- VEGA in Citylink is MUCH cheaper than everywhere else
- Mid-segment hit Lenovo K5 Note is 30% cheaper at Citylink
- !!! GTX 1070 Gigabyte Stack 3x at an even MORE SUPER price
You can mark fragments of text that interest you,
which will be available via a unique link in the browser address bar.
Testing LGA 1155 processors, socket FM1, socket FM2 in 2D
Dmitry Vladimirovich 03/09/2013 00:00 Page: 1 of 5| | print version | | archive
- Page 1:
- Page 2: Test configurations, level of electricity consumption, test results: video playback Jellyfish 35/60/120 Mbit
- Page 3: Test results: WinRAR 4.2, Java Micro Benchmark, Excel BenchMark, XnView
- Page 4: Test results: Xilisoft Video Converter Ultimate, Xilisoft Audio Converter Pro, Pinnacle Studio 16, Adobe Photoshop CS6 (x64)
- Page 5: Test results: Cinebench (x64), final results: performance to price and power consumption ratio, performance to price ratio, summary table, conclusion
Twenty-one processors, three platforms, thirty-six modes - what is this, the construction of the century? Not at all. This is just the start of testing Intel models and AMD in various software conditions, which was made possible thanks to the help of our reliable partner - Regard. The emphasis is on everyday use. SPEC, Maya, SolidWorks, Lightwave, VRay, Autodesk - you won’t see any of the above in the review. But results await you in completely ordinary programs, those that we, I, and they use daily, weekly, at work and at home. And with what we constantly do for a number of reasons beyond our control.
More advanced entry-level chips
Click to find out more. Currently available chassis tested to be compatible with 95W processors should meet the requirements. We strongly recommend using a thermally enhanced chassis. First, make sure your desktop system's socket is compatible with the processor package type. You can obtain this information from the manufacturer of your motherboard for desktop PCs.
Once your desktop system model number is found, go to the appropriate manufacturer's support page. Fan box heaters work through baseboards because the space between baseboards is uniform. It must match the thermal design profile of the respective processors in order to properly respond to heatsinks suitable for the respective processors.
Are you shooting home video, no, you thought wrongly about Home Video, everything is simple to the point of banality, this is the video that we shoot in nature, while boarding or skiing, cycling or rollerblading. We take photographs with a camera and then process the photographs using the tools we are familiar with in programs. We work in the Office package and calculate in Excel. Download files from the Internet and unpack them. All those things that many of us do on our home and work computers.
Processors have previous versions and future compatibility with some important exceptions. Check with your desktop board manufacturer to ensure they are compatible with the processor you want to use and the recommended upgrade process.
64-bit operating systems are not compatible with 16-bit components, 16-bit applications, or 16-bit applications. Contact your software vendor for the 32-bit or 64-bit versions that you use with this operating system and the latest equipment. None of the voltages listed are likely to damage the processor or significantly reduce life cycle processor.
And since we are talking about entertainment, we need to see at what speed we are transcoding films into “tablet format” or preparing audio files excellent quality FLAC to the popular MP3 format. I will not focus your attention now on the comparison between Intel solutions and AMD, but I will conduct a test without being tied to a specific manufacturer. This is not least due to the price of processors.
This is a technology used in integrated circuit sockets that are designed so that the force required to insert or remove a microchip is low. So many were released motherboards, if any, who used it, especially since the standard socket 3 was sufficient. This bus includes data, address and control signals, as well as clock signals that synchronize its operation.
- Both manufacturers have excellent options.
- Will you be playing in what and at what resolution?
- But it's still a dead platform.
The fact is that I did not have a sufficient range of CPUs for a fair comparison. Intel products are several times more expensive than AMD models, but this does not mean that they will not take part in comparisons in the future; on the contrary, I am really looking forward to them. For now let's return to brief description processors.
Technical characteristics of processors with integrated graphics core
Intel model specifications (LGA 1155, Ivy Bridge).
Understanding Alphabet Soup
However, the same doesn't happen with desktops that are constantly plugged in - where productivity takes a higher priority. But some of them have the same model number as others. The suffix identifies certain characteristics of the chips.
Technical characteristics of processors with integrated graphics core
Ultrabooks with upgraded hardware should take several months to complete. You can also buy processors from hardware stores. The above figures were obtained from Kaboom without discount. In some stores the price may be 20% lower, depending on the form of payment. The branch block accesses it as soon as a new branch is predicted, and if successful, most of the front end processing becomes useless.
