As you know, Intel microprocessor architectures change every two years. Computing power is constantly growing, the flagships of the recent past are turning into outsiders, giving way to the strongest representatives of the new architecture. With the launch of processors based on the Nehalem architecture in November 2008, Intel significantly strengthened its position in the Hi-End desktop PC sector. And the recent top models in the Core 2 Quad and Core 2 Duo lines could no longer compete with Core i7 processors, so they had to shift to the middle price niche, giving way to high-performance newcomers in the Hi-End segment. Intel's future plans include expanding the presence of representatives of the new architecture in all market segments. However, the Core i7 line in its original form is in no way able to fit into the budget of mid-range and budget desktop PCs. That is why, for the general public, the company’s engineers have developed a “lightweight” series of CPUs based on the Nehalem architecture. Today, Intel officially introduced three new microprocessors - Core i7 870, Core i7 860 and Core i5 750, designed to work in the Socket LGA 1156 processor socket. The first representatives of the Core i7 family were designed for installation in the Socket LGA 1366 processor socket, and motherboards for these processors were built on the basis of the only available system logic set - Intel X58. The entry of new members of the Core family into the market required the development of a new chipset and motherboards based on it. The new chipset is the Intel P55 chipset. Before we look in detail at the differences between the new solutions for Socket LGA 1156 and the old LGA 1366, let's take a look at the summary table of characteristics of Core i5/i7 central processors and Intel P55 and X58 system logic sets.

Main characteristics
Intel Core Processor	i5-750	i7-860	i7-870	i7-920	i7-940	i7-950	i7-965 Extreme	i7-975 Extreme
Core	Lynnfield			Bloomfield
Technical process	45 nm
Connector	Socket LGA 1156			Socket LGA 1366
Chipset	Intel P55			Intel X58
Kernel stepping	B1			C0/D0	C0/D0	D0	C0	D0
Core frequency, GHz	2.66	2.8	2.93	2.66	2.93	3.06	3.2	3.33
Factor	20	21	22	20	22	23	24	25
Multiplier step with Turbo Boost*	1 - 4	1 - 5	1 - 5	1 - 2	1 - 2	1 - 2	1 - 2	1 - 2
L1 cache, KB	32/32
L2 cache, KB per core	256
L3 cache, MB	8
Bus type "Processor-chipset"	DMI			QPI
Integrated PCI-Express controller	Yes			No
TDP, W	95			130
Maximum memory bandwidth of the processor-chipset bus, GB/s	2			25
RAM channels	2			3
Physical cores	4
Supported technologies
Hyper-Threading	No	Yes
VT-x	Yes
VT-d	No	Yes
TXT	Yes
EIST	Yes
Intel 64	Yes

*The frequency step is determined by the step of the processor multiplication factor from the original one, depending on the load on the cores. From the above table it follows that the differences in the internal structure of the LGA 1366 and LGA 1156 processors are not limited to the lack of support for a three-channel memory controller in Lynnfield. In fact, the difference is much more significant. Let's take a more detailed analysis of the differences between these CPUs.

Design

Intel Core i7 and Core i5 processors based on the Lynnfield core are designed to work with the Socket LGA 1156 processor socket, which, in fact, is not very different from the Socket LGA 775/LGA 1366 sockets. The only difference is that the CPU locking mechanism has changed slightly, as well as the location of the holes for cooling system mountings. Next we will take a closer look at the new connector.

Memory controller

All processors, designed to work in motherboards with Socket LGA 1366, have a three-channel integrated DDR-3 memory controller, providing extremely high memory bandwidth. Core i5 and Core i7 processors designed for Socket LGA 1156 have a dual-channel integrated memory controller, which may slightly reduce its throughput. However, testing the memory subsystem will show how big the difference in memory bandwidth is.

Hyper-Threading Technology

This technology first appeared back in the days of Pentium 4 processors with NetBurst architecture. All Intel Core i7 processors, regardless of design, support HT, which allows them to perform up to 8 computational threads simultaneously. Intel Core i5 series processors do not support Hyper-Threading.

Turbo Boost Mode

The essence of this mode is to increase the operating frequency of one or more processor cores, depending on the computing load, by increasing the processor multiplier. Intel Core i7 processors for Socket LGA 1366 are capable of increasing the operating frequency by 1 or 2 steps (by step we mean the CPU multiplier step). While processors designed to work in Socket LGA 1156, depending on the load, can be overclocked by 1-5 steps for the Core i7 series and 1-4 steps for the Core i5 series. It is obvious that Turbo Boost technology has reached a certain maturity, and new Intel processors are able to increase the frequency significantly more than before. In addition, it is worth noting an interesting trend. Modern Intel technologies allow processors to “intelligently” distribute their forces to achieve maximum results depending on the type of tasks being performed.

Bundle "Lynnfield - P55"

Core i7 processors for Socket LGA 1366 interact with the Intel X58 system logic set using the QuickPath Interconnect (QPI) bus, providing throughput up to 25 GB/s. In turn, Core i7 and Core i5 processors, developed for Socket LGA 1156, “communicate” with the Intel P55 chipset via the DMI (Direct Media Interface), first used by Intel back in 2004 in conjunction with the ICH6 southbridge. It's no secret that the DMI interface cannot provide the same high throughput as the QPI bus. Judge for yourself, the bandwidth of the DMI interface is ~2 GB/s versus ~25 GB/s for QPI. And how, in this case, to “pump” huge amounts of data between the processor and devices connected to the PCI-Express 2.0 bus, for example, video cards that require data transfer rates of up to 16 GB/s. But there are also less demanding devices, such as network controllers, hard drives, etc. Intel engineers solved the problem quite elegantly. The PCI-Express controller and DMI interface, along with the memory controller, are now integrated into the CPU, which largely solves the bottleneck. Why largely and not completely? The fact is that the integrated PCI-Express 2.0 controller supports up to 16 lanes, which will be entirely occupied by one or a pair of graphics accelerators. For a single video card, all 16 PCI-Express lanes are allocated; when installing two video cards, the lines are distributed as 2x8. It turns out that for other devices the capabilities of the integrated PCI-Express controller are no longer enough. However, this problem has been successfully solved! Thanks to the integration of part of the control units on the CPU substrate, the Intel P55 chipset is just one chip, which has received a new name. Now this is not just a south bridge, it is the so-called Platform Controller Hub (PCH), which, along with the standard set of south bridge functions, also received support for a PCI-Express 2.0 controller to meet the needs of peripheral devices.

VT-d

Virtualization technology for directed I/O is an input/output virtualization technology created by Intel as an addition to the existing Vanderpool computing virtualization technology. The essence of this technology is to allow a remote OS to work with I/O devices connected to PCI/PCI-Ex directly at the hardware level. All modern Intel Core i7 processors, regardless of the processor socket used, support this technology, but Core i5 series processors do not.

TDP

Thanks to optimization of production technology and a modified CPU core, Intel managed to reduce the TDP value for Core i7/i5 series processors for Socket LGA 1156 to 95 W, versus 130 W for Intel Core i7, designed for the Socket LGA 1366 platform.

From theory to practice. Test platform

Before moving on to testing, let's look at the components of the test platform based on Socket LGA 1156, and also consider the nuances in the operation of the Lynnfield + P55 combination. An engineering sample of the Intel Core i5 750 processor arrived in our laboratory. Unfortunately, modern engineering CPU samples are in no way different from production units, even the available multiplication factors are the same as those of ordinary representatives of this series. The sizes of processors with the Socket LGA 1156 design are significantly smaller than the sizes of the CPUs of their older brothers, designed to work in Socket LGA 1366, compare:

Core i5 750 on the left, Core i7 920 on the right

As the basis for our test bench, we used the MSI P55-GD65 motherboard, kindly provided by the Russian representative of MSI. We will definitely publish a detailed review of the MSI P55-GD65 a little later, but for now we will focus on describing the key features of the board:

• Processor support for Socket LGA1156
• 4 slots for DDR-3 memory
• Supports 7 SATA II connectors
• SLI and CrossFireX technology support
• Supports proprietary MSI OC Genie technology

RAM manufactured by Apacer. The kit consists of three modules with a capacity of 1 GB each and is designed to work in three-channel mode with Core i7 processors. Of course, to test the Core i5 750 processor we used only two modules from the kit.

Now is the time to look at the Core i5 in action and talk about the features of overclocking new Intel processors based on the Lynnfield core.

Features of Core i7 and Core i5 processors on the Lynnfield core

CPU Clock - CPU cores operate at this frequency. unCore Clock (UCLK)- operating frequency of the north bridge integrated into Core i7/i5 processors. The integrated third-level cache operates at this frequency, as well as the Core i7/i5 RAM controller. QPI bus frequency. The frequency at which the QPI interface operates, connecting the Core i7 9xx to the Intel X58 chipset. Overclocking of non-extreme Core i7 processors of the 9xx family very often rested on the frequencies of UCLK, QPI and DDR-3 memory (to a lesser extent). The fact is that the processor frequency multiplication factor for conventional Core i7s is strictly limited from above. Therefore, to increase the CPU frequency, it is necessary to increase the base frequency (BCLK), and an increase in BCLK entails an increase in the UnCore, UCLK and DDR-3 frequencies. It was possible to “cope” with the increase in RAM frequency using dividers, but there was no way to tame the increase in QPI and UCLK frequencies, because the requirement that the UCLK frequency must be at least twice the DDR-3 frequency contributed. It was precisely because of the instability of one of these CPU units at higher frequencies that CPU overclocking was limited to values ​​slightly exceeding 200 MHz BCLK. With the arrival of Lynnfield, some of the problems for overclockers have been solved. Now the UCLK frequency is locked, and the dividers for the QPI bus frequency are smaller, so in theory we can get a higher stable BCLK frequency.

A little more than a year has passed since the appearance of the Nehalem platform, but prices for new processors still cannot be called affordable. The expansion of the modern line of CPUs through models based on the Lynnfield core under LGA1156 did not in any way affect the pricing of their older brothers, and they themselves were not affordable. Until recently, the most economical processor based on the new architecture was the Core i5-750, which led to the fairly great popularity of this model. And even the recent appearance of Clarkdale processors from the same series is unlikely to shake the position of the “old man,” which has four real cores versus four “virtual” cores in the new products. But we will have a separate article dedicated to Clarkdale, and in this article, as you may have guessed, we will focus specifically on the Core i5 750.

