Review of the energy efficient Intel T.

17.04.2019 Internet, Wi-Fi, local networks

Test bench configuration
motherboards	Intel DP67BG; MSI 890GXM-G65; ASUS Maximus III Extreme
Processors	AMD Phenom II 1100T; AMD Phenom II 1075T; AMD Phenom II 975; Intel Core i5-2500K; Intel Core i7-870
CPU cooling systems	Thermalright Silver Arrow; Intel BOX Cooler
RAM	2 x 1024 MB DDR-3 Apacer
Video adapter	NVIDIA GeForce GTX 580
HDD	Seagate Barracuda 7200.10 750 GB (ACHI Mode)
Power Supply	IKONIK Vulcan 1200 W
Frame	Cooler Master test bench 1.0
Operating system	Windows 7 Ultimate x64

Before overclocking the processor and testing its performance, let's take a look at how AMD Turbo Core technology works. To do this, we will load from one to six Phenom II 1100T cores one by one and monitor their frequency using the AMD Overdrive program, and using the task manager, we will set a strict correspondence between each thread and each physical processor core.

As you can see, in all cases, except for the last one (with all six cores loaded), the frequency of at least one core dropped to a value lower than the nominal one. At the same time, during loading from one to three cores, their frequency increased by one step to 3.7 GHz. Actually, Turbo Core does not provide for the next stages.

The operation of AMD Turbo Core technology can be called quite clear and consistent with the stated conditions of its operation. However, compared to the competitor's solution, the principle of operation of AMD Turbo Core is implemented a little rough and simple. And, of course, that Turbo technology Core is available only to owners of six-core processors so far, we can safely blame AMD marketers. Apparently, the company was afraid of possible internal competition for its products. So, if automatic overclocking was available for quad-core solutions, the sales of Phenom II 1100T could be hindered by the increased interest in the same AMD Phenom II 975, since often using a higher-frequency quad-core processor is much more efficient than using a six-core one.

Testing and overclocking of AMD processors were carried out on the motherboard MSI board 890GXM-G65.

The maximum frequency that was achieved at air cooled with a Thermalright Silver Arrow cooler and with the system running completely stable, it was 4077 MHz. The result is not very impressive, especially against the background of Intel processors, for which overclocking in air to 4 GHz is not a special achievement. With the release of the new processor, the frequency limit of six-core AMD chips has not increased.

The main parameters of the overclocked system are shown in the table:

Parameter	Nominal value	Manual overclocking value
CPU Frequency, MHz	3330	4077
DRAM Frequency, MHz	1333	1551
HT Frequency, MHz	2000	932
СPU NB Ratio	10	8
Active Processor	All	All
Cool "n" Quiet	Enabled	Disabled
C1E Support	Enabled	Disabled
AMD Turbo Core	Enabled	Disabled
BUS Frequency, MHz	200	233
CPU Ratio	16,5	17,5
Memory Timings (Ch1 / Ch2)	9-9-9-24/9-9-9-24	9-9-9-24/9-9-9-24
PCI-E Frequency, MHz	100	100
NB Voltage, B	1,30	1,337
CPU Voltage, V	1,48	1,531
CPU VTT, V	1,118	1,48
CPU-NB VDD	1,050	1,45
DRAM Voltage, V	1,60	1,60
Turbo Core	Enabled	Disabled

Below is a summary table of the results of testing the AMD Phenom II 1100T processor at the nominal frequency and with manual overclocking, as well as the test results of the subject's rivals.

	AMD Phenom II 1100T @ 3.3 GHz		AMD Phenom II 1075T @ 3.0 GHz	AMD Phenom II 975 @ 3.6 GHz	Core i5-2500K @ 3.3GHz	Core i7-870K @ 2.93 GHz
Everest, Reading from memory, MB / s	8560	8541	8504	8177	16020	15644
Everest, Write to memory, MB / s	7080	6620	6933	6824	17536	10892
Everest, Copy in memory, MB / s	11078	10681	10956	9817	18020	15601
Everest, Memory latency, ms	51,1	48,9	49,5	55,4	54,0	52,2
Everest, CPU Queen, points	32455	39630	29610	26397	38624	36735
Everest, CPU PhotoWorxx, points	26238	29292	26454	29973	42150	35695
Everest, CPU Zlib, MB / s	131576	162254	119837	95312	118037	101694
Everest, CPU AES, points	35728	44039	32515	25977	330508	25525
Everest, FPU Julia, points	14226	17509	12947	10547	14018	13431
Everest, FPU Mandel, points	8208	10065	7479	6068	7321	7296
Everest, FPU SinJulia, points	4318	5343	3919	3139	4837	5989
Super Pi 1M, s	19,969	17,472	19,952	19,328	10,265	12,277
wPrime 32M, s	8,488	6,597	9,578	11,388	9,377	8,443
wPrime 1024M, s	251,129	205,546	277,602	340,953	287,054	255,211
Cinebench R10 X64 Multi CPU Score	18836	22329	17325	14640	20012	18630
Fritz Chess Benchmark, Relative Speed	11250	13735	10511	8648	10142	11719
x264 HD Benchmark 3.0, FPS	75,34	86,31	70,62	74,18	96,39	74,18
7-Zip 9.20 x64, MIPS	17963	21542	16815	13421	14073	18614
WinRAR x64 4.00 Beta4, KB / s	2857	3200	2809	2414	3073	3194

3D tests
Test name, mode, units	AMD Phenom II 1100T @ 3.3 GHz	AMD Phenom II 1100T @ 4.08 GHz	AMD Phenom II 1075T @ 3.0 GHz	AMD Phenom II 975 @ 3.6 GHz	Core i5-2500K @ 3.3GHz	Core i7-870K @ 2.93 GHz
Crysis x64 v1,2 CPU Benchmark, 1280x1024, FPS	66,99	74,24	64,21	64,44	96,45	85,02
Crysis x64 v1,2 GPU Benchmark, 1920x1200, FPS	36,34	36,54	36,24	35,84	37,49	37,37
DiRT 2, 1280x1024 no AA / AF, FPS	56,88	59,52	53,14	83,39	105,45	123,3
DiRT 2, 1920x1080 4xAA / 16xAF, FPS	51,02	59,48	51,51	78,65	93,58	98,55
Resident Evil 5, 1280x1024 no AA / AF, FPS	88,6	99,8	84,2	88,0	124,9	128,6
Resident Evil 5, 1920x1080 4xAA / 16xAF, FPS	81,8	95,4	80,6	84,0	115,8	113,4
Far Cry 2 DX10, 1280x1024 no AA / AF, FPS	81,95	84,97	82,43	87,65	117,53	108,48
Far Cry 2 DX10, 1920x1080 8xAA / 16xAF, FPS	84,89	88,85	85,20	92,22	98,81	99,97
3DMark Vantage, Performance, Overall / CPU	20236 16888/	21822 20405/	19477/15456	18424/12358	21956/ 17353	22457/ 19831
3DMark 2011, Performance, Overall / CPU	5281/5335	5474 5841/	5225/5067	5156/4367	5664/ 6383	5731/6562

Below are the comparative diagrams of processor performance based on pivot tables for each of the test groups.

Working with RAM:

Testing the speed of reading, writing and copying in RAM clearly shows the advantage of an Intel processor memory controller over an AMD controller. Despite the higher latencies when accessing memory, the speed of working with RAM in Intel processors is almost twice as compared to AMD processors when using the same modules in dual-channel mode.

Integer and floating point performance.

Everest synthetic tests show a rather diverse picture. Some algorithms, such as PhotoWorxx, CPU Queen, are better given to Intel processors, and some, on the contrary, prefer AMD (ZLib) processors. Here, apparently, the number of physical cores in the CPU plays an important role.

Testing the speed of work with AES demonstrated a significant advantage of the Core i5-2500K processor, which is quite natural, since only this processor out of all test participants has a block hardware acceleration work with AES.

