High-End Conroe

Evgenie Rudometov

authors@rudometov.com

Realising the newest developments in the field of microarchitecture of superdifficult chips, Intel corporation has released a new generation high-powered and energoekonomichnyh the processors which set has received the name of Intel Core 2 Duo

For a long time productivity of computers, was carried out basically at the expense of increase in internal clock rate of the processors which have reached almost 4 GHz. The method was considered so obvious and perspective, that predictions for fast reaching of boundaries in 10 and even 30 GHz already became . However, as it has appeared, this path step-by-step carried on computer branch in the lockup linked with extremely high energopotrebleniem of high-frequency chips. Really, semiconductor chips on energy density promptly came nearer to nuclear reactors. Moreover, approximating of the given growth showed possibility of reaching of levels of the energy corresponding to a surface of our star. About it leading experts of Intel at sessions IDF repeatedly warned.

One of the solutions providing further overgrowth of productivity without a dangerous thermal overvoltage of semiconducting crystals of processors, hardware support of parallel calculations is. Estimating the given possibility, it is necessary to mark, that usage within the limits of one system of several processors is not optimal. It is linked to price growth, the common energopotreblenija and level of acoustic noise from operation of numerous systems of cooling.

Realised as alternative to multiprocessing systems, processing technique Hyper-Threading, provided a demanded multithreading, giving to systems two logical kernels at one physical. Implantation of the given processing technique increased productivity, but did not reduce levels energopotreblenija and teploobrazovanija for a processor kernel. Moreover, the given processing technique even increased them, aggravating a problem of thermal load on a semiconducting crystal.

Implementation dvujadernosti — two kernels allocated in one processor tank was a following logical stage . For a transition expedition the first two-nuclear models have been fulfilled on the basis of microarchitecture NetBurst well debugged in the last generation of traditional one-nuclear processors of Intel.

For further improvement of parametres and the extension of functionality of computer systems experts of Intel have offered the new microarchitecture providing considerable lowering of expenses of energy on performance of one instruction. As criterion of efficiency experts it was offered to use metric Energy per Instruction — the energy spent for performance of one instruction. Advanced design, has received the name of Intel Core Microarchitecture. On its basis variety of processors among which models as for notebooks are presented, and and for desktop PCs recently has been released.

In table 1 first five models of processors for the desktop PCs created on the basis of new microarchitecture and known earlier under name Conroe are resulted.

Table 1. Models of processors Conroe for desktop PCs

Models	Internal clock rate, GHz	Frequency of the system bus, MHz	The Cache memory 2nd level, Mb
Intel Core 2 Extreme X6800	2,93	1066	4
Intel Core 2 Duo E6700	2,66	1066	4
Intel Core 2 Duo E6600	2,40	1066	4
Intel Core 2 Duo E6400	2,13	1066	2
Intel Core 2 Duo E6300	1,86	1066	2

Intel Core Microarchitecture has absorbed in itself all reachings which have been developed, realised and debugged in products of prior generations. Among them there is a set of Intel Core Duo (Yonah), actually being prototype Conroe. Main differences Conroe from Yonah can be formulated as follows:

·         the Advanced decoder of instructions expanded to 4 decoders of x86-macrooperations (for the nearest clones, Intel of Pentium M / Core Duo — 3),

·         Speed of fulfilment of 128-bit SIMD-instructions is finished to 1 instruction for clock tick in each final control device (in 2 times faster, than for Yonah),

·         Advanced operating principle with memory and hardware presampling (prefetch),

·         the Cache memory of the second level (L2) is the common for both computing kernels, its size is reallocated between them dynamically, depending on load,

·         Further improvement of processing techniques energosberezhenija,

·         support of new set SIMD of the instructions which have received name SSE4 Is entered.

The specified differences are realised in the form of a number of processing techniques among which first of all it is necessary to mark Intel Wide Dynamic Execution, Intel Advanced Smart Cache, Intel Smart Memory Access, Intel Advanced Digital Media Boost, Intel Intelligent Power Capability.

The microarchitecture of a kernel and the main processing techniques are resulted in a Fig. 1.

Fig. 1. Microarchitecture of kernel Conroe and the main processing techniques

The processing technique of Intel Wide Dynamic Execution raises productivity and a processor overall performance, allowing each kernel to execute to four instructions for clock tick with usage effective 14-etapnogo the pipeline.

The processing technique of Intel Advanced Smart Cache is linked to presence of the common cache memory of 2nd level that reduces energopotreblenie, reducing to a minimum the "traffic" size in a memory subsystem, raises productivity of system, providing one of kernels access to all cache memory at idle time of other kernel.

The processing technique of Intel Smart Memory Access raises productivity of system by lowering of delays at memory access and thus optimises usage of accessible capacity thanks to what the processor obtains data when they are required.

The processing technique of Intel Advanced Digital Media Boost doubles speed of performance of the commands often used in multimedia and graphic applications.

The processing technique of Intel Intelligent Power Capability activates separate sites of the chip only as required, that considerably reduces energopotreblenie systems as a whole. In addition to it in the architecture the mechanisms reducing clock rate of operation and a supply voltage of kernels at lowering of computing load are realised.

To the resulted innovations it is necessary to add as well Intel 64 Technology, providing support of 64-bit calculations. It gives to the processor access to essential bolshemu to a memory size in comparison with a marginal level in 4 Gbytes for traditional 32-bit models.