For retail sale, the Intel Core i5 750 is supplied in a boxed version, but sometimes you can also find tray options that are provided with a 12-month warranty from the seller.

The standard cooler has fairly compact dimensions and a low radiator height; the core is made of copper. The design is no different from the cooling systems of processors with LGA775 design.

The architecture of Lynnfield processors was reviewed in detail by us in one of our previous materials. The Northbridge is completely integrated into the processor, which itself provides support for 16 PCI Express 2.0 lanes. This, by the way, leads to a small drawback of the platform associated with the limited bandwidth of the interfaces of two video cards operating in CrossFireX mode. Unlike their predecessors for Socket LGA1366, the new CPUs have only a dual-channel DDR3 memory controller. Thanks to the x6 multiplier (effective x12), new Core i7 processors in nominal modes can work with DDR3-1600 (not an officially supported standard), and the younger Lynnfield, Core i5 750 in particular, with a x5 multiplier (effective x10) with DDR3-1333. Higher memory frequencies can only be used by raising the base frequency (BCLK), and if you are using high-frequency memory, then for its X.M.P. profile. The board will automatically raise BCLK and lower the processor multiplier when the voltages are adjusted accordingly. For DDR3-2000, the reference frequency will be set to 200 MHz, and the multiplier on the Core i7 750 processor will be set to x14 instead of x20. If the memory does not have X.M.P. profiles. for LGA1156 processors, then the user will need to make all adjustments manually. The frequency of the Uncore block, which includes a memory controller and a shared third-level cache, is fixed relative to the base frequency due to a x16 multiplier at 2130 MHz. The QPI bus now connects the processor only with the PCI Express controller; its frequency is formed as the product of BCLK by x18 (x36), which gives 2400 MHz (4800 GT/s). You can manually set a lower multiplier x16 (x32).

The processor frequency in nominal mode is 2.66 GHz with a x20 multiplier. The quad-core Core i5 750 does not have Hyper-Threading support.

Thanks to Turbo Boost technology, when running applications that are poorly optimized for multi-threading, the frequency of individual cores can be increased. This overclock can be up to 4 points (133 MHz) for one of the cores. More precisely, in single-threaded applications, the loaded core will operate at 3.2 GHz. If the load falls on two cores, then their frequency rises to intermediate values, and even if all cores are loaded, the frequency of all of them will rise by one point. In the latter case, we actually get a quad-core CPU at 2.8 GHz (with a x21 multiplier) instead of 2.66 GHz. By the way, such a multiplier can initially be set manually for the Core i5 750 in the BIOS of almost all LGA1156 motherboards and without activating the Turbo Boost mode.

For tests in nominal mode, we used a 4 GB memory kit (Team TXD34096M2000HC9DC-L), which worked with timings of 7-7-7-20. All other delays and settings are shown below in the screenshot of the CPU-Tweaker utility.

Well, a few words about overclocking. It is carried out by increasing the base frequency. Since the frequencies of other blocks and DDR3 memory depend on it, the corresponding multipliers are reduced if necessary. So for DDR3 you can set the minimum multiplier x6, which will give a nominal frequency of 800 MHz, and when overclocking BCLK to 200 MHz, it will already be 1200 MHz. Reducing the QPI frequency of Lynnfield processors has no practical benefit for overclocking (at least with air cooling). But reducing the Uncore frequency during overclocking will not work at all, and at 200 MHz according to BCLK, this unit will already operate at 3200 MHz. However, increasing the frequency of the L3 cache will only have a positive effect on performance.

With air cooling, all Core i5 processors achieve a BCLK frequency of about 200-220 MHz. Having several budget motherboards for Socket LGA1156, we found out that the limit of our CPU in terms of base frequency (with air cooling) is 220 MHz. At higher values, significant system instability was observed. Thus, with a maximum x21 multiplier “in the air,” it is theoretically possible to get even 4620 MHz. In fact, we settled on 4066 MHz, at which full stability was maintained in stress tests (OCCT, LinX, etc.). Note that this result was achieved on the Gigabyte GA-P55M-UD2 board with a CPU Vcore voltage of 1.4 V and a QPI/Vtt Voltage of about 1.35 V. Further overclocking required a significant increase in voltages for stability, which entailed overheating under stress. tests

All memory settings during overclocking are shown in the following screenshot:

As you may have noticed above, the overclocked memory frequency was only 642 MHz (effective 1284 MHz). In fact, the Team memory kit itself is designed for 2000 MHz, but with the Gigabyte GA-P55M-UD2 board, when overclocking the processor, it was simply impossible to set the memory to a more productive mode. At a higher multiplier, the system froze before loading the operating system, and raising the corresponding voltages did not help. And in nominal mode, the board had problems with the operation of the X.M.P. profile, but we will cover these nuances in a separate article on this board. Due to the “incompatibility” of high CPU frequencies and high memory multipliers (by the way, we encountered something similar in some AMD Phenom II units), we had to limit ourselves to a low DDR3 frequency, but with latencies of 6-6-6-16, which must somehow compensate for the lag even from the nominal 1333 MHz. To slightly increase the memory frequency with its minimum multiplier, the CPU multiplier was specially reduced so that the BCLK frequency could be raised even higher. Comparative characteristics

To compare the performance of the Intel Core i5-750 in question, we selected the following quad-core processors:

• Intel Core 2 Quad Q8300;
• Intel Core 2 Quad Q9505;
• Intel Core 2 Quad Q9450;
• Intel Core 2 Quad Q9550;
• AMD Phenom II X4 810;
• AMD Phenom II X4 940 BE;
• AMD Phenom II X4 955 BE.

All of these models appeared in our latest big processor test, where you can glean details about them. We have a “virtual” Core 2 Quad Q9450, it was obtained from the Core 2 Quad Q9550 by reducing the multiplier from x8.5 to x8 and was added to the tests specifically so that we could clearly evaluate the advantages of the Lynnfield architecture over the Yorkfield-12M at the same frequency 2.66 GHz. It will also be quite interesting to see how much the performance of the junior quad-core CPU of the new generation has increased relative to the junior representative of the previous generation from Intel (Core 2 Quad Q8300) and the junior representative of AMD (Phenom II X4 810). To determine the benefits of Turbo Boost, our Intel Core i5 750 was tested at a fixed standard frequency of 2.66 GHz, i.e. with this technology disabled, and, accordingly, when it is activated.

		Intel Core 2 Quad Q9550	Intel Core 2 Quad Q9450	Intel Core 2 Quad Q9505	Intel Core 2 Quad Q8300	AMD Phenom II X4 955 BE	AMD Phenom II X4 940 BE	AMD Phenom II X4 810
Core	Lynnfield	Yorkfield	Yorkfield	Yorkfield	Yorkfield	Deneb	Deneb	Deneb
Connector	LGA1156	LGA775	LGA775	LGA775	LGA775	AM3	AM2+	AM3
Technical process, nm	45 high-k	45 high-k	45 high-k	45 high-k	45 high-k	45 SOI	45 SOI	45 SOI
Number of transistors, million	774	820	820	820	820	758	758	758
Crystal area, sq. mm	296	214	214	214	214	258	258	258
Frequency, MHz	2666 (up to 3200 in Turbo Boost)	2833	2666	2833	2500	3200	3000	2600
Factor	x20 (up to x24 in Turbo Boost)	x8.5	x8	x8.5	x7.5	x16	x15	x13
Base frequency, MHz	133	-	-	-	-	200	200	200
Bus QPI/FSB/HT, MHz, GT/s*	4800	1333	1333	1333	1333	4000	3600	4000
L1 cache, KB	(32+32) x 4	(32+32) x 4	(32+32) x 4	(32+32) x 4	(32+32) x 4	(64+64) x 4	(64+64) x 4	(64+64) x 4
L2 cache, KB	256 x 4	6144 x 2	6144 x 2	3072 x 2	2048x2	512 x 4	512 x 4	512 x 4
L3 cache, KB	8192	-	-	-	-	6144	6144	4096
Supply voltage, V	0,65—1,4	0,85—1,3625	0,85—1,3625	0,85—1,3625	0,85—1,3625	0,875—1,5	0,875—1,5	0,875—1,425
TDP, W	95	95	95	95	95	95	125	125

* — for QPI buses (Intel Core i5-750) and HyperTransport (AMD Phenom II) the speed is indicated in GT/s.

Test configurations

Test configuration Intel LGA1156:

• Motherboard: Gigabyte GA-P55M-UD2;
• Memory: Team TXD34096M2000HC9DC-L (2x2GB DDR3);
• Video card: Point of View GF9800GTX 512MB GDDR3 EXO (@818/1944/2420 MHz);
• Sound card: Creative Audigy 4 (SB0610);
• Hard drive: WD3200AAKS (320 GB, SATA II);
• Power supply: FSP FX700-GLN (700 W);

• Operating system: Windows Vista Ultimate SP1 x64;
• Video card driver: ForceWare 190.62.

Now let's look at the differences in the test benches of the other platforms that were used for comparison with the Core i5-750.

Intel LGA775 test configuration:

• Cooler: Thermalright Ultra-120 eXtreme;
• Motherboard: ASUS Rampage Formula (Intel X48, Socket LGA775);
• Memory: OCZ OCZ2FXE12004GK (2x2GB DDR2-1200);

AMD AM2+/AM3 test configuration:

• Cooler: Thermalright Ultra-120 eXtreme;
• Motherboards: MSI 790XT-G45 (AMD 790X, Socket AM2+), MSI 790FX-GD70 (AMD 790FX, Socket AM3);
• Memory: OCZ OCZ2FXE12004GK (2x2GB DDR2-1200), Kingston KHX1600C9D3K2/4G (2X2GB DDR3-1600);

Windows Defender, User Account Control and Superfetch were disabled in the operating system. The page file was fixed at 1024 MB. As noted above, the Core i5-750 processor was tested in two nominal modes - with Turbo Boost technology disabled and enabled. The mode with active Turbo Boost is indicated in the diagrams as “Core i5-750 TB”. The main characteristics of the test benches and memory operating modes for nominal modes and overclocking for each processor are given below in the form of two tables. In them you can see that the frequency data for some CPUs and their units may differ by a couple of megahertz relative to standard specifications, which is due to overestimation or underestimation of the reference frequency and FSB directly by the boards themselves.