The performance of the AMD Phenom II 1100T in the WinRAR archiver turned out to be slightly lower than in the quad-core Intel processors. Only after overclocking the AMD flagship managed to catch up with the Core i7-870, which in this test helped by the presence of Hyper Threading technology. 7-zip turned out to be more sensitive to the number of executable threads, so processors with six cores left quad-core solutions behind. However, the same Core i7-870, which has eight processing threads (due to HT), outperformed the competitor's six-core processors operating at nominal frequencies. Overclocking AMD Phenom II 1100T changes the balance of power and allows the new product to outperform its competitors, but let's not forget that Intel chips are overclocked to much higher frequencies.

The Fritz Chess Algorithm Test shows a similar picture to 7-zip. In the Cinebench R10 test, the advantages of the new Intel architecture Sandy bridge, which allows the Core i5-2500K processor, with only four execution threads, to bypass not only the Core i7-870 with its eight virtual cores, but all six-cores from AMD. However, overclocking the AMD Phenom II 1100T processor again gives a noticeable performance boost and makes it the leader.

Encoding HD content using the x264 codec is insensitive to a large number of streams executed by the processor; the frequency and architectural features of the chips play the main role here. All processors are on par in this test, and only the Core i5-2500K won an absolute victory. The overclocking of the AMD Phenom II 1100T processor could not change the alignment of forces either.

The WPrime test, on the other hand, scales well with the number of cores and threads, so all the "pure" quad-cores are behind, although the Core i5-2500K is close to the AMD Phenom II 1075T, and the Core i7-870 is almost on par with the overclocked AMD Phenom II 1100T. The acceleration of our subject sharply brought him to the leading position.

Beloved by enthusiasts since ancient times, the Super Pi test, which uses only one processor core, gives clear preference to Intel processors, and no "amateur" overclocking of AMD processors will change the distribution of the pedestal.

Performance in 3D applications

The situation here turned out to be quite unambiguous and not very rosy for AMD. In all test games, Intel processors take the leading position in both light and heavy graphics modes for a video card. If systems based on Intel processors react strongly enough to the transition from light graphics modes to heavy ones, that is, a video card becomes a bottleneck here, then a system based on AMD processors practically does not notice this transition. This behavior can only mean one thing: the processor limits the possibilities graphics adapter in both hard and light modes. Increasing the CPU frequency partially corrects the situation, but does not fundamentally change it. Only in the hard modes of Crysis does the performance limit precisely to the possibilities NVIDIA graphics cards GeForce GTX 580.

A similar picture can be seen in the 3DMark 2011 and Vantage test suites. If in processor tests after overclocking the AMD Phenom II 1100T still manages to catch up, and in Vantage even to overtake Intel processors, then due to the graphic tests of the video card according to the final rating, Intel processors are in the lead.

Such a lag of AMD processors, most likely, may be due to the strong advantage of the Intel memory controller over a similar solution from AMD, which can no longer provide the required modern games memory bandwidth, as well as the general weakness of the K10.5 architecture.

In conclusion, it is worth saying a few words about temperature and power consumption.

The temperature range of the AMD Phenom II 1100T processor is comparable to that of the previous models from the company. Without overclocking and with the Cool "n" Quiet technology enabled, the fan speed rarely rose above the minimum values under the Thermalright Silver Arrow cooler, and the heatsink was a little warm all the time, but it was worth turning off Cool "n" Quiet and overclocking the processor to 4077 MHz as it heats up increased significantly. There were some problems with determining the temperature values of the processor cores, since different utilities produced different indicators. Even the AMD Overdrive utility often showed temperatures below room temperature, which, in principle, cannot be, while the radiator was noticeably warmer than the human body.

With regard to energy consumption, then everything is quite clear:

	AMD Phenom II 1100T @ 3.3 GHz	AMD Phenom II 1100T @ 4.08 GHz	AMD Phenom II 1075T @ 3.0 GHz	AMD Phenom II 975 @ 3.6 GHz	Core i5-2500K @ 3.3GHz	Core i7-870K @ 2.93 GHz
Idle, W	104	137	103	99	65	110
OCCT CPU Load, W	239	386	229	212	124	206
Far Cry 2 DX10, 1280x1024 no AA / AF, W	290	343	278	274	239	280
Far Cry 2 DX10, 1920x1080 8xAA / 16xAF, W	330	428	321	313	256	308
Crysis x64 v1.2 CPU Benchmark, 1280x1024, W	307	370	304	298	260	318
Crysis x64 v1.2 GPU Benchmark, 1920x1080, W	340	458	334	333	278	325
7-Zip 9.20 x64, W	195	310	182	172	115	180
WinRAR x64 4.00 Beta4, W	156	226	151	150	100	150
x264 HD Benchmark 3.0, W	206	350	193	180	116	183

In almost all system modes, the Intel Core i7-870 processor and all tested AMD processors show comparable power consumption values. And the clear leader here is the Intel Core i5-2500K processor, which, paired with the Intel P67 logic set, turns out to be much less power hungry, in comparison with both the previous generation Intel platform and the current AMD platform. At the same time, overclocking the AMD Phenom II 1100T noticeably raises the system's power consumption in all operating modes.

⇡ Conclusions

The new processor from AMD did not revolutionize the market, it is only a logical continuation, and maybe the end of the line based on the K10.5 architecture. With almost equal prices for motherboards for AMD and Intel processors, AMD platform does not have any advantages over competitor solutions. Despite the fact that in a number of tests AMD Phenom II 1100T is on an equal footing with a competitor's processors in one price category, and after overclocking it breaks out to the first place, this six-core is still able to compete only with four-core Intel processors. This gives the latter the opportunity to set exorbitant prices for its own six-core chips. Of course, in light of the recent events with the recall of motherboards based on the Intel P67 chipset, the advantage of the main processor giant of this planet may fade slightly, but no one canceled the previous proven platform of the same company, which still looks more preferable than AMD solutions in the same price range.

In such a situation, all AMD's hope is for the fast-moving Bulldozer, which, if it does not "level the competitor" to the ground, then, we hope, will give it a worthy rebuff. After all, good competition in the market is a guarantee of progress and correct prices.

When choosing a processor from Intel, the question arises: which chip from this corporation to choose? Processors have many characteristics and parameters that affect their performance. And in accordance with it and some features of the microarchitecture, the manufacturer gives the appropriate name. Our task is to highlight this issue. In this article, you will find out what exactly the names of Intel processors mean, as well as learn about the microarchitecture of chips from this company.

Indication

It should be noted in advance that solutions earlier than 2012 will not be considered here, since technologies are moving at a fast pace and these chips have too low performance at high power consumption and it is also difficult to buy them in new condition. Also, server solutions will not be considered here, since they have a specific scope and are not intended for the consumer market.

Caution The nomenclature below may not be valid for processors older than the above period.

And also if you have any difficulties, you can visit the site. And read this article, which tells about. And if you want to know about integrated graphics from Intel, then you should.

Tick-tock

Intel has a special strategy for releasing its "stones" called Tick-Tock. It consists of annual incremental improvements.

Tick means a change in microarchitecture, which leads to a change in socket, better performance and optimization of power consumption.
This means that it leads to a decrease in energy consumption, the possibility of location more transistors on a chip, a possible increase in frequencies and an increase in cost.

