Hot answers tagged

25

In both cases (EEPROM/flash and DRAM) a small (femtofarads) capacitor is used. The difference is the way the capacitor is connected. In the case of DRAM it is connected to the source or drain of a MOSFET. There is a tiny bit of leakage through the transistor channel and the charge will leak off in a relatively short period of time (seconds or minutes at ...


11

Here is a (slightly dated) paper that discusses the differences: http://www.ece.neu.edu/faculty/ybk/publication/ASSESSING_MERDRAM_ELSEVIER.pdf Basically, it boils down to a few important differences. Leakage current. The pass transistors for the DRAM cells must be extremely low leakage, otherwise the leakage current will affect the bit stored in the ...


10

It's basically because the leakage current of a transistor in a DRAM cell is temperature dependent. Here's a typical DRAM cell schematic (taken from that book): And here's a generic graph of how the leakage current varies with temperature for IC-level FETs The latter graph is from another book. This is in fact an exponential dependence on temperature (note ...


9

Why is there an additional capacitor needed in DRAM cell then? Because a memory cell needs at least 1 switch and 1 memory element for it to be addressable, a minimum of two elements. There are memories like SRAM can use only transistors in a bi-stable configuration. From wikipedia "Memory cells that use fewer than four transistors are possible – but, such ...


7

A dedicated memory controller is mandatory for several reasons (built into the CPU or not). Because it is dynamic memory each memory cell has to be 'refreshed' with a strobe every 16mS or so. With some Dram chips having gigabytes of storage, the CPU would spend all of its time refreshing the Dram. Using timed latency or handshake signals (still a known ...


7

This is addressed in the paper, in the second paragraph of section 3: Over time, charge will leak out of the capacitor, and the cell will lose its state or, more precisely, it will decay to its ground state, either zero or one depending on whether the fixed conductor of the capacitor is hard-wired to ground or power. As charge leaks from the ...


6

Simplifying a little bit, think of DRAM as being a 2D array of memory cells1. Each cell in the array is a minuscule capacitor. Along one edge of that array, we have a set of sense amps. There's one sense amp for each cell along that dimension. For either a refresh cycle or a normal read, we activate an entire row (or column, if you prefer to look at it the ...


6

I think that the origin of the name comes from the convention for flip flops. On a D flipflop the data in pin is generally called D, and the data out pin, Q. Since the memories data bus is bidirectional, a data pin can be D when it is input or Q when it is output hence the name DQ.


6

There are complete schematics for the IBM PC/XT in the IBM Personal Computer XT technical reference manual (Appendix D), which you may be able to find on line. The problem here is that, given a strobe line which is activated upon a memory read or write, you wish to generate RAS, CAS and a control line (call it MUX) for the address multiplexer. For ...


5

You sort of answer your own question. How exactly does a computer wait a certain number of clock cycles to receive information? It doesn't know. It must have a memory controller that knows about the available memory. The CPU typically¹ just tries to access an external memory address, and waits for the memory controller to signal that the access is ...


5

It's two inverters that use positive feedback to retain a bit. Once the bit is setup and there remains power, the two inverters will forever be stuck in that same state. To change the bit, you have to override the internal positive feedback of the system. You would do that through M5 and M6.


5

Your question is complicated enough that I'm not even sure what your actual problem is, but I'll try! The "cleanest" 6502-based DRAM design I could find is from the Commodore PET 2001-N. It has a 6502 running at 1 MHz, but the DRAM logic is clocked at 16 MHz, likely to generate all the timings. I have not analyzed the details, but the main action seems to ...


5

It seems that the latest NVDA GPU chips have >10GB of on-chip memory. On-package != on-chip. The HBM are stacks of several chips right next to the main GPU. Consumer grade GPUs have the memory soldered to the PCB in normal BGA packages. GPU memory works "better" because there are no connectors that could impact the signal path between chips, allowing ...


4

We're talking about an SODIMM module here. It has multiple chips on it, and has an overall format of 1G (230) locations of 64 (26) bits each. (Total of 236 bits.) The module contains 16 (24) chips that contain 4G (232) bits each. (Total of 236 bits.) The memory in each chip is organized as 8 (23) banks, each with 64K (216) rows and 1024 (210) columns of ...