System characteristics in nominal modes:

CPU	Processor frequency, MHz	Memory type	Memory frequency, MHz
Intel Core i5 750 Turbo Boost	2660-3198	DDR3	1330	7-7-7-20	2128	-
	2660	DDR3	1330	7-7-7-20	2128	-
Intel Core 2 Quad Q9550	2839	DDR2	1069	5-5-5-18	-	1336
Intel Core 2 Quad Q9450	2672	DDR2	1069	5-5-5-18	-	1336
Intel Core 2 Quad Q9505	2839	DDR2	1069	5-5-5-18	-	1336
Intel Core 2 Quad Q8300	2505	DDR2	1069	5-5-5-18	-	1336
AMD Phenom II X4 955	3200	DDR3	1600	8-8-8-22	2000	-
AMD Phenom II X4 940	3000	DDR2	1067	5-5-5-18	1800	-
AMD Phenom II X4 810	2600	DDR3	1600	8-8-8-22	2000	-

System characteristics during overclocking:

CPU	Processor frequency, MHz	Memory type	Memory frequency, MHz	Basic delays (CL, tRCD, tRP, tRAS)	Uncore frequency for Intel, NB for AMD, MHz	FSB frequency for Intel LGA775, MHz
	4066	DDR3	1284	6-6-6-16	3424	-
Intel Core 2 Quad Q9550	3962	DDR2	1165	5-5-5-16	-	466 (1864)
Intel Core 2 Quad Q9505	4004	DDR2	1178	5-5-5-16	-	471 (1884)
Intel Core 2 Quad Q8300	3548	DDR2	1183	5-5-5-16	-	473 (1892)
AMD Phenom II X4 955	3793	DDR3	1640	8-8-8-22	2255	-
AMD Phenom II X4 940	3675	DDR2	1120	5-5-5-18	2100	-
AMD Phenom II X4 810	3725	DDR3	1589	9-8-7-20	2384	-

Testing methodology

The testing methodology is described in the previous material. POV-Ray was excluded from the list of tests, since the built-in performance test in the version 3.7 beta 27 we used did not work correctly on the LGA1156 platform, and in newer versions the results have changed significantly on older processors. Due to the lack of the opportunity to repeat the test again in the new version of POV-Ray on processors from our list, we had to do without this program. For general information, we can only note that in POV-Ray 3.7 beta 35 the Intel Core i5 750 processor demonstrated a result that was almost 10% lower than the Core 2 Quad Q9550, and with Turbo Boost enabled it was 5% lower. Resident Evil 5 was excluded from gaming tests due to the strange behavior of the “fixed test” and “limitation” of performance on quad-core CPUs after running the application on dual-core configurations.
Test results

Synthetics. Application software

PCMark Vantage

The first synthetic test demonstrates the unconditional superiority of the Core i5-750 over the rest of the test participants, surpassing even the Phenom II X4 955, operating at 3.2 GHz. Compared to Core 2 Quad based on Yorkfield, Lynnfield has an advantage of about 13% at one frequency.

In this test the difference is not so great, although again the advantage of Lynnfield over the older Yorkfield tends to 10%. Unlike the previous overclocking test, the Core 2 Quad Q9505 and Core i5-750 show identical results.

In the Productivity Suite test, we again see Lynnfield's advantage over Yorkfield with 12MB cache of about 10%. If the older AMD processor outperforms Intel's previous generation rivals in this test, then the Core i5 is no longer a match for it.

In this archiver, Lynnfield has a huge advantage over its predecessors - more than 30%. Activating Turbo Boost helps you gain a couple more percent, but no more. The Core i5's leadership position with overclocking only strengthens, and at a frequency of 4066 MHz this processor already demonstrates an advantage of 40% over the Q9550 and 47% over the Phenom II X4 955. However, the results of the performance test in WinRar strongly depend on the performance of the memory subsystem, and in real archiving, the difference may no longer be so stunning.

The 7-Zip archiver has a rather cool attitude towards the Lynnfield processor. The performance of the Core i5 is only slightly higher than that of the Core 2 Quad Q9450. It manages to bypass the Q9550 when Turbo Boost is activated. In the same mode, the processor in question falls only 0.6% short of the performance of the Phenom II X4 940, operating at 3 GHz. With overclocking, the Core i5-750 is again ahead of everyone.

Paint.Net

In this test, Lynnfield at 2.66 GHz was only 1% more productive than Yorkfield with 12 MB cache at the same frequency. In Turbo Boost mode, our processor is already on par with the Core 2 Quad Q9550. With overclocking, the Core i5 is quite traditionally superior to other rivals; the difference with the Core 2 Quad is again not great, but already more than 3%.

Adobe Photoshop

In Adobe Photoshop, the younger Lynnfield confidently outperforms all other Intel rivals even without Turbo Boost, losing 11 seconds only to the AMD Phenom II X4 955. In turbo mode, the Core i5 is beyond competition, overtaking the older Phenom II processor by more than a minute. When overclocked, the Core i5-750 copes with the task almost two minutes faster than the older Core 2 Quad, operating at frequencies of about 4 GHz, and almost three minutes faster than competitors from AMD overclocked to 3.7-3.8 GHz.

CineBench

At the same frequency, the difference between Lynnfield and Yorkfield with 12 MB cache reaches 13% in favor of the first. In Turbo Boost mode, the Core i5 processor demonstrates better results than its steel rivals. Without turbocharging, the CPU is second only to the Phenom II X4 955, and that’s less than one percent. And at a frequency of 4066 MHz, the processor in question is completely out of competition: Core 2 Quad at 4 GHz is inferior to it by up to 19%, and Phenom II X4 at frequencies of 3.7-3.8 GHz is up to 33%.

Encoding Xvid Video to VirtualDub

Again, no surprises. Core i5 copes with the task faster than anyone else. Only without Turbo Boost, only the Phenom II X4 955 demonstrates an identical level of performance (and this is at a higher frequency of 540 MHz). With the same frequency, Lynnfield wins almost a minute over Yorkfield. When overclocked to 4.07 GHz, the advantage of the Core i5-750 over other competitors at higher frequencies is calculated in even greater numbers. Interestingly, the younger Core 2 Quad Q8300, even at 3.5 GHz, is slightly inferior in performance to the Core i5-750 with Turbo Boost. And the older Phenom II X4, only overclocked to 3.8 GHz, beats the processor in question in this mode by only seven seconds.

X264 Benchmark

In nominal modes, the Core i5-750 is inferior to the Phenom II X4 955, and even then, not by much. The advantage of Lynnfield over Yorkfield at one frequency reaches 12%. With overclocking, not a single processor is simply able to adequately compete with the CPU in question, which outperforms its predecessors by almost 16%, and AMD representatives by 20% or more.

PHP Benchmark

In this test, which is mainly sensitive only to the frequency of the processor itself, the Core i5-750 also did not lose face, and in Turbo Boost mode it turned out to be no worse than the high-frequency Phenom II X4 955. With overclocking, the processor again copes with the task faster than anyone else, although the difference with Core 2 Quad is already minimal.

Fritz Chess Benchmark

Core i5 is slightly more productive than Core 2 Quad Q9550 only in Turbo Boost mode. At 2.66 GHz, it is slightly inferior to the older quad-core CPUs of the previous generation, outperforming the Core 2 Quad Q9450 by only 2.8%. With overclocking, the younger Lynnfield strengthens its position, beating its closest competitors (Core 2 Quad Q9505 and Q9550) by about 7%.

Super Pi

In this test application, the Core i5-750 demonstrates a very impressive advantage over all processors in nominal mode, even without activating Turbo Boost. Compared to the Core 2 Quad on the Yorkfield core with a 12 MB cache at the same frequency, Lynnfield has an advantage of almost 23%. The rest of the overclocked competitors, at best, show the same results as the Core i5 without overclocking, but with Turbo Boost. Gaming applications

The first gaming test demonstrates the complete superiority of the Core i5-750 over other competitors. The younger Lynnfield manages to outperform the Core 2 Quad Q9550 and Phenom II X4 955 even without activating Turbo Boost. And when this mode is enabled, Core i5 demonstrates the same results as overclocked AMD Phenom II X4. Intel's predecessors for Socket LGA775 are not so sad, but they also cannot compete with the overclocked Lynnfield, despite the fact that with overclocking they all reached frequencies close to 4 GHz.

Battlestations: Pacific

In this game, despite the high fps, we were limited by the video card's capabilities, and, as a result, the difference in results is minimal. This is also explained by the peculiarity of the selected script scene, which creates a minimal load on the CPU. In any case, the Core i5 along with the Core 2 Quad Q9550 demonstrate the highest results in this game. When Turbo Boost is activated, a minimal drop in performance is noticeable, but it is difficult to talk about anything specific with such a small difference.

X3 Terran Conflict

In this game, the Core i5-750 doesn't even need Turbo Boost to beat its opponents. When activated, the result of the CPU in question turns out to be 5-10% higher than that of the older Core 2 Quad and 9-17% higher than that of the Phenom II X4 955. With overclocking, the lag of AMD processors reaches a huge 25-28%, and the Q9550 with its 3.96 GHz lags behind the leader with a frequency of 4.07 GHz by 8-10%. The younger Core 2 Quad and Phenom II X4 with overclocking only reach the performance of an unoverclocked Core i5 with Turbo Boost.

H.A.W.X.

One of the few gaming applications in which AMD processors are noticeably more productive than the old Intel Core 2 Quad, and even then only in low resolution. But the newer Core i5-750, unlike its predecessors, is not inferior to competitors from the “green camp”, outperforming their older processor with a frequency of 3.2 GHz by as much as 15% at 2.66 GHz. The superiority of Lynnfield over older Yorkfield at one frequency reaches almost 35%! But the Turbo Boost mode has almost no effect on the result - only plus 3%. When accelerating, the leader's gap from other competitors is no less impressive.

But with maximum image quality, the balance of power changes. So fast in weaker mode, the Core i5-750 suddenly takes last place. And what’s interesting is that the Turbo Boost mode no longer affects performance in any way, and overclocking is of little use.

