

Fab time will not just be a rare ressource, every goddamn chipmaker will be in 100% on AI. So there’s no form of competition rising from this void that I can see. AMD will have AI’s dick in their mouth too, they already have eyes on it

 

[Bernoulli]

Yeah, the gold rush comparison is very on point. They earned it, though. NVIDIA pushed CUDA support hard almost 15 years ago, and AMD is still playing catch-up. HPC might be a bit different with ROCm, but anything AI is usually easier to launch with a NVIDIA GPU.

AMD Launches Instinct MI300X AI GPU Accelerator, Up To 60% Faster Than NVIDIA H100

AMD Launches Instinct MI300X AI GPU Accelerator, Up To 60% Faster Than NVIDIA H100

AMD has announced the launch of its flagship AI GPU accelerator, the MI300X, which offers up to 60% better performance than NVIDIA's H100.

wccftech.com

 

[Aces High]

I'm expecting Nvidia to dominate cloud streaming in the future.

Geforce Now is technically superior to Sony's and Microsoft's solution by a huge margin and I'm expecting this advantage to grow with ai being used in rendering and network technology.

I never liked Nvidia, but their innovations in the past years have been impressive. They're pushing like crazy.

 

[SportsFan581]

Hopefully they at least leave the 50 series cards priced below $1k.

 

[winjer]

I would have thought they already are.

This news is already a month old. And it's about NVidia surpassing Intel by the end of 2023.
For whatever reason no one cared about this article a month ago. But better late then never, I guess.

 

[Buggy Loop]

AMD Launches Instinct MI300X AI GPU Accelerator, Up To 60% Faster Than NVIDIA H100

AMD Launches Instinct MI300X AI GPU Accelerator, Up To 60% Faster Than NVIDIA H100

AMD has announced the launch of its flagship AI GPU accelerator, the MI300X, which offers up to 60% better performance than NVIDIA's H100.

wccftech.com

Ah, AMD "up to"™

They forgot a little asterisk where they didn't use Nvidia's TensorRT-LLM nor show real-world inferences per second results

While AMD has yet to catch up to H100 in real results, they're already on H200

 

[Magic Carpet]

Stock was around 500 the other day. It keeps climbing.