7
\$\begingroup\$

Historically, when compared to DRAM (dynamic ram), SRAM (static ram) is more expensive with a lower density, but it's much faster and more efficient. As a result, SRAM is reserved for devices, such as processor caches and microcontrollers, where maximum efficiency and or performance are critical. For devices requiring large amounts of memory, such as a PC, SRAM is too expensive despite its advantages over DRAM.

Has any of today's major advances in semiconductor manufacturing carried over to SRAM technology to mitigate some of its issues? Apple Silicon, for instance, integrates the system DRAM into the chip, yielding improved performance and efficiency, but why not use SRAM instead? With transistors nearing the sub-3nm scale, is memory density still an issue? What challenges are holding SRAM back?

\$\endgroup\$
1
  • \$\begingroup\$ Are you talking about SDRAM (a type of DRAM) or SRAM? \$\endgroup\$
    – Hearth
    Commented Jun 24, 2023 at 21:06

2 Answers 2

6
\$\begingroup\$

Has any today's major advances in semiconductor manufacturing,carried over to SRAM technology to mitigate some of its issues?

No they have not, and in fact SRAM density has not scaled down with smaller transistor nodes in recent generations. TSMC for example has had essentially the same SRAM cell size for the last two generations:

https://fuse.wikichip.org/news/7343/iedm-2022-did-we-just-witness-the-death-of-sram/

As SRAM stopped scaling it becomes relatively more expensive, so there is a need to use it judiciously. Usually this means relying on dram as much as possible.

What challenges are holding SRAM back?

SRAM density is mostly limited by the poly pitch (CPP), which hasn't scaled much. Perhaps some future semiconductor technology will address that. Otherwise there will be a need to use alternative memory technologies as much as possible.

\$\endgroup\$
3
  • \$\begingroup\$ Kinda sucks, huh? Nature always throws innovation a curveball. I wonder if GaN or silicone and graphene address this issue. \$\endgroup\$
    – user148298
    Commented Jun 25, 2023 at 19:10
  • 2
    \$\begingroup\$ @user148298 Silicone is not a semiconductor, it's quite a strong insulator. Are you thinking of silicon? In any case, current technology can't produce structures nearly as small on GaN or SiC as we can on silicon; that may change in the future, but right now a single transistor in SiC would be larger than an entire SRAM cell in silicon. \$\endgroup\$
    – Hearth
    Commented Jun 25, 2023 at 19:35
  • \$\begingroup\$ I meant silicene and autocorrect botched it. \$\endgroup\$
    – user148298
    Commented Jun 26, 2023 at 4:24
6
\$\begingroup\$

The simple answer is that SRAM requires more transistors to implement one bit of memory. Several designs exist, and they tend to use four or more MOSFETs per bit.

DRAM on the other hand, only requires one transistor per bit. So it can be made much more dense in terms of bytes per area of silicon, and therefore cheaper.

\$\endgroup\$

Not the answer you're looking for? Browse other questions tagged or ask your own question.