This is how this strategy looks for desktop and laptop models:

MODEL "TIK-TAK" IN DESKTOP PROCESSORS

MICROARCHITECURA	STAGE	EXIT	TECHPROCESS
Nehalem	So	2009	45 nm
Westmere	Teak	2010	32 nm
Sandy bridge	So	2011	32 nm
Ivy bridge	Teak	2012	22 nm
Haswell	So	2013	22 nm
Broadwell	Teak	2014	14 nm
Skylake	So	2015	14 nm
Kaby lake	So +	2016	14 nm

But at low-power solutions(smartphones, tablets, netbooks, nettops) platforms are as follows:

MICROARCHITECTURE MOBILE PROCESSORS

CATEGORY	PLATFORM	CORE	TECHPROCESS
Netbooks / Nettops / Laptops	Braswell	Airmont	14 nm
	Bay Trail-D / M	Silvermont	22 nm
Top tablets	Willow trail	Goldmont	14 nm
	Cherry trail	Airmont	14 nm
	Bay Tral-T	Silvermont	22 nm
	Clower trail	Satwell	32 nm
Top / mid-range smartphones / tablets	Morganfield	Goldmont	14 nm
	Moorefield	Silvermont	22 nm
	Merrifield	Silvermont	22 nm
	Clower Trail +	Satwell	32 nm
	Medfield	Satwell	32 nm
Medium / budget smartphones / tablets	Binghamton	Airmont	14 nm
	Riverton	Airmont	14 nm
	Slayton	Silvermont	22 nm

It should be noted that Bay Trail-D is made for desktops: Pentium and Celeron with index J. And Bay Trail-M for is mobile solution and will also be denoted among Pentium and Celeron by its letter - N.

Judging by the latest trends of the company, the performance itself is progressing rather slowly, while energy efficiency (performance per unit of energy consumed) is growing from year to year, and even soon, laptops will have the same powerful processors as on large PCs (although such representatives there is now).

At the dacha, I use an Acer Extensa 5630g laptop as desktop computer, with external monitor and keyboard.

The stock processor Core2Duo T5800 and 3 GB of memory were not enough, almost at the limit of comfortable work.

It was decided to carry out a radical upgrade: change the processor, add memory and replace the hard drive with an SSD.

Bios-mod, SetFSB, TME-unlock, Q9xxx quads, stepping, overclocking, Throttlestop.

Read on about the thorny path and results. Lots of pictures, lots of text.

The camera broke down while writing the article, so the photos did not work out as we would like.

Actually, this is the lower half of the laptop, the upper part was in my way and I just took it off.

Almost the only task of the device is to browse the Internet via a 3G modem. The speed is quite low, a couple of megabits per second (although I even made an antenna from a Soviet aluminum electric heater with a diameter of about 25 cm, + 5 dB), and the traffic is paid. So far, I fit into 1GB per month and do not plan to pay the telecom operator for large volumes.

For such harsh conditions, the assembly of the Opera browser - OperaAC based on the old Presto kernel is ideal. In it, you can use a separate button to enable the download of images, disable it, or download images from the browser cache for each tab. You can give a command to download a particular picture. Flash animation is disabled by default, the required flash element can be downloaded separately. Adblock cuts unnecessary ads, primarily text ads. The service complements all this Opera Turbo.

In fact, everything turns out pretty quickly and comfortably. The web pages look clean and tidy :). For the rare cases when Opera can no longer cope with some complex sites, usually actively interacting with the user, Chrome and Firefox are installed.

About the laptop: Acer Extensa 5630g, 2008, business line. Discrete video ATI HD3470 256MB, Intel PM45 chipset, DDR2 memory, Intel T5800 processor - 2GHz on 800MHz bus, 2MB cache memory.

The laptop is pretty tormented by life: someone reballated or changed the Northman (why? There is a problem-free Intel PM45), the touchpad is flooded and does not work, the battery lasts for 5-10 minutes, the power plug is reinforced with electrical tape, the matrices were poked in the cable, the case is broken in places, dvd -rw The drive is not very confident in reading discs. And nevertheless - everything works.

Why a laptop? - takes up little space, lightweight, mobile, equipped with built-in Wi-Fi and uninterruptible power supply, consumes little electricity. Mega performance is not required from him ...

O Acer laptops

For several years I have held and disassembled several dozen laptops. Many of them were Aspire Eisers from 2005-2008. All of them made a bad impression: cheap badly painted plastic, minimum metal, dubious technical solutions, pop design. Inside, it was the same as outside. The maximum cost reduction was evident in everything, simultaneously with the desire to look cooler outwardly. The result was some kind of cheap pop. I wouldn't buy it myself. Even Asus was made in a completely different way, not to mention HP, Toshiba, etc.

Surely, now something has changed, but then it was like that. Owners of Acer - do not be offended, please.

However, picking up the Acer Extensa 5630, I suddenly realized that this is not the same Acer. Obviously not cheap plastic, rigid construction, excellent keyboard, well thought-out technical solutions inside - a massive radiator, a large Sunon cooler with maglev technology, connectors, hinges - everything is done at a qualitatively different level. Everything turned out to be simple: it's not Acer, it's Homa. I have not found any information on this Chinese contract manufacturer, but judging by Google, it is strongly associated with Acer.

A definite plus is the presence of a full discrete graphics card from AMD, the ability to normally use external video outputs and the Catalyst control program.

With many other laptops with embedded video, usually from Nvidia I couldn't get normal work with an external monitor. Everywhere there were some pitfalls, as if the conclusion on external monitor this is a secondary functionality.

Everything is fair and transparent right there: one analog, two digital outputs, the ability to shuffle them as you like in a catalist, as if I were using discrete video on a desktop. True, one of the digital pins is "tightly" tied to the built-in digital interface. In addition to it, you can connect one external digital, one analog and use three screens at the same time. The video card can decode HD video normally. The latest version of the catalyst for this platform is 13.9. There are zero problems in working on an external monitor, as if it were a desktop.

The PM45 chipset supports processors with a bus frequency up to 1066 MHz - the maximum for this platform, supports mobile quad-core (!) Processors, 8GB DDR2 or DDR3 memory. However, more on that later.

A maximum of 8GB of memory can be installed in two DDR2 slots. But 4GB slats are grown-up, so affordable option- 2x2GB DDR2-800 or 667.

Bus acceleration

Initially, I just wanted to overclock my T5800 on the bus, from 800 to 1066 MHz and in frequency from 2 to 2.66 GHz. After all, the multiplier for ordinary Intel processors has been locked for a long time, you can only drive it by frequency.

Who is not going to drive with a tire, feel free to scroll through.

The SetFSB program has been downloaded. My clocker, Realtek RTM875N-606 (this is a chip that generates frequencies) was not in the list of supported ones, similar ones were found. After all, it is not the clocker's model that is important, but the data written to the registers; they often coincide in cloakers with similar markings. For example, I chose a similar clocker RTM875T-606 and changed the value of the registers directly. Did not help…

Choosing a processor

Having considered the offers on Ali from the price-quality segment, the following list of candidates with a FSB frequency of 1066 MHz was compiled, the store chose one of the cheapest and with a good reputation:

P8700 - 2.53Ghz, 3mb cache, TDP 25W for $ 3.28
T9400 - 2.53Ghz, 6mb cache, TDP 35W for $ 4.2 (now it has risen in price by a dollar)
T9550 - 2.66Ghz, 6mb cache, TDP 35W for $ 6.35
T9600 - 2.8Ghz, 6mb cache, TDP 35W for $ 11.05 (now cheaper by a dollar)

The cheapest quad is Q9100, 2.26Ghz, 12mb cache, $ 21.5. Unfortunately not compatible with my laptop.

The T9800 and T9900 cost over twenty dollars with minimal overclocking. Cross out.

It is easy to see that the characteristics of the P8700 and T9400 differ only in the thermal package and the size of the L2 cache, 3 versus 6 megabytes.

I wonder what kind of speed increase is given by the additional 3MB of cache?

The local guru, kizwan, fulfilled the wish, posted the mod. MZWiZard flashed the BIOS into its extenza and got a brick. Everything. No way to restore the BIOS, even using a special floppy disk (!), Helped. After a couple of weeks, MZWiZard dropped his hands. It turned out that kizwan made a mistake in the bios mode. Then it was fixed and a working version of the BIOS was uploaded.

However, six months later, the comrade MZWiZard returned to the idea of installing a quad. As follows from his posts, he had to buy a new motherboard in a service center ...

In general, the firmware of any BIOS is a responsible business. Moreover, the modified one.

