Dual Core Processing: Simplified, Demystified and Explained

SEMINAR BY SARITHA.V Roll No. 7420 BRANCH-IT

What is a dual core processor?

A dual-core processor combines two independent processors and their respective caches and cache controllers onto a single silicon chip. They fall into the architectural class of tightlycoupled multiprocessor. IBMs POWER4 was the first dual core processor. AMD, INTEL, Sun have since brought out their own models.

Why a dual core processor?

Microprocessor performance scaling has been sustained by increasing clock frequency and instruction-level parallelism (ILP). However, these two factors are both reaching the point of diminishing returns. Network-based applications, such as online transaction processing, are rich in thread-level parallelism which require high computing throughput to execute multiple threads/processes simultaneously rather than high single thread performance. Power consumption is also critical. The critical requirements for these types of applications are high computing throughput, high memory bandwidth, large addressing space, high reliability, low power and low cost. To address the above design targets, the optimum solution is an on-chip dual-core processor.

Architecture

The general architecture was as shown in slide 3. Here I will explaining dual-core processor based on the UltraSPARC I/II micro-architecture. It has a 1MB L2 cache, DDR-1 memory controller and symmetric multiprocessor bus (JBus) interfaces. This core provides an efficient performance per watt having balanced hardware complexity with 4-issue superscalar, 9-stage pipeline and in-order execution/out-of-order completion. Predominantly static circuit design styles and the balanced H-tree clock distribution also contribute to achieving low power. Typical power dissipation at 1.2GHz and 1.3V is 23Watts, which is the lowest published figure for 64b server processors.

Architecture contd

The memory controller supports up to 16GB of physical memory. JBus controllers allow low-cost multiprocessing systems with configurations of up to four-chips (eight threads). The chip is fabricated in Texas Instruments' advanced 0.13m CMOS process with 7 layers of Cu and a low-k dielectric. The transistor count is 80M.

Chip Block Diagram

Applications and benefits

There are a number of categories of applications and middleware that can benefit from dual-core processors. The following tables summarize the categories and representative applications that should see significant benefit, moderate benefit, or little benefit from a switch from single-core to dual-core processors.

Table 2 :The benefits for dualcore servers

The benefits for dual-core servers contd

Table 3 : The benefits for

workstations.

Performance Comparisons

The given below are tests IBM performed with a number of industry benchmarks to compare the fastest single- and dual-core Opteron processors available for the e326 and the LS206. These benchmarks can be useful for comparing
processor performance under varying workloads.

Linpack:Graph 1 shows the results of the Linpack benchmark

Table 4 compares published SPECint_rate2000

servers versus dual-core LS20 and e326 servers.

and SPECfp_rate2000 scores for single-core LS20

STREAM :Table 5 compares results for singlecore e326 and LS20 servers versus dual-core servers.

Hyper Threading Vs Dual Core Processing

or is it?

In a single-core dual-threaded processor, each thread is assigned its own set of registers. This makes the processor appear as two (virtual) processors. But this rarely happens because those virtual processors are still part of one physical processor, requiring the threads to time-share common resources such as integer units, floating-point units and cache. The effective throughput is somewhat less than the theoretical, with one thread waiting for the other to release a resource. In addition, if the software stack is largely single-threaded, the second virtual processor may be idle much of the time, rendering its value moot.

Hyper Threading Vs Dual Core Processing

or is it? Contd.

With a dual-core dual-threaded processor, you actually have two physical processors on one chip Because each core has its own cache, registers and other resources, there is less resource contention than you might see with a simple dual-threaded, single-core processor. Two separate single- or multithreaded programs can be running simultaneously, for up to twice the throughput of a same-speed single-core processor

Hyper Threading Vs Dual Core Processing

or is it? Contd.

For hyper-threading its strength is also its weakness. On the one hand, a deeply pipelined CPU is able to take care of multiple actions within a processing cycle. On the other hand, each and every operation will cause the CPU to go through most of these stages, which can easily end up wasting valuable clock cycles. To compensate in part, Intel implemented logic that can provide a higher average utilization of the Pentium 4 pipeline - which has grown to 31 stages with the current Prescott architecture

Drawbacks

The availability of Dual core processors. The need for the software to be aware of dual cores present. Need for new motherboards for some of the newer processors in case of a upgrade.

Future Trends

Beyond dual-core processors, there are examples of chips with multiple cores. Such chips include network processors which may have a large number of cores or microengines that may operate independently on different packet processing tasks within a networking application. One option is to widen the registers and/or the bus interface of an existing processor architecture. Widening the bus interface alone leads to superscalar processor architectures, and widening both usually requires new programming models. Other options include including multiple levels of memory cache, and developing system-on-a-chip solutions.

Conclusion
A server with one DC processor is much like a server

with two SC processors in one respect. In order to see the full benefits of dual-core processors, especially with 64-bit software, it is necessary to provide ample requirements for the software. Given the performance advantages of dual-core processors, does this mean the end of single-core processors? Eventually, yes For now concurrency especially in the form of simultaneous multithreading(SMT) and dual-core processors on the same die is the design seems to provide the best performance per square millimeter of silicon.

References

A Dual-Core 64b UltraSPARC Microprocessor for Dense Server Applications By Toshinari Takayanagi, Jinuk Luke Shin, Bruce Petrick, Jeffrey Su and Ana Sonia Leon, Sun Microsystems, Inc. The Benefits of Dual-Core Processors in High-Performance Computing By Mark T. Chapman, IBM Systems and Technology Group VIS Instruction Set Users Manual, Sun Microsystems, Inc. HOT CHIPS 15SCALING THE SILICON MOUNTAIN, IEEE MICRO, MARCHAPRIL 2004 www.wikipedia.com, www.amd.com, www.intel.com, www.ibm.com, www.sun.com