DDR II ram   











VIA Brings us Quad band DDR Memory.

Date: 17.09.2002

The next step in Memory Performance.

Quadband? Where has this sprung from?

With many memory manufacturers fighting to release faster and faster DDR Memory but with escalating manufacturing costs and low yields and a tough international market things are not quite as good as many would like.

For some background you can read VIA's press release about the new licencing deal HERE

Lets take a small step back before showing this new technology.

DDR SDRAM was originally developed to keep pace with newer, more powerful PC processors like the AMD Athlon™ XP and Intel® Pentium® 4 which require faster access to the system memory to deliver to their full performance potential. Compared to PC100/133, DDR doubles the effective clock rate by transferring data on both the rising and falling edges of the clock. Also, due to its evolutionary, parallel technology, the latency of DDR is quite low compared to competing serial memory technologies like RDRAM.

The basic specifications of JEDEC ratified SDRAM technologies are summarised below.
Jedec is the memory industry standards body.


Currently, DDR SDRAM cannot keep pace with the demands of these new CPU’s and high performance system components. The PC industry has been planning for this eventuality by developing a new specification for DDR memory enabling it to scale to higher frequencies. This successor named DDR II is still a few years away with samples expected in 2004. The DDR ‘II’ specification already agreed by JEDEC, redesigns both the DDR device and the module interface. These modules will be 240 pin, unlike both DDR I and initial QBM modules, meaning that DDR ‘II’ mainboards will not be backwards compatible with standard DDR modules. This leaves us with a current technology that cannot keep up and a future technology that requires a major redesign and upgrade.

Enter stage left… Quadband DDR.

QBM is a module based technology that uses existing DDR devices (DDR266/333/400) to deliver the bandwidth expected from future memory devices such as DDR II. The innovative technology provides the bandwidth of the yet to be available DDR II modules (expected 2004/2005) and will be available for sampling in December of 2002 with production modules becoming available in Q1 2003. QBM667 (PC5400) modules will be available in Q2 of 2003 to coincide with the release of 667 FSB Processors. QBM800 (PC6400) modules based on DDR I-400 devices will be available in the second half of 2003 followed by DDRII based QBM modules ranging from QBM800 (PC6400) – QBM1066 (PC8500) in the 2004-2005 timeframe, with significant cost savings and backward compatibility with DDR modules.

QBM, using DDR266 devices, will deliver a price/performance ratio improvement over DDR333 and DDR400 modules, as well as other non-DDR compatible memory technologies like RDRAM. Let's us assume for one moment that by the time QBM modules are available, that DDR I modules have reached DDR533. Being the latest high speed stepping of DDR I technology, DDR533 chips will be at a premium price. DDR266 on the other hand will have been available for well over 18 months by that stage and will therefore be very cost effective to use. With QBM technology, you can take DDR266 chips, for example, and make a QBM533 module. Think what can happen if we then take DDR533 modules, we could make DDR1066. Now we are talking huge increases in performance but not in costs.

The table below compares the data transfer rate of current DDR modules against QBM.

The QBM technology consists of at least two independent DDR memory devices running simultaneously. Both memory devices share the same command and address lines and each device receives a pair of differential clock signals at frequency to control all synchronous operations. The clock of the first memory device is in-phase with the controller clock, while the second device’s clock has a phase difference of 90-degrees from the first clock (created by the on DIMM PLL). The result is the data packed in 4-bits per cycle (double DDR I) versus today’s DDR at 2-bits per cycle.

Why is this so good for us?
Well other than becoming the perfect transition between the current DDR and the future DDR II standards, it increased bandwidth, keeps costs down and it doesn’t increase the latency we have seen in other memory types (such as RDRAM) as QBM uses the existing DDR device latency. QBM is the same as DDR because it uses the same “no-op cycles” for BUS turnaround between reads and writes. The packing and unpacking of the data through the switch happens in real time. This Licence between VIA and Kentron is the first time that PC based systems will be able to interface to high speed DDR533 and DDR667 modules featuring QBM technology. New systems using QBM enabled chipsets from VIA and S3 Graphics will combine the fastest speed memory in the industry with the fastest processor bus speeds and maintain backward compatibility with today’s DDR modules.