Model | Clock frequency | Intel® Turbo Boost | If- core quality | If- stream quality | Cache memory | Graphics model chetical system | Graphics frequencies logical core | Recommended bathroom price, $ | ||
Celeron G1610 | 2.6 GHz | - | 2 | 2 | 2 MB | Intel® HD Graphics | 650-1050 MHz | 55 Watt | 42 | 1500 |
Celeron G1620 | 2.7 GHz | - | 2 | 2 | 2 MB | Intel® HD Graphics | 650-1050 MHz | 55 Watt | 52 | 1900 |
Pentium G2010 | 2.8 GHz | - | 2 | 2 | 3 MB | Intel® HD Graphics | 650-1050 MHz | 55 Watt | 64 | 2100 |
Pentium G2020 | 2.9 GHz | - | 2 | 2 | 3 MB | Intel® HD Graphics | 650-1050 MHz | 55 Watt | 64 | 2100 |
Pentium G2120 | 3.1 GHz | - | 2 | 2 | 3 MB | Intel® HD Graphics | 650-1050 MHz | 55 Watt | 75-82 | 2600 |
Pentium G2130 | 3.2 GHz | - | 2 | 2 | 3 MB | Intel® HD Graphics | 650-1050 MHz | 55 Watt | 86-93 | 2900 |
Core i3-3210 | 3.2 GHz | - | 2 | 4 | 3 MB | Intel® HD Graphics 2500 | 650-1050 MHz | 55 Watt | 117-120 | 3700 |
Core i3-3220 | 3.3 GHz | - | 2 | 4 | 3 MB | Intel® HD Graphics 2500 | 650-1050 MHz | 55 Watt | 117-125 | 3800 |
Core i3-3225 | 3.3 GHz | - | 2 | 4 | 3 MB | Intel® HD Graphics 4000 | 650-1050 MHz | 55 Watt | 134 | 4400 |
Core i3-3240 | 3.4 GHz | - | 2 | 4 | 3 MB | Intel® HD Graphics 2500 | 650-1050 MHz | 55 Watt | 138-147 | 4500 |
Core i5-3330 | 3.0 GHz | 3.2 GHz | 4 | 4 | 6 MB | Intel® HD Graphics 2500 | 650-1050 MHz | 77 Watt | 182-187 | 5600 |
Core i5-3450 | 3.1 GHz | 3.5 GHz | 4 | 4 | 6 MB | Intel® HD Graphics 2500 | 650-1100 MHz | 77 Watt | 184-195 | 5800 |
Core i5-3470 | 3.2 GHz | 3.6 GHz | 4 | 4 | 6 MB | Intel® HD Graphics 2500 | 650-1100 MHz | 77 Watt | 184-195 | 5800 |
Core i5-3550 | 3.3 GHz | 3.7 GHz | 4 | 4 | 6 MB | Intel® HD Graphics 2500 | 650-1150 MHz | 77 Watt | 205-213 | 6400 |
Core i5-3570 | 3.4 GHz | 3.8 GHz | 4 | 4 | 6 MB | Intel® HD Graphics 2500 | 650-1150 MHz | 77 Watt | 205-213 | 6400 |
Core i5-3570K | 3.4 GHz | 3.8 GHz | 4 | 4 | 6 MB | Intel® HD Graphics 4000 | 650-1150 MHz | 77 Watt | 225-235 | 7600 |
Core i7-3770 | 3.4 GHz | 3.9 GHz | 4 | 8 | 8 MB | Intel® HD Graphics 4000 | 650-1150 MHz | 77 Watt | 294-305 | 9100 |
Core i7-3770K | 3.5 GHz | 3.9 GHz | 4 | 8 | 8 MB | Intel® HD Graphics 4000 | 650-1150 MHz | 77 Watt | 332-342 | 10300 |
Specifications of AMD models (socket FM1, Llano).
