test four models / processors and memory

It is very likely that right now, when you read these lines, the Russian currency in its fall pierces the next bottom. Unfortunately, most of us can not do anything with economic trends, and the only thing that remains for us is to watch with gloomy growth in the prices of computer components. No import substitution among them, either now or in the foreseeable future, is obviously impossible. We would not like to over-dramatize the situation, but, for example, the senior Devil’s Canyon managed to add more than twice to the ruble price since the summer of last year: from 14 to 30 thousand, and this is probably not the limit . It is obvious that such a rapid rise in the price of computer components will affect not only the mood, but also the preferences and behavior of buyers and collectors of PCs this year.

Obviously, the demand will now shift towards medium or even lower price components, as many people simply can not afford to buy flagship parts. If in recent years the domestic computer market aspired in its structure to the European one, now, most likely, it will be more like Chinese. The differences are dramatic: the average salary in China is about 400-500 dollars, so the best sales in this country are achieved by far from universally recognized leaders, and manufacturers and suppliers of low-cost computer components. A good example: the products of AMD, which is content with a rather small share in the global market, is popular in China to such an extent that it even decided to put this country in the center of its entire desktop processor business.

Along with the shift in the interest of PC users to the segment of inexpensive components, it can also be predicted that they will begin to pay more and more attention to overclocking in its classic, rather than sporting display. The acquisition of an inexpensive platform and its subsequent overclocking is an excellent and time-tested method of obtaining good performance with a minimum of material costs. With this in mind in the very near future in the focus of interests of sophisticated users should not be too popular to date low-cost processors with overclocking functions.

Before, we paid little attention to such proposals, believing that for enthusiasts they are not particularly interesting. Indeed, until recently, even in ordinary home desktops in most cases, installed at least a Core i5 processor, the speed of which is more than enough for work and entertainment. However, new realities make it necessary to abandon old principles and look at what worthy chips exist in the lower price segment. And fortunately, even among low-cost processors, you can find quite interesting options that allow you to build platforms that are quite acceptable in terms of performance, which are suitable for both professional activities and games. Preparing for this “anti-crisis” testing, we asked ourselves a very strict framework and limited the maximum cost of the CPUs of interest to us by a sum of $ 100. But even so, there were at least three (or even four) interesting processors that not only have promising characteristics, but they also allow themselves to be overclocked, that is, they can eventually interest even sophisticated users. It is about such finds that will be discussed in this material.

⇡ # AMD FX-6300

Unable to upgrade the processor architecture for the flagship Socket AM3 + platform, last year AMD dramatically reduced the cost of its old solutions related to the generation of Vishera. As a result, among the processors with a cost below $ 100, several very interesting variants appeared at once, the most attractive of which is the FX-6300. It’s no joke, this processor, built on the basis of three compute modules Piledriver, received the official price at the level of Intel Pentium and junior Core i3. Of course, this is only the youngest current six-core AMD, and its clock frequency is only 3.5 GHz, but no other offers with such a number of cores and a similar price simply does not exist. In addition, for all the other characteristics of the FX-6300 – this is a full-fledged design carrier Vishera. It has an 8-megabyte L3-cache, supports Turbo Core 3.0 technology, thanks to which it can accelerate to 4.1 GHz with low-current load, and also has almost all modern extensions of a set of x86-64 commands, including AVX, AES and FMA4.

Of course, any carrier of the processor design Vishera, and FX-6300 including, seems somewhat outdated at least due to the fact that it is produced by 32-nm technology. However, the same can be seen and plus: the use of mature technology provides a good frequency potential, which can be disclosed in overclocking. Since AMD FX-6300 belongs to the Black Edition family, its multiplication factors are completely unblocked, and increasing the frequency above the nominal values ​​does not cause any difficulties. In particular, we were able to easily increase the frequency of our FX-6300 test device to 4.8 GHz, and this only required raising the supply voltage to 1.55 V and turning on the Load-Line Calibration function in the BIOS of the motherboard.

Yes, AMD FX-6300 does not differ coldly. Its rated design heat dissipation is set at 95 W, and when overclocked, it seriously exceeds this value. However, the single-tower cooler Noctua NH-U14S used to remove heat from the FX-6300 operating at 4.8 GHz was enough.

