SPRT 597
SPRT 597
SPRT 597
2
• Management data input/output (MDIO) Total performance
module (also part of the EMAC) that Coprocessor
(@1.2-GHz core frequency)
continuously polls all 32 MDIO addresses
1,908 MSPS @ 256-FFT
in order to enumerate all PHY devices in FFT/DFT
1,683 MSPS @ 192-DFT
the system LTE – 582 Mbps @ 6144 block size, 6 iterations
Turbo decode
• Packet coprocessor that provides L2 to L4 WCDMA – 371 Mbps @ 5114 block size
classification functionalities and the Turbo encode
LTE – 2572 Mbps @ 6144 block size
processing power of up to 1.5 Gbps WCDMA – 2556 Mbps @ 5114 block size, 6 iterations
• Security accelerator block capable of Viterbi decoder >38 Mbps (K = 9) Mbps = 9)
wire-speed processing on 1-Gbps Ethernet Rake Search Accelerator 32-bit multiplication per cycle
traffic on IPSec, SRTP, and 3GPP air inter WCDMA despreading 256 AMR users supported @ eight fingers
face security protocols WCDMA spreading 256 users supported with two radio links and diversity
• Embedded Ethernet switch that allows Encryption/decryption IPSec 2.8 Gbps
multiple devices to be connected through LTE – DL 2.2Gbps, UL 1.1Gbps
BCP
SGMII, eliminating the need for a board level WCDMA – DL: 800 Mbps, UL 400 Mbps
Ethernet switch TCI6618’s coprocessors
The TCI6618 has twelve high-performance
embedded coprocessors to perform intensive As wireless radio standards evolve and decoder coprocessors (VCP2_A, VCP2_B,
signal processing functions common to wire- related implementations become standardized, VCP2_C, and VCP2_D), three third-generation
less base station applications. The result is each evolution of TI’s wireless SoC devices turbo decoder coprocessors (TCP3d_A,
increased overall system performance, yielding has included more and more radio accelera- TCP3d_B, and TCP3d_C), turbo encoder
40 percent spectral efficiency over conven- tion/coprocessing, and as such provides a coprocessor (TCP3e), three fast Fourier trans-
tional decoding techniques. compelling roadmap to lower power and costs form coprocessors (FFTC_A, FFTC_B, and
while delivering higher performing base station FFTC_C) and a bit rate coprocessor. Together,
Bit rate coprocessor for solutions for our customers. TI’s SoC strategy they significantly accelerate channel encod-
increased spectral efficiency of integrating DSP cores with coprocessors is ing/decoding operations. Also included in the
The bit rate coprocessor (BCP) is a multi- the simplest and most economical approach to TCI6618 are four tightly coupled rake/search
standard acceleration engine that offloads all wireless base station solutions and continues accelerators (RSAs) for code division multiple
bit rate processing in the wireless signal chain. to be the market-leading solution today. TI’s access (CDMA) assistance with chip-rate
The BCP contains the modulator, demodulator, coprocessors eliminate external FPGAs and processing.
interleaver/de-interleaver, turbo and convolu- ASICs that were previously needed to deliver the
tion encoding, rate matcher/rate de-matcher, performance needed for base stations, lowering Delivering full multicore
correlator for block code decoding, and CRC the system cost and design complexity. entitlement
engine. The BCP enables turbo interfer- The TCI6618 has multiple, dedicated TI’s TCI6618 is based on the KeyStone
ence cancellation for MIMO equalization and high-performance embedded coprocessors to multicore SoC architecture, the first of its
enables high-performance PUCCH format 2 perform intensive signal processing functions kind to provide full multicore entitlement. This
decoding. It offloads approximately 15 GHz common to wireless base station applications. provides non-blocking access to all process-
of CPU MIPS with a maximum LTE downlink The coprocessors are four enhanced Viterbi ing cores, peripherals, coprocessors, and I/
throughput of 2.2 Gbps and uplink through-
put of 1.1 Gbps. For WCDMA, the maximum
downlink throughput is 800 Mbps and uplink
Traffic manager
Navigator interface
throughput is 400 Mbps. These techniques,
coupled with the powerful MIMO processing DIO interface
128 bit packet streaming switch
3
Os. Innovations that unleash full multicore and other IP blocks, eliminating memory management, segmentations and reassembly,
entitlement are: Multicore Navigator, TeraNet, contention. Shared memory access for code is and delivery across multiple cores and devices
Multicore Shared Memory Controller (MSMC), nearly identical in latency to local L2 access, are all supported by Multicore Navigator. Layer
and HyperLink. with highly effective prefetch mechanisms for 2 data-plane and transport-plane overhead
Multicore Navigator – TI’s Multicore Naviga- code and data. can be reduced by 10-15 times due to the
TI’s TCI6618’s DDR3 external memory inter- fast-path and zero-copy processing enabled
tor is an innovative packet-based manager
face (EMIF) is a 1,600-MHz, 64-bit bus with 8 by the Multicore Navigator.