Model | Clock frequency | Turbo Core | If- core quality | If- stream quality | Cache memory | Graphics model chetical system | Graphics frequencies logical core | Maximum rated power | Recommended bathroom price, $ | Average retail price, rubles |
A4-3300 | 2.5 GHz | - | 2 | 2 | 1 MB | HD 6410D | 443 MHz | 65 Watt | 35-40 | 1200 |
A4-3400 | 2.7 GHz | - | 2 | 2 | 1 MB | HD 6410D | 600 MHz | 65 Watt | 40-50 | 1200 |
A6-3500 | 2.1 GHz | 2.4 GHz | 3 | 3 | 3 MB | HD 6530D | 443 MHz | 65 Watt | 60-70 | 1900 |
A6-3600 | 2.1 GHz | 2.4 GHz | 4 | 4 | 4 MB | HD 6530D | 443 MHz | 65 Watt | 80-90 | 2400 |
A6-3620 | 2.2 GHz | 2.5 GHz | 4 | 4 | 4 MB | HD 6530D | 443 MHz | 65 Watt | 80-90 | 2400 |
A6-3650 | 2.6 GHz | - | 4 | 4 | 4 MB | HD 6530D | 443 MHz | 100 Watt | 70-80 | 2400 |
A6-3670K | 2.7 GHz | - | 4 | 4 | 4 MB | HD 6530D | 443 MHz | 100 Watt | 80-90 | 2400 |
A8-3800 | 2.4 GHz | 2.7 GHz | 4 | 4 | 4 MB | HD 6550D | 600 MHz | 65 Watt | 90-100 | 3000 |
A8-3820 | 2.5 GHz | 2.8 GHz | 4 | 4 | 4 MB | HD 6550D | 600 MHz | 65 Watt | 90-100 | 3000 |
A8-3850 | 2.9 GHz | - | 4 | 4 | 4 MB | HD 6550D | 600 MHz | 100 Watt | 85-95 | 2800 |
A8-3870K | 3.0 GHz | - | 4 | 4 | 4 MB | HD 6550D | 600 MHz | 100 Watt | 80-90 | 2800 |
Specifications of AMD models (socket FM2, Trinity).
Performance and tests
This results in less front end utilization and ultimately a significant increase in performance per power consumption. Modern processors use out-of-order execution mechanisms, i.e. capable of processing instructions in a different order from assembly code written by the programmer or generated by the compiler.
These results are eventually copied into a retirement register file that corresponds to a post-processing set of architectural registers. The ring that controls them. Difficulty responding to simultaneous requests from four cores. The ring plays the role of an arbiter with simple design and several advantages: short paths and a simple increase in the number of agents who can connect to it. This type of arbitration sometimes arises in GPUs, some resources of which are distributed among many units.
Model | Clock frequency | Turbo Core | If- core quality | If- stream quality | Cache memory | Graphics model chetical system | Graphics frequencies logical core | Maximum rated power | Recommended bathroom price, $ | Average retail price, rubles |
A4-5300 | 3.4 GHz | 3.6 GHz | 2 | 2 | 1 MB | HD 7480D | 723 MHz | 65 Watt | 50-60 | 1600 |
A6-5400K | 3.6 GHz | 3.8 GHz | 2 | 2 | 1 MB | HD 7540D | 760 MHz | 65 Watt | 65-75 | 2200 |
A8-5500 | 3.2 GHz | 3.7 GHz | 4 | 4 | 4 MB | HD 7560D | 760 MHz | 65 Watt | 105-115 | 3400 |
A8-5600K | 3.6 GHz | 3.9 GHz | 4 | 4 | 4 MB | HD 7560D | 760 MHz | 100 Watt | 100-110 | 3200 |
A10-5700 | 3.4 GHz | 4.0 GHz | 4 | 4 | 4 MB | HD 7660D | 760 MHz | 65 Watt | 130-140 | 3700 |
A10-5800K | 3.8 GHz | 4.2 GHz | 4 | 4 | 4 MB | HD 7660D | 800 MHz | 100 Watt | 120-130 | 4200 |
It was not without reason that the FM1 and FM2 platforms were tested in full. Interest in APUs is constantly growing, and AMD's developments in this area will soon appear in a multi-platform set-top box. And it is even more interesting to check how much the gap between FM1 and FM2 has grown, and maybe even disappeared. They will be joined Intel processors. True, on this moment only a small proportion of them take part in testing, but with what we have on hand, we can draw certain conclusions.
This positioning, which requires small optimizations at the interconnect level, avoids increasing chip size. The challenge when a new set of instructions comes out is getting it accepted by developers. The first condition that a developer output instruction set must satisfy is a noticeable performance improvement.