World in Conflict

Intel Core i5 once again demonstrates a level of performance unattainable by its competitors. The advantage over Yorkfield is about 30%. All processors except the Core 2 Quad Q9550 with overclocking only approach the performance of the leader operating at nominal. And the Core 2 Quad Q9550 at 3.96 GHz does not have a particularly impressive advantage over the Core i5-750 with Turbo Boost, given the huge difference in frequency.

Higher resolution and heavier graphics settings slightly temper the ardor of the “unstoppable” Core i5-750, and now all overclocked Core 2 Quads manage to beat its result in nominal mode. In terms of minimum fps, the leader loses ground to the older Core 2 Quad even more noticeably, and even at nominal value it does not outperform the Core 2 Quad Q9550 in this parameter.

Unreal Tournament 3

In Unreal Tournament 3, a permanent leader pushes all opponents into the background. For AMD processors, everything is completely sad - even when overclocked to 3.8 GHz, they cannot demonstrate the same results as the Core i5-750 at 2.66 GHz. And over its predecessor, the Core 2 Quad Q9450, the advantage reaches almost 30%, while the Core 2 Quad Q9550 is inferior by a significant 20%. Turbo Boost mode increases Lynnfield performance by no more than 4%. With overclocking, the balance of power between Intel processors remains almost unchanged, but the gap between AMD and them only increases.

S.T.A.L.K.E.R.: Clear Sky

Unlike the previous game, in this domestic project the Core i5-750 secures its leadership without any reservations. Its advantage over older models Core 2 Quad and Phenom II X4 reaches almost 30% in low resolution and 23% in high resolution. And even with overclocking, competitors are struggling to somehow catch up with such a gap. Traditionally, AMD processors, when overclocked to 3.7-3.8 GHz, do not reach the Core i5 performance at a nominal 2.66 GHz.

Far Cry 2

At low resolutions, the Core i5-750, as usual, turns out to be the “fastest” of all, and “poor” AMD processors again cannot achieve the same results when frequencies increase to 3.7-3.8 GHz.

But at maximum settings, completely unexpectedly, the Core i5 again becomes an outsider, as it was in H.A.W.X. And again, Turbo Boost does not provide any advantages, nor does overclocking (mainly an increase in minimum fps).

In low resolution everything is quite predictable and the leadership position of the Core i5-750 is undeniable. The advantage of Lynnfield over Yorkfield with 12 MB cache at the same clock speed of 2.66 GHz is 26%. With Turbo Boost activated (which brings only 3%), the advantage over the older Core 2 Quad Q9550 and Phenom II X4 955 reaches 21-22%, and when overclocked, these rivals reduce their gap to only 17-20%.

In high resolution in nominal modes, the leadership of the Core i5 also does not raise questions, even though in this mode the performance is already noticeably limited by our video adapter. But with overclocking the CPU for some reason shows results slightly lower than the older Core 2 Quad. The difference is, of course, negligible, but still it is not an error, which, based on the results of several test runs, usually falls within a much smaller range.

Crysis Warhead

Crysis Warhead does not present any surprises and in all resolutions the Core i5 is the undisputed leader, and identical results with the Q9550 at 1280x1024 when overclocked are fully explained by the insufficient power of the video card, which played the role of a “limiter”. In low resolution, the advantage of Lynnfield over Yorkfield at a single frequency of 2.66 GHz reaches 17.5%. Activating Turbo Boost helps increase the result by 4.5%, and rivals from AMD cannot achieve such figures even when overclocked. The Core 2 Quad Q9550, which took second place on the “pedestal,” is inferior to the leader by 10% (without Turbo Boost) to 16% in nominal and 10% when overclocked.

Grand Theft Auto 4

Based on the testing results in this extremely processor-dependent game, it is clear that the requirements for the video subsystem are also quite high, despite the far from advanced graphics. As a result, both in low and high resolutions, we have hit a certain “ceiling” and the differences between processors are calculated in very negligible values, which, given the instability of the built-in benchmark itself, can often be attributed to measurement errors. True, this does not prevent the Core i5-750 from quite confidently taking the place of the leader at a resolution of 1024x768 at medium settings, but at higher settings it is already slightly inferior to the Phenom II X4 955. But in the same mode (at a resolution of 1280x1024) with overclocking, when The results of all processors, it would seem, hit the limit of 56 frames and above, the video card no longer “allows”, the Core i5 suddenly demonstrated a higher (almost 1 frame) result. And this clearly goes beyond the margin of error, and once again demonstrates the powerful potential of Lynnfield.

Armed Assault 2

We have already noted the poor results of AMD processors in this test application in a recent article. Let us remind you that we are using a pre-release demo version of the game, which is equipped with its own play test. It is quite possible that in the full version of the game, which has been overgrown with a huge number of patches, the performance of Phenom II has increased significantly.

The object of our review, the Intel Core i5-750, is quite expectedly the leader, but the Core 2 Quad Q9550 is literally a few percent behind it. With overclocking, the Core i5 at 4.07 GHz outperforms the Core 2 Quad Q9550 at 3.96 GHz by a more significant 10%.

Cryostasis: Sleep of Reason (Cryostasis)

In this application, poorly optimized for multi-core processors, the Core i5-750 manages to outperform the older Core 2 Quad Q9505 and Core 2 Quad Q9550 only when Turbo Boost is activated. With overclocking, Lynnfield's most significant advantage is in minimum fps (which is more relevant for this benchmark when using NVIDIA PhysX software processing), and in average fps it is on par with the overclocked older Core 2 Quad.

conclusions

It's time to sum up some results of our testing. The Intel Core i5-750 we reviewed turned out to be out of competition compared to other processors of the previous generation and compared to AMD solutions. In almost all applications, it demonstrated a level of performance higher than the higher-clocked Core 2 Quad Q9550, sometimes even without activating Turbo Boost. The very benefit of this technology of auto-overclocking different cores brings on average an increase of no more than 5%, although in rare single-threaded tasks (for example, in the SuperPi test) it can reach as much as 15%.

The youngest representative of Lynnfield had the most significant advantage in gaming tests, but it must be admitted that in a number of applications the situation was ambiguous. With a significant advantage over all other CPUs at low settings, the Core i5-750 could be slightly inferior to them with high-quality graphics at a higher resolution. This was most clearly demonstrated in FarCry 2, when at a resolution of 1024x768, Lynnfield's lead over its closest competitors was almost 17-20%. But at the same time, at 1280x1024 and rendering in DirectX 10, these same competitors demonstrate results that are 15% higher. In similar applications, overclocking the CPU itself brings minimal benefit, and activating Turbo Boost has almost no effect on the result at all. The mechanism for this decrease in performance is not entirely clear; we can only state that the Core i5-750 is not always good at high resolutions and at high graphics settings. But this does not diminish the advantages of this processor. It may be inferior to its competitors in certain conditions, but in most games it demonstrates performance unattainable for them, often at the same frequency superiority over its predecessors on the Yorkfield core (with a maximum of 12 MB of L2 cache) reaches 30 % and more! It is also significant that the younger Yorkfield with 4 MB of cache memory in a number of applications achieves a comparable level of performance only when overclocked to 3.5 GHz. But the Core i5-750 is also the youngest representative of its family. Progress, as they say, is obvious.

However, the older Core 2 Quads are also not impressive compared to the Core i5-750 in low resolutions, but thanks to overclocking to 4 GHz they are even more or less comparable to the newcomer in some gaming applications. As for overclocking the object of our article itself, its frequency potential has increased slightly relative to its predecessors. The 4.07 GHz we received does not seem to differ much from the 4 GHz of the Core 2 Quad Q 9505 or 3.96 GHz of the Core 2 Quad Q 9550, but further overclocking of Lynnfield was limited mainly due to the insufficient performance of the Thermalright Ultra-120 eXtreme cooler . If we take into account that we used a powerful fan at maximum speed, then when working in quiet modes with air cooling systems in everyday use, the frequency limit for all these processors will be approximately the same. But SBO users can easily count on great results from overclocking the Core i5-750.

Due to Intel's pricing policy aimed at promoting new products, there is no point in buying the older Core 2 Quad Q9550 now, because the Core i5-750 on the local market will cost you at least $65 cheaper with higher performance. And Core 2 Quad Q9500 or Core 2 Quad Q9505 are also not particularly attractive in price. This situation makes many Core 2 Duo users think about a complete change of platform instead of upgrading to Core 2 Quad. And the Core i5-750 in this case will be the ideal choice, because with its level of performance it is the best processor for $200-220.

AMD processors generally look depressing compared to the Core i5-750, especially in gaming applications. In particular, the Phenom II X4 955, with a frequency difference of about 500 MHz, is almost always inferior to the younger Lynnfield in games. At the moment, it is simply impossible to consider AM3 processors as the basis for a promising gaming platform, and this is sad. You can argue that the cost of AMD products is lower and for the price of an Intel solution you can take the top-end Phenom II X4 965 with a frequency of 3.4 GHz. But will these additional 200 MHz help, if 500 MHz did not really help the Phenom II X4 955?.. I would still like to see more worthy and competitive solutions from AMD that could withstand not only the previous generation Intel processors, but also newer models. Let's hope that the upcoming Phenom II X6 will live up to our expectations.

Test equipment was provided by the following companies:

• AMD - AMD Phenom II X4 940 and Phenom II X4 955 processors;

DCLink – Intel Core i5-750, Core 2 Quad Q9550, Core 2 Quad Q9505, Core 2 Quad Q8300 processors, Gigabyte GA-P55M-UD2 board and Team TXD34096M2000HC9DC-L memory;

• MSI - AMD Phenom II X4 810 processor, MSI 790XT-G45 and 790FX-GD70 boards;
• SerOl - Point of View GF9800GTX 512MB GDDR3 EXO video card;
• Special educational equipment - memory Kingston KHX1600C9D3K2/4G;
• —hard drive WD3200AAKS.

Introduction

The launch of the Intel LGA 1156 platform was very successful, with online publications and user opinions being very positive. Our first articles about Core i5 covered processor and platform technologies, and gaming performance. Now is the time to explore the possibilities of overclocking new processors. How well can you overclock the latest Intel platform? What will be the impact of Turbo Boost technology? What about power consumption at higher clock speeds? We will try to answer all these questions in the article.

P55: “Next BX?”

