Dr. Chang-Gyu Hwang, President of Samsung Electronics' Device Solution Network Memory business, predicts that the advent of an IT era will usher in a new round of high growth for memory devices. His "Theory of New Memory Semiconductor Growth in the IT Era" offers new challenges and hope for both businesses and academics involved in semiconductors. Memory bit capacity has grown an average of 70% annually in the decade ending in 2000. However, the growth curve will begin to skyrocket in approximately 2005.
Growth in memory demand between 1990 and 2005 is mainly being driven by increased memory capacity in separate systems such as PCs, servers and mobile phones. Around 2005, almost all major electronic systems will have been transformed to the digital format and the paradigm will shift. The theoretical basis for growth in memory capacity will explode.
Memory consumption will expand from increased PC functions, game product applications and 3G mobile phones with video capabilities. From around 2005 to 2010, digital TVs, home networking systems and other digital consumer electronic products are expected to become commonplace, boosting memory consumption sharply.
There will be an upcoming change in memory storage media. Currently the magnetic tape used to record audio and video data is being replaced by CDs, DVDs and other disc formats. In the future, memory card use will be expanded from digital cameras and MP3 players to include all digital products. Around 2005, digital video will be stored in 5-gigabyte memory cards, and memory cards will completely replace CDs sometime thereafter.
Turning to technology, the gap in operation frequency would narrow between the memory and the CPU. Memory, which currently runs at 400MHz (or less), is to increase almost fourfold to reach 1.5GHz by 2010. Storage density will also go from the current 350Mb/cm2 to 20-40Gb/cm2, an increase of 80 times. At the same time, transmission speed will accelerate about tenfold, from 0.8Gbps to 12Gbps.
In addition, it is forecast that the future progression of semiconductor memory functions would evolve through three stages, from the "universal standard memory" of the past to "system solution memory" of the present and onto to "fusion memory" in the future.
By "universal standard memory", this means synchronous DRAMs and other singular devices that serve commonly as the main platforms for PCs, servers and elsewhere. Growth in this area continues to slow.
"System solution memory" will be the mainstream between now and 2005. This refers to various memory devices that have been optimized for specific applied systems. Examples include high-capacity memory for servers, power-saving memory for mobile phones and extra-fast memory for network systems. Also predicted is the development of "fusion memory," which will integrate the high capacity of DRAM, the high speed of SRAM, the non-volatile storage capability of flash memory plus a logic element. "Fusion memory" would be a total memory solution, boosting all the system functions that users want. "Fusion memory" is user friendly.
Predictions about fusion memory and theory on new memory growth are linked with absolute theories such as Moore's Law. Gordon Moore, one of the Intel founders, contends that chip functions have been doubling about every 18 months since 1965. What is presented is a departure from the view that semiconductor technology and the semiconductor image are developed around CPUs. The paradigm is changing, from the memory perspective, to accommodate diverse users in the future IT era. Memory sales will enter a new high growth phase of 20% per annum. Such a change would have major repercussions on the industry.