Register management has been optimized to manipulate 256-bit operands, as we saw above. The most common operations involve 1 cycle latency on 256-bit operands: multiply, add, reorder, add a second 128-bit read port to double the power of 256-bit blocks.
Testing tools and methodology
The level of energy consumption is measured in three quantities.
- First, during idle time: all energy-saving functions of the motherboard (not the processor) are disabled.
- Second: 100% CPU load is carried out by running Prime x64.
- Third: 100% CPU+GPU utilization – in addition to Prime x64, EVGA OC Scanner X appears.
It is worth talking a little about the programs used in testing and the reasons for their choice.
The second condition is the ability of the proposed instructions to facilitate the work of developers and, in particular, data formatting. Conversely, in the “scalar” mode, the register contains only one data, which naturally reduces the interest in the instruction set. Initialization, which consists of optimally grouping data into registers, is still poorly done automatically by compilers, and the best results are always obtained through manual, time-consuming and often complex work.
Indeed, when all cores are not fully utilized, the processor suffers from its thermal performance. The principle is based on a very simple observation: when the processor is requested, it starts to heat up, and the temperature rise is not instantaneous, it takes the entire package some time to reach cruising temperature - if the load lasts long enough. Of course, the whole difficulty lies in the duration of use of this device. The longer it is applied, the more noticeable the acceleration, but the more the processor heats up beyond its specifications, protection.
Playing a video file using the classic player Media Player Classic with KLite codecs installed by default. The original video was taken from the work Jelly Fish from at different speeds flow from 35 Mbit to 120 Mbit. The first is typical for HD movies, those that we are used to buying and downloading from the Internet. 60 Mbit – the quality corresponds to Blu-ray, and 120 Mbit is equivalent to the developing 4K standard. The core load (average and maximum) is assessed, as well as the number of dropped frames. On modern processors There is practically no effect of frame twitching, and the number of frames in 99 cases out of 100 is easily maintained at 24.9-25.
It is difficult to choose the optimal value since the processor temperature rise obviously depends on the cooling mechanism used, a parameter external to the processor and therefore not quantifiable in one way.
This may seem paradoxical, since desktop computers usually win more effective solutions for cooling than laptops. We'll have to wait until next February 20th to see the dual-core models we'll be returning to on that date.
The square range of the kernel is decomposed as follows. The last letter then differentiates the processor ranges. There are also 45 and 65 W versions reserved for integrators. It will undoubtedly take some time to get used to this new nomenclature.
WinRAR 4.2 x64– built-in performance test is used. The program itself is located on a disk partition located on SSD drive, thereby eliminating low performance classic HDD. The test result is the average value obtained after three runs of the program. WinRAR appears in this review for a reason, because we often have to download and unpack files. Moreover, RAR is very common among archivers and supports multithreading well.
In this way, energy management can be unified through the power control unit. On the other hand, a limit will be required for models with less cache. A feature that is used on a case-by-case basis in pilots through a list pre-established by the manufacturer.
However, some changes have been forced in relation to the execution units, as we reported at the time. Increased number of managed registers, improved management of complex connections, and native support for more instructions. Only the maximum frequency varies depending on the model.
Java Micro Benchmark. An atypical test among processor reviews. Java Micro Benchmark allows you to compare system performance metrics on different platforms.
Excel BenchMark an even rarer guest. Initially, the task was to test the speed of work in the Office package. Converting from Word to PDF works well, but there is too much dependence on the rest of the system configuration, especially the HDD. And productivity growth is often higher from changing frequencies random access memory than from an additional 100-200 MHz processor frequency. Therefore, I had to look for a more adequate test that loads the processor-memory-chipset combination. Fortunately, such a test was found, and I saw additional attractiveness after the update test file. The author of the benchmark is definitely not sitting idle. So what is an Excel test? Initially, this is a table with data from which a dynamically changing graph is built during the benchmark process.
From our point of view this makes no sense. The difference mainly comes from the memory chips and other components present on the graphics card. However, at this frequency stability was not optimal, graphic applications sometimes they suddenly disappeared. At this frequency, throttling was still present. More strange than this segmentation. This allows us to maintain a constant processor frequency during these tests.
The magnification is identical in the three modes tested. The graphics load for the execution units increases depending on the resolution. Because if the memory controller is shared, the last level cache also passes through the ring bus.
