The creation of the world’s fastest HBM3e DRAM has been verified by SK hynix, and samples are now being sent to NVIDIA and other clients for testing purposes.
The HBM3e DRAM from SK Hynix is not only faster, but it also provides higher capacities, improved heat dissipation, and easy compatibility.
It was reported in June that NVIDIA had made a request to SK hynix to sample their next-generation HBM3e DRAM. This request became a reality when NVIDIA introduced its GH200 GPU with enhanced HBM3e DRAM, giving up to 5 TB/s bandwidth per chip. SK hynix was able to fulfill this request.
SK Hynix Inc. has issued the following press release to indicate that it has successfully created HBM3E, the next-generation of the highest-specification DRAM for AI applications that is presently available, and that a customer’s examination of samples is now under way.
On top of its expertise as the industry’s single mass supplier of HBM3, the business claimed that the successful creation of HBM3E, the extended version of HBM3 which offers the world’s finest specifications, comes on top of its experience as the extended version of HBM3. With its expertise as the supplier of the industry’s biggest volume of HBM products and the mass-production readiness level, SK hynix wants to mass-produce HBM3E in the first half of next year and strengthen its unparalleled leadership in the AI memory market. This will be accomplished by using its experience as the supplier of the industry’s largest volume of HBM goods.
The business claims that its most recent product not only satisfies the industry’s most stringent criteria for speed, which is the most important need for artificial intelligence memory devices, but also excels in all other areas, such as capacity, heat dissipation, and user-friendliness.
In terms of processing speed, the HBM3E is capable of handling data at a rate of up to 1.15 terabytes (TB) per second. This is the same as being able to process more than 230 Full-HD movies that are each 5GB in size in a single second.
The cutting-edge technology of the Advanced Mass Reflow Molded Underfill, also known as MR-MUF**, has been integrated onto the most recent version of the product, which has resulted in a 10% increase in the rate at which heat is dissipated. In addition to this, it offers backward compatibility, which makes it possible to implement the most recent product onto systems that have already been designed for HBM3 without having to make any modifications to their structures or designs.
MR-MUF: a process of attaching chips to circuits and filling the space between chips with a liquid material when stacking chips instead of laying a film to improve efficiency and heat dissipation
Backward Compatibility: an ability to allow interoperability between an older and a new system without modification to the design, especially in information technology and computing spaces. A new memory product with backward compatibility allows continued use of the existing CPUs and GPUs without modifications to design
“We have a long history of working with SK hynix on High Bandwidth Memory for leading-edge accelerated computing solutions,” said Ian Buck, Vice President of Hyperscale and HPC Computing at NVIDIA. “We look forward to continuing our collaboration with HBM3E to deliver the next generation of AI computing.”
According to Sungsoo Ryu, who serves as the Head of DRAM Product Planning at SK hynix, the business has enhanced its market leadership via the creation of HBM3E. This was accomplished by further expanding the completeness of the HBM product range, which is in the limelight as a result of the growth of AI technology. “SK hynix will also work toward achieving a rapid turnaround in its business by increasing the supply share of the high-value HBM products,”