1) the battery must be fully charged (the laptop may refuse to flash if the battery is faulty or not charged)
2) flash strictly from DOS
3) the flash drive should be of the minimum size, very desirable up to 2GB, FAT32
4) it is highly recommended to make an emergency flash drive or emergency floppy disk.

In case of failure, using rescue disk you can try to restore the BIOS. By the way, in the service manual for the Acer Extentsa 5630 it is prescribed to use a floppy disk via a USB-floppy drive.

I personally did not find a USB floppy drive for an adequate price, so I took a chance and stitched like this.

The 32MB flash drive was formatted with the HP USB Disk Storage Format Tool", Threw the boot files and firmware there. The firmware took about 5 minutes, after which I saw “Intel Menu” in the BIOS with sections CPU, MCH, ICH, etc.

If in the CPU settings section everything is more or less clear and comparable to desktops, then in MCH there is such a jungle that I would not touch the parameters from there at all.

Actually, I just needed to turn off Intel Speedstep.

Overclocking with Throttlestop

After loading the system, energy saving naturally did not work anymore, the frequency is 10x266 = 2.66 GHz. Load the Throttestop program.

We remove all the jackdaws from the checkboxes. Press the Turn on button. Click on Options and check the Start Dual IDA checkbox.

We leave to the main window. Set the checkbox in the EIST checkbox. We have enabled manual energy saving management and can manage it.

We set the maximum available multiplier (10.5 in my case) and voltage. Without load, the multiplier rises to 10.5, and when loaded, it automatically decreases to the standard 10.

And now we REMOVE the checkbox from EIST and ... the multiplier 10.5 and the frequency 2.8 GHz were and remain. Now you can load the processor as you like - it works in turbo mode on an ongoing basis :)

add Throttlestop to startup and it will do everything when the computer is turned on: tick the EIST checkbox, wait for the multiplier to increase, and uncheck it.

Throttlestop can change multiplier and voltage depending on load, just like in RMClock. However, I did research and found out that lowering the frequency and voltage during idle has almost no effect on power consumption and there is no point in it.

To change the supply voltage or the processor multiplier, you need to enable EIST in the program (check the box, the changes take place instantly). Now changing the values in the VID and Set Multiplier windows actually lead to their change, see the result in Aida, CPU-Z or similar programs. Set the right one - EIST was disabled. Those. any manipulations with the multiplier and voltage must be carried out with the EIST turned on in the Throttlestop, and then turn it off.

The program is somewhat confused, or maybe Intel is confused with its mega-technologies. If something is not completely clear, swing the jackdaws for a few minutes and everything will clear up.

Power consumption and downvolting.

Since I purchased a wattmeter in Bangud, now everywhere I measure the electricity consumption of everything that consumes it. Consumption - no display, it is not physically. Together with the original 90W PSU.

The results on this Acer Extensa 5630g / HD3470 / 3Gb / SSD are as follows:

T5800 first

Simple, frequency 800 Mhz, voltage 0.9v - about 20W
Simple, frequency 2 Ghz, voltage 1.25v - about 22W
Load Aida, frequency 2 Ghz, voltage 1.25v - about 47W

Additional Information

T9550

Simple, frequency 800 Mhz, voltage 0.9v - about 23W
Simple, frequency 2.66 Ghz, voltage 1.25v - about 25W
Load Aida, frequency 2.66 Ghz, voltage 1.25v - about 52W
Load Aida, frequency 2.66 Ghz, voltage 1.1v - about 45W

Additional Information

Frequencies 800 and 1600 are achieved with nominal power saving, while the automatic setting is the minimum multiplier for these processors: 4 for the T5800 and 3 for the T9550.

If you delve into the numbers, you can see that lowering the frequency and voltage during idle almost does not give any savings, literally a couple of watts. At the same time, performance suffers a little, the "response" of the system drops: the energy saver needs time to track the increase in load and switch to productive mode.

Now about downvolting. The vast majority of mobile processors from Intel can successfully operate at full frequency with a noticeably reduced voltage. Which one exactly?

This is determined experimentally, using the knobs in RMClock or ThrottleStop, we reduce the voltage and carry out stability tests.

In my case, the T9550 was able to run stably at 2.8 Ghz at 1.05v instead of 1.25v. I set it higher though, 1.1v for guaranteed stability.

Consumption at full load dropped by almost 7W, while idle remained about the same. But this is the noise from the cooler, its wear, the heating of the video card - it sits on the same heat pipe with the processor. The temperature on the processor dropped by 5 degrees under load, about 65 degrees.

Thus, in the ideal case, the processor frequency is always maximum (we turn off all power saving nafik), and the voltage is minimum.

SSD testing.

Previously, I had WinXP x32, the hard drive was working in IDE mode. The results of its testing in CrystalDiskMark x32 are shown below. And this is a great fast 7200 rpm hard drive, usually much dull drives are installed in laptops.

After installing SSD and Win7, tested it in IDE modes and ACHI, 1GB mode. This SSD works without a drop in speed across the entire volume.

IDE on the left, ACHI on the right

We can see a slight acceleration of linear read and write in ACHI mode, a very slight deceleration in random read and write of small blocks. And there is a huge difference in reading and writing queued blocks in favor of ACHI.

Conclusion - enabling the ACHI mode has an extremely beneficial effect on the speed of work with small portions of data.

Comparison of processor performance.

I ran the benchmark on CPU-Z x32 and all Aida 1.60.1300 benchmarks on Windows7 x64 Ultimate

There were two gigahertz, two megabytes of cache, the 800 bus became 2.66 gigahertz, 6 meters of the cache and the 1066 bus.

The growth in processor frequency is 33%, everything that is higher is the architecture, cache and bus. The memory also began to move faster, especially the recording.

Considering that I also overclocked the processor by 5% (not taken into account in the table), we get almost a 1.5-fold increase in performance. Very noticeable, by the way.

Interesting - in WinXP x32 CPU-Z showed 281 points, and in Win7 x64 - 306. So much for you
xp.

For comparison - a desktop quad at the same frequency Q9400 2.66 / 6mb / 1333 p35 / ddr3 produces 411 in a single thread and 1350 in a multi thread. Those. one T9550 core is at least as fast as the Q9400.

Screenshots

T5800 WinXP x32

T5800 Win7 x64

T9550 Win7 X64

DDR2 Memory Comparison: 667 vs 800

The memory controller takes its parameters from the SPD, a special module on the memory strip. In my case, the speed was 800 MHz, the delays were 5-6-6-6-18.

If one of the two modules is marked as 667, then the controller switches to 667 mode, in my case - 5-5-5-15.

How much faster is 800MHz 5-6-6-18 mode 667 5-5-5-15?

As for the speed of working with memory, everything is clear, DDR2-800 is faster by 10 percent.

But the fact that there is a real difference in the speed of the cache memory of the first and second levels, about 5% - is news to me. Although the memory controller sits in the north bridge of the chipset, and the L1 and L2 caches are in the processor and, in theory, should be independent of the memory frequency.

However, all tests - except for working with memory from Aida Benchmark showed the same results.

Conclusion - DDR2-800 is better, of course, but there is practically no difference with DDR2-667.

Miscellaneous notes

Installing more than 4GB of memory requires support from the BIOS, without it even in x64 systems (!) Less will be seen, on Extense - 3.25GB. Moreover, in the first versions of the BIOS for my 5630, this support was, then it was removed (!) And users in the forums were indignant - why in the x64 system they do not see their legitimate 4GB. In the latest bios, support has been added again. Such are the squiggles.

WinXP x32 uses about 300MB of memory for its needs, win7 x64 - about 500-600MB. That is, it makes sense to put seven x64 with 4GB of memory, the memory gain in comparison with winXP x32 is about 500MB. Win7 x32 does not make sense to put.

You can replace that sellers sometimes have strange prices - the T9500 is much more expensive than the T9600, and so on. The answer is simple: powerful processors that can be installed on older motherboards (with a 667, 800 bus, etc.) are much more expensive than for more modern motherboards (with a 1066 bus). Users of, say, the PM965 chipset have to pay extra for a powerful processor on a slow 800 bus.

