The basis of any processor is hardware addition and subtraction of numbers, represented in binary form. For the most general understanding, it consists of an arithmetic-logic device in which the indicated operations are performed, as well as a number of control circuits and memory elements. Thanks to the efforts of developers to continuously improve processors and increase their performance, the real structure of the CPU is much more complex. Apart from the RISC architecture, manufacturers are following the path of using specialized information processing units. Examples include instruction sets of the SSE family for parallelizing a computing process or an extension of the AVX instruction set to support high-definition video hardware encoding.
From the point of view of an ordinary user, a processor is a large microcircuit designed to control a computer. Unfortunately, their different generations are incompatible with each other in a number of indicators, so when choosing another CPU, you have to take into account several points.
MAIN SELECTION CRITERIA
The main selection criterion is, of course, the platform. In a broad sense, it is a hardware implementation of an entire computing system, and in a narrow sense, it is a set of a processor with a specific internal microarchitecture and a corresponding set of system logic soldered on the motherboard. Of the current Intel platforms, two deserve close attention:
- LGA1150 for the updated Haswell family;
- LGA2011 for Ivy Bridge-E architecture.
The life cycle of the previous generation of Ivy Bridge and the LGA1155 platform based on this microarchitecture is coming to an end, although it may still be considered in the event of an upgrade of the existing system. The updated version of LGA2011-v3 on Haswell-E is promising, but still quite expensive, not least due to the transition to a new type of DDR4 memory.
AMD’s situation is generally the same:
- The Richland and Trinity families leaving the scene for socket FM2;
- Vishera lineup for AM3 +;
- Processors with Kaveri microarchitecture for FM2 + socket.
New developments have been announced, but they should not be expected for wide sale until 2016.
The next thing to look for when choosing a processor is the number of cores… The criterion is rather ambiguous since under different conditions the CPU performance will be determined not only by their number but also by the operating frequency, as well as by the nature of the applications being executed. Not to mention the fact that eight cores of AMD processors are usually inferior to four “stones” from Intel, because in most cases they share the total space of internal memory in pairs. It should also be borne in mind that a feature of the Turbo Boost mode, which automatically increases the clock frequency within certain limits, is the alternate shutdown of cores when it rises above certain threshold values, so their “standard” number does not always work productively. It follows from those tasks, the execution of which on the computer will prevail:
- For budget gaming systems and most office PCs, 2-core Intel or 4-core AMD is sufficient;
- Powerful workstations that do not deal with regular multimedia processing, and the bulk of gaming machines will need 4-core Core i5 processors;
- It makes sense to build high-end gaming systems based on four to six-core Core i7 processors;
- Dedicated workstations for professional 3D modeling, HD video editing, and other demanding tasks need the most powerful multi-threaded solutions. The 8-core flagship processors of both companies will be appropriate here.
When selecting CPUs for systems with a predominantly single-threaded load (which is also generated by most games), their clock frequency is put forward in second place in terms of importance… Last but not least, AMD’s lag in the segment of high-performance gaming processors is caused by more modest limit values of this indicator, and comparable values are achieved only due to a significant increase in power consumption. If you plan to overclock in manual mode, you will have to look for reviews of different models and analyze their respective potential. Otherwise, the maximum operating frequency the processor achieves with Turbo Boost enabled can serve as a guideline. It should be noted that additional overclocking is not always possible, and even in the most promising chips, its potential is rather modest. Also, keep in mind that the motherboard must support Turbo Boost for overclocking.
OTHER PROCESSOR CHARACTERISTICS
- Integrated graphics make it possible to build a system without a discrete graphics card. This is mainly true for office cars and entertainment centers. The older series of integrated accelerators HD Graphics, GT *, and Radeon HD can be considered as the entry and lower mid-end level of cards in the gaming segment, but their capabilities will not satisfy the demanding gamer. One cannot ignore the idea of hybrid processors (APUs), actively promoted by AMD, whose graphics subsystem is capable of performing part of non-graphics computing. From this point of view, the attitude towards integrated solutions changes somewhat.
- The unlocked multiplier is theoretically the main leverage for speeding up the CPU. By increasing it in the BIOS, for some models, it is possible to significantly increase the clock frequency relative to the nominal value. In any case, this situation was in the first round of overclocking. Then Intel began to block the possibility of such overclocking at the hardware level. At the moment such processor functionality is presented by her as a bonus and a sign of overclocking status. In reality, given the value of the base frequency and the nominal value of the multiplier, changing it within small limits does not give a special increase in performance, but costs money. AMD is more loyal to attempts to over-speed up the processor, but the overall effect is comparable.
- At present, the thermal package is the indicator that unambiguously divides the processors into very power-hungry (and this is not Intel) and energy-efficient. The TDP of direct competitors differs almost at times and requires the use of powerful cooling systems and power supplies for AMD products. It is not for nothing that many models of this company are supplied without standard coolers.
- The used version of the PCI – Express controller is important if this bus is expected to be used intensively, for example, a pair of video cards operating in x16 mode or another set of devices with such an interface that requires a large total number of communication lines. For home gaming systems with one accelerator, as well as for the vast majority of workstations and office machines, this parameter is not critical.
- L3 cache and Hyper-Threading support. The L3 buffer size directly affects the overall performance of the processors of both companies (bigger is better). Hyper-threading technology is an exclusive feature of Intel and allows you to split the instructions processed inside one core into two threads and execute them in parallel. The operating system perceives such a processor as having a doubled number of cores.
In fact, there is no particular opposition from manufacturers now. Company AMD is completely lost to Intel not only a segment of high-performance processors, but also the category of upper mid-end, focusing on hybrid APU-decisions, and the situation is unlikely to change in the foreseeable future. Competition between the two corporations remains only in the group of multicore CPUs focused on resource-intensive applications and multithreaded computing.
Any mistakes are frustrating. It is especially annoying when you are counting on significant savings, and you get additional costs and unnecessary trouble. In such a situation, it is easy to find yourself as a result of an error with the platform, for example, during an upgrade or when choosing a set from a processor and a motherboard on your own. Upgrading an outdated system requires the closest attention – a chipset soldered to the “mother” may not support the selected CPU, even if there is formal compatibility in the processor socket.
When choosing a processor, you should not rely solely on the magic of numbers in the characteristics. A model with a seemingly impressive number of cores in modern games will be inferior to a 2-core one, but with a higher nominal frequency and the ability to noticeably accelerate by standard means or manually. Conversely, for video encoding, the advantage is the ability to parallelize the process, not the speed of the processor. Be sure to consider the nature of the applications that you intend to run on the build system most of the time.
Don’t expect a significant performance boost by targeting a processor with an unlocked base-clock multiplier. Currently, this is more of a marketing ploy than a real overclocked product. At a minimum, you will have to equip the system with a bulky cooling system and a reliable power supply unit with a large margin of power. In many cases, it is more expedient to direct the additional costs for overclocking to increase the budget for the processor itself and choose a faster model from the usual series.