Spring 2002EECS150 - Lec19-memory Page 1 EECS150 - Digital Design Lecture 20 - Memory April 4&9, 2002 John Wawrzynek

Spring 2002EECS150 - Lec19-memory Page 2 Memory Basics Uses: –data & program storage –general purpose registers –buffering –table lookups –CL implementation –Whenever a large collection of state elements is required. Types: –RAM - random access memory –ROM - read only memory –EPROM, FLASH - electrically programmable read only memeory Example RAM: Register file regid = register identifier sizeof(regid) = log2(# of reg) WE = write enable

Spring 2002EECS150 - Lec19-memory Page 3 Register File Internals Functionally the regfile is equivalent to a 2-D array of flip-flops: Cell with write logic: How do we go from "regid" to "SEL"?

Spring 2002EECS150 - Lec19-memory Page 4 Regid (address) Decoding

Spring 2002EECS150 - Lec19-memory Page 5 Standard Internal Memory Organization Special circuit tricks are used for the cell array to improve storage density. (We will look at these later) RAM/ROM naming convention: –examples: 32 X 8, "32 by 8" => 32 8-bit words –1M X 1, "1 meg by 1" => 1M 1-bit words

Spring 2002EECS150 - Lec19-memory Page 6 Read Only Memory (ROM) Functional Equivalence: Of course, full tri-state buffers are not needed at each cell point. Single transistors are used to implement zero cells. Logic one’s are derived through precharging or bit-line pullup transistor.

Spring 2002EECS150 - Lec19-memory Page 7 Column MUX in ROMs and RAMs: Controls physical aspect ratio In DRAM, allows reuse of chip address pins

Spring 2002EECS150 - Lec19-memory Page 8 Cascading Memory Modules (or chips) example 256 X 8 ROM using 256 X 4 parts: example: 1K X * ROM using 256 X 4 parts: each module has tri-state outputs:

Spring 2002EECS150 - Lec19-memory Page 9 Definitions Bandwidth: Total amount of data accross out of a device or across an interface per unit time. (usually Bytes/sec) Latency: A measure of the time from a request for a data transfer until the data is received. Memory Interfaces for Acessing Data Asynchronous (unclocked): A change in the address results in data appearing Synchronous (clocked): A change in address, followed by an edge on CLK results in data appearing. Somtimes, multiple request may be outstanding. Volatile: Looses its state when the power goes off.

Spring 2002EECS150 - Lec19-memory Page 10 Example Memory Components: Volatile: –Random Access Memory (RAM): DRAM "dynamic" SRAM "static" Non-volatile: – Read Only Memory (ROM): Mask ROM "mask programmable" EPROM "electrically programmable" EEPROM "erasable electrically programmable" FLASH memory - similar to EEPROM with programmer integrated on chip

Spring 2002EECS150 - Lec19-memory Page 11 Volatile Memory Comparison SRAM Cell Larger cell  lower density, higher cost/bit No refresh required Simple read  faster access Standard IC process  natural for integration with logic DRAM Cell Smaller cell  higher density, lower cost/bit Needs periodic refresh, and refresh after read Complex read  longer access time Special IC process  difficult to integrate with logic circuits word line bit line word line bit line

Spring 2002EECS150 - Lec19-memory Page 12 In Desktop Computer Systems: SRAM (lower density, higher speed) used in CPU register file, on- and off-chip caches. DRAM (higher density, lower speed) used in main memory Closing the GAP: Innovation targeted towards higher bandwidth for memory systems: –SDRAM - synchronous DRAM –RDRAM - Rambus DRAM –EDORAM - extended data out SRAM –Three-dimensional RAM –hyper-page mode DRAM video RAM –multibank DRAM

Spring 2002EECS150 - Lec19-memory Page 13 Important DRAM Examples: EDO - extended data out (similar to fast-page mode) –RAS cycle fetched rows of data from cell array blocks (long access time, around 100ns) –Subsequent CAS cycles quickly access data from row buffers if within an address page (page is around 256 Bytes) SDRAM - synchronous DRAM –clocked interface –uses dual banks internally. Start access in one back then next, then receive data from first then second. DDR - Double data rate SDRAM –Uses both rising (positive edge) and falling (negative) edge of clock for data transfer. (typical 100MHz clock with 200 MHz transfer). RDRAM - Rambus DRAM –Entire data blocks are access and transferred out on a highspeed bus-like interface (500 MB/s, 1.6 GB/s) –Tricky system level design. More expensive memory chips.

Spring 2002EECS150 - Lec19-memory Page 14 Non-volatile Memory Mask ROM –Used with logic circuits for tables etc. –Contents fixed at IC fab time (truly write once!) EPROM (erasable programmable) & FLASH –requires special IC process (floating gate technology) –writing is slower than RAM. EPROM uses special programming system to provide special voltages and timing. –reading can be made fairly fast. –rewriting is very slow. erasure is first required, EPROM - UV light exposure Used to hold fixed code (ex. BIOS), tables of data (ex. FSM next state/output logic), slowly changing values (date/time on computer)

Spring 2002EECS150 - Lec19-memory Page 15 FLASH Memory Electrically erasable In system programmability and erasability (no special system or voltages needed) On-chip circuitry (FSM) to control erasure and programming (writing) Erasure happens in variable sized "sectors" in a flash (16K - 64K Bytes) See: for product descriptions, etc.