The 1066 bus seems to be abnormally supported by the PM965 chipset, with the help of special technicians it is suggested to him that the system operates at 800 frequency. Strictly unofficial mode of operation.

TDP of a processor is not its heat dissipation, at least in Intel's understanding. These are the requirements for the heatsink of the cooling system !!! Therefore, to say that the T8700 emits 25W, and the T9600 35W is wrong.

In principle, one could try to make a TME-unlock, somehow solder an smd resistor there. Then order a processor on the 800 bus, something like a 2.5 GHz T9300 for $ 19. Overclock on bus with setFSB 2, [email protected], 3 plus add 200MHz using Throttlestop. Most likely you would have worked at standard voltage. Or better yet, the $ 6 T8300, it would be 2, [email protected], 2 + 200MHz. If it worked out.

What happened in the end

The transition from quantity to quality. The SSD especially helped with this. Ordinary tasks suddenly began to fly, quite comparable to a multi-core desktop, also on an SSD and 8 g of memory.

I was already thinking of buying myself a more modern laptop, something on i5 or i7 + 8GB of memory. Now I see that the Extensa will still serve.

At least until a normal Internet appears in the country. You can visit the neighbors :)

The final

There are cats, there are a lot of them, they are hairy, energetic, they demand to eat, in the house, to iron and do some kind of skoda. There will be no photo. Instead of cats ...

There will be a little tin, no need to repeat

... although this can be called a natural, normal course of events.

In fact, I installed the processor several times: first I installed the T9550 and the SSD, installed Win7 instead of WinXP, and tested it. I installed the T5800, tested it. I returned the T9550 back.

And then the devil pulled me to pull up the thermal pad on the north bridge. Thermal grease is not suitable there, the radiator is much higher than the level of the northern one and, in addition, goes at a slight angle. The gasket turned out to be practically disposable and immediately fell apart. Okay, there was another one. Then I decided ... to bend the radiator a little so that it touches the northern one better.

I had a difficult day, but I did not pay attention to it.

The radiator in that place is a thick aluminum plate and a thick copper thermotube. As soon as I pressed well on it, the thermal tube bent over the processor area. Nyvopros, let's align. Rrraz! The thermal tube bent in a different place. Align it like this, pliers went into action: D The next crash site is near the graphics chip, I flattened it near the processor. H'm. I checked it with a lighter, everything seemed to be heating up. But the geometry was all violated, both 9 attachment points and three planes should coincide in the right places.

One kind of leveled, the second was broken. I checked the alignment by eye, illuminating with a BLF-348 flashlight, on the Nichiya diode, which was "smashed" in its time in the style of the above (Dedom means removing the plastic cap from the LED, the protection and the optical system at the same time. Changes the nature of the glow and is associated with the risk of ditching the LED. It is somewhat reminiscent of processor scalping).

Okay, sort of aligned by eye. Tightening the screws. Crack - the embedded part on the motherboard near the socket fell off. Another crack - the second mortgage has fallen off: D they are made of some kind of light fragile alloy, sealed on one side into the board. Then I leveled the radiator by eye and the screws were tightened with an off-design interference.

In addition, in the process, I cleared the corner well both on the processor and on graphics chip... Famously.

Then I tried to pull myself together and somehow start to solve the issue on the merits :) instead of the embedded part, you can take an elongated screw with a similar thread, insert it on the other side of the board into the hole, and use the torn-off key as a nut. The board had to be pulled out of the case, in order to tighten this ersatz mount, access is needed from both sides of the board at the same time.

And that was done.

The system booted, but during the stability test, the temperature jumped almost immediately to 60-70 degrees and quickly reached 100. At 105, the laptop said "that's enough!" and passed out. Throttling by only a few% could not bring down the temperature and started at 100 degrees ... The heat pipe was warm.

In general, the radiator is ready for replacement, it can no longer be aligned. It is possible that I also severely squeezed it and the transfer of heat by the liquid was disturbed. I didn’t feel 100 degrees on the processor with my finger, so there were also specific problems with adhesion, although everything was even by eye.

A spare radiator ... yes, I found it. Why not find a spare radiator for the Extensa 5630g in the bins of the Motherland. True, not where I thought, but in a completely different box, I was already starting to get nervous.

This time, the dashing mood passed me, the working radiator was correctly installed calmly, everything coincided, with the KPT-8 the temperature did not exceed 65 degrees, the heat pipe heats up like a frying pan in hell.

Fortunately, the chips did not lead to any problems. Even if they did, in fact, the losses would be moderate.

I plan to buy +58 Add to favorites I liked the review +129 +240

Introduction Sandy Bridge processors have earned the right to be called the revolutionary development of the Core microarchitecture not only with their increased speed - at the same time they also offered users a higher specific performance in terms of each watt of energy expended. This immediately has a beneficial effect on the battery life of modern mobile computers, bringing the dream of notebooks that do not require recharging all day long to come true. Moreover, it is the Sandy Bridge microarchitecture that should give life to a new class of portable devices - ultrabooks, which will combine the main advantages of tablets and classic laptops: compactness, lightness, versatility and low cost. In other words, the influence of the modern processor microarchitecture on the development of the mobile market was more than noticeable.

But the energy efficiency of Sandy Bridge is reflected not only in the properties of today's laptops. It also played a role in the desktop segment. So, it was thanks to her that Intel introduced a whole large family of desktop processors with lower power consumption. These processors were able to obtain registration in a separate class home computers called "Lifestyle PC" and combining HTPCs, compact and quiet home systems, all-in-ones and so on. Of course, we can't say that until Sandy's release Bridge Intel could not offer users anything like that, but before, desktop processors with low heat dissipation were represented only by exceptional and rare models. Now the situation has changed dramatically: in parallel with conventional desktop 95- and 65-watt CPUs, Intel's product range has expanded due to two full-fledged processor lines with lower thermal packs of 65 and 45/35 W. Moreover, these processors, like their "normal" counterparts, have a quite acceptable integrated graphics core of the Intel HD Graphics series, which in many economical systems allows you to do without a discrete graphics card.

Of course, economical models are somewhat inferior in performance to conventional processors, which do not prioritize low power consumption and heat dissipation. But, nevertheless, it is impossible to characterize their performance with any offensive epithet, since by modern standards they are quite productive. The following table shows how the nominal clock frequencies of current processors are distributed in conventional and economy lines.

Frequencies of conventional processors are shown against a pink background. Against a light green background, the frequencies of the economical S-series processors are indicated, which have a typical heat dissipation reduced to 65 W. The blue background highlights the frequencies of the T-series, which are among the most energy efficient models - with a thermal package of 35 or 45 W.

Roughly speaking, the S-Series offers economical versions of the most powerful Sandy Bridge processors and delivers a 30% reduction in heat dissipation at the expense of a 20% reduction in clock speed. The T-series provides more radical savings, but at the same time, the clock frequency reduction relative to conventional models can be up to 25-30%.

In this article, we decided to pay attention to the most interesting line of economical processors - the T series. Their calculated heat dissipation is so low that it allows using such CPUs without any tricks in the smallest Mini-ITX cases and build on their basis quiet fanless systems. Since the integrated video core of the Intel HD Graphics series in these processors in many cases allows you to do without an external graphics card, and the power consumption of the chipset required for Sandy Bridge is only 6.1 W, a complete system with a T-series processor can easily get along with a 60-watt one. power supply, in terms of energy parameters closely approaching mobile platforms... However, the question arises - is it not too significant sacrifices in terms of performance will have to be made by the users focused on economy? It is these doubts that we will try to dispel with this study.