This phrase is often used to describe a new chipset or platform that has the potential to become the de facto standard, that is, to dominate all direct competitors for longer than the life cycle of a conventional product would imply. Once upon a time, the 440BX chipset that powered the second generation Pentium II became the most popular chipset, although some competitors offered better specs on paper. The BX provided a lot for its price, and journalists often recall the name of this product.

Many users are still running a Pentium 4, Pentium D or Athlon 64/X2 or even the first generation Core 2 systems - and they want to upgrade to four cores and perhaps install Windows 7. The Core i5 is one of the most attractive options in terms of price/performance ratio today, especially for users with serious overclocking ambitions.

Does the P55 platform have the potential to become the next BX? Yes and no. On the one hand, Intel will be promoting the LGA 1156 socket interface for at least a couple of years, although the pinout and electrical specifications may change. From what we know today, we can assume that the base platform will survive until 2011, and this socket will be able to install all 32nm Westmere processors. So yes, he has good prospects.

However, there are some functions that promise to become relevant soon and which the P55 platform does not support today. The first is USB 3.0. The second is SATA with a 6 Gbit/s interface. Of course, the accelerated SATA interface will only have a significant impact on flash-based SSDs and eSATA snap-ins that connect multiple drives through a single eSATA interface. But USB 3.0, it seems to us, should become a mandatory standard when it appears, since most external drives are usually limited to a throughput of only 30 MB / s due to the bottleneck of the USB 2.0 interface.

Acceleration: good speeds, but some obstacles

For our project, we used the MSI P55-GD65 motherboard, planning to overclock the entry-level Core i5-750 processor to 4.3 GHz. However, we were able to reach frequencies just above 4 GHz by disabling some important processor functions.

Choosing the best LGA 1156 processor for overclocking

Click on the picture to enlarge.

Intel has so far released three different processors, all based on the LGA 1156 interface: the Core i5-750 at 2.66 GHz, the Core i7-860 at 2.8 GHz, and the fastest Core i7-870 at 2.93 GHz. These processors differ not only in their standard clock speed, but also in the implementation of the Turbo Boost acceleration function. The 800 series processors can accelerate individual cores more aggressively than other models. Let me give you a small table.

Turbo Boost: available steps (within TDP/A/Temp limits)
Processor model	Standard frequency	4 cores active	3 cores active	2 cores active	1 core active
Core i7-870	2.93 GHz	2	2	4	5
Core i7-860	2.8 GHz	1	1	4	5
Core i5-750	2.66 GHz	1	1	4	4
Core i7-975	3.33 GHz	1	1	1	2
Core i7-950	3.06 GHz	1	1	1	2
Core i7-920	2.66 GHz	1	1	2	2

Many people expect that faster processor models will overclock better, but this is not always confirmed in practice. Since the cores of all existing LGA 1156 processors are the same, we decided to first analyze the prices. And the price when purchasing in a batch of 1000 pieces from the Core i7-870 is $562. We think this is a bit pricey for enthusiasts looking for the best price/performance ratio, so we decided to look at the remaining models: the Core-i7-860 for $284 and the i5-750 for $196.

Since in our review at the time of the processor's launch and related articles we usually used faster models, we initially decided to take an entry-level processor in the overclocking project. Indeed, this model will be the most attractive to most of our readers.

We'll start with a stock clock speed of 2.66 GHz, and this model's Turbo Boost implementation can increase the clock speed to a maximum of 3.2 GHz. Since the Core i7-870 tops out at 3.6GHz at maximum single-core Turbo Boost, we decided to start overclocking at 3.6GHz and then see what the highest frequency the most affordable Core i5 processor can reach.

Platform Description

Click on the picture to enlarge.

On the Internet you can find many results of successful overclocking of different platforms on the LGA 1156 architecture (there are also results that are best avoided; we provided additional details in review of entry-level motherboards based on the P55 chipset). All major motherboard manufacturers consider the P55 chipset to be a key product, so they all invest a lot of money in development. We've already used three different P55 chipset motherboards in article dedicated to the release of the processor, so for overclocking we decided to take the flagship model MSI P55-GD65. There is also a P55-GD80 model on the market, which has a larger heatpipe cooling system, as well as three x16 PCI Express 2.0 slots instead of two. However, the three P55-GD80 slots are limited to 16, 8 and 4 lanes, while the P55-GD65 operates in 16 and 8 lane configurations.

MSI has implemented a seven-phase dynamic voltage regulator, a heatpipe cooling system and many other features that motherboard manufacturers usually install on models for overclockers. One small feature that sets this MSI board apart from many others is the OC Genie Overclocking System, a simple solution that automatically overclocks your system by increasing the base frequency once activated. MSI claims that the system manages all the necessary settings itself, but this feature requires high-quality platform components. But for this review, we decided to abandon all the unusual features and chose the traditional overclocking method.

We installed the latest BIOS, which allows us to disable Intel Overspeed protection, and then began our overclocking project. The largest multiplier we could select corresponded to the maximum Turbo Boost mode with four cores active - that is, one step above the default 20x (21 x 133 = 2.8 GHz). We got a higher clock speed by increasing the base frequency to 215 MHz.

Click on the picture to enlarge.

The i5-750's stock voltage is 1.25V - and with it we were able to achieve exactly the same maximum clock speed that Intel specifies for the Core i7-870 processor with maximum Turbo Boost mode with a single core: 3.6 GHz.

3.6 GHz idle.

3.6 GHz - memory settings.

The result is quite impressive, but we didn't expect anything less. We were able to overclock Core i7 processors on the LGA 1366 socket in exactly the same way without raising the voltage too much.

3.7 GHz idle.

3.7 GHz under load.

3.7 GHz - memory settings.

We reached the 3.8 GHz frequency without any problems. However, we had to increase the voltage in the BIOS from 1.25 to 1.32 V.

3.8 GHz idle.

3.8 GHz under load.

3.8 GHz - memory settings.

3.9 GHz idle.

3.9 GHz under load.

3.9 GHz - memory settings.

4.0 GHz idle.

4.0 GHz under load.

4.0 GHz - memory settings.

We were able to reach 4.0 GHz with a further increase in voltage to 1.45 V. We also increased the PCH chipset voltage (P55) to ensure stability, but our first problems did not manifest themselves until 4.1 GHz.

Remember that it was the 1.45 V voltage that turned out to be problematic when we carried out tests of inexpensive motherboards. Three P55 models (ASRock, ECS and MSI) failed. We plan to release a story next week in which we will look at the steps each manufacturer has taken to address the identified deficiencies.

4.1 GHz idle.

4.1 GHz under load.

4.1 GHz - memory settings.

We were able to run the Core i5-750 at 4.1 GHz with the BIOS Vcore set to 1.465 V, but the system was unable to return from peak load to idle without crashing. Further increasing the processor or platform voltage did not help either. We were able to further increase clock speeds when we turned off C-state support in the BIOS.

Unfortunately, the power consumption of the system after this step in idle mode increased by a significant 34 W. Of course, we were able to achieve higher clock speeds, but we also had clear evidence that it is better to keep the processor in the lowest possible idle state, so that transistors and entire functional blocks are turned off when they are not needed.

4.2 GHz idle.

4.2 GHz under load.

4.2 GHz - memory settings.

To achieve stable operation at 4.2 GHz, we had to increase the voltage to 1.52 V.

4.3 GHz idle.

4.3 GHz under load.

4.3 GHz - memory settings.

By increasing the voltage of our Core i5-750 to 1.55 V, we were able to reach 4.3 GHz, but this setting no longer made a difference. The system was stable enough to run Fritz tests and take CPU-Z readings, but we were unable to complete the entire suite of tests. However, we still don't recommend this setting for everyday use, as power consumption in idle mode increases to 127 W. Let's see what level of performance we can get after overclocking to 4.2 GHz, and how such a frequency will affect efficiency.

Table of clock frequencies and voltages

Overclocking Core i5-750	3600 MHz	3700 MHz	3800 MHz
Factor	20	20	20
	74 W	75 W	77 W
	179 W	190 W	198 W
BIOS Vcore	1.251 V	1.301 V	1.32 V
CPU-Z VT	1.208 V	1.256 V	1.264 V
CPU VTT	1.101 V	1.149 V	1.149 V
PCH	1.81 W	1.81 W	1.85 W
Memory	1.651 V	1.651 V	1.651 V
Fritz Chess test results	10 408	10 698	10 986
C-states	Included	Included	Included
Stable work	Yes	Yes	Yes

Overclocking Core i5-750	3900 MHz	4000 MHz	4200 MHz
Factor	20	20	20
System power consumption when idle	78 W	79 W	125 W
System power consumption under load	221 W	238 W	270 W
BIOS Vcore	1.37 V	1.45 V	1.52 V
CPU-Z VT	1.344 V	1.384 V	1.432 V
CPU VTT	1.203 V	1.25 V	1.303 V
PCH	1.9 W	1.9 W	1.9 W
Memory	1.651 V	1.651 V	1.651 V
Fritz Chess test results	11 266	11 506	12 162
C-states	Included	Included	Off
Stable work	Yes	Yes	Yes

Overclocking Core i5-750	4100 MHz	4100 MHz	4300 MHz
Factor	20	20	20
System power consumption when idle	80 W	114 W	127 W
System power consumption under load	244 W	244 W	282 W
BIOS Vcore	1.465 V	1.463 V	1.55 V
CPU-Z VT	1.384 V	1.384 V	1.456 V
CPU VTT	1.25 V	1.25 V	1.318 V
PCH	1.9 W	1.9 W	1.9 W
Memory	1.651 V	1.651 V	1.651 V
Fritz Chess test results	11 785	11 842	12 359
C-states	Included	Off	Off
Stable work	No	Yes	No