4

Yes, they really have that many capacitors in that small of an area. There are two dominant technologies to do this: stacked capacitor DRAMs and trench capacitor DRAMs. Stacked capacitors basically use a number of layers of metal and insulator to build a capacitor of reasonable capacity in a small surface area. Trench capacitor DRAMs basically etch a "...


4

An SRAM cell is actually two crossed feedback loops. Look at its schematic and you will immediately see it. As long as the SRAM has power, the loops are running and maintain their state, even without any input signal.


3

You add the cost of: Connectors The actuall RAM module that might be more expensive than the individual components The fact that if you use a connector, someone will have to manually insert the modules, as I don't think this can be done by machine (in contrast to placing the components directly on the motherboard, as you would do when you don't use modules)....


3

The reason DRAM needs a large storage capacitor is that it has to be able to charge up the bit lines. The bit lines have relatively large parasitic capacitance since they connect all of the transistors in a column. DRAM cells is arranged in a grid. The rows address lines are connected to the gates of the MOSFETS, and the column lines are connected to ...


3

Yes there are standards for physical layout, impedance controlled tracks and track latency skew, physical, functional layout etc. https://www.jedec.org/document_search?items_per_page=60&search_api_views_fulltext=ddr3 free downloads with registration.


3

Well, you're never going to get that exactly. That's the theoretical max bandwidth. However, you can get pretty close. DRAM is set up so that rather large blocks can be read out sequentially with no wait states. Well, you might need some wait states to set it up and a few to move to the next block, but within the block it will transfer at the full rate. ...


3

http://isca2016.eecs.umich.edu/wp-content/uploads/2016/07/8A-1.pdf There is a new research in ISCA, a prestigious conference on DRAM-based FPGA recentely.


3

Yes, that's right. The capacitors are VERY small!


3

Both Static and Dynamic RAMs contain arrays of memory cells with rows and columns, and in theory, there's nothing to stop you from implementing a Static RAM using a multiplexed address bus and hence needing the RAS/CAS signals to manage access. But in practice, nobody does this for two main reasons: The vastly greater complexity of an SRAM cell means that ...


3

Block Ram is a dedicated Ram that does not consume any additional LUT in your design whereas distributed Ram is built up with LUT. In terms of speed the distributed RAM is faster than Block Rams. Generally speaking, if not much Ram is needed you can consider to implement it as a distributed Ram. Some synthesizers may even use distributed Ram if you ...


3

The biggest difference is that SSDs are based on NAND flash cells with the serial structure shown below. Where DRAM accesses all of the bits in a row at the same time, NAND flash serially accesses the row. This significantly slows down read access versus DRAM. Write access is significantly slowed because flash uses a floating gate which requires orders of ...


3

During a read T1 is off and T3 is on. Then T2 controls the read line from the charge on the capacitor. Since it’s gate is very high impedance, no current is drawn from C to do that. That leaves C’s charge, hence the state of the bit, unchanged. T2 is both the strength and weakness of this cell design. Yes, it makes readout nondestructive. But it also ...


3

No. Besides the fact that what you have drawn there is neither an EEPROM nor a DRAM cell, DRAM and EEPROM function in completely different ways; EEPROM relies on charge stored in a floating gate (which can be put there or removed only by tunnelling current), while DRAM is just a (very tiny) capacitor connected to the data line by a FET which turns on or off ...


2

Your question is frankly a bit too broad (although you may not realize that yet). But since you're literally asking for any SPICE model for any DRAM circuit, the Micron page for one of their many DRAM chips [MT41J256M16HA-093] has a HSPICE model. HSPICE is a dialect/implementation of SPICE.


2

Looks to me like it is hybrid memory cube: http://www.micron.com/products/hybrid-memory-cube Just on package. HMC has a high speed serdes controller interface, that connects to a stack of dram dies. You can read more about it there. HMC was my first guess and it's mentioned here: http://www.anandtech.com/show/8217/intels-knights-landing-coprocessor-...


2

The configuration of an FPGA is stored in special SRAM cells (less transistors and lower static current) or in flash memory. This memory needs to be 'read' at every time otherwise the path transistors won't work. DRAM can't be read continously. Producing normal CMOS logic and DRAM logic are different processes. DRAM needs other machines and materials. This ...


Only top voted, non community-wiki answers of a minimum length are eligible