3-Dimensional Read-Only Memory

- (3D-ROM)

 

U.S. Patent 5,835,396 & Chinese Patent ZL98 1 19572.5

 

3D-ROM is a new non-volatile semiconductor memory with lower cost (~10x), higher capacity(>4x) and comparable bandwidth (Table 1). It is compatible with standard CMOS process. More importantly, 3D-ROM can be readily integrated with RAM/flash ROM. 3-dimensional integrated memory(3DiM, 3DiM=3D-ROM+RAM) will enable "Computer-on-a-Chip" and other applications.

 

Technical details about 3D-ROM can be found in the article "3D-ROM - A First Practical Step towards 3D-IC" published in Semiconductor International, July '00.

 

 

Fallacies

 

 

Fallacy #1:

Flash offers unlimited programmability, there is no need to develop another ROM with limited programmability!
    rebuttal...

 

 

Fallacy #2:

Because it has a long latency (~ms) (Table 1), 3D-ROM cannot be used in high-speed applications!
    rebuttal...

 

 

Table 1.

3D-ROM Prosposed Spec. vs. Conventional Storage Devices targeted for portable applications.

 

 

 

 

Cost

Cell Area

Capacity

Latency

Bandwidth*

3D-ROM

~$0.03/MB

4F2

120MB

0.5-5ms

>0.5GB/s

SRAM

~$10/MB

~100F2

1MB

3-10ns

~5GB/s

DRAM

~$0.25/MB

8F2

32MB

~50ns

~0.5GB/s

Flash

~$0.5/MB

10F2

16MB

~50ns

~0.5GB/s

mask-ROM<$0.3/MB>

 

8F2

32MB

~50ns

~0.5GB/s

IBM MicroDriveTM

~$0.3/MB

-

1GB

~12ms

~13MB/s

 

* Bandwidth refers to the applications where embedded memory is used.

 

 

Fallacy #1:

Flash offers unlimited programmability, there is no need to develop another ROM with limited programmability!

 

       Ideally, an IC designer would desire an infinitely large non-volatile memory (NVM) with unlimited programmability. In principle, flash alone can satisfy this desire. Against this backdrop, the idea of developing another ROM (particularly a ROM with limited programmability) might seem odd at first. WHY Bother? Before we answer this question, let's take a look at the memory hierarchy in a typical computer system. Ideally, one would desire unlimited amounts of fast memory. In principle, SRAM alone can fulfill this need. Then, WHY would we bother to develop DRAM and hard-disk drive (HDD)? The answer is simply cost. By carefully balancing the memory size at different levels, the cost and performance of a computer system can be optimized. This is a simple case of trading speed for cost.

 

       In the case of NVM, we can also trade programmability for cost. ROM, even though with limited programmability, is included in an IC to reduce overall system cost. This sounds against the current industry trend. As any industry observer can point out, more and more IC manufacturers are replacing mask-ROM with embedded flash. To understand this trend, we need to compare the price of the mask-ROM and flash. With the continuous price drop of the flash memory, the price differential between them is ~2x (Table 1). Thus, the mask-ROM can no longer provide significant cost-saving over the flash. Weighing cost and programmability, industry would naturally choose the flash. However, with a price differential of >10x against the flash, it is a different scenario for the 3D-ROM. By carefully balancing the amount of the 3D-ROM and flash, the cost and programmability of an IC system can be optimized.

 

 

 

       In conjunction with flash, 3D-ROM forms the core NVM of an IC. 3D-ROM is particularly suitable to store system software, which is fixed and requires a large memory space. With most information stored in the 3D-ROM, the requirement on the flash capacity can be eased. This memory architecture allows low-cost, large-capacity storage in the 3D-ROM while maintaining the flexibility and field-upgradibility of the flash. It is particularly suitable for object-oriented softwares such as Java. Java has a large library that contains most of its functionality. It progresses and extends its facilities by the addition of libraries. While the Java library demands large memory space, Java programs are generally compact and can be frequently updated from the internet. To embed Java in an IC, the original release of the Java library can be hard-coded in the 3D-ROM, while new Java library release and Java programs can be stored in the flash. Combining 3D-ROM with flash, embedded software can stay current and the IC can have an extended field lifetime.

 

 

 

Fallacy #2:

Because it has a long latency (~ms) (Table 1), 3D-ROM cannot be used in high-speed applications!

 

 

 

       Most IC designers are accustomed to semicondcutor memory with a latency of tens of nanoseconds. At this latency level, there is no speed mismatch between the processor and memory, thus rendering a simple IC design. The latency of the 3D-ROM is ~ms (Table 1), far longer than any other semiconductor memory. Intuitively, one would imagine that the 3D-ROM is not suitable for high-speed applications. Is this true?

 

 

 

       Again, we refer back to hard-disk drive (HDD). With a latency of ~10ms, can the HDD not be used in high-performance computers? The answer is obvious. How does the HDD manage to handle high-speed applications and how can the same engineering philosophy be applied to the 3D-ROM? We will leave this as a mental exercise for the interested readers.

 

 

 

 

 

 

| HOME | About 3D-ROM | Other Patented Technologies | Contact Us |