Sandy Bridge processors, T-series

Any energy efficient processors Intel gets in a very simple way. The manufacturing process makes no distinction between standard Sandy Bridge semiconductor chips and chips for lower power and lower heat dissipation processors. Only at the final stage of production, processors, which subsequently should be distinguished by lower heat dissipation and power consumption, are assigned lower clock frequencies and, in addition, a reduced supply voltage is set for them. These two elementary actions are quite enough to separate the second generation Core processors into different groups, in which the typical heat dissipation differs significantly.

Although the described approach to creating energy-efficient CPUs seems primitive, it not only works great with Sandy Bridge semiconductor crystals, but also allows them to keep their costs low. It is because of this that the prices of T-series processors are only slightly higher than those of conventional models. So Intel does not put any economic obstacles in the way of their distribution, which further encourages their widespread distribution.

Intel currently has four T-series processors with typical heat dissipation reduced to 45W or 35W. All these processors belong to different lines and differ not only in clock frequency. They offer a different number of cores and a different set of supported technologies. In other words, their variety is quite enough to make it possible to choose the most suitable option based on the required level of performance and functionality.

Let's take a closer look at the representatives of the T series.

Core i5-2500T

The Core i5-2500T is the only quad-core in the T-series. Obviously, squeezing this processor into such a narrow power consumption frame was not easy, the typical heat dissipation for it was set at 45 W, while all other representatives of the T-series have a TDP of 35 W. Therefore, the fact that the nominal frequency of this model is only 2.3 GHz, that is, below the frequency of a full-fledged Core i5-2500 by a whole gigahertz, does not cause any surprise.

However, the concept of the nominal clock speed for the Core i5-2500T is very relative. This processor supports the technology Turbo Boost which in in this case works very aggressively. The maximum frequencies to which the Core i5-2500T can auto-overclock when loaded on a different number of cores are given in the table. For comparison, in the same table, we have placed data on the operation of Turbo Boost technology in conventional Core i5 processors.

As you can see, comparatively low frequency inherent in the Core i5-2500T only when loaded with three or four cores. In a state of less intensive load, this processor is capable of significant auto-overclocking, reaching one gigahertz. As a result, an economical processor catches up with its full-fledged counterparts and, with two or three cores passive, is capable of delivering even higher performance than a Core i5-2300 or Core i5-2310.

The voltage value shown in the above screenshot of CPU-Z cannot be trusted. In fact, our Core i5-2500T was powered by 1.080 volts, which is about 0.1 volts less than typical quad-core Core i5 processors. So the efficiency of the Core i5-2500T stems not only from its trimmed clock speeds, but also from its operation at reduced voltage.

Particularly curious in this regard are the characteristics of the graphics core built into the Core i5-2500T. In this case, it is Intel HD Graphics 2000 with six execution units, which is present in most second generation Core processors for desktops. However, the frequency of this core in the Core i5-2500T can vary much wider than that of other processors. The nominal value is 650 MHz (versus the usual 850 MHz), but the "graphic" Turbo Boost can increase this frequency to 1.25 GHz (versus 1.1 GHz in standard version). In other words, if the load on the graphics is not accompanied by the full employment of processor cores, then the Core i5-2500T will surpass even the 95-watt Core i5-2500 in 3D performance.

Core i5-2390T

Although Intel has classified the Core i5-2390T as part of the Core i5 family, this processor is fundamentally different from the rest of this family. While all other Core i5s are quad-core CPUs, the Core i5-2390T is a dual-core processor. However, it also would not fully fit into the Core i3 family, since it has support for Turbo Boost technology, which is inherent only in the Core i5 and Core i7 lines. In other words, it would be most correct to single out the CPU in question into a non-existent "intermediate" Core i4 group, but, obviously, Intel did not want to confuse the already complicated nomenclature for the sake of one product.

Similar to the Core i3 processors, the Core i5-2390T supports Hyper-Threading technology, that is, in the operating system it looks like a quad-core, just like the "real" Core i5. However, Hyper-Threading cannot be an alternative to physical processor cores, so that their absence has to be compensated for by clock frequencies. For example, full-time Core frequency The i5-2390T is 2.7 GHz, while the Core i5-2500T is 400 MHz lower.

The Core i5-2390T and Turbo Boost technology are quite energetic. Let's compare the clocks of the 35W Core i5-2390T with those of the other non-economy series Sandy Bridge dual cores with Hyper-Threading Technology, which have a TDP of 65W.

Despite the fact that the nominal clock speed of the Core i5-2390T is significantly lower than the frequency of the Core i3 series processors, in real work it can overclock and surpass them in speed, because the other dual-core Sandy Bridge technology does not support Turbo Boost at all. However, being implemented in an economical dual-core processor, this technology is clearly not formal, it can significantly increase the frequency of this CPU.

The operating voltage of the Core i5-2390T turned out to be 1.092 V, and this is higher Core voltage i5-2500T. But, nevertheless, due to the reduced number of processing cores, the dual-core processor has a lower calculated typical heat dissipation of 35 watts.

As for the graphics core, in this case Intel HD Graphics 2000 is built into the processor with a nominal frequency of 650 MHz lower than that of conventional CPUs. However, the Turbo Boost technology for the GPU compensates for this disadvantage - graphics auto-overclocking is provided up to 1.1 GHz, that is, to the same level to which the graphics core in a 95-watt Core i5 can autonomously overclock. At the same time, this means that in terms of graphics speed, the Core i5-2390T is inferior to its quad-core economical counterpart, the Core i5-2500T.

Core i3-2100T

The Core i3-2100T has no special secrets. This is a regular dual-core Core i3 with support for Hyper-Threading technology, but without Turbo Boost technology, which has a reduced clock frequency for the sake of lower heat generation and power consumption. However, the magnitude of this decrease is not so significant. Even regular Core i3 processors are quite economical, so in order for the Core i3-2100T to fit into the 35-watt thermal package, Intel had to slow down its frequency relative to the 65-watt Core i3-2100 by only 600 MHz.

It should be noted that in this case it was possible to get by with less deceleration. For example, the Core i5-2390T, similar in number of cores and threads, successfully operates at a higher clock frequency without going beyond the 35-watt frame. So just 2.5 GHz Core i3-2100T frequency is partly a marketing move aimed at ensuring that there is a noticeable performance gap between the Core i5-2390T and the Core i3-2100T. Moreover, the operating voltage of the Core i3-2100T and Core i5-2390T is the same and is 1.092 V.

The graphics core in the Core i3-2100T does not differ from the graphics in the Core i5-2390T. Used Intel HD Graphics 2000 with six executive units and frequency from 650 MHz to 1.1 GHz in auto overclocking mode.

Delving deeper into the study of the differences between the Core i3-2100T and the Core i5-2390T, it should be noted that the processor of the younger series is deprived of support for the set AES instructions... But this is a feature of all Core i3s that has nothing to do with energy conservation.

Pentium G620T

Closing the four processors with reduced heat dissipation is the Pentium series processor. This is Pentium G620T - a budget dual-core CPU without Hyper-Threading and without Turbo Boost support. It differs from the usual Pentium G620 with a 400 MHz clock frequency and a calculated heat dissipation reduced from 65 to 35 W.

As we saw in previous tests , even standard Pentium processors do not differ too much from the Core i3-2100T in terms of their consumption. So the creation of a 35-watt model in the ranks of this family is not so much an improvement. However, the power supply voltage of an economical budget CPU turned out to be reduced to 1.056 V, which is about 0.05 V lower than the voltage used by conventional Pentiums.

The graphics core of the Pentium G620T is Intel HD Graphics. Compared to the graphics built into the second generation Core processors, it lacks support for Quick Sync technology, and this is a feature of any Pentium. What distinguishes the energy-efficient model is the operating frequencies of this GPU. The nominal value is 650 MHz, not 850 MHz. However, the maximum frequency during auto-overclocking reaches 1.1 GHz, that is, there are no differences from 65-watt Pentiums in this parameter.

Like everyone else Pentium line, the energy efficient model does not have support for AES and AVX instructions. Also, the specifications of this processor do not include support for DDR3-1333 SDRAM, so in practice this CPU has to be used with slower memory.