From the processors of the set of Intel Core released for desktop PCs 2 Duo high model is Intel Core 2 Extreme X6800. It is characterised not only the greatest performance level, but also rather low energopotrebleniem, and, therefore, and teploobrazovaniem — 75 W at maximum computing load. As to numerical estimations of productivity they are traditionally defined in the course of testing though and it is known, that at sessions IDF the corporation lips of the heads promised 40 % of a gain.

As the testing object has been used engineering sempl Intel Core 2 Extreme which appearance is resulted in a Fig. 2.

Fig. 2. Kernel Conroe and engineering sempl the two-nuclear processor of Intel Core 2 Extreme X6800

In a role of the measurement standard for the researched processor in tests the model of Intel of Pentium D 820 used  .

Pattern of the system applied in testing:

·         the Motherboard of Intel D975XBX (a chip set i975X),

·         the Processor of Intel Core 2 Extreme X6800 (Conroe, Dual Core, 65 nanometers, 2,93 GHz, the bus — 1066 MHz, L2  — 4 Mb, TDP  — 75 W),

·         the Processor of Pentium  of Intel of Pentium D 820 (SmithField, Dual Core, 90 nanometers, 2,8 GHz, the bus — 800 MHz, L2  — 2х1 Mb, TDP  — 89 W),

·        Video subsystem ATI X600,

·         the Disk drive on hard disks Seagate ST3400832AS,

·         Dynamic storage Apacer 2x1 a Gbyte, DDR2-677,

·         the Operating system of Microsoft Windows Server 2003 Enterprise Edition SP1.

During testing the given model of the processor was identified as the two-nuclear processor of Intel Core 2 Extreme X6800. Its key parametres defined by program CPU-Z, are presented in a Fig. 3.

Fig. 3. Key parametres sempla the two-nuclear processor of Intel Core 2 Extreme X6800

Results of performance of test SiSoftware Sandra 2004 (CPU Arithmetic Benchmark, ALU) are presented in Table 3 and in a Fig. 4.

Table 3. Results of test SiSoftware Sandra 2004 (CPU Arithmetic Benchmark, ALU)

Models of processors	Results of the test
Intel OF PENTIUM D 820	12906
Intel Core 2 Extreme X6800	32688

Fig. 4. Results of test SiSoftware Sandra 2004 (CPU Arithmetic Benchmark, ALU)

Results of performance of tests SiSoftware Sandra 2004 (CPU Multi-Media Benchmark) are presented in Table 4 and in a Fig. 5.

Table 4. Results of tests SiSoftware Sandra 2004 (CPU Multi-Media Benchmark)

Models of processors	Results of the test
Intel OF PENTIUM D 820 (Integer)	26683
Intel Core 2 Extreme X6800 (Integer)	72389
Intel OF PENTIUM D 820 (Float)	31633
Intel Core 2 Extreme X6800 (Float)	81009

Fig. 5. Results of tests SiSoftware Sandra 2004 (CPU Multi-Media Benchmark)

The resulted results of testing illustrate considerable advantage of Intel Core Microarchitecture over prior microarchitecture NetBurst.

So, for example, for Intel Core 2 Extreme X6800 internal frequency of operation of kernels most of all on 4,6 % in comparison with Intel of Pentium D 820. However in test SiSoftwareSandra 2004 (CPU Arithmetic Benchmark, ALU) the new processor shows essentially big productivity — growth on 153 %.

In tests SiSoftware Sandra 2004 (CPU Multi-Media Benchmark) excess of productivity of the processor of Intel Core 2 Extreme X6800 also rather big: in the test of floating-point operations excess has made 156 %, and in the test of integer calculations excess — 171 %.

If to consider a metric of productivity Energy per Instruction, that, considering smaller energopotreblenie new processors in comparison with their predecessors, advantage in used tests becomes still big. The given circumstance only raises attractiveness of the new set of processors.

As to competing models from AMD, how show the numerous researches resulted on the Internet, the systems created on the basis of AMD Athlon 64 FX-62 essentially concede on productivity to systems from Intel Core 2 Extreme X6800. Especially it concerns multimedia tasks where positions of processors of Intel are traditionally strong. Well and if to estimate systems on metric Energy per Instruction advantage of Intel Core 2 Extreme X6800 with maximum teploobrazovaniem 75 W over AMD Athlon 64 FX-62 about 125 W becomes still big. Quite often it expresses any more in tens percent, and in times. The given circumstance strengthens positions of corporation of Intel in sector of high-powered systems with an acoustic noise low level.

In summary it is necessary to remind, that in tests used engineering sempl models of Intel Core 2 Extreme X6800 and as shows experience, in process of perfection of processing technique and internal microarchitecture parametres of processors essentially improve. It allows to hope for conquest of new boundaries of productivity by two-nuclear processors of the set of Intel Core 2 Duo.

And, nevertheless, despite already shown, and also the expected performance levels, the resulted two-nuclear set not long can carry the crown of the leader. It is linked with rather fast, is possible this year, appearance chetyrehjadernyh models of processors of Intels grounded on Intel Core Microarchitecture, transited as it was shown by results of testing, successful approbation in two-nuclear models of the set of Intel Core 2 Duo.

The processor of Intel Core 2 Extreme X6800 and the motherboard of Intel D975XBX have been given by the Moscow office of corporation of Intel