Pricing is also a major factor in memory. Because QBM Modules use standard off the shelf DDR chips, the modules will reflect the market price of DDR memory chips just like standard DDR DIMMs. QBM technology will lower the cost of extra bandwidth when compared with dual-channel solutions. QBM533 is expected to be priced on par with current DDR Modules

Physically the difference between an unbuffered DDR I DIMM and an unbuffered QBM DIMM is that the QBM DIMM (non-ECC mode) has eight, low cost QBM-10(R) switches and one PLL. These components are in standard packages and require no special manufacturing considerations. Both DDR I and QBM DIMMs fit into standard 184-pin connectors unlike the 240-pin DDR II which is not backward compatible with DDR I.


What about performance?

QBM modules double the peak bandwidth (transfer rate) between the memory and the chipset. The improvement in bandwidth reduces time needed to transfer data back and forth; thus performing the operation in less time. The system level performance gain will be application dependent. QBM will reduce memory access times by an average of 8%, which translates to a certain range of increased performance at the system level. As soon as we can get hold of some QBM benchmarks showing system performance will be released to the general public.

Who are the major players involved in QBM ?

The QBM Alliance is an international alliance aimed at positioning Quad Band Memory (QBM) technology as an industry standard. Key industry vendors, representing best-in-class complementary solutions to Kentron’s QBM memory technology, participate in the evaluation and implementation of products using QBM memory related technology and receive the most current updated technical design assistance. Alliance members include: ALI, Acuid Corporation, Avant Technology, CST, Denali Software, ICS, Kentron Technologies, Netlist, Peripheral, PNY Technologies, SIS, SiSoft, STMicroelectronics, Terarecon and VIA. STMicroelectronics and ICS have provided Kentron with key components for QBM technology. STMicroelectronics is manufacturing the QBM 10(R) switch, the cornerstone of QBM; and ICS has designed the PLL.

Will QBM become a JEDEC standard?

Kentron is an active member of JEDEC and participates on numerous committees, including devices, modules and support components. Kentron has taken the initiative and has understood that for a product or architecture to become an industry standard, it should co-exist well with the industry’s other principal standards and architectures. Kentron is bringing forward the QBM technology for standardization consideration in the upcoming JEDEC meetings.

How does QBM compare to dual-channel DDR chipset solutions like the forthcoming VIA Apollo P4X600

The VIA Apollo P4X600 will solve the bandwidth gap by offering up to 5.4GB/s of bandwidth with DDR333. All dual-channel configurations like P4X600 require a minimum of two DIMM modules, and a more complex motherboard implementation which has cost implications for the platform. VIA is extending that choice by offering both dual-channel and QBM solutions to the market will benefit our customers and ultimately the end user. VIA is the first QBM Alliance member to license QBM and incorporate it to their chipsets.

Is this going to be another licensing issues with QBM?

Kentron has decided to adopt a royalty-free license approach to make it easier for the chipset manufacturers to enable their chipsets and the module manufacturers to provide QBM modules to the marketplace on a global basis.


DDR ‘II’ is very firmly on the PC industry’s roadmap and will be the implemented by all the major manufacturers. The current problem is the technology is too expensive to produce and yields are too low. QBM is the obvious choice to bridge this gap in the transition of DDR I to DDR II. It has been chosen by VIA to allow the existing DDR SDRAM infrastructure to maximize its potential and build a performance bridge to DDR ‘II’ while remaining cost competitive. It was VIA that really took the DDR and entrenched it as the best replacement for the aging SDRAM and VIA's adoption of this new technology from Kentron are showing they are ready to step up to the plate again and show there competitors how its done.

Startside ] Opp ] [Søk]

Copyright © 2002 Øyvind Haugland
Sist endret:  13 januar 2019

  Interested in this stuff? Please write to:

HTML Counter            stats counter