Download the authoritative guide: Enterprise Data Storage 2018: Optimizing Your Storage Infrastructure
Static RAM (SRAM) and dynamic RAM (DRAM) are different types of RAM, with contrasting performance and price levels. Both play a key role in today's technology.
SRAM: is a memory chip that is faster and uses less power than DRAM.
DRAM: is a memory chip that can hold more data than an SRAM chip, but it requires more power.
First, some background. Random access memory (RAM) is a semiconductor device placed on a processor that stores variables for CPU calculations. RAM provides memory locations for requested data (registers). The CPU receives a data read instruction with the data’s memory address or location. The CPU sends the address to the RAM controller.http://o1.qnsr.com/log/p.gif?;n=203;c=204655439;s=10655;x=7936;f=201806121855330;u=j;z=TIMESTAMP;a=20400368;e=i
The controller, in turn, sends the address to the proper pathway, opening path transistors and reading each capacitor value. The read data is transmitted back to the CPU’s cache.
The speed of this read/write operation is called timings. Faster timings with less lag between them result in faster access times and low latency. Slower timings result in lower performance and higher latency. Bandwidth also affect performance: the larger the bandwidth, the more data per second RAM can process and the faster the timings.
In short, if you're shopping for the best SSD for your price point, make sure you know if you're getting SRAM or DRAM.
RAM Structure: SRAM and DRAM
SRAM and DRAM processes data in different ways, depending on the data’s requirements.
There are many combinations and next-generation memory components that build on these two technologies. But it’s important to understand the basics of SRAM and DRAM before delving into newer technologies built on top of them.
• What is SRAM? Each SRAM cell stores a bit using a six-transistor circuit and latch. (DRAM uses transistors and capacitors.) SRAM is volatile but if the system is powered, SRAM retains data values without recharging cells. It is fairly insensitive to electrical noise, which is unwanted electrical signal that interferes with a desired signal. Since it is faster and costs more than DRAM, it normally operates as CPU memory caches or on high-end, high-performance servers. SRAM system memory is typically 20-40ns (nanoseconds).
• What is DRAM? Each DRAM cell stores a bit using a single paired capacitor and transistor. Since single component pairs can create a cell, and billions of them can fit on a single chip, DRAM is capable of very high densities. Like SRAM, DRAM if volatile. But unlike SRAM, each cell must be periodically refreshed since capacitors leak power. It is sensitive to electrical noise. DRAM speeds usually range between 60ns and 100ns – still fast but slower than SRAM. Typical per-second speeds are 20-40GB/s, and continual cell recharging results in higher latency and bandwidth delays than SRAM.
SRAM vs. DRAM in Computers
|Performance||Slower: Off-chip memory with longer access time||On-chip memory with minimal access time; can run at the speed of the host microprocessor|
|Use case||Main memory||Level 1 and Level 2 microprocessor caches|
|Density||Less density per cell (1 transistor per chip) but can pack more cells into space||Denser (6 transistors per chip) but can fit fewer cells into space|
|Power||Generally higher: Capacitors leak power thanks to imperfect insulation, requiring regular power refreshes.||Generally lower: No charge leakage since it changes direction of current through switches instead of leaking power through the capacitor. However, this depends on the application environment and SRAM can consume as much or more power as DRAM.|
|Storage capacity||Larger: Connects directly to CPU bus, volatile storage measured in GBs||Smaller: Acts as cache; storage measured in MBs|
|Volatility||Volatile: Must have active power supply plus frequent charges while active.||Volatile: Does not require additional charges while it is receiving power, but eventually loses data without it.|
|Physical placement||Motherboard||Processors or between processor and main memory|
SRAM and DRAM Evolution
Computing developments have impacted computer memory. Let’s look at some major SRAM and DRAM improvements and developments.
- SDRAM – Synchronous Dynamic Random Access Memory (SDRAM) is a type of DRAM that synchronizes with the CPU’s clock cycles, so the memory controller knows exactly when requested data will be ready to access. This cuts access time and improves memory performance.
- DDR – The latest generation of SDRAM is Double Data Rate (DDR). DDR increases speed, reduces power consumption, introduces refresh mechanisms, and adds security functions such as CRC. For example, DDR3 transfers IO data 8 times faster than the speed of its own cells, which enables higher throughput and faster speeds. (Yet it does not lower latency.) Single chip capacities can reach 8GB, which in practice effectively doubles to 16GB.
- SGRAM – Synchronous Graphics RAM is a clock-synchronized DRAM. SGRAM opens two memory pages simultaneously, which emulates dual porting at lower cost
- MDRAM – Modern high-performance VRAM includes Multibank Dynamic RAM (MDRAM). Traditional VRAM presents the entire frame buffer for data access. MDRAM divides memory into 32KB banks for concurrent access.
- VRAM – Video RAM is a type of DRAM that stores image data to send to an LED computer display. VRAM acts as a frame buffer between the user’s display monitor and the processor. The processor initially reads the video data from main storage RAM and writes it to a video RAM format. The frame buffer converts the digital video data into analog signals and send them to the display. Older VRAM is dual-ported, which means that as the CPU processor writes a new frame into video RAM, the monitor is reading from the video and updates its display.
- EDRAM – Enhanced DRAM, or EDRAM, combines SRAM and DRAM to serve Level 2 caches. Typically 256 bytes of SRAM pairs with DRAM. Data read operations check SRAM first for the requested data, and then DRAM if the data is not stored on SRAM.
- WRAM – Window RAM (not related to Microsoft Windows) is a high-performance, dual-ported VRAM. Its architecture yields about 25% greater bandwidth than traditional VRAM, at a lower cost. It accomplishes this with high performance data reads for operations like text drawing and block fills. It uses true color (24-bit color), which makes it ideal for high resolution graphics monitors.
- EDO DRAM – Extended Data Output DRAM (EDO DRAM) pre-fetches the next block of memory while sending the previous block to the CPU. This results in speeds up to 25% faster than standard DRAM.
RAM Research and Development
There are many other iterations of SRAM and DRAM available on the market, as well as disruptive memory technologies that might replace them. This has not yet happened on any wide scale, but some companies are pouring resources into computer memory R&D.
For example, phase-shifting computer memory (PCRAM) is an interesting experiment that seeks to replace DRAM. PCRAM sounds like it’s straight out of science fiction. It switches between two states, one a low-conductive atomic structure and the other a high-conductive crystalline state. By recording 0’s and 1’s according to which state they are currently in, the processor can write and rewrite data.
In earlier experiments, these elements proved unreliable when subjected to strong electric currents. Recently, researchers at China’s Academy of Sciences in Shanghai reported that they have boosted PCRAM performance by a factor of 10, engineered it to be non-volatile, and improved reliability.
The key to this advance turned out to be a shape memory alloy (SMA) composed of magnesium and scandium. Despite progress, PCRAM cannot yet match DRAM’s record of writing and rewriting data many times. (And by “many,” researchers mean “trillions.”) As of now, SRAM and DRAM – and their updated variations – still rule the CPU memory roost.