⇡ # AMD FX-4350

Approximately to the same $ 100 level as the FX-6300, the other design carrier Vishera, FX-4350, fell in price last year. This is another AMD processor for the Socket AM3 + platform, but, unlike its fellow, it only has four Piledriver computing cores. As compensation, the FX-4350, which is the top four-core AMD, can offer a much higher passport clock frequency: 4.2 GHz – in nominal and 4.3 GHz – in turbo mode. In this case, the FX-4350 retains all the typical features of the Vishera design: a 2MB cache in 2MB capacity for each dual-core processor module, a common 8MB L3 cache, and a dual-channel memory controller that is formally compatible with DDR3-1866, but in fact supports and faster modules. There are no restrictions on extensions of instructions: all vector and cryptographic instruction sets are fully supported.

It should be noted that, unlike FX-6300, FX-4350 is designed for a more powerful heat package – 125 watts. However, in fact, the consumption of quad core is more economical. And what is important for enthusiasts, this can ensure its overclocking to more distant borders, the benefit of AMD FX-4350 also belongs to the Black Edition series. We tested this with our own experience – with an increase in the supply voltage to 1.5 V, the test specimen of the processor was able to function stably at 4.9 GHz.

The frequency achieved in overclocking exceeds the passport frequency by 17 percent, and this is not an impressive achievement in relative scale, because the FX-6300 allowed to raise its frequency by 37 percent above the nominal value. Nevertheless, in absolute terms 4.9 GHz is one of the best results for modern processors. With the use of rather ordinary air cooling, such boundaries are not subject to any other carriers of the Piledriver / Steamroller microarchitecture, or Haswell / Haswell-E processors.

⇡ # AMD Athlon X4 860K

Along with the FX series, AMD develops another line of classic CPUs (without the built-in graphics core) – Athlon X4. Such processors are oriented to use in the Socket FM2 + platform, which, it must be said, looks fresher than the released Socket AM3 + in 2012. At least in it there is an in-built support for PCI Express 3.0 and USB 3.0, which is absent in AMD 990FX / 990X / 970 chipsets. A newer microarchitecture than the carriers of the Vishera design has the Athlon X4 itself. Remember, soon after the release of desktop Kaveri, AMD fans turned to AMD with a petition about transferring the FX series to the more progressive microarchitecture Steamroller? Athlon X4 860K is just such a discrete processor on a new microarchitecture, which, we recall, is fundamentally different from Piledriver by the appearance of each core of its own decoder. Although, of course, in fact, the Athlon X4 860K is not a development of the FX series, but a trivial desktop Kaveri (A10 or A8 series) with a disconnected graphics core.

That is, the Athlon X4 860K is a quad-core processor, manufactured using a 28-nm process technology and based on the Steamroller microarchitecture. Which, incidentally, just makes it an attractive offer on the background of the FX-4350, especially when you consider that the official price of the Athlon X4 860K is set at $ 82, and the real value in stores has already dropped below $ 70. The clock frequency of this CPU is set at 3.7 GHz, and in turbo mode it can rise to 4.0 GHz. The total cache size of the second level, like that of the FX-4350, is 4 MB, but the L3-cache for the Athlon X4 860K is not: it’s the legacy of Kaveri. In the rest, the Athlon X4 860K is at least not inferior to the four-core Vishera, including the overclocking capabilities. The multipliers of this processor are unlocked and can be easily overclocked.

However, unfortunately, the frequency potential of the chips with the Steamroller cores, which are produced in the GlobalFoundries factories using SHP-technology with 28-nm standards, turned out to be slightly worse than their predecessors. Despite the fact that the thermal package of the Athlon X4 860K is limited to a 95-watt value, and in practice this processor is heated much less than its counterparts for Socket AM3 +, in overclocking it is not as good as the same FX-4350 . So, having resorted to increasing the supply voltage to 1.45 V, we were able to raise the frequency of our Athlon X4 860K to only 4.5 GHz. Any more violent acceleration led to instability.

However, all this does not mean that the Athlon X4 860K is worse than the FX-4350. Do not forget: the Athlon X4 860K has a newer microarchitecture, the specific performance of which, according to AMD itself, is better by 15-20 percent. Therefore, the overclocked Socket FM2 + platform with the Athlon X4 860K may be a worthy option for an inexpensive but fast system.