Test configuration

System hardware
Performance tests
Motherboard (Socket LGA 1156)	MSI P55-GD65 (Rev. 1.0), chipset: Intel P55, BIOS: 1.42 (09/08/2009)
CPU Intel I	Intel Core i5-750 (45 nm, 2.66 GHz, 4 x 256 KB L2 and 8 MB L3, TDP 95 W, Rev. B1)
CPU Intel II	Intel Core i7-870 (45 nm, 2.93 GHz, 4 x 256 KB L2 and 8 MB L3, TDP 95 W, Rev. B1)
DDR3 memory (two channels)	2 x 2 GB DDR3-1600 (Corsair CM3X2G1600C9DHX) 2 x 1 GB DDR3-2000 (OCZ OCZ3P2000EB1G)
Cooler	Thermalright MUX-120
Video card	Zotac Geforce GTX 260², GPU: Geforce GTX 260 (576 MHz), memory: 896 MB DDR3 (1998 MHz), stream processors: 216, shader frequency: 1242 MHz
HDD	Western Digital VelociRaptor, 300 GB (WD3000HLFS), 10,000 rpm, SATA/300, 16 MB cache
Blu-ray drive	LG GGW-H20L, SATA/150
power unit	PC Power & Cooling, Silencer 750EPS12V 750 W
System software and drivers
operating system	Windows Vista Enterprise Version 6.0 x64, Service Pack 2 (Build 6000)
Intel Chipset Drivers	Chipset Installation Utility Ver. 9.1.1.1015
Intel Storage Subsystem Drivers	Matrix Storage Drivers Ver. 8.8.0.1009

Tests and settings

3D games
Far Cry 2	Version: 1.0.1 Far Cry 2 Benchmark Tool Video Mode: 1280x800 Direct3D 9 Overall Quality: Medium Bloom activated HDR off Demo: Ranch Small
GTA IV	Version: 1.0.3 Video Mode: 1280x1024 - 1280x1024 - Aspect Ratio: Auto - All options: Medium - View Distance: 30 - Detail Distance: 100 - Vehicle Density: 100 - Shadow Density: 16 - Definition: On - Vsync: Off Ingame Benchmark
Left 4 Dead	Version: 1.0.0.5 Video Mode: 1280x800 Game Settings - Anti Aliasing none - Filtering Trilinear - Wait for vertical sync disabled - Shader Detail Medium -Effect Detail Medium - Model/Texture Detail Medium Demo: THG Demo 1
iTunes	Version: 8.1.0.52 Audio CD ("Terminator II" SE), 53 min. Convert to AAC audio format
Lame MP3	Version 3.98 Audio CD "Terminator II SE", 53 min convert WAV to MP3 audio format Command: -b 160 --nores (160 Kbps)
TMPEG 4.6	Version: 4.6.3.268 Video: Terminator 2 SE DVD (720x576, 16:9) 5 Minutes Audio: Dolby Digital, 48000 Hz, 6-channel, English Advanced Acoustic Engine MP3 Encoder (160 Kbps, 44.1 KHz)
DivX 6.8.5	Version: 6.8.5 == Main Menu == default == Codec Menu == Encoding mode: Insane Quality Enhanced multithreading Enabled using SSE4 Quarter-pixel search == Video Menu == Quantization: MPEG-2
XviD 1.2.1	Version: 1.2.1 Other Options/Encoder Menu - Display encoding status = off
Main concept Reference 1.6.1	Version: 1.6.1 MPEG-2 to MPEG-2 (H.264) MainConcept H.264/AVC Codec 28 sec HDTV 1920x1080 (MPEG-2) Audio: MPEG-2 (44.1 kHz, 2-channel, 16-bit, 224 Kbps) Codec: H.264 Mode: PAL (25 FPS) Profile: Settings for eight threads
Adobe Premiere Pro CS4	Version: 4.0 WMV 1920x1080 (39 sec) Export: Adobe Media Encoder == Video == H.264 Blu-ray 1440x1080i 25 High Quality Encoding Passes: one Bitrate Mode: VBR Frame: 1440x1080 Frame Rate: 25 == Audio == PCM Audio, 48 kHz, Stereo Encoding Passes: one
Grisoft AVG Anti Virus 8	Version: 8.5.287 Virus base: 270.12.16/2094 Benchmark Scan: some compressed ZIP and RAR archives
Winrar 3.9	Version 3.90 x64 BETA 1 Compression = Best Benchmark: THG-Workload
Winzip 12	Version 12.0 (8252) WinZIP Commandline Version 3 Compression = Best Dictionary = 4096KB Benchmark: THG-Workload
Autodesk 3D Studio Max 2009	Version: 9 x64 Rendering Dragon Image Resolution: 1920x1280 (frame 1-5)
Adobe Photoshop CS 4 (64-Bit)	Version: 11 Filtering a 16MB TIF (15000x7266) Filters: Radial Blur (Amount: 10; Method: zoom; Quality: good), Shape Blur (Radius: 46 px; custom shape: Trademark sysmbol), Median (Radius: 1px), Polar Coordinates (Rectangular to Polar)
Adobe Acrobat 9 Professional	Version: 9.0.0 (Extended) == Printing Preferred Menu == Default Settings: Standard == Adobe PDF Security - Edit Menu == Encrypt all documents (128-bit RC4) Open Password: 123 Permissions Password: 321
Microsoft Powerpoint 2007	Version: 2007 SP2 PPT to PDF Powerpoint Document (115 Pages) Adobe PDF-Printer
Deep Fritz 11	Version: 11 Fritz Chess Benchmark Version 4.2
Synthetic tests
3DMark Vantage	Version: 1.02 Options: Performance Graphics Test 1 Graphics Test 2 CPU Test 1 CPU Test 2
	Version: 1.00 PCMark Benchmark Memories Benchmark
SiSoftware Sandra 2009	Version: 2009 SP3 Processor Arithmetic, Cryptography, Memory Bandwith

All the games we tested showed impressive benefits. Left 4 Dead scales especially well with clock speed. 3DMark Vantage doesn't run much faster because it's a test that relies more on graphics performance.

Application performance also improves significantly after overclocking.

The same can be said about audio and video encoding tests. Higher processor clock speeds have a noticeable effect.

System power consumption remains virtually unchanged even if you increase the processor frequency and voltage. The processor's power-saving features provide excellent power efficiency by turning off blocks and cores when they are not needed. However, we had to disable C-state support to overclock the processor above 4 GHz, a move that had a noticeable impact on system idle power consumption.

The difference in energy consumption at peak load is also noticeable. Power consumption almost doubles when moving from 2.66 to 4.2 GHz. Of course, performance does not double, meaning system efficiency will suffer from overclocking.

Total energy consumed per PCMark Vantage run (Wh).

Average power consumption per PCMark Vantage run (power, W).

Efficiency: result in points per average power consumption in watts.

As you might expect, stock clock speeds with Turbo Mode active provide the best efficiency (performance per watt). Increasing clock speeds and voltages in the old-fashioned way improves performance, but increases power consumption even further. If you need an efficient machine, then it is better to avoid serious overclocking.

Our expectations for productivity gains were high but realistic. Intel's Nehalem architecture is unparalleled in terms of performance per clock today; we expected it to scale nicely with each megahertz added to the clock speed. In fact, our MSI P55-GD65 motherboard-based test system delivered a significant and almost linear increase in performance all the way up to 4GHz, where we had to turn off the processor's internal power-saving system (C-states) to reach the maximum clock speed. Of course, we don't recommend taking this step if you want to keep power consumption low during idle mode.

Knowing that there are many examples on the Internet demonstrating 4.5 GHz and above, our results seem disappointing. But remember that we used Intel's entry-level Core i5-750 processor in this project, which has a stock clock speed of 2.66 GHz. If we take the reasonable maximum of 4 GHz, we still get a clock speed increase of 1.33 GHz, or 50 percent. Additionally, we didn't care too much about the choice of cooling system. The Thermalright MUX-120 air cooler performed well, but liquid or more powerful air solutions can give even higher overclocking limits.

The Core i5-750 is a great processor for overclocking, but you still shouldn't get too carried away with the process to avoid excessive power consumption. Yes, you can get 4.2GHz frequencies similar to many LGA 1366 platforms, which have about the same overclocking potential - and for a lot less. But, again, we can't help but notice that the usual "rough" overclocking is no longer as attractive as it used to be.

Intel today is changing the very concept of overclocking, as it is changing processor specifications from clock speed to thermal package. As long as the processor does not exceed certain thermal and electrical thresholds, it can run as fast as possible. In fact, it is precisely this model that future AMD and Intel processors may be based on. The Core i5 processor and our overclocking project clearly show that static frequencies are no longer so interesting. What really matters is the clock speed range and thermal/electrical limits within which the processor can operate. And overclocking in the future may be about changing those limits rather than hitting any maximum clock speed.

We don't know if the P55 platform can be called the "next BX", but the Core i5/i7 processors for Intel's new LGA 1156 interface have great practical value whether you overclock them or not.

This material opens a series of notes in which I will tell you about the overclocking potential of interesting pieces of hardware. Processors, video cards, RAM - these are the three main components that every overclocker overclocks. The idea of ​​​​creating an overclocking database has existed for quite a long time, but statistical data is too scarce, so we will tell you about our impressions of the overclocking of our charges.

We start with perhaps the most interesting processors from Intel at the moment – ​​the Core i5 750. The cheapest processors of the current generation will now face each other, and we will find out which of the 8 copies will be the best.

Test bench

To study the platform for socket 1156, we chose the following configuration:

• Asus P7P55D Deluxe motherboard
• Cooler Scythe Ninja 2
• RAM 2x2Gb OCZ Flex 1600MHz CL6 1.65v
• Saphire 4890 OC video card (PCI-E plug required)
• Chiftec 1200W power supply
• Seagate 7200.12 250Gb hard drive

This is the first time I've encountered a motherboard from Asus on the P55 chipset and I want to note that the first acquaintance can be considered successful. The board worked easily and without problems with all set voltages. Among the features, I would like to note that the voltage set for the processor in the BIOS matched the readings with CPU-Z, which is very pleasing.

Testing methodology

All eight processors were tested at three frequencies:

• max valid frequency – maximum validated CPU-Z frequency.
• max bench frequency – the frequency at which the processor can be forced to operate in light benchmarks; the Super Pi1M test is taken as an indicator.
• max stable frequency – the frequency at which the processor will work 24 hours, 7 days a week, 365 days a year, without turning off for a second. Naturally, I'm joking - in our express testing conditions it is difficult to find a truly stable frequency. But as an estimate, we will take the test frequency of Hyper Pi 32M - the same Super Pi32M only multi-threaded.

From the settings in the BIOS the following were used:

• CPU Voltage: 1.35-1.45 V;
• CPU PLL:1.9-2.0V;
• IMC Voltage:1.4V;
• Dram Bus Voltage: 1.65 V.

