SK hynix is aiming the tape-out of its HBM4 memory by October and will go on to power NVIDIA's next-gen Rubin AI chips.
SK hynix's HBM4 memory for the next-gen NVIDIA AI Chips will be ready to tape out in October
SK Hynix, a key player in supplying HBM (High Bandwidth Memory) to NVIDIA has set its plans to 'tape-out' the HBM4 memory for next-gen NVIDIA AI chips. As per the report, the company has also prepared for HBM4 tape-out for AMD AI chips but only after a few months. ZDNet reports that the 6th generation HBM memory is nearing completion & the design will be sent for manufacturing in October.
'Tape-Out' essentially refers to the final stage of design, which involves the completion of the complete chip design, HBM in this case, to ensure fully functional memory. The design process is followed by validation and verification before the HBM memory can be sent for manufacturing and looks like SK Hynix is keeping pace with its rival, Samsung.
SK Hynix is the main player supplying NVIDIA with its HBM memories for its AI chips and it already took the lead to supply its 5th gen HBM3E a few months back. HBM4 memory will be another significant step in the company's HBM development timeline since it's the fastest DRAM, delivering exceptional power efficiency and higher bandwidth.
HBM4 will offer twice the channel width compared to HBM3E i.e., 2048 bit vs 1024 bit, expanding the route for faster data transfer, which is crucial for higher performance. HBM4 allows stacking 16 DRAM dies compared to 12 on HBM3E, offering support for 24Gb- and 32Gb layers. This will result in a capacity of 64GB in a single stack compared to 32GB on HBM3E.
SK Hynix already promised to offer 20 to 30 times higher performance than its existing HBM memory recently, and this further intensifies the competition as Samsung has already planned to tape out their own HBM4 memory in the next quarter. While Samsung didn't achieve success with passing quality checks for its HBM3E as quickly as SK Hynix, it looks like SK Hynix has a big rival ahead as the company may be looking forward to providing AMD and NVIDIA its HBM4 memory chips.
SK Hynix has reportedly formed development teams to supply HBM4 chips to NVIDIA and AMD and the memory is supposedly entering mass production by the end of the next year. The mass production is coinciding with Samsung's plan to mass produce its HBM4 at the end of 2025, which means both companies are looking forward to supplying newer HBM memory for NVIDIA's Rubin-based AI chips.
News Source: zdnet
About the author: Sarfraz Khan is a hardware reporter with a focus on PC components and the builder community. With years of experience writing about PC hardware and laptops, his work has been featured on several reputable technology publications. Sarfraz's hands-on experience is demonstrated through his first-person accounts of using and comparing different hardware configurations, providing practical and relatable insights for everyday users. His technical analysis is respected by peers in the enthusiast community and has been cited by specialized hardware sites such as Germany's Igor's Lab.
Follow Wccftech on Google to get more of our news coverage in your feeds.
