"H100" content on TweakTown - Page 1

NVIDIA A100 7936SP AI GPU hits China with more CUDA cores and more HBM memory than the regular A100 with 80GB versus 96GB on the new A100 in China.

SK hynix has high hopes for its US operations, with next-generation memory products will be delivered with next-gen advanced packaging technologies.

NVIDIA's Hopper H100 AI GPU shortages are getting better: lead times are dropping from around 4 weeks, to AI GPU customers now waiting 8-12 weeks.

Meta's next-gen MTIA AI processor is made on TSMC 5nm, up to 1.35GHz frequency, PCIe Gen5 x8 interface, to fight NVIDIA in the cloud business.

AMD's upgraded MI350X AI GPU teased: newer 4nm process node at TSMC, upgraded HBM3E memory that will better fight NVIDIA H200, B200 AI GPUs.

Intel's new Gaudi 3 AI accelerator is official: 64 Tensor Cores, 128GB of HBM2e memory with up to 3.7TB/sec memory bandwidth, and 900W of power on air.

SpaceX and Tesla CEO Elon Musk says that AGI being smarter than the smartest human is going to happy 'probably next year, within two years'.

Elon Musk said that training Grok 2 used around 20,000 NVIDIA H100 AI GPUs, while Grok 3 training will require 100,000 H100 AI GPUs.

Tesla has between 30,000 and 350,000 of NVIDIA's H100 AI GPUs, while Elon Musk is also training xAI with a huge stack of high-end AI GPUs.

OpenAI video tool 'Sora' compute estimates for large-scale deployment: 720,000 x NVIDIA H100 AI GPUs, worth over $21 billion in total.

Micron says that all of its HBM memory is sold out for 2024, and most of its HBM supply is disappearing for 2025... insatiable AI GPU demand driving force.

Meta has reportedly placed early orders for NVIDIA's new Blackwell B200 AI GPU, where it expects initial shipments later this year.

TSMC is considering building advanced packaging capacity in Japan, with new CoWoS packaging tech plant in Japan being the first outside of Taiwan.

NVIDIA GTC 2024 kicks off tomorrow, where we will be introduced to the next-gen Blackwell GPU architecture, and new B100 and B200 AI GPUs.

Samsung will have its HBM inside of NVIDIA's upcoming H200 AI GPU, also receiving certification for AMD's new Instinct MI300 AI GPU.

NVIDIA revenue expected to grow to a huge $130 billion in 2026 thanks to the future of AI GPUs, which the company absolutely dominates.

By the end of 2024, Meta plans to have 350,000 NVIDIA H100 GPUs powering its AI infrastructure - it's all part of its 'ambitious' AI roadmap.

NVIDIA is the 'kingmaker' according to analyst firm Omdia, which predicts that NVIDIA will make $87 billion from data center GPUs in 2024 alone.

SK hynix reportedly provided initial samples of its new 12-layer HBM3E memory to NVIDIA last month, offering 36GB HBM3E per stack.

AMD has been halted by the US Commerce Department, banning its AI processors from shipping to China, even though it was cut-down AI GPU.

NVIDIA CEO Jensen Huang asked on when AI computers will think like humans, he said AGI will arrive soon, should do every human test or exam in 5 years.

SK hynix Vice President Son Ho-young says he wants to see his company become the 'total AI memory provider' with its HBM on next-gen AI GPUs.

NVIDIA's super-expensive H100 AI GPUs are being offloaded as AI GPU customers, reportedly more cost-effective to rent AI processing through cloud providers.

AMD confirms it's working on beefed-up Instinct MI300 AI GPU with ultra-fast HBM3e memory, regular Instinct MI300X AI GPU features HBM3 non-e memory.

NVIDIA reportedly delaying data center, AI GPU shipments if they find out a customer is talking to AI GPU competitors, ex-Radeon GPU boss chimes in.

OpenAI's surprise release of text-to-video AI superstar, Sora, has poured 'cold water' on China's advancements into AI, as it's now years away from the US.

NVIDIA is reportedly gearing up to release two new AI GPUs for the Chinese market, which the company wants to see continued market share growth in China.

AMD's next-gen Instinct MI400X AI accelerator expected in 2025, will compete directly against NVIDIA's next-gen Blackwell B100 AI GPU.

NVIDIA says supply of its Hopper AI GPUs is improving, with demand 'very strong' while expecting next-gen B100 to be 'supply constrained'.

Baird raises the target price of NVIDIA shares from $750 to $1050: AI GPU shipments are ramping up, new H200 and B100 AI GPUs are around the corner.

NVIDIA briefly flew over the market cap of Alphabet (Google) and Amazon, with $650 billion in market cap added since January 1... incredible stuff.

Tesla will eventually spend 'several billion dollars' on AI GPUs, with $500M+ being spent on its Dojo AI supercomputer buying 10,000 x NVIDIA H100 AI GPUs.

NVIDIA will continue to struggle doing AI GPU business in China, with the Indian market now opening up with a huge $1 billion+ order for AI GPUs.

The semiconductor industry has the largest boost in over 10 years, largely driven by AMD and NVIDIA... especially NVIDIA with its massive AI GPU sales.

NVIDIA reportedly plans mass production of its new H20 AI GPU designed exclusively for China, in Q2 2024, meeting the updated US sanctions.

NVIDIA has placed advanced payments for next-gen HBM3e memory with SK hynix and Micron for its next-gen H200 'Hopper' and B100 'Blackwell' AI GPUs.

NVIDIA will have over 3.5 million H100 AI GPUs in 2024, using around 13,000 GWh of electricity per year, as much as countries like Guatemala/

Intel CEO Pat Gelsinger: NVIDIA was 'extraordinarily lucky' with AI GPU dominance, adds that 'they didn't even want to support their first AI project'.

Intel says its next-gen Gaudi 3 AI accelerator will compete against NVIDIA's current-gen H100 AI GPU, and AMD's new Instinct MI300X in AI performance.

Moore Threads introduces the new MTT S4000 AI GPU with 48GB GDDR6 memory, 25 TFLOPs of compute performance, on the PCIe Gen5 x16 standard.

The battle between two AI GPUs heats up with AMD updating its benchmarks in response to NVIDIA. Showing that yes, the MI300X is faster than the H100.

NVIDIA responds to AMD's showcasing the MI300X being faster than the H100 GPU. Here are the real results where the H100 is up to 2X faster.

Intel CEO Pat Gelsinger has fighting words for NVIDIA and AI GPUs: 'you know, the entire industry is motivated to eliminate the CUDA market'.

Analyst expects NVIDIA AI GPU shipments to surge another 150% year-over-year in 2024, led by the new H200 GPU and next-gen Blackwell B100 GPUs.

AMD has officially launched its new flagship AI GPU accelerator, the new Instinct MI300X, which is up to 60% faster than NVIDIA's H100 AI GPU.

AMD's new Instinct MI300X accelerator will be unveiled soon, takes the AI accelerator fight to NVIDIA against its H100 and H200 Hopper AI GPUs.

Here's one way to get around AI regulations - build an ocean-based platform that's its own 'state' and is packed with NVIDIA H100 GPUs (10,000 of them).

The NVIDIA H100 Tensor Core GPU is not built for gaming, but that hasn't stopped people from using the $40,000 GPU to play PC games.

It's starting to look like we're getting pretty close to seeing a new GPU draw 1KW of power as standard, with AMD's new Instinct MI300X rated at 750W.

GPU demand exceeding supply, sound familiar? This time it's all about high-performance chips like NVIDIA's A100 and H100 - and TSMC is trying to keep up.