There are six subtests in total.
- The first one creates five columns * 65,535 rows of random data.
- The second one displays the time required to calculate the indicator across five columns of the 65,535 rows of data created in the first test.
- The third one shows the speed at which price changes are displayed in 30 seconds.
- Fourth - 63,000 values with changed prices are converted to OHLC data.
- The fifth is a test using several conditions (wait until all price change values are formed) and calculation formulas. As a result, you can see the result of the number of price changes that can be completed with the recalculation formula within 30 seconds.
- Sixth - This test is identical to the fifth, except that all formulas are simultaneously dependent on a change in the cell, E5000.
XnView A fairly common program for viewing photographic material. It's free and easy to use. Additionally, it has built-in simple functions for converting formats, making changes, and others. I was interested in the everyday view of the test, or more precisely, how long it would take for the program to make changes and save 35 NEF format files. Typical requirements of an amateur photographer. But the task is complicated not just by changing the JPG format, but also by the requirement to make changes to the graphic files. The simplest and most obvious things were chosen: changing the color balance, changing the temperature, leveling the horizon, removing the bulge, adding sharpness, changing the size to 1900 pixels on the larger side. I won’t say that all resources are used during testing, but the result depends 85% on the speed of the processor. The remaining 15% is affected by the hard drive.
The values represent the percentage of threads relative to the number of physical cores on the chip. We watched the show on three platforms. While this may explain the loss of productivity, it doesn't explain everything. Several hypotheses are possible, and multiple factors are likely at play.
It should also be noted that, in the absence of a cause, the throttling we observed could also be due to another problem. Indeed, the given frequency may well be the result of insufficient graphics load, which will not force the graphics core to increase to its maximum frequency.
Xilisoft Video Converter Ultimate is a popular video converter. I used it for one reason - it knows how to load the processor well, using its capabilities 100%. From the entire list of possibilities, my choice fell on a 20-minute video file with one episode of the series in MKV 720p format, and the output should be a convenient file for viewing on a tablet. A task common among tablet owners who are buying more and more users. Of course, over the years, the number of CPU cores and the power of GPUs in them has been growing, but still not all copies can play back unconverted video.
The power management unit probably plays a role in this situation. According to sources, very good quality The differences are subtle and depend on the noise canceling choice more than anything else. Disabled by default, the option seems rather ineffective. His interest is even more so because once again, with good quality sources, the need for corrections is limited. In both cases it's not really about coding High Quality. Software Doesn't allow much wiggle room when it comes to setup additional parameters coding.
Xilisoft Audio Converter Pro. We convert an artist's album from FLAC to MP3, suitable for use on phones, tablets and players. The FLAC file is monotonous and filled with all the songs sequentially; we need to split it into compositions and save each one as MP3. A simple action for the user, but not simple for the system. The problem is that most audio converters do not load all the cores, meaning they are single-threaded jobs. Unfortunately, I couldn't find it suitable program, which adequately loads the processor, but it will be interesting to check how single-core acceleration technologies work on processors from different companies.
No, accelerated decoding, accelerated encoding and encoding decoding are accelerated. Before entering the results, we must insist that we have obtained four files of almost comparable, but not the same size. The same goes for quality, which we will discuss below. Therefore, encoding time should be considered with caution.
Note also that when it comes to accelerated encoding, the processor always contributes. Qualitatively, the results are certainly not identical. Knowing that the source video has a slightly higher bitrate than the one that was selected, in four cases it cannot be said that the quality of the final image is at the expected level. Color chromatic aberration appears on the right side of the image in three accelerated versions. In the end, the pure processor version clearly seems to us to be of a different level.
Pinnacle Studio 16 – a new version the most famous platform for processing video materials. This is Pinnacle Studio's first release since Corel acquired Avid's consumer video editing business. Theoretically, during the final assembly of video material, the program uses all processor technologies, but most importantly, it is multi-threaded! The program itself is very common among amateur editing systems, but we don’t need much. I decided to combine several fragments from the action camera into one, provide them with smooth transitions and bring the picture to the same temperature, as well as color balance and sharpness.
Adobe Photoshop CS6 (64 Bit)– there’s no need for a lot of words here. The latest Adobe product available. The test result is the time it takes to apply filters to one picture. I did not become like many people and took an ordinary JPG file of medium size. And then I went through filters, resizing, gamma settings, and so on. Quite a typical set for the program. Unlike video encoding, Photoshop never became multi-threaded; rather, it can be called a moderately CPU-intensive program.