Spring 2002EECS150 - Lec19-memory Page 16 Relationship between Memory and CL Memory blocks can be (and often are) used to implement combinational logic functions: Examples: –LUTs in FPGAs –1Mbit x 8 EPROM can implement 8 independent functions each of log 2 (1M)=20 inputs. The decoder part of a memory block can be considered a “minterm generator”. The cell array part of a memory block can be considered an OR function over a subset of rows. The combination gives us a way to implement logic functions directly in sum of products form. Several variations on this theme exist in a set of devices called Programmable logic devices (PLDs)

Spring 2002EECS150 - Lec19-memory Page 17 A ROM as AND/OR Logic Device

Spring 2002EECS150 - Lec19-memory Page 18 PLD Summary

Spring 2002EECS150 - Lec19-memory Page 19 PLA Example

Spring 2002EECS150 - Lec19-memory Page 20 PAL Example

Spring 2002EECS150 - Lec19-memory Page 21 Memory Blocks in FPGAs LUTs can double as small RAM blocks: –5-LUT is a 16x1 memory –achieves 16x density advantage over using CLB flip-flops Newer FPGA families include additional on chip RAM blocks (usually dual ported) –Called “block-rams” in Xilinx Virtex series

Spring 2002EECS150 - Lec19-memory Page 22 Memory Specification in Verilog Memory modeled by an array of registers: reg[15:0] memword[0:1023]; // 1,024 registers of 16 bits each //Example Memory Block Specification // //Read and write operations of memory. //Memory size is 64 words of 4 bits each. module memory (Enable,ReadWrite,Address,DataIn,DataOut); input Enable,ReadWrite; input [3:0] DataIn; input [5:0] Address; output [3:0] DataOut; reg [3:0] DataOut; reg [3:0] Mem [0:63]; //64 x 4 memory (Enable or ReadWrite) if (Enable) if (ReadWrite) DataOut = Mem[Address]; //Read else Mem[Address] = DataIn; //Write else DataOut = 4'bz; //High impedance state endmodule

Spring 2002EECS150 - Lec19-memory Page 23 Error Correction Codes (ECC) Memory systems generate errors (accidentally fliped-bits) –DRAMs store very little charge per bit –“Soft” errors occur occasionally when cells are struck by alpha particles or other environmental upsets. –Less frequently, “hard” errors can occur when chips permanently fail. Where “perfect” memory is required –servers, spacecraft/military computers, … Memories are protected against failures with ECCs Extra bits are added to each data-word –extra bits are used to detect and/or correct faults in the memory system –in general, each possible data word value is mapped to a unique “code word”. A fault changes a valid code word to an invalid one - which can be detected.

Spring 2002EECS150 - Lec19-memory Page 24 Simple Error Detection Coding Each data value, before it is written to memory is “tagged” with an extra bit to force the stored word to have even parity: Each word, as it is read from memory is “checked” by finding its parity (including the parity bit). Parity Bit b7b6b5b4b3b2b1b0pb7b6b5b4b3b2b1b0p + b7b6b5b4b3b2b1b0pb7b6b5b4b3b2b1b0p + c A non-zero parity indicates an error occurred: –two errors (on different bits) is not detected (nor any even number of errors) –odd numbers of errors are detected.

Spring 2002EECS150 - Lec19-memory Page 25 Hamming Error Correcting Code Use more parity bits to pinpoint bit(s) in error, so they can be corrected. Example: SEC on 4-bit data –use 3 parity bits, with 4-data bits results in 7-bit code word –3 parity bits sufficient to identify any one of 7 code word bits –overlap the assignment of parity bits so that a single error in the 7-bit work can be corrected Group parity bits so they correspond to subsets of the 7 bits: –p 1 protects bits 1,3,5,7 –p 2 protects bits 2,3,6,7 –p 3 protects bits 4,5,6, p 1 p 2 d 1 p 3 d 2 d 3 d 4 Bit position number 001 = = = = = = = = = = = = 7 10 p1p1 p2p2 p3p3

Spring 2002EECS150 - Lec19-memory Page 26 Hamming Code Example Example: c = c 1 c 2 c 3 = 101 –error in 4,5,6, or 7 (by c 3 =1) –error in 1,3,5, or 7 (by c 1 =1) –no error in 2, 3, 6, or 7 (by c 2 =0) Therefore error must be in bit 5. Note the check bits point to 5 By our clever positioning and assignment of parity bits, the check bits always address the position of the error! c=000 indicates no error p 1 p 2 d 1 p 3 d 2 d 3 d 4 –Note: parity bits occupy power-of- two bit positions in code-word. –On writing parity bits are assigned to force even parity over their respective groups. –On reading, check bits (c 1,c 2,c 3 ) are generated by finding the parity of the group along with its parity bit. If an error occurred in a group, the corresponding check bit will be 1, if no error the check bit will be 0.

Spring 2002EECS150 - Lec19-memory Page 27 Hamming Error Correcting Code Overhead involved in single error correction code: –let p be the total number of parity bits and d the number of data bits in a p + d bit word. –If p error correction bits are to point to the error bit (p + d cases) plus indicate that no error exists (1 case), we need: 2 p >= p + d + 1, thus p >= log(p + d + 1) for large d, p approaches log(d) Adding on extra parity bit covering the entire word can provide double error detection p 1 p 2 d 1 p 3 d 2 d 3 d 4 p 4 On reading the C bits are computed (as usual) plus the parity over the entire word, P: C=0 P=0, no error C!=0 P=1, correctable single error C!=0 P=0, a double error occurred C=0 P=1, an error occurred in p 4 bit Typical modern codes in DRAM memory systems: 64-bit data blocks (8 bytes) with 72-bit code words (9 bytes).