NVIDIA, IBM, Intel And SK Hynix Reveal New AI Chips And More

Hot Chips 2024, the globally renowned symposium on high-performance chips, was held between August 25 and August 27. The event witnessed a wide range of new and upcoming semiconductors presented by the biggest chip manufacturers, along with updates about other related products.

This included names such as industry-leader NVIDIA, IBM, Intel and more.

It’s no surprise that the products on display were centered around and dominated by artificial intelligence (AI) technology, especially in the wake of the growing popularity of generative artificial intelligence (GenAI).

So, what did the chipmakers announce and what benefits will their products bring users? Let’s explore!
 

What’s New With NVIDIA?

 

• At Hot Chips 2024, NVIDIA focused on its upcoming Blackwell GPU, however, more than spending loads of time on the chip itself, the company spoke about the AI ecosystem it powers.

• It spoke about how Blackwell brings together multiple chips, systems and its NVIDIA CUDA software to power the next generation of AI across use cases, industries and countries, just as it planned.

• NVIDIA announced the launch of new CUDA libraries to expand accelerated computing, which helps reduce energy consumption and costs in data processing, AI data curation, 6G research, AI-physics and more.

• As such, the company now provides over 400 libraries through CUDA.

• Coming back to the powerful, mighty and highly anticipated Blackwell chip, the company published a blog post detailing the latest round of MLPerf industry benchmarks (Inference v4.1), where Blackwell set a new standard for generative AI in its debut.

• The NVIDIA Blackwell platform outperformed the NVIDIA H100 Tensor Core GPU on MLPerf’s biggest LLM workload, Llama 2 70B, offering 4x more performance.

• At the same time, the NVIDIA H200 Tensor Core GPU also delivered outstanding results on every benchmark.

• Hot Chips also saw NVIDIA share updates on research on liquid cooling for data centers, including details about a multi-node, liquid-cooled, rack-scale solution connecting 72 Blackwell GPUs and 36 Grace CPUs.

• This solution is said to bring 30x LLM inference, 4x better performance for LLM training and 25x improved energy efficiency in comparison to the NVIDIA H100 Tensor Core GPU.

• NVIDIA also announced its financial results for Q2 2025, which saw a revenue gain of $30.0 billion, up 15% from the previous quarter and up 122% from a year ago.

• For Q3 2025, the company is expecting its revenue to be $32.5 billion (+/- 2%).

• However, NVIDIA’s forecasts didn’t meet the expectations of investors as the company lowered AI-linked chip stocks, leading to a 5-6% drop in its shares losing out around $150-$175 billion in market value. Its shares also dropped 7% in Frankfurt.

• This was joined by the drop in shares for AMD (4%), Broadcom (4%), SK Hynix (4.5%) and Samsung (2.8%).

 

What’s New With IBM?

 

• At the 2024 edition of Hot Chips, IBM revealed architecture details for its upcoming IBM Telum II Processor and IBM Spyre Accelerator.

• The new products are set to scale processing capacity across next generation IBM Z mainframe systems, boosting the use of traditional AI models and LLMs through a new ensemble method of AI.

• IBM’s Telum II Processor is designed to power next-generation IBM Z systems and comes with increased frequency, memory capacity, 40% stronger cache and integrated AI accelerator core and a new, coherently attached Data Processing Unit (DPU).

• The processor will be manufactured by long-time partner Samsung Foundry on its high performance, power efficient 5nm process node.

• The IBM Spyre Accelerator adds AI compute capability to the Telum II processor.

• Both products are expected to be available in 2025.

• Tina Tarquinio, VP, Product Management, IBM Z and LinuxONE, said, “Our robust, multi-generation roadmap positions us to remain ahead of the curve on technology trends, including escalating demands of AI.”

• “The Telum II Processor and Spyre Accelerator are designed to deliver high-performance, secured, and more power efficient enterprise computing solutions. After years in development, these innovations will be introduced in our next generation IBM Z platform so clients can leverage LLMs and generative AI at scale.”

 

What’s New With Intel?

 

• Intel’s campaign at Hot Chips 2024 covered advancements across AI use cases, spanning data center, cloud, network, edge computing and PC enhancements, presenting four technical papers about its new offerings.

• The company shared details about its Xeon 6 system-on-chip (SoC) design and how it helps overcome edge-specific challenges such as unreliable network connections and limited space and power. It is touted to be the company’s most edge-optimized processor to date.

• The new chip enhances performance, power efficiency and transistor density compared to previous technologies.

• Intel’s showcase included the industry’s most advanced and first-ever fully integrated optical compute interconnect (OCI) chiplet for high-speed AI data processing.

• Ahead of this, Intel showcased its Lunar Lake, which is poised to improve x86 power efficiency while delivering leading core, graphics and client AI performance.

• This comes with 40% lower power usage in comparison to previous generations with up to 4x faster performance.

• Intel also spoke about the Intel Gaudi 3 AI accelerator, which helps beat the cost and power challenges observed in the training and deployment of generative AI models that require extensive compute power.

• More details of the Gaudi 3 AI accelerators and Xeon 6 will be shared during a launch event in September.

• This was followed by an announcement conveying the expansion of its Xeon workstation processor family with the new Intel Xeon W-3500 and Intel Xeon W-2500 desktop workstation processors.

• These processes can deliver the compute power and reliability that professional creators, researchers, engineers and software developers require.

 

What’s New With SK Hynix?

 

• Although it wasn’t a part of Hot Chips 2024, the world’s second-largest memory chip maker, SK Hynix, announced it has developed the industry’s first 10-nm, 16Gb DDR5 chip.

• This consists of the sixth-generation 10-nanometre DRAM (dynamic random access memory) chip, which comes with cost and power-efficiency benefits, reducing costs by around 30% and power efficiency by 9%.

• The chip will be ready to mass produce within a year and start volume shipments in 2025.

• Kim Jonghwan, Head of DRAM Development, said, “We are committed to providing differentiated values to customers by applying the 1c technology equipped with the best performance and cost competitiveness to our major next-generation products including HBM1, LPDDR62, and GDDR73.”

• “We will continue to work towards maintaining the leadership in the DRAM space and position as the most-trusted AI memory solution provider.”

Do you think these moves by the chip makers will propel them to the top of the AI chip industry?

Let us know in the comments below!