The system was overclocked from Windows using a utility from Asus - TurboV. The operating system Windows XP SP2 was used for tests.

№	Max valid frequency, MHz	Max bench frequency, MHz	Max stable frequency, MHz	Butch	Voltage on the core, B	Validation CPU-Z	Screenshot Super Pi1M	Screenshot Hyper Pi32M
1	4577	4465	4274	L922B943	1,432
2	4535	4442	4233	L922B943	1,432
3	4527	4380	4213	L922B943	1,400
4	4577	4400	4256	L922B943	1,408
5	4527	4360	4214	L924B920	1,440
6	4600	4535	4337	L930B637	1,448
7	4536	4464	4256	L922B943	1,440
8	4577	4442	4274	L922B943	1,440

conclusions

Eight processors from three weeks of release took part in the testing: six copies from the 22nd week, one copy from the 24th week, and one copy from the 30th week. Based on the results, we can identify the winner of our testing: it was the copy with serial number 6, released in the 30th week of 2009. This processor is the coldest, and it is the only one that achieved the coveted numbers of 4.6 GHz. The processors of the 22nd week of release can be called strong middle peasants; half of the processors showed results close to 4600 MHz, but at the same time, the other half overclocked 50 MHz worse. And the most unfortunate, in my opinion, was the processor released in the 24th week of 2009; its distinctive features were its hot temper and zero response to voltage increases higher than 1.4 V.

The frequency at which the processors were able to withstand Super Pi1M was on average 4400-4450 MHz, the best percentage was able to pass 1M at 4535 MHz, and the worst only at 4380 MHz. 100 MHz means a lot in benchmarking. But in terms of stability, the frequency spread of all processors is not that high. Everyone withstood 4200 MHz, the winner even 4300 MHz. With confidence, you can set your home system to 4 GHz and operate the computer for your pleasure.

At present, the opinion has already been established, formed under the influence of system requirements, that a productive desktop computer focused on modern demanding games should have a powerful quad-core processor and a high-performance video card of the latest generation, and often a pair of video cards. However, given the prices for new processor models, such a computer can cost a pretty penny. For example: the most affordable processor of the latest generation, Intel Core i7-920, at the time of writing, costs more than $300. An entry-level motherboard based on the Intel X58 Express chipset (more details in the ASUS P6T review) compatible with this processor will cost about $200, and a modest three-channel RAM kit from $75. In total, for the combination “processor + motherboard + memory” you will need to pay an amount that is enough to buy a full-fledged ready-made computer based on AMD products, and the processor in such an assembly will also be quad-core, and the video card will be of the latest generation. To resolve this incident, Intel, whose brainchild the above proposed “expensive” system is, presented, in its opinion, more affordable proposals: Intel Core i7-860; Intel Core i7-870 and Intel Core i5-750 on the same Nehalem microarchitecture. Also, to reduce the cost of the finished system, the new Intel P55 Express system logic was introduced (more details in the GIGABYTE GA-P55M-UD2 review), on the basis of which you can create more affordable motherboards than on the Intel X58 compatible with the Intel Core i7-920. In this review, we will try to figure out how much more accessible high-performance solutions from Intel have become, and indeed, have they remained high-performance? We will judge by the Intel Core i5-750 processor, which at the time of writing is offered at a price of about $240 and is the most affordable offer on the revolutionary Nehalem microarchitecture.

Package

The CPU-Z program, although the latest version 1.52.1, is inherently unable to convey all the information about the capabilities of the processor. The fact is that the Intel Core i5-750 contains several innovative technologies that can only be seen during system operation, and a screenshot of the program can display the state of affairs only at one point in time. Naturally, all innovations will be examined and analyzed in detail, but a little later, since it is simply impossible to describe such a volume of information in one paragraph. At this stage, it should be noted that the processor in nominal mode operates at a frequency of 2.66 GHz, the voltage supplied by the motherboard in the “AUTO” mode is 1.232 V (with Turbo Boost technology enabled 1.304 V). It is also worth noting the QPI value of 2.4 GHz, which indicates the frequency of the bus of the same name. This bus, one might say, plays the role of an FSB, by analogy with processors for the Socket LGA 775 platform. However, unlike the “classic” FSB, which connected the processor with the north bridge of the motherboard, the QPI bus connects the processor core with the RAM controller and the bus controller PCI-E, it is noteworthy that the latter are built into the processor, and the northbridge is completely absent in Socket LGA 1156 motherboards.

To better understand the above image and innovations in the Socket LGA 1156 platform, you should track the evolution of Intel platforms and changes in the corresponding processors.

We should start with the Socket LGA 775 platform, which appeared on the market as a result of the improvement of Pentium 4 series processors. But it makes no sense to consider all stages of evolution, so let’s start with the still popular Intel P45 chipset today.

As can be seen from the block diagram of the Intel P45 chipset, the processor communicates with the north bridge (MCH) via the FSB bus (with a bandwidth of 10.6 GB/s). The north bridge, in turn, is capable of communicating with two channels of RAM (bandwidth 6.5 GB/s when using DDR2 or 12.5 GB/s with DDR3 modules), the south bridge (ICH) via the DMI bus (2 GB/s) and one PCI-E x16 v2.0 port or two PCI-E x8 v2.0 ports.

In such an “assembly” all elements are balanced and do not infringe on each other, with the exception of the limitation on PCI-E lines. The two video cards will operate in x8 mode instead of x16 and will lose a little performance due to the halving of the PCI-E x16 v2.0 port bandwidth.

The Intel X48 chipset is the latest and most productive for the Socket LGA 775 platform. It differs from the Intel P45 in the presence of as many as two PCI-E x16 v2.0 lanes, which, when using two video cards with the appropriate interfaces, will not be “impaired” in performance, because the bandwidth The PCI-E x16 v 2.0 port capacity is 5 GB/s.

Processors with the Nehalem microarchitecture brought with them the Intel X58 chipset and the Socket LGA 1366 platform, which over the years have rearranged the layout of the controllers. From now on, the memory controller has moved into the processor itself (similar to AMD solutions), thereby allowing the latter to communicate with the memory bypassing the north bridge. The processor itself began to communicate with the northbridge via the QPI bus. Its throughput is 25.6 GB/s, which is twice as much as that of the Socket LGA 775 platform (in the best case scenario, the FSB bus can provide a throughput of 12.8 GB/s). The north bridge, in turn, provided two PCI-E x16 v2.0 ports and communicated with the south bridge via the DMI bus. This arrangement of “forces” made it possible to more fully use a video system consisting of two video adapters with a PCI-E x16 v2.0 connection interface, a disk subsystem consisting of at least ten drives, a pair of network adapters, a powerful sound card, etc.

Such features could not be cheap, so it is not surprising that a set of a motherboard and a Socket LGA 1366 platform processor will cost about $500.

This is why Intel recently announced the “people’s” Nehalem and the accompanying Socket LGA 1156 platform with the only chipset supporting the Intel P55 Express.

Yes, the Intel P55 chipset is not full of “cosmic numbers”, but the absence of a north bridge is immediately noticeable. In the Socket LGA 1366 platform, the northbridge, by and large, served only as a QPI => 2xPCI-E x16 v2.0 + DMI switch. Moving it, after the memory controller, into the processor itself was simply a revolutionary move. Now the processor communicates with the RAM and video card practically without “intermediaries,” which will naturally affect the performance of the system as a whole. But, since the Socket LGA 1156 platform was released under the slogan: “people's Nehalem,” there are also some simplifications in comparison with the Socket LGA 1366 platform.

Firstly, the memory controller lost one channel and became dual-channel, like the Socket LGA 775 platform, but did not undergo any other changes, as evidenced by the Memory tab of the CPU-Z program. In all cases (using Intel Core i7-920 and Intel Core i7-860 processors), the timings and operating frequencies were the same.

Secondly, the number of PCI-E bus lanes was reduced to 16, which returned the video system throughput to the level of the Intel P45 chipset (one PCI-E x16 v2.0 or two PCI-E x8 v2.0).

Returning to the main topic, I would like to note that when buying a processor, you now have to, willy-nilly, buy part of the chipset (northbridge), which we discussed a little higher. Let's not forget about the processor characteristics itself, which are not limited to the clock frequency and QPI bus.

The Caches tab revealed to us the identity of both the volume and organization of the cache memory of the Intel Core i5-750 and Intel Core i7-9*0, and Intel Core i7-8*0 processors.

For a more clear comparison of all the above changes, we suggest that you familiarize yourself with the following table, which presents the most “bright” models of all four generations.

Kernel codename
Number of cores, pcs
Clock frequency, GHz
Level 1 cache, MB
L2 cache, MB
Level 3 cache, MB
Multiplier (nominal)
System bus, MHz / GB/s
Technical process, nm
Power dissipation, W
Supply voltage, V	0,8500 – 1,3625
Maximum memory capacity, GB
Memory type, MHz	determined by chipset		DDR3-800/1066/1333	DDR3-800/1066/1333
Number of memory channels, pcs
Crystal dimensions, mm
Crystal area, mm 2
Number of transistors, million pieces
Platform, Socket
Virtualization technology
Turbo Boost Mode
Multiplier for a single-threaded task / final clock frequency, MHz
Multiplier for a two-threaded task / final clock frequency, MHz
Multiplier for three-threaded and four-threaded tasks / final clock frequency, MHz
Hyper-Threading Technology

Speaking of the Intel Core i5-750, we see an updated implementation of the Nehalem architecture, which involves the use of a high-speed QPI bus and communication with RAM and a video adapter without any “intermediaries,” which is a definite plus, not to mention a more pleasant price. Moreover, motherboards for this processor cost only a little over ~$100 (for example, GIGABYTE GA-P55M-UD2). This platform is noticeably more affordable than a combination of Intel Core i7-920 and even an inexpensive motherboard based on the Intel X58 chipset.