How we tested

When testing the low-power T-series processors, we decided to compare its performance with the speed of conventional LGA1155 processors. This will answer the question posed at the very beginning - do 45- and 35-watt CPUs lose much in speed compared to typical Sandy Bridge. Therefore, along with the four processors with the T suffix in the name, Core i5-2310, Core i3-2120, Core i3-2100, Pentium G850 and Pentium G620 took part in the tests.

When testing energy-efficient Sandy Bridges, we tried to recreate their typical "habitat", and therefore refused to use an external high-performance video card, preferring it to the graphics core built into the processor. As a basis for the test platform, we chose the popular Mini-ITX motherboard on the Intel H61 chipset, ASUS P8H61-I.

As a result, the composition of the test systems included the following hardware and software components:

Processors:

Inlel Core i5-2500T (Sandy Bridge, 4 cores, 2.3 GHz, 6 MB L3, 45 W);
Inlel Core i5-2390T (Sandy Bridge, 2 cores, 2.7 GHz, 3 MB L3, 35 W);
Inlel Core i5-2310 (Sandy Bridge, 4 cores, 2.9 GHz, 6 MB L3, 95 W);
Intel Core i3-2120 (Sandy Bridge, 2 cores, 3.3 GHz, 3 MB L3, 65 W);
Intel Core i3-2100 (Sandy Bridge, 2 cores, 3.1 GHz, 3 MB L3, 65 W);
Intel Core i3-2100T (Sandy Bridge, 2 cores, 2.5 GHz, 3 MB L3, 35 W);
Intel Pentium G850 (Sandy Bridge, 2 cores, 2.9 GHz, 3 MB L3, 65 W);
Intel Pentium G620 (Sandy Bridge, 2 cores, 2.6 GHz, 3 MB L3, 65 W);
Intel Pentium G620T (Sandy Bridge, 2 cores, 2.2 GHz, 3 MB L3, 35 W).

CPU cooler: stock Intel cooler.
Motherboard: ASUS P8H61-I (LGA1155, Intel H61, Mini-ITX).
Memory - 2 x 2 GB DDR3 SDRAM (Kingston KHX1600C8D3K2 / 4GX):

DDR3-1067 7-7-7-21 with Pentium G620 and Pentium G620T processor;
DDR3-1333 9-9-9-27 when using other processors.

Hard disk: Kingston SNVP325-S2 / 128GB.
Power supply: Tagan TG880-U33II (880 W).
Operating system: Microsoft Windows 7 SP1 Ultimate x64.
Drivers:

Intel Chipset Driver 9.2.0.1030;
Intel HD Graphics Driver 15.22.1.2361;
Intel Management Engine Driver 7.1.10.1065;
Intel Rapid Storage Technology 10.5.0.1027.

Performance

Overall performance

To assess the performance of processors in common tasks, we traditionally use the Bapco SYSmark 2012 test, which simulates the user's work in common modern office programs and applications for creating and processing digital content. The idea of the test is very simple: it produces a single metric that characterizes the weighted average speed of a computer.

The results in SYSmark 2012 are quite expected. The Core i5-2500T loses to the Core i5-2310 by about 9%, the Core i3-2100T is 17% behind its 65-watt counterpart, and the Pentium G620T is 13% behind the regular Pentium G620. At the same time, the 45-watt quad-core Core i5-2500T outperforms all 65-watt dual-core processors, and the same can be said about the dual-core Core i5-2390T, which is well supported by its Turbo Boost technology. The slower dual-core economical processor Core i3-2100T in terms of performance is on a par with the Pentium G850, but the Pentium 620T turns out to be a completely unhurried product, which, obviously, can only compete with the representatives of the Celeron series in LGA1155 that have not yet been released.

A deeper understanding of the SYSmark 2012 results can provide insight into the performance scores obtained in various system use cases. The Office Productivity script simulates typical office work: preparing word, processing spreadsheets, working with e-mail, and surfing the Internet. The script uses the following set of applications: ABBYY FineReader Pro 10.0, Adobe Acrobat Pro 9, Adobe Flash Player 10.1, Microsoft Excel 2010, Microsoft Internet Explorer 9, Microsoft Outlook 2010, Microsoft PowerPoint 2010, Microsoft Word 2010, and WinZip Pro 14.5.

In the script Media creation the creation of advertising video using pre-shot digital images and videos. For this purpose, popular packages from Adobe are used: Photoshop CS5 Extended, Premiere Pro CS5 and After Effects CS5.

Web Development is a scenario within which the creation of a website is modeled. Applications used: Adobe photoshop CS5 Extended, Adobe premiere Pro CS5, Adobe Dreamweaver CS5, Mozilla Firefox 3.6.8, and Microsoft Internet Explorer 9.

Data / Financial Analysis Scenario is dedicated to statistical analysis and forecasting of market trends that are performed in Microsoft Excel 2010.

The 3D Modeling script is all about creating 3D objects and rendering static and dynamic scenes With using Adobe Photoshop CS5 Extended, Autodesk 3ds Max 2011, Autodesk AutoCAD 2011 and Google SketchUp Pro 8.

The last script, System Management, performs backups and installs software and updates. Several different versions of Mozilla Firefox Installer and WinZip Pro 14.5 are involved here.

Application performance

To measure the speed of processors when compressing information, we use the WinRAR archiver, with which we archive the folder with the maximum compression ratio. various files with a total volume of 1.4 GB.

We measure performance in Adobe Photoshop using our own benchmark, which is a creatively reworked Retouch Artists Photoshop Speed Test including typical processing of four 10-megapixel images taken with a digital camera.

When testing the audio transcoding speed, the Apple iTunes utility is used, with the help of which the contents of a CD-disc are converted to AAC format. Note that a characteristic feature of this program is the ability to use only a couple of processor cores.

To measure the speed of video transcoding into H.264 format, the x264 HD test is used, which is based on measuring the processing time of the original MPEG-2 video recorded in 720p resolution with a 4 Mbps stream. It should be noted that the results of this test are of great practical importance, since the x264 codec used in it underlies numerous popular transcoding utilities, for example, HandBrake, MeGUI, VirtualDub, etc.

Testing the final rendering speed in Maxon Cinema 4D is performed using the specialized Cinebench benchmark.

Looking at the given diagrams, you can once again repeat everything that has already been said in relation to the SYSmark 2011 results. In general, the Core i5-2500T and Core i5-2390T seem to be very productive, but at the same time economical processors. In most cases, their speed is in the range between the speed of the quad-core 95-watt Core i5 and dual-core 65-watt Core i3. Therefore, it is these processors that are of primary interest if you want to build a powerful and economical system.

As for the performance of the Core i3-2100T and Pentium G620T, they should first of all be considered from the position of a favorable price. Frankly speaking, they show low results, but no one promised that inexpensive solutions would shine with dizzying performance.

In addition to our tests, to test Intel Quick Sync technology, we measured the transcoding speed of a 3GB 1080p H.264 movie (which was a 40-minute episode of the hit TV series) at downsampled resolution for viewing on an iPhone 4. For The popular commercial Cyberlink MediaEspresso 6.5 utility, which supports Quick Sync technology, was used for transcoding.

Here the results are divided into two large groups... The first includes the Core i5 and Core i3 processors, in which there is support for the Quick Sync technology, the second is the Pentium, which lacks this technology. The difference in transcoding time for these groups is approximately fourfold. The second factor that can affect the speed of MediaEspresso is the frequency of the graphics core. That is why the economical Core i5-2500T unexpectedly takes the lead in this test. Its graphics core is capable of dynamically overclocking up to 1.25 GHz, while in all other processors the maximum graphics frequency is limited to 1.1 GHz.

Gaming performance

The 3D gaming benchmark group opens with 3DMark Vantage, which was used with the Entry profile.