Cinebench x64– a common processor test in rendering. Initially, I would like to provide the results in Autodesk 2013 packages, but due to the tight binding to the system configuration, changing the processor requires new registration product. And even after re-registration, the package does not work properly, as a result, I had to abandon it. Having the results of one system with different processors in Autodesk, I compared them with the difference according to the Cinebench testing results and did not find a significant difference.
The results are presented both as a table and as a summary graphical table. In order to compare processors, just select the model you are interested in horizontally, and vertically the CPU being compared, in the cell you will see the percentage difference in performance between the first and second CPU.
- Page 1: Introduction, specifications processors with built-in graphics core, tools and testing methods
- HTC U Play in Citylink is cheaper than from gray sellers
- !!! GTX 1070 Gigabyte Rock drain, super price!
- 4k flagship Sony XZ Premium - check the price!
- Processors Core i3-2100, Penium G850 and Pentium G620;
- Maternal ASUS board P8P67 Deluxe;
- Processor Pentium G6960;
- ASUS P7P55D PRO motherboard;
- Memory 4 GB DDR3-1067 7-7-7-21 (Kingston KHX1600C8D3K2/4GX).
- Processors Athlon II X2265 and Athlon II X3 455;
- Motherboard M4A89TD PRO;
- Memory 4 GB DDR3-1600 9-9-9-27 (Kingston KHX1600C8D3K2/4GX).
- CPU-Z 1.58;
- Futuremark PCMark 7 1.0.4;
- Futuremark 3DMark Vantage 1.1.0;
- Futuremark 3DMark 11 1.0.2;
- SuperPi mod 1.5 XS;
- wPrime 2.05;
- WinRAR 4.01 x64;
- 7-zip 9.20 x64;
- Fritz Chess Benchmark 4.2;
- MAXON Cinebench Release 11.5 x64;
- TechARP x264 HD Benchmark 4.0;
- Crysis 2 1.9;
- Far Cry 2 1.0.3;
- Metro 2033: The Last Refuge 1.2;
- DiRT 3 1.1.
- In 1280x1024 resolution with High quality settings and full-screen anti-aliasing disabled;
- In a resolution of 1920x1080 with Ultra High quality settings and the maximum available level of full-screen anti-aliasing.
Description of test systems
In testing, we compared the new LGA1155 Pentium processors with both higher-class processors, Core i3, and with the products they replace - Pentium processors in the LGA1156 version. A performance study wouldn't be complete without also including AMD processors. Therefore, a couple was added to the list of rivals Athlon processors II with two and three cores, which can be purchased for about the same amount as the Pentiums in question.
In the end, three different test platforms were used.
LGA1155 platform:
LGA1156 platform:
Socket AM3 platform:
In all these platforms the graphics card remained constant NVIDIA GeForce GTX 560 Ti and HDD Western Digital VelociRaptor WD3000HLFS. Testing was carried out in operating system Microsoft Windows 7 SP1 Ultimate x64.
Formal characteristics of the processors that took part in testing:
Core i3-2100 | Pentium G850 | Pentium G620 | Pentium G6960 | Athlon II X2 265 | Athlon II X3 455 | |
---|---|---|---|---|---|---|
Microarchitecture | Sandy Bridge | Sandy Bridge | Sandy Bridge | Westmere | Stars | Stars |
Frequency, GHz | 3,1 | 2,9 | 2,6 | 2,93 | 3,3 | 3,3 |
Cores/Threads | 2/4 | 2/2 | 2/2 | 2/2 | 2/2 | 3/3 |
L2 cache, KB | 2x256 | 2x256 | 2x256 | 2x256 | 2x1024 | 3x512 |
L3 cache, MB | 3 | 3 | 3 | 3 | No | No |
Technical process, nm | 32 nm | 32 nm | 32 nm | 32/45 nm | 45 nm | 45 nm |
TDP, W | 65 | 65 | 65 | 73 | 65 | 95 |
Official price, $ | 117 | 86 | 64 | 89 | 69 | 76 |
Software used:
In games, testing was carried out twice:
⇡ Test results in applications
Summarizing the numbers shown in the graphs, several conclusions can be formulated. Firstly, although the new Pentiums belong to the Sandy Bridge generation of processors, it is very difficult to call them close relatives of the Core i3. There is a whole chasm between the results of these processors, which is deeper, the better an application is optimized for multi-threading. Still, the lack of support for Hyper-Threading is a very serious flaw, which, together with limited clock frequencies, results in an average lag of 30% between the Pentium G850 and the Core i3-2100.