⇡ # Intel Pentium G3258

Until recently, the release of cheap processors, suitable for overclocking, dealt exclusively with AMD, and this, incidentally, in no small measure fueled the popularity of its products. Intel had overclockers processors only in the expensive series Core i5 and Core i7, so their prices began with a 242-dollar mark. But this summer everything changed: Intel finally decided to turn its attention to the audience of economical enthusiasts, specifically for which the Pentium Anniversary Edition was introduced. This chip, which received the model number G3258, became an analog of the Pentium G3250 processor, but with fully unlocked multipliers, which made it the first over the last few years to budget Intel’s overclocking offer. The cost of the Pentium G3258 is only $ 8 higher than that of its ordinary counterpart, which means that it is possible to buy the overclocked Intel processor almost three times cheaper than before – for $ 72.

In terms of the formal characteristics of the Pentium Anniversary Edition, the G3258 is a dual-core Haswell processor for the LGA1150 platform. Unlike the representatives of the Core i3 family, it is devoid of Hyper-Threading technology, its cache memory of the third level is cut down to 3 MB, and in addition, the AES, AVX and AVX2 instruction sets are additionally disconnected. The clock frequency of the Pentium G3258 is 3.2 GHz, there is no support for turbo mode. The official specifications of the Pentium G3258 also mention the compatibility of its memory controller with only two-channel DDR3-1333, but in fact this processor is able to work with much faster modules – there are no artificial limitations here. Like all other processors of the Haswell family, the overclocker Pentium has a built-in graphics core, which allows to use it even without a discrete video card. However, with the GT1 design, integrated graphics provide only rudimentary level of performance in 3D, so for most users it has no interest.

In the overclocking Pentium G3258 demonstrates a very good potential. Thanks to the relative simplicity of this 22-nm chip and the rather low heat dissipation (TDP is set at 53 W), the frequencies achievable for Pentium G3258 are even higher than those of Devil’s Canyon. For example, our specimen was able to function stably at 4.7 GHz without any problems, for which it was only necessary to increase the supply voltage to 1.375 V.

Thus, the Pentium Anniversary Edition was the most overclocked processor in today’s compilation in relative terms. With our cooler Noctua NH-U14S, its frequency was raised by 47 percent above the nominal. However, in this case, the Pentium G3258 is also the only dual-core that has come to our attention. Of course, Haswell microarchitecture provides an excellent specific performance of one core, but will there be enough two such cores to ensure that the Pentium G3258 can withstand equivalent AMD processors with twice and three times as many computing cores?

⇡ # Specification comparison

So, the main characters of today’s article are four cheap overclockers for three different platforms. Before proceeding to the test results, we will reduce their passport characteristics into a single table.

AMD FX-6300 AMD FX-4350 AMD Athlon X4 860K Intel Pentium G3258
The code name Vishera Vishera Kaveri Haswell
The processor socket Socket AM3 + Socket AM3 + Socket FM2 + LGA1150
The number of nuclei / flows 6/6 4/4 4/4 2/2
Hyper-Threading Technology No No No No
The clock speed 3.5 GHz 4.2 GHz 3.7 GHz 3.2 GHz
The maximum frequency in the turbo mode 4.1 GHz 4.3 GHz 4.0 GHz
The unlocked multiplier There are There are There are There are
L2-cache 3 × 2 MB 2 × 2 MB 3 × 2 MB 2 × 256 KB
L3-cache 8 MB 8 MB No 3 MB
Support for DDR3 1600/1866 1600/1866 1600/1866 1333
Extensions of the instruction set SSE4.2, AVX, AES SSE4.2, AVX, AES SSE4.2, AVX, AES SSE4.2
The built in graphics No No No HD Graphics
The technology of production 32 nm 32 nm 28 nm 22 nm
TDP 95 W 125 W 95 W 53 W
The official price $ 101 $ 97 $ 84 $ 72

If you notice an error – select it with the mouse and press CTRL + ENTER.

Tags: , , ,
Previous Post

at the crossroads of two worlds / Processors and memory

Next Post

Future of AI in Social Media

Leave a Reply

Your email address will not be published. Required fields are marked *