Study on grading standard of computer Products evaluation technology

Source: Internet
Author: User
Tags benchmark prefetch

Evaluation criteria

In the computer procurement, people first think of the Big Three is: CPU performance, memory capacity, display. Because the monitor update speed is relatively slow compared to the CPU, at the same time, the use of KVM makes multiple hosts share the display more and more, so people in the procurement of computers, especially server class products, often only evaluate the performance of the host, therefore, this article has skipped the content related to the display.

(i) CPU performance indicators

CPU is the core of the whole computer system, it is often synonymous with various grades of computers, the performance of the CPU generally reflects the performance of the computer, so its performance indicators are very important. The main indicators are:
(1) Frequency: CPU clock frequency (CPU clock speed). Generally speaking, the higher the frequency, the faster the CPU speed. However, since the internal structure is different, not all CPUs with the same clock frequency have the same performance.
(2) Speed of communication between memory bus speed (Memory-bus)--cpu and Level Two (L2) cache and memory.
(3) Extended bus speed (Expansion-bus)-the operating speed of a local bus such as VESA or PCI bus interface card installed on a microcomputer system.
(4) Address bus speed (Expansion-bus)-Determines the physical address space that the CPU can access.
(5) Data bus width-determines the amount of information that is transmitted between the CPU and the two-level cache, memory, and input/output devices.
(6) Built-in coprocessor-a CPU with a built-in coprocessor that can speed up certain types of numerical calculations, and some software systems that require complex calculations, such as high-version auto CAD, require coprocessor support.

(ii) Memory performance indicators

(1) Speed-the memory speed is generally used to access data once the time required (in general NS) as a performance indicator, the less time, the faster.
(2) Capacity--memory is the main part of the computer, it is relative to the external memory. Memory is * * *, but to be limited by the maximum capacity of the motherboard support, and is the current computer, this limit is still blocked.
(3) Number of memory lines-refers to the number of points of contact between the memory and the motherboard, these contact points are gold fingers, there are 72 lines, 168 lines and 184 lines. 72-, 168-, and 184-wire memory bar data widths are 8-bit, 32-bit, and 64-bit respectively.

--CPU Package

The so-called "CPU packaging technology" is a kind of integrated circuit insulation of nutrients or ceramic materials packaging technology. Taking the CPU as an example, we actually see the size and appearance of the actual CPU core is not the sizes and faces, but the CPU core and other components after the encapsulated product. CPU encapsulation is a must for the chip and is also crucial. Because the chip must be isolated from the outside, in order to prevent impurities in the air to the chip circuit corrosion caused by electrical performance degradation. On the other hand, the encapsulated chip is also easier to install and transport. As the encapsulation technology plays a direct role in the performance of the chip itself and the PCB (printed circuit board) connected to the design and manufacture, it is essential. As each manufacturer in each package in the packaging of the number of chips and packaging technology, resulting in a CPU represents a different meaning, the operation speed is also a great difference. With the development of CPU technology, the number of CPUs on each computer is increasing rapidly, and the 32-core and 64-core machines are becoming more and more popular. Because of the need for synchronization between different chips, so the performance of the computer and the number of chips can not be a linear relationship, so the same number of chips can not be a linear relationship, so there is the same chip number of the computer performance gap.

--Streamline instruction set complex instruction sets

The CPU itself is simply an integrated VLSI on a block of silicon wafers, so why is a sophisticated chip capable of controlling a large and complex computer system? This is the set of instructions that are integrated in the CPU. The so-called instruction set is a set of instructions that the CPU uses to compute and control a computer system, and each new type of CPU specifies a series of instruction systems that are compatible with other hardware circuits at design time. and the instruction set of advanced or not, but also related to the performance of the CPU, it is also an important indicator of CPU performance. The instruction set of the CPU is divided into the compact instruction set and the complex instruction set from the mainstream architecture, and the complex instruction set is basically used on the ordinary computer processor. When the streamlined instruction set appeared, everyone was aware of its superiority, but Intel, which was in the market at the absolute advantage, was able to achieve its own product by improving its level of research and development by making its products more capable of using a complex set of instructions than a streamlined instruction set. Thus relying on the market rather than technology to ensure their own product advantages. But the rivalry between the two has continued.

--processing technology and market factors

Intel is an absolute advantage in the current CPU market, and AMD's products can only harvest small orders at low cost. Since the processing of AMD products is generally considered to be worse than that of Intel, the products that simply evaluate both from a frequency perspective cannot meet the wishes of the bidders.

(iii) Performance index of mechanical hard disk

1, rpm (rpm,revolutions per Minute): This indicator represents the speed of the hard drive spindle motor (drive disk), such as 5400RPM means that the spindle speed of the drive is 5400 rpm per minute.