Secondly, the new Pentiums actually turned out to be significantly faster than the old ones - this is quite expected and natural.
But in confrontation with AMD processors there is no such certainty. The Pentium G620 is not always ahead of the Athlon II X2 265. One can even say that a certain parity has been established between these processors, with the exception of SuperPi, where representatives of the Pentium family performed with phenomenally high results. The Athlon II X3 455 turns out to be a powerful competitor even for the older of the new Pentiums. Its three cores provide a crushing advantage if the application is well optimized for multi-threading. In the absence of explicit optimization, the Pentium G850 and Athlon II X3455 demonstrate approximately equal performance.
⇡ Test results in 3D games
We would like to highlight the performance indicators we obtained in the popular 3DMark series packages. These are still not real games, but some synthetic tests that simulate them. Moreover, those who do this are clearly biased in favor of Intel processors, since the results of the Athlon II here clearly do not agree with what we saw earlier. Even the full-fledged three-core Athlon II X3 455 here fails to outpace the new Pentiums, which have only two cores and do not support Hyper-Threading.
However, despite the above, the new Pentiums are certainly faster than the old ones in games, so some progress is obvious. Moreover, they absolutely work and are better than the dual-core Athlon II.
In passing, I would like to note this. We have long been accustomed to the fact that gaming performance depends primarily on power graphics card. The processor affects the number of frames per second very little, especially when using high resolutions and high quality settings. However, in the case of weak CPUs, such as the Pentiums we are considering today, this message turns out to be incorrect. Even at maximum settings, the dependence of performance on the processor is present, and sometimes it is even significant. So when creating an inexpensive gaming system, you need to pay close attention to balancing the capabilities of the CPU and GPU.
⇡ Overclocking
Overclocking is the weak point of any LGA1155 processors, except for special overclockers Core models i7-2600K and Core i5-2500K. Due to the unification of the processor clock circuit and interface elements, overclockers have lost the ability to increase the BCLK frequency almost completely. Increasing the processor multiplier is allowed only on selected processors, and cheap Pentiums, of course, are not one of them.
Therefore, all hopes that Pentium performance can be increased through overclocking should be abandoned. The limit for increasing the BCLK frequency on most motherboards is in the region of 104-107 MHz, which means you should not count on an increase in the final CPU frequency of more than 5-7%.
For example, when experimenting with the Pentium G850 in our test system, we were able to increase the BCLK frequency without compromising stability only up to 105.6 MHz, which resulted in an increase in the processor frequency from the stock 2.9 GHz to a modest 3.06 GHz.
This means that if you are serious about overclocking experiments and believe that the processor should function at the limit of its capabilities, you clearly should not opt for the Pentium for LGA1155. Any of inexpensive processors for alternative platforms that do not block overclocking, will allow you to get better performance after an abnormal promotion clock frequency. In other words, the new Pentiums are contraindicated for overclockers.
⇡ Conclusions
The appearance of Pentium processors for the LGA1155 platform can only be welcomed. Their main advantage is the reduction in the minimum cost of the LGA1155 system, and this overshadows all their disadvantages. Given the promise of the LGA1155 platform, the new Pentium processors open up the possibility of creating low-cost computers that can subsequently be easily improved to achieve the most top level productivity.
However, systems based on Pentium G850/G840/G620 are not so bad even if you don’t think about upgrading. These processors are significantly faster than their predecessors, sold under the same name, and even competing ones dual core processors They are superior to AMD. The only frustrating thing is their serious lag behind full-fledged Core i3s, but such is the fate of this family. The characteristics of the Pentium compared to other Sandy Bridges are cut down too much to be put on par with Core processors second generation.
The real alternative for the new Pentium can only be the competitor's three-core processors, which show better performance under multi-threaded loads. In addition, the weak point of the new product is the almost complete lack of overclocking. Otherwise, the Pentium G850/G840/G620 are quite normal offers, the capabilities of which, in our opinion, are quite consistent with their cost.