• Reduces cost, power and area of a CDR
• Improves jitter performance in high-speed links

What is Phase Interpolator-Based CDR Technology?

Buffered modules introduce a memory buffer between the memory controller and the DRAM devices on each module, isolating the DRAM from the memory bus and enabling an increase to the width of the memory without increasing the pin count of the controller. This also reduces the effective capacitive load on the memory bus enabling support for multiple modules at high speed.

A phase-interpolator based CDR is an alternative circuit architecture developed by Rambus which provides multiple advantages compared to PLL-based CDRs. This type of CDR uses a PLL or DLL to implement a reference loop which accepts an input reference clock signal and produces a set of high speed clock signals, used as reference phases, spaced evenly across 360 degrees. These reference phases are then fed to a CDR loop which includes circuitry for selecting pairs of reference phases and interpolating between them to provide clocks for recovering the data from the data signal.

Because of the separation between the reference loop and the CDR loop, the designer of a phase interpolator based CDR can separately optimize both the noise suppression of the reference loop and the tracking agility of the CDR loop. Additionally, the reference loop is not affected by the contents of the data signal, potentially allowing this type of CDR to track a wider variety of data signals. Furthermore, the relatively long locking time of the reference loop applies only at start-up when initially locking to the reference clock signal. After the initial locking time, interpolator-based CDRs can provide much faster re-locking compared to PLL-based CDRs whenever the data signal returns after being interrupted.

Another benefit of phase interpolator based CDRs is that the data sampling point can be precisely adjusted by a digitally controlled offset. This allows the cancellation of offsets from device mismatches and other causes, and enables in-system measurements of the timing margin available for reliably extracting data from the data signal.

Lastly, although the reference loop can occupy the majority of the area and dissipate the majority of the power in a phase interpolator based CDR, its reference phases can be shared among several CDR loops on chips receiving multiple data signals. In this way, the average size and power required for the CDR functionality per data signal can be greatly reduced.

Who Benefits?

The use of phase interpolator based CDRs benefits many different groups. By designing ASICs including Rambus IO cells that utilize phase interpolator based CDRs, ASIC vendors benefit from the smaller area, lower power, and more stable operation of the IO cells. These benefits are magnified when dual, quad, or other multi-lane IO cells are used since these cells use one reference loop to drive multiple CDR loops for implementing multiple CDRs. The area and power savings can be significant compared to using a PLL per lane, as required by other CDR designs. The ability to digitally offset the data sampling clock when using a phase interpolator based CDR allows in-system testing of timing margins in the actual operating environment. Such system-level testing increases the reliability of manufactured systems for system integrators. Finally, the cost, power, performance, and testability benefits from using phase interpolator based CDRs is passed along to products purchased by consumers in the form of lower prices, longer battery life, and improved reliability.

• Improves data rates and system voltage margin
• Increases DRAM yield
• Compensates for variations in trace impedance and termination voltage
• Improves system reliability over a wide range of operating conditions

What is Output Driver Calibration Technology?

Variations in process, voltage, and temperature can reduce the size of data eyes. Data eyes reveal characteristics of the quality of the signaling environment such as timing and voltage margins. Robust signaling relies on having wide (good timing margin) and tall (good voltage margin) data eyes. Output drivers are designed to drive signals between high and low voltage levels, shown as Voh and Vol in the previous illustration. Variations in process, voltage, temperature, and other factors can cause output drivers to overshoot and/or undershoot the desired signaling voltage levels, resulting in reduced margins that impact signal integrity. Reduced timing margins limit the maximum signaling speed because the window of time over which the data is valid (width of the data eye) is smaller. Reduced voltage margins can require larger IO voltage swings to ensure accurate transmission of data, but such larger swings result in increased IO power and can increase the sensitivity of the system to cross talk. In order to increase signaling rates and reduce IO power, output driver overshoot and undershoot must be managed.

Output driver calibration allows for optimal signaling levels to be established and maintained using adjustable output drive strengths to compensate for variations in process, voltage, and temperature. Calibrating the output drivers during normal operation allows for drive strength adjustments to respond to changes in voltage and temperature which can fluctuate while a system is in use.