But the good news does not end on these optimistic notes. Intel Turbo Boost technology is simply revolutionary. And its version, which was implemented in the Intel Core i7-9*0 line of processors, simply looks frivolous compared to the implementation of the latter in the Intel Core i7-8*0 and Intel Core i5-7*0 line. Let us recall that processors of the Intel Core i7-9*0 line, when activating Intel Turbo Boost technology, could dynamically (independently) increase their multiplier by one, thereby increasing the clock frequency of all cores by 133 MHz. Here's what the new interpretation of this technology looks like:

When a processor performs a single-threaded task, it on one's own changes its multiplier from 20 (clock frequency 2.66 MHz) to 24 and ultimately gets the resulting clock frequency of one of the cores 3200 MHz, which is 540 (!) MHz is higher than nominal. What is this if not legalized overclocking? For some games where, due to the use of an old-style engine, only one core is used, this processor mode will be a real gift. Further, technicians and marketers apparently decided that single-threaded tasks are nothing more than an antiquity, and it was a long time ago, and in general it’s not true. But two-threaded tasks, i.e. optimized for dual-core processors are precisely a ubiquitous relic of the past. So why not speed up the work of two-threaded tasks? Therefore, when loading only two cores, the processor independently increases the multiplier, as in the first case, from 20 to 24, which ultimately makes it possible for two cores to operate at the same coveted clock frequency of 3.2 GHz (!) . Fabulous!

Operation of Intel Turbo Boost processor

To test the operation of Intel Turbo Boost technology, the processor was initially started in nominal mode without turning it on. The specialized program CPUID TMonitor monitored the operation of all cores separately.

As can be seen from the screenshot of the CPU-Z program, all cores operate at the standard x20 multiplier and remain in this mode regardless of the load. But this is not entirely true and you should not trust the CPU-Z program from now on. The Enhanced Halt State (C1E) power saving technology in idle mode reduced the clock frequency to 1200 MHz on all processor cores and this is already the true value, which the CPUID TMonitor program modestly proved to us.

The next step in the motherboard BIOS was disabled three cores for a more visual and unambiguous representation of the operation of Intel Turbo Boost. To put it simply, the Intel Core i5-750 processor has been converted to a single-core processor, and Intel Turbo Boost technology has been activated.

From the very beginning and without stopping, the processor worked at 3.2 GHz, regardless of the level and complexity of the task.

By switching the Intel Core i5-750 processor to dual-core mode (disabling two cores in the BIOS), the effect was similar to the previous one. Regardless of the type of task, both cores operated at 3.2 GHz. Fritz Chess Benchmark, running in dual-threaded mode, served as an excellent test suite.

Next, it's time to run the Intel Core i5-750 processor at full power. With all four cores enabled, he was given a clean single-threaded task using Fritz Chess Benchmark. To our great surprise, Intel Turbo Boost technology not only worked clearly and without any “jags”, increasing the multiplier of one core to x21, but also cleverly transferred the task from one core to another.

Deciding to repeat the previous experience, the once popular Super Pi program was adopted. The result turned out to be completely identical. Intel Turbo Boost technology still cleverly played with a single-threaded process, transferring it from a relatively more loaded core to an idle one. If the operating system, for personal needs, loaded one of the cores with the execution of some system service, then the Super Pi process “quickly jumped” to a freer core.

To be sure, the experiment was repeated a third time. Now the Lame Explorer utility, which is a shell for the corresponding codec, was taken as the “load”. Once again we were pleased with the effect! One of the cores serving compression worked properly at a clock frequency of 2.8 GHz.

No matter how much I would like to move on to testing on this optimistic note, there was still a “fly in the ointment” in this “barrel of honey”...

Cooling and power consumption

Important performance characteristics of the processor, and the entire system, of course, are power consumption and heat dissipation. It is doubly interesting to check the performance characteristics, because the processor under study has a declared thermal package of up to 95 W, and is equipped with a rather modest cooler. Therefore, we measured the power consumption of the entire system and the temperature of the Intel Core i5-750 in various modes using a “boxed” cooler and an ASUS Maximus III Formula motherboard.

	Core supply voltage, V	Core clock frequency, MHz	Energy consumption of the system as a whole, Watt	CPU heating, C°
Idle, Intel Turbo Boost Technology disabled
Under load, Intel Turbo Boost technology disabled
Under load, Intel Turbo Boost Technology enabled

As a result, we got very interesting results. Firstly, it is worth paying attention to power consumption - 165 watts at the very peak of load seems an implausibly small value. This is exactly how the architectural features of this platform affect it. After all, the main consumer is now the processor, which also acts as the north bridge, and the Intel P55 Express chipset consumes only 5 W. It also uses cost-effective DDR3 RAM. As a result, if you subtract all low-consuming components from the total power consumption of 165 W, it turns out that more than half of the energy is “eaten up” by the processor. And it is from the processor that the cooler will have to dissipate this energy in the form of heat.

Secondly, when using a “boxed” cooler, we recorded significant heating of the Intel Core i5-750 processor. Moreover, the system was assembled in a fairly well-ventilated CODEGEN M603 MidiTower case with a pair of 120 mm intake/exhaust fans. This is the “fly in the ointment”. When the processor was operating at maximum load, even with Intel Turbo Boost technology deactivated, its temperature exceeded the stated maximum of 72.7 C°. To be confident in the measurement results, we carried out repeated tests with different motherboards. The result turned out to be approximately the same, but with one caveat - different motherboards set the core supply voltage differently in the “AUTO” mode, although not in a very wide range. Depending on the supply voltage, there was a dependence on power consumption and processor heating, but with not a very large scatter. Thus, the advisability of using a “boxed” cooler, as well as its presence in the package, is doubtful. That is why the supplied “boxed” cooler E41759-002 was replaced with Scythe Kama Angle.

During testing we used Processor Test Stand No. 1

Motherboards (AMD)	ASUS M3A32-MVP DELUXE (AMD 790FX, sAM2+, DDR2, ATX)GIGABYTE GA-MA790XT-UD4P (AMD 790X, sAM3, DDR3, ATX)
Motherboards (AMD)	ASUS F1A75-V PRO (AMD A75, sFM1, DDR3, ATX)ASUS SABERTOOTH 990FX (AMD 990FX, sAM3+, DDR3, ATX)
Motherboards (Intel)	GIGABYTE GA-EP45-UD3P (Intel P45, LGA 775, DDR2, ATX)GIGABYTE GA-EX58-DS4 (Intel X58, LGA 1366, DDR3, ATX)
Motherboards (Intel)	ASUS Maximus III Formula (Intel P55, LGA 1156, DDR3, ATX)MSI H57M-ED65 (Intel H57, LGA 1156, DDR3, mATX)
Motherboards (Intel)	ASUS P8Z68-V PRO (Intel Z68, sLGA1155, DDR3, ATX)ASUS P9X79 PRO (Intel X79, sLGA2011, DDR3, ATX)
Coolers	Noctua NH-U12P + LGA1366 KitScythe Kama Angle rev.B (LGA 1156/1366)ZALMAN CNPS12X (LGA 2011)
RAM	2x DDR2-1200 1024 MB Kingston HyperX KHX9600D2K2/2G2/3x DDR3-2000 1024 MB Kingston HyperX KHX16000D3T1K3/3GX
Video cards	EVGA e-GeForce 8600 GTS 256 MB GDDR3 PCI-EASUS EN9800GX2/G/2DI/1G GeForce 9800 GX2 1GB GDDR3 PCI-E 2.0
HDD	Seagate Barracuda 7200.12 ST3500418AS, 500 GB, SATA-300, NCQ
power unit	Seasonic SS-650JT, 650 W, Active PFC, 80 PLUS, 120 mm fan

Choose what you want to compare Intel Core i5-750 with

Alas, the miracle did not happen... Although there was hope for the Intel Core i5-750 thanks to Intel Turbo Boost technology, synthetic tests showed another “vinaigrette” of results, giving preference either to one of the models - representatives of the Nehalem generation, or to the already outdated Intel Core 2 Quad Q9550. The AMD Phenom II X4 955 was a complete fiasco in synthetic tests, despite its clock frequency of 3.2 GHz and a total cache size of 8 MB, just like the Nehalem representatives.

Game tests showed a more linear picture. Resource-intensive games Word in Conflict, Far Cray 2 and Race Driver:GRID gave preference to representatives of the Nehalem architecture, placing them according to price requests. The now “outdated” Intel Core 2 Quad Q9550 lags behind the top three quite significantly, although it is in a higher price category than the Intel Core i5-750. The exception was the demo version of Tom Clancy's H.A.W.X., which gave preference to the AMD Phenom II X4 955 and Intel Core 2 Quad Q9550. In her opinion, Intel Core i5-750, Intel Core i7-860 and even Intel Core i7-920 have insufficient performance. Apparently, this application is primarily important to the processor clock speed.

In general, given the cost of the new Intel Core i5-750 processors, they quite successfully compete with junior solutions for the LGA1366 platform and older processors for LGA775. Therefore, when equipping a new productive system, you should pay attention to the LGA1156 platform.

The Efficiency of Intel Turbo Boost Technology

Having received not quite the test results that were expected, it was decided to evaluate the effectiveness of Intel Turbo Boost technology in terms of its impact on performance.

Test package		Result		Productivity gain, %
	Rendering CB-CPU
	Shading, CB-GFX
	DirectX 9, High, fps
	DirectX 10, Very High, fps

Oddly enough, the average performance increase in all test programs and games was only 2.38%, but it was completely free and without a noticeable increase in power consumption. Let's assume that this became possible due to a mismatch in the type of load, because to enable the mechanism for increasing the multiplier from x20 to x24, a strictly single-threaded or dual-threaded load is required. Achieving this from test programs turned out to be extremely problematic. But even under such conditions there is some acceleration, resulting in 1-6% additional performance. Therefore, we recommend that you do not forget to activate Intel Turbo Boost technology in the BIOS.

Overclocking

Method for overclocking Intel Core i5-750 processors; Intel Core i7-860 and Intel Core i8-870 (Socket LGA 1156 platform, Lynnfield core) are slightly different from the Intel Core i7-920 line (Socket LGA 1366 platform, Bloomfield core). The fact is that the ratio of the BCLK frequency (similar to the FSB on the Socket LGA 775 platform) and the RAM frequency is set by the corresponding multiplier, which can take a value from x2 to x6. Thus, the processor operating in normal mode (without overclocking) can theoretically work with memory, the frequency sometimes ranges from 533 MHz (133 * 2 * 2) to 1600 MHz (133 * 6 * 2). In turn, this makes it possible to overclock the processor to the desired level without using too high-frequency, and as a result, expensive memory. For example: when overclocking a processor to 4.0 GHz, you