that controls 8,192 queues and abstracts the
connections between the various subsystems GB of addressable memory space. Tied directly
to the MSMC, the DDR3 EMIF reduces latency Lowest power consumption
on the TCI6618. With a unified interface for
associated with external memory fetches for performance
communication, data transfer, and job man-
and provides the speed increase and support TI has a history of providing the lowest power
agement, Multicore Navigator enables higher
needed for larger applications that operate on wireless base station SoCs on the market. TI is
system performance with fewer interrupts and
large amounts of data, which is essential for able to achieve its ultimate low power through
reduced software complexity with a “fire and
advanced 3G and 4G base stations. the combination of its process technology,
forget” paradigm.
SmartReflex™ technology, and the proactive
Multicore Navigator instructs the next
HyperLink – HyperLink, with four lanes at up use of power management techniques (such
free DSP core to read the job and process it.
to 12.5 Gbps/lane, is a proprietary high-speed as adaptive voltage scaling) in every wireless
Multicore Navigator simplifies the software
interconnect that allows low protocol and base station semiconductor device to keep ac-
architecture and improves performance of
high-speed communication and connectivity tive power at a minimum. TI’s latest technology
base stations with: in the TCI6618 has brought the SoC power
to other KeyStone devices providing OEMs
• Dynamic resource/load sharing a seamless path to scalable solutions. The consumption for macro and compact base
• Offloading CPU overhead/delay related to HyperLink on the TCI6618 works in conjunc- stations down to the industry’s lowest levels at
inter-subsystem communications tion with the Multicore Navigator to dispatch 0.075 mW/GMAC of power.
• Hardware-based task prioritization tasks to multiple devices transparently, so
• Dynamic load balancing they execute as if they are running on local Complete tools and support
• Common communication methodology for resources. TI provides a full suite of best-in-class Eclipse-
all IP blocks (software, I/O, and accelerators) based development and debugging tools with
the TCI6618; these include a new C compiler,
TeraNet – TeraNet provides a hierarchal TCI6618 as Layer 2 and an assembly optimizer to simplify program-
switch fabric that combines to deliver more transport processing engine ming and scheduling, and a Windows debug-
than two terabits of bandwidth for data TCI6618 combines the unmatched PHY ger interface for visibility into source-code
transfer within the SoC. This virtually guaran- processing capabilities with dedicated copro- execution. TI’s compiler generates highly effi-
tees that the cores or coprocessors are never cessors for the Layer 2 and transport layer cient code that is “first-pass efficient” so there
starved for data and can deliver the entitled processing. This enables designers to create is less need to optimize it. TI’s debugging tools
processing horsepower. Because the switch base stations without a separate network pro- help developers visualize problems and resolve
fabric is hierarchical instead of flat crossbar, cessor, thereby reducing board complexity and them quickly, so designers can get products
overall power consumption is much lower cost without compromising performance. to the field faster while saving development
in idle states and also delivers systems with The network coprocessor enables fast-path resources. In addition, TI will offer a TCI6618
minimal latency, which is a key requirement in processing in the transport network layer and evaluation module (EVM) to help customers
next-generation base stations. deep into the Layer 2 of the radio network. prototype quickly with the TCI6618.
Multicore Shared Memory Controller Within the network coprocessor inside the
(MSMC) – TI’s TCI6618 includes a unique TCI6618, the Packet Accelerator and the For more information
memory architecture for enhanced per- Security Accelerator perform fully-accelerated To learn more about the TCI6618 SoC
formance. TI’s Multicore Shared Memory autonomous packet-to-packet processing. visit www.ti.com/tci6618. Discover how the
Controller (MSMC) allows the cores to directly They leverage the Multicore Navigator, which TCI6618 can add performance to your next
access shared memory without having to use uses a zero-copy approach to optimize data wireless base station design.
any TeraNet bandwidth. The MSMC arbitrates processing at all layers. Classification and or-
access to shared memory between the cores dering, multicore-accessible storage, memory
Important Notice: The products and services of Texas Instruments Incorporated and its subsidiaries described herein are sold subject to TI’s standard terms and
conditions of sale. Customers are advised to obtain the most current and complete information about TI products and services before placing orders. TI assumes no
liability for applications assistance, customer’s applications or product designs, software performance, or infringement of patents. The publication of information
regarding any other company’s products or services does not constitute TI’s approval, warranty or endorsement thereof.
The platform bar, C6000, and SmartRelex are trademarks of Texas Instruments. A122010
All other trademarks are the property of their respective owners.
Mailing Address: Texas Instruments, Post Office Box 655303, Dallas, Texas 75265
Copyright © 2011, Texas Instruments Incorporated