Output Driver Calibration uses feedback that is provided to the output driver circuitry to adjust the output impedance of the output driver circuitry, thereby controlling the circuit’s drive strength.in order to achieve optimal signal performance. The driver’s output impedance is compared to a reference resistor RZQ that is placed off the device. The output impedance is then calibrated to be equal to or proportional to the reference precision resistor.

The circuit above depicts how an Output Driver Calibration circuit may be configured. The voltage dropped across the topmost array of resistors is dependent upon the state of the respective transistors in series with those resistors and the value of the RZQ resistance on the line. The states of the transistors in the transistor array are individually controlled by the Drive Strength Register and are set so that the value of Vterm = Vref. The reference voltage, Vref , is representative of the desired output signal level. When the Vterm = Vref condition is achieved, the impedance in the top of the network divider is optimized for the driver. The values used for configuring the transistor array can be stored in the register and may be updated as needed.

The figure above illustrates the effect that Output Driver Calibration has on the magnitude of overshoot and undershoot along the transmission line. The reduction in overshoot and under-shoot results in increased voltage and timing margins.

The reference resistor RZQ and much of the circuitry for Output Driver Calibration can also be utilized for on-die termination ODT calibration.

Who Benefits?

Output driver calibration provides benefits from the device up through the system. By increasing DRAM yield and allowing DRAM output drivers to automatically compensate for process variation, output driver calibration improves margin and device testability, saving design and test time. Output driver calibration also allows board designers to compensate for variations in trace impedance and termination voltage caused by manufacturing and assembly processes. This ability to compensate for manufacturing tolerances of some components enables test specifications to be relaxed and saves component and tester costs.

At the system level, Output driver calibration enables system integrators to use one DRAM in multiple designs that utilize different trace impedances and that operate in different environments. Output driver calibration also increases voltage and timing margins, resulting in higher system reliability over a wider range of operating conditions. In addition, adjustable drive strengths help compensate for variations in temperature which allows system integrators to more effectively manage their system power and thermal budgets, thereby decreasing overall system cost.

• Supports high-speed operation with reduced IO signaling power
• Improved signal integrity
• Reduced design complexity
• Eliminates the need for thick-oxide transistors on memory controller
• Reduced voltage is better matched to advanced processes

What is Near Ground Signaling Technology?

Near Ground Signaling (NGS) is a single-ended, ground-terminated signaling technology that enables high data rates at significantly reduced Input Output (IO) signaling power and design complexity, while maintaining excellent signal integrity. With a VDDIO of 0.5V, Near Ground Signaling has a reduced signal swing compared to traditional Stub Series Terminated Logic (SSTL) signaling and lowers IO power on both the DRAM and controller.

This lower IO voltage is better matched to the operating voltage of advanced CPU’s and GPU’s and reduces the cost and complexity of integrating the memory controller on the processor chip.

Who Benefits?

Memory designers benefit from the reduced design complexity, lowering their integrated memory controller implementation costs. They are also able to achieve increased data rates at significantly reduces IO signaling power.

Commercial server managers and consumer end users benefit from the reduced cost of ownership and increased battery life for their end systems and devices.

• Improves memory throughput up to 50%
• Reduces power consumption by 20% for equivalent workloads versus conventional modules
• Enables full utilization of memory IO bandwidth
• Utilizes conventional DRAM

What is Module Threading Technology?

Multi-thread computing is driving up memory bandwidth, but needs smaller access granularity due to the random nature of the data accesses. However, small transfers of data are becoming increasingly difficult with each DRAM generation. Although the memory interface has become faster, the frequency of the main memory core has remained relatively the same. As a result, DRAMs implement core prefetch where a larger amount of data is sensed from the memory core and then serialized to a faster off-chip interface, effectively increasing the access granularity. This discrepancy between the interface speed and the core speed translates to a core-prefetch ratio of 8:1 in current DDR3 DRAMs, and is forecasted to reach 16:1 in future DRAM. This larger prefetch ratio and transfer size can lead to computing inefficiency, especially on multi-threaded and graphics workloads with the need for increased access rate to smaller pieces of data.