2. average Seek time (Average seek times): If no special instructions generally refer to the seek time when reading, the unit is MS (MS). The meaning of this indicator is the average time required for the drive to move to a specified track (which should be a cylinder, but a track for a specific head) after it has received a read/write instruction. In addition to the average seek time, there is also an inter-lane seek time (track-to-track or cylinder Switch time) and the full-track-seek-times (fully-tracks or complete Stroke), which refers to the time required for the head to move above the adjacent track from the current The latter refers to the time required for the head to move above the outer (or most inner) lap track to the top (or most outer) lap track, which is substantially more than the average seek time. We generally only care about the average seek time for actual work.

3. average incubation period (Average Latency): This indicator refers to how long the specified read/write sector will be moved to the bottom of the head (the disc is rotated) when the head is moved to the specified track, and the faster the disk, the shorter the incubation period. The average incubation period refers to the time it takes for the disk to turn half a circle. Obviously, the average incubation period for the same speed drive is fixed. Approx. 5.556ms at 7200RPM at approx. 4.167ms,5400rpm.

4. Average access time (Average access time): Also known as average access times, is generally not provided in the manufacturer's published specifications, which is generally a test score, which means the average time taken from the read/write instruction to the first data read/write, including the average seek time, The average incubation period and related housekeeping time (such as instruction processing), because the house operation time is generally very short (generally around 0.2ms), negligible, so the average access time can be approximately equal to the average seek time + average incubation period, and thus also known as the average addressing time. If the average seek time for a 5400RPM hard drive is 9ms, the average access time in theory is 14.556ms.

    6, buffer capacity (buffer Size): Many people also call the cache capacity in megabytes. The cache Buffer is also written in some vendor profiles. The basic function of the buffer is to balance the internal and external DTR. To reduce the host's wait time, the hard disk will read the data into the buffer, and so on all read or buffer filled, and then quickly sent to the host at the interface rate. With the development of technology, the manufacturers later added caching capabilities to SCSI hard disk buffers. This is mainly reflected in three aspects: Prefetch (Prefetch), experiments show that in typical cases, at least 50% of the read operation is continuous reading. The prefetch function simply means that the hard disk "privately" expands the read range, after the buffer sends the specified sector data to the host (that is, the head has already read the specified sector), the head then reads the contiguous slices of the sector data and feeds into the buffer, if the subsequent reads exactly point to the adjacent sectors that have been prefetch, The access speed is improved by reading from the buffer without the head re-addressing. Write cache, usually in the case of write operations, the data is written to the buffer before sending to the head, when the head is written and then report the host write, the host will start processing the next task. A hard disk with write cache will report write completion to the host after the data is written to the buffer, allowing the host to "free" processing of other transactions in advance (the remaining head write operations master does not wait), improving overall efficiency.

In order to further improve the performance, the current manufacturers have basically applied the segmented cache technology (multiple Segment cache), the buffer is divided into a number of small pieces, storage of different write data, rather than waste the entire buffer space for small data, You can also wait for all segments to write and write uniformly, with better performance. Read cache, which temporarily stores the read data in a buffer, which can be provided directly from the buffer if the host is needed again, to speed up. Read caching can also use segmented technology to store multiple disparate chunks of data, cache multiple read data, and further improve cache hit ratios.
7, noise and temperature (Noise & temperature): These two belong to non-performance indicators. For the noise, before the manufacturers do not care, but from 2000 onwards, for the needs of the market (such as OEMs want to produce a quieter computer to increase the selling point) manufacturers through a variety of means to reduce the work of the hard disk noise, the third version of the ATA-5 specification also added a subset of automatic acoustics (noise) management (AAM, Automatic acoustic Management), so all new hard disks currently support AAM functionality. The noise of the hard drive mainly comes from the spindle motor and the voice coil motor, noise reduction is also from the two points (the increase in the disk will also increase the noise, but there is no way). As for heat, in fact each manufacturer has its own standards, and claimed that the performance of the hard drive is what they expected, completely within the scope of security, no problem. This is not to worry about, but the key is that the hard disk is a part of the chassis, its high heat will increase the overall temperature of the chassis, perhaps the hard disk itself is nothing, but may not be able to withstand the accessories around, other do not say, if it is two high-heat hard disk installed very close, then it can withstand nearly double heat? So the heat of the hard drive still needs to be noticed by manufacturers.

Performance evaluation Criteria

In order to evaluate the performance of a computer as a whole, many international organizations have also put forward the relevant evaluation standards, which are recognized by everyone including the following two kinds.