The score in the popular 3DMark Vantage benchmark is primarily affected by graphics performance. Therefore, the first place here is occupied by the Core i5-2500T, in which the Intel HD Graphics 2000 core operates at a higher frequency than the rest of the test participants. The rest of the processors are located in a relatively dense group, in which the differences in readings are determined primarily by their computing capabilities. At the same time, we note that, in contrast to the results in SYSMark 2012 and in applications, the Core i5-2390T looks somewhat disappointing in 3DMark Vantage. Here its speed drops to the level of the Core i3-2100 due to the fact that it is a dual-core processor, although the manufacturer attributes it to the Core i5 series.

To study the performance in real games, we selected Far Cry 2, Dirt 3 and Starcraft 2. These games are characterized by the fact that they show an acceptable level of performance on the integrated graphics core Intel HD Graphics 2000. However, to achieve it, we ran the tests in 1280x800 mode, and the quality settings were set to Low.

And again, for obvious reasons, the Core i5-2500T is in the lead. It would seem that the HD Graphics 2000 graphics core of this processor can only be overclocked a little more than in other CPUs, but even this is enough for quite tangible gaming superiority. The rest of the Core i5 and Core i3 processors are in a tight group on the diagrams. Only the Core i3-2100T lagged somewhat behind in Starcraft 2. Like the younger Pentium CPUs, it clearly lacks the computational performance to fully support the graphics in this highly processor-dependent game.

Energy consumption

Testing has shown that T-series processors are significantly inferior to "conventional" modifications in terms of computing performance. This is a side effect of reducing power consumption, which is achieved, among other things, by reducing those frequencies. However, so far about low level consumption, we spoke only in a theoretical way, based on the official specifications. Now is the time to evaluate practical energy efficiency.

The graphs below show two energy consumption values. The first is the total system consumption (without a monitor), which is the sum of the energy consumption of all components involved in the system. The second is the consumption of only one processor through a dedicated 12-volt power line. In both cases, the efficiency of the power supply is not taken into account, since our measuring equipment is installed after the power supply and records the voltages and currents entering the system via 12-, 5- and 3.3-volt lines. During the measurements, the load on the processors was created by the 64-bit version of the LinX 0.6.4 utility. The FurMark 1.9.1 utility was used to load the graphics cores. In addition, to correctly estimate idle power consumption, we have activated all available energy-saving technologies, as well as Turbo Boost technology.

The differences in consumption between conventional and low-power processors are already noticeable when they are idle. 45- and 35-watt processor modifications, even when idle, can provide a total savings of 1-2 watts, which is achieved through lower idle processor voltage when Enhanced Intel Speedstep technology is activated.

A very interesting picture is observed when the load is only on one computational core of the processor. Here, the 45- and 35-watt Core i5-2500T and Core i5-2390T processors do not show their economy in any way. The reason for this lies in their very aggressive implementation of the Turbo Boost technology. With part-time employment, they briskly lift up the clock frequency, choosing the entire resource of the heat pack and closely approaching in speed to their 95- and 65-watt counterparts, which do not dare to overclock equally decisively. As for the Core i3-2100T and Pentium G620T, they do not have Turbo Boost technology, and therefore their consumption turns out to be several watts lower than that of the 65-watt Core i3-2100 and Pentium G620.

Interesting results are obtained even with the maximum load on the processing power of the processors. In general, systems built using representatives of the T-series demonstrate significantly lower consumption than platforms using standard processors of the same class. But, nevertheless, in the practical indicators of consumption, you can notice some funny inconsistencies. For example, the Core i5-2500T processor, which has a maximum calculated heat dissipation of 45 W, turns out to be more power hungry than the Core i3-2120, whose TDP is 20 W higher. It is clear that this is due to the different number of cores, but the fact remains. The same way The Core i5-2390T has higher power consumption than the Pentium G850.

All this suggests that T-series processors in real life are not always more economical than their "normal" counterparts. They are better in terms of specific performance per watt of energy consumed, but when comparing absolute values of power consumption, they can lose to significantly slower CPUs with a higher declared TDP level. And this must be borne in mind.

When testing consumption with graphics load, the results do not give any particular surprises. The Intel HD Graphics 2000 graphics core is much less power hungry than the computing cores, so the manufacturer did not particularly optimize this part of the CPU. The result of this approach is a slight discrepancy in the actually measured consumption in this case. Only the Core i5-2500T stands out, in which the graphics core is overclocked to more high frequency than in all other cases.

A similar picture is observed when using processors as the basis of a media center. Load in the form of decoding video content high resolution results in slightly different power consumption for systems with 95-, 65-, 45-, and 35-watt TPUs.

conclusions

The microarchitecture of Sandy Bridge is striking in its versatility. We have admired more than once how powerful processors built on it can be, and today we are convinced that it is equally suitable for creating attractive proposals for quiet, small-sized and economical systems. However, the economical T-series processors based on Sandy Bridge have revealed a number of peculiar features that are not reflected in the specifications in any way and can somewhat change the general impression of these products.

Let's take a look at the power consumption. Despite the fact that T-series processors have half the TDP compared to conventional processors, this does not mean that in reality they consume half as much. Firstly, economical processors get close enough to the border of their thermal package, while CPUs that are not burdened by any obligations in terms of maximum heat release often demonstrate much less power consumption and heat release than indicated in the specification. Therefore, in reality, the difference in practical consumption between T and non-T processors there is no one double class. Secondly, a serious difference in consumption between economical and conventional processors manifests itself in only a small number of scenarios, while in most situations they generally demonstrate very similar energy appetites. In fact, the energy efficiency of the T-series models is only found out under heavy multi-threaded computational workload. In idle states, with single-threaded operation, or when using the graphics core, the T-series processors do not give any serious advantages in terms of consumption.

All this means that there is practically no point in using special energy-efficient Sandy Bridge options for the sake of energy savings alone. Considering that in real life the maximum processor load is sporadic, T-series processors will not give a significant benefit when paying for electricity bills.

The real and indisputable advantages of these processors are manifested in something else - when, for some reason, it is necessary to limit the maximum consumption or heat dissipation from above. For example, if the system is assembled in a case that makes it possible to accommodate only a low-efficiency cooling system or in the case when a low-power power supply is forcedly used, Intel T-series processors can be really indispensable.

However, the limitations in power consumption and heat dissipation significantly affect the speed. From the point of view of peak computing performance, processors with a thermal package reduced to 45 and 35 W work on average 15-20% slower than conventional CPUs of the same class and similar cost. However, in the case of the Core i5-2500T and Core i5-2390T, such a significant lag is manifested only under heavy multi-threaded load, in other situations, these economical processors are seriously helped by the aggressively tuned Turbo Boost technology. Another pair of T-series processors, Core i3-2100T and Pentium G620T, does not have Turbo Boost support and lags far behind full-fledged counterparts in any scenario.

But it is not all that bad. The Core i5-2500T and Core i5-2390T are unique in terms of performance and are able to outperform conventional 95W and 65W processors in a number of ways. In particular, the Core i5-2500T has the fastest modification of the Intel HD Graphics 2000 graphics core, which provides a higher performance of this processor in 3D and using Quick Sync technology compared to most LGA1155 brothers. The Core i5-2390T can be called the fastest dual-core desktop processor based on the Sandy Bridge microarchitecture.

As a result, we come to the conclusion that the T series processors, and especially those that belong to the Core i5 family, are very curious products that sometimes have even completely unexpected advantages. However, in general, we can talk about Core i5-2500T, Core i5-2390T, Core i3-2100T and Pentium G620T only as niche products that are really interesting only in limited number situations. At the same time, we should not forget that in many cases, instead of T-series processors, you can generally do with 65-watt Pentiums, which in reality often demonstrate comparable or even lower power consumption than 45- and 35-watt processors of the Core i5 and Core i3.

In other words, choosing the right CPU for an energy efficient system is a very difficult question, and there is no single recipe for the answer. The variant proposed by Intel with special modifications with low thermal packets should not be dismissed, of course, but we cannot say that it will be the only correct one in any case.