The memory subsystem in today’s computing platforms are typically implemented with DIMMs that have a 64bit-wide data bus and a 28bit command/address/clock bus. On a standard DDR3 DIMM module, all the devices within a module rank are accessed simultaneously with a single Command/Address (C/A). An example module configuration places eight (x8) DDR3 components assembled in parallel onto a module printed circuit board and has a minimum efficient data transfer of 64Bytes.

Greater efficiency with a multi-threaded workload and smaller transfers can be achieved by partitioning the module into two separate memory channels, and multiplexing the commands across the same set of traces as a traditional module but with separate chip selects for each respective memory channel. In threaded module, each side of the module is accessed independently, thereby reducing the minimum transfer size to one-half the amount of a standard single-channel module.

Threaded modules can lower the power of main memory accesses. For a conventional eight-device module, all eight DRAMs are activated (ACT) followed by a read or write (COL) operation on all eight devices. A threaded or dual-channel module can accomplish the same data transfer by activating only four devices and then performing two consecutive read or write operations to those devices. Since only four devices are activated per access instead of eight devices, a threaded or dual-channel module achieves equivalent bandwidth with one-half the device row activation power. On a memory system, this translates to approximately 20% reduced total module power.

Another benefit that threaded modules offer is increased sustained bandwidth at high data rates. Many modern industry-standard DRAMs have limited bandwidth due to power restrictions on the DRAM devices. On DRAMs starting with the DDR3 generation, only a limited number of banks may be accessed in order to protect the on–DRAM power delivery network and maintain a stable voltage for the memory core. This parameter, know as tFAW (Four Activate Window period) allows only 4 banks to be activated in the rolling tFAW window.

For a computing system, tFAW restricts the memory controller from issuing additional row activate commands once four activates have already been issued in a given tFAW period. This stalls the memory controller and results in lost data bandwidth. A DDR3 DRAM running at 1600Mbps data rates loses up to 50% of its sustained data bandwidth due to this and other restrictions. Since the DRAMs in a threaded module are activated half as often as those in a conventional module, the sustained bandwidth of a threaded module is not limited by the core parameters.

Who Benefits?

Module threading delivers system designers the benefits of increased bandwidth from improved transfer efficiency and smaller access granularity, while maintaining the commodity cost structure of the module. End users can benefit from the 50% improvement in throughput performance as well as the 20% reduction in total memory power.

• Simplifies high-speed system design
• Eliminates requirements for trace length matching and reduce routing area requirements
• Optimizes IO signal timing for improved timing margins
• Complements Fly-by command/address system architectures

What is FlexPhase Technology?

FlexPhase technology anticipates the phase difference between signals on different traces and manages the transmission of data bits so that the data arrives at the memory device with a known timing relationship with respect to the command and address signals sent to the memory device. It can also be used to enhance conventional DRAM architectures by managing the variation in signal propagation times due to variations in trace lengths.

In DRAM systems, FlexPhase circuits can be used to optimize data and strobe placement. FlexPhase circuits can also be used to finely tune the timing relationships between data, command, address and clock signals. In conventional DRAM architectures, FlexPhase circuits can be used to deskew incoming signals at the controller in order to compensate for uncertainty in the arrival times of signals. Further, FlexPhase circuits can be used to intentionally inject a timing offset – “preskew” data such that the data will arrive at the DRAM devices coincident with the command/address or clock signal. FlexPhase minimizes the systematic timing errors in typical memory systems by adjusting transmit and receive phase offsets at each pin or pin-group.