TPCC is a specification specifically for online transaction processing systems (OLTP systems), and in general we refer to these systems as business processing systems. Almost all of the foreign mainstream vendors that provide soft and hard platforms in the OLTP market have released corresponding TPC-C test results, and these test results are constantly being refreshed as computer technology continues to evolve.

Standard origin

TPC (Transaction Processing Performance Council, transaction Performance Committee) is a non-profit organization created by dozens of member companies, TPCC members are mainly computer hardware and software manufacturers, not computer users, Its function is to develop standard specifications, performance and price metrics for business application benchmarking programs, and to manage the release of test results.

TPC does not give the code of the benchmark program, but only the standard specification of the benchmark program. Any manufacturer or other tester can optimally construct their own test system (test platform and test program) according to the specification. To ensure the integrity of the test results, the tester (usually the manufacturer) must submit to tpc-a complete set of reports (full disclosure report), including the detailed configuration of the system under test, the classified price and the total price including the 5 year maintenance cost. The report must be verified by the TPC authorized Auditor (TPC itself does not audit).

Test model

The model used in the TPC-C test is a large wholesale sales company with several commodity warehouses distributed in different regions. When the business expands, the company adds a new warehouse. His warehouse is responsible for the supply of 10 points of sale, each of which has 3,000 customer care services, each of the orders submitted by customers, an average of 10 products per order, all orders of about 1% of the products in their direct ownership of the warehouse is not in stock, must be supplied by other regional warehouses. At the same time, each warehouse to maintain the company's sales of 100000 kinds of goods inventory records.

Test metrics

The results of TPC-C test mainly have two indicators, namely Flow index (throughput, abbreviation TPMC) and price/performance ratio (price/performance, abbreviation PRICE/TPMC).

Flow indicator (throughput, abbreviation TPMC): According to the definition of TPC organization, the flow indicator describes how many new order transactions can be processed per minute while performing support operations, order status inquiries, shipping and inventory status inquiries for 4 types of transactions. The response time of all transactions must meet the requirements of the TPC-C test specification, and the proportion of the number of transactions should also meet the requirements of the TPC-C test specification. In this case, the larger the value of the traffic indicator, the higher the online transaction processing power of the system.

Price/performance ratio (price/performance, abbreviation PRICE/TPMC): That is, the overall price of the test system and the ratio of flow indicators, in the case of obtaining the same TPMC value, the lower the prices good.

--spec CPU

Spec is the abbreviation of standard performance evaluation Company (performance Evaluation Corporation). Spec is a nonprofit organization consisting of a number of companies, such as computing manufacturers, system integrators, universities, research institutes, and consultants, and the goal of this Organization is to establish and maintain a set of standards for evaluating computer systems.

Spec CPU 2006 is the latest version of the CPU subsystem evaluation software introduced by the spec organization, and in the early years, the industry was using its previous version of Spec CPU2000. Like spec CPU2000, spec CPU2006 includes CINT2006 and C FP2006 two sub-projects, the former for measuring and contrasting integer performance, while the latter is used to measure and compare floating-point performance, Spec CPU2006 Some of the tests in CPU2000 were upgraded, and some tests were discarded/added, so the two-version test scores were not comparable.

Spec CPU test, the test system's processor, memory subsystem and the use of the compiler (SPEC CPU provides the source code, and allows test users to do some compilation optimization) will affect the final test performance, and I/O (disk), network, operating system and graphics subsystem for the SPEC The effect of CPU2006 is very small.

Recommendations for scoring standards

In the specific bidding process, computer performance is the main evaluation index, however, due to different projects on the hardware has different actual requirements, the system's reliability, compatibility and so must be fully considered.

Reliability --also known as robustness or robust--is the key to the survival of the system in exceptional and dangerous situations. For example, if the computer software in the input error, disk failure, network overload or intentional attack situation, can not panic, do not crash.

compatibility -in the actual project execution process, because the user itself has been running the relevant system, in order to protect the user's prior equipment investment, need to consider the compatibility with the prior equipment.

Requirements for Software systems --sometimes, in order to satisfy the operating system or software application, the hardware product must be required accordingly.

In short, with the development of computer hardware and software technology, the application scenario is more and more complex, simple evaluation indicators can not meet user requirements. For the requirements of computer performance, the simplest way is to query some third-party machine performance indicators, and in the bidding documents to be qualified. For the user-specific needs, must be in the scoring standard to give a certain proportion, so as to ensure that the procurement of products can be achieved customer satisfaction.


As a means of choosing an agreed-upon business, the tendering work has been paid more and more attention, and the quantity of the computer products is increasing with the popularity of paperless methods. The functional refinement of related products, the formulation of technical scoring standards must also be more targeted.

Study on grading standard of computer Products evaluation technology

Related Article

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.