When using a Fly-by architecture, the amount of time required for the data, strobe, command, address and clock signals to propagate between the memory controller and DRAMs is primarily affected by the lengths of the traces between the controller and the DRAM devices over which the signals propagate. In a Fly-by system, the command, address and clock signals arrive at each DRAM at different times, which in turn results in the data signals being transmitted from each DRAM device at different times. FlexPhase can be used at the controller to deskew those data signals to eliminate the offset due to the Fly-by architecture in addition to any inherent timing offsets of the system. Similarly, because the command, address and clock signals arrive at each DRAM at different times, the data for write operations to the memory devices needs to be preskewed by the controller to account for the difference in when the memory devices will be expecting the write data. FlexPhase can accomplish that preskewing while still eliminating inherent timing offsets in the system.

FlexPhase is a departure from traditional serial link technologies in which timing deskew is performed using an embedded clock. Such deskewing techniques, which typically rely on 8b/10b encoding to ensure adequate transition density for clock recovery, require more chip area, have added power consumption, increase latency, and suffer a 20 percent bandwidth penalty associated with the 8b/10b encoding.

FlexPhase includes in-system timing characterization and self-test functionality that enables aggressive timing.

During READ access operations, a memory controller incorporating FlexPhase technology determines and stores the “receive” phase difference between the transmitted control signals and the data received from each memory device. The phase difference corresponding to each memory device is subsequently used to deskew the data signals which arrive at the memory controller at different times, thereby allowing proper reconstitution of the data accessed from each of the memory devices.

During WRITE operations, a similar process is performed where a “transmit” phase difference is determined for each memory device and stored within the memory controller. Those transmit phase differences are then used to modify (preskew) the timing delay between the transmitted command/address signals and the data sent to each memory device.

Who Benefits?

FlexPhase circuit technology brings flexibility, simplicity, and savings to memory system designers. At the device level, FlexPhase technology helps to compensate for the manufacturing variations that degrade timing windows and operational performance of the memory. The FlexPhase approach allows memory interfaces to operate at GHz rates without the power, area and latency penalties incurred in systems using Clock and Data Recovery (CDR) techniques. FlexPhase also provides for improved testability by using digital phase offsets for margin testing of the high speed chip interfaces—saving design time and cost.

At the system level, FlexPhase technology relaxes PCB trace length matching requirements by anticipating and calibrating the signaling phase offsets caused by variations in trace lengths and impedances. FlexPhase timing adjustments allows much simpler, more compact and cost-efficient memory layouts. FlexPhase timing adjustments provide for in-system test and characterization of key data signals, thereby enabling performance testing of the high speed links.

• Increases transfer rate without increase system clock rates
• Improves memory system bandwidth

What is Double Bus Rate Technology?

In a read transaction for a single data rate DRAM, the address, control, and data is transferred on one edge of each clock cycle. Memory bandwidth can be improved by applying double bus rate technology and increasing the per-pin data signaling rate of a DRAM. Double Bus Rate Technology allows data to be transferred more quickly, increasing the bandwidth that a DRAM can supply.

Doubling the data rate of the data transfers affects the relationship between address/control information and data for a Read transaction. When transactions are interleaved, a problem can occur when the amount of time that data occupies the memory bus is smaller than the amount of time that address and control information occupy the bus. In this situation, the insufficient address/control bandwidth leads to bubbles in the data transfer on the bus, resulting in reduced memory bandwidth and loss of performance.

The issue of performance loss can be addressed by applying Double Bus Rate Technology to the address and control pins as well. Double Bus Rate Technology is used to balance address, control, and data bandwidth, thereby eliminating the concerns relating to insufficient address and control bandwidth. As a result, bandwidth is increased by 50% compared to the interleaved transactions with double bus rate technology. Another example of where increased control bandwidth can be useful is in systems that use write masking. In systems that utilize write masking, increasing the amount of data being transferred to memory requires that more byte masking control information be specified in order to maintain support for data masking at byte granularities. By balancing address, control, and data transfer rates on the bus with Double Bus Rate Technology, performance losses due to insufficient address and control bandwidth are eliminated.

Who Benefits?

Many groups can benefit from double bus rate technology. By balancing address, control, and data bandwidth, system designers are able to achieve the highest levels of memory bandwidth in their systems. This in turn helps to reduce the number of DRAMs necessary to achieve a given level of memory performance, reducing component count and easing system component placement, routing concerns, and thermal dissipation. System designers and integrators benefit from the reduced component count needed to achieve a given level of memory bandwidth, resulting in lower system cost and smaller form-factor systems.

[/fusion_text][/one_full][/fullwidth]

• Enables high data rates at very low IO power consumption
• Improves signal integrity

What is Very Low-Swing Differential Signaling Technology?

VLSD signals are point-to-point and use an ultra-low 100mV signal swing (50 to 150mV) and 100mV common-mode voltage, which results in a 200mV peak-to-peak differential signal swing. This swing is less than 1/10th the signaling swing of commodity memory interfaces. VLSD enables high data rates with very low IO power consumption.

Who Benefits?

VLSD enables system designers to achieve high-speed operation through the robust signaling characteristics inherent to differential signaling, while minimizing IO power consumption through the use of a ground-referenced low-voltage-swing signaling system. This combination of high-bandwidth and low-power operation improves mobile device performance and battery life for consumers.

• Calibrates ODT termination impedance
• Reduces signal reflections
• Compensates for variations across PVT and operating conditions

What is On Die Termination Calibration Technology?

Conventional DRAM memory module architectures typically include line termination resistors on the motherboard. Although the termination resistors on the motherboard reduce some reflections on the signal lines, they are unable to prevent reflections resulting from the stub lines that connect to the DRAMs on the module. A signal propagating from the memory controller to the DRAM encounters an impedance discontinuity at the stub leading to the DRAM on the module. The signal that propagates along the stub to the DRAM will be reflected back onto the signal line, thereby introducing unwanted noise into the signal. The introduced noise and the consequential signal degradations that are not addressed by such off-die termination become more pronounced with higher data rates and longer stub lengths. Larger, multi-drop systems containing multiple DRAM modules introduce even more reflections and consequently add more reflective noise, thereby resulting in further signal degradation.

By placing the termination resistance on the die itself rather than the motherboard, the reflections resulting from discontinuities in the line are significantly reduced, thus producing a cleaner signal and enabling faster data rates.

ODT calibration is a technique that involves calibrating the termination impedance in order to optimize the reduction of signal reflections. ODT calibration allows an optimal termination value to be established that compensates for variations in process and operating conditions.

A calibrated ODT value significantly reduces unwanted signal reflections while only minimally attenuating the magnitude of the signal swing due to the added resistive loading. The resulting cleaner data signal allows for higher data rates.

ODT calibration is achieved by establishing an ODT impedance that is proportional to an external precision resistor. The same external resistor can also be used for Output Driver Calibration.

The ODT calibration controller, compares the voltage drop across the ODT resistor network with a voltage drop across an external resistor represented. The controller modifies the resistor network with coarse tuning and fine tuning to achieve an impedance value that closely approximates the external, reference resistance.

Who Benefits?

ODT calibration delivers benefits at the device, subsystem and system level. By implementing ODT calibration, devices are able to achieve enhanced signal performance and higher data rates, which enables designers to achieve superior DRAM device and module performance.

In addition, placing the termination components on the DRAM devices removes these elements from the PCB. In doing so, the number of components and signal lines on the motherboard is reduced, lowering the cost and complexity while increasing reliability.

Finally, the system benefits from the superior data rates and module performance that are enabled through the improved signal integrity achieved with ODT calibration.

• Improves transfer efficiency for multi-core computing applications
• Doubles DRAM core data rate versus conventional techniques
• Maintains high sustained bandwidth while lowering power consumption

What is Micro-threading Technology?

Access granularity is a function of the accessibility of data within a memory architecture. A typical DRAM is comprised of eight storage banks. Within such DRAMs, each bank is typically further subdivided into two half banks, “A” and “B”. For such a DRAM with 32 data pins, each A half bank is connected to 16 data pins and each B half bank is connected to 16 pins. The banks are in opposite quadrants of the physical die, and each quadrant has its own dedicated row and column circuitry – each bank half operating in parallel in response to the row and column commands.

A row command selects a single row in each bank half of the bank being addressed, thereby sensing and latching that row. Physical timing constraints impose a delay (i.e., tRR) before a row in another bank can be accessed. Column commands are similarly constrained (i.e., tCC). However, the row timing interval is typically twice the column timing interval; therefore two column commands can be issued during the mandatory delay required for a single row activation.

The column prefetch length, the amount of data delivered per transaction, is determined by the respective column and row timing delays and bit transfer rate, where:

Prefetch = timing delay/bit transfer rate

A core of a mainstream DRAM typically operates up to 200MHz, whereas a core of a high performance industry standard DRAM can typically operate up to 400MHz. Core frequencies exceeding 400MHz are difficult to achieve using modern industry standard DRAM technologies without sacrificing production yields or increasing costs. Therefore, a column prefetch of 16bit is required for such a high performance DRAM core to support external data rates exceeding 3200 MHz since the DRAM cores is organized with each half-bank operating under the same row or column operation

In addition:

Column granularity = (column prefetch) x (number of data pins per half bank) x (number of half banks per access)

Or:

For a 32-bit wide DRAM with 16 data pins per half bank:

Column granularity per access = 16 x 16 x 2 = 512 bits or 64 bytes.

Moreover, during the row timing interval, in order to maintain peak bandwidth, at least two column operations must be performed. This is typically described as two column address strobes per row address strobe (two CAS per RAS). This results in a minimum row granularity of 128 bytes. This large access granularity translates into inefficient data and power utilization for applications such as 3D graphics.

Using largely the same core resources as in the previous example, a sample micro-threaded DRAM core has 16 banks, each bank in the micro-threaded DRAM being equivalent to a half bank in the typical DRAM core. The even numbered banks connect to the A data pins and odd numbered banks connect to the B data pins (again with 16 pins in each case). However, unlike a typical core, each four-bank quadrant can operate independently, through the use of independent row and column circuitry for each quadrant. Moreover, interleaving, simultaneous access to more than one bank of memory, allows concurrent accesses to the lower quadrant on the same physical side of the core as the previous access.

Micro-threading enables four independent accesses to the DRAM core simultaneously. Although the same time interval as a typical core must still elapse before accessing a second row in a particular bank or bank quadrant, the three banks in the other quadrants remain separately accessible during the same period. Columns in rows in other quadrants can be concurrently accessed even though a column timing interval must pass before a second column is accessible in the previously activated row. The net effect of this quadrant independence and interleaving is that four rows (one in a bank of each quadrant) and eight columns (two in each row) are accessed during the row timing interval (compared to a single row and two columns with the typical DRAM technique).

Timings are similar to the typical DRAM core, but each column only sends data for half the column timing interval. The interleaved column sends data for the other half of the interval. Micro-threading reduces minimum transfer granularity size while maintaining a high-yielding and cost effective core frequency. By interleaving the column accesses from four different banks, a micro-threaded DRAM core (of a given column prefetch length and core frequency) can support data rate two times higher than that of a conventional DRAM core. Conversely, micro-threading of the column operation enables a DRAM core to cost-effectively sustain a specific data transfer and granularity while relaxing the column cycle time (tCC) by up to two times compared to those of a conventional DRAM core.

With micro-threading, column granularity is now:

Column prefetch/2 x 16 pins = 16/2 x 16 = 128 bits or 16 bytes (one quarter of the previous value).

The row granularity is 32 bytes (again one quarter of the previous value). Micro-threading’s finer granularity results in a performance boost in many applications. For example, in a graphics application with 8 byte micro-threaded column access granularity, computational and power efficiency increased from 29 percent to 67 percent after introducing the technique.

Who Benefits?

Micro threading enables twice the data rate from a DRAM core over conventional techniques, providing memory system designers high sustained bandwidth while lowering power consumption. In addition, micro threading benefits DRAM designers and manufacturers by providing an alternative approach to improve efficiency and reduce access granularity using largely the same DRAM core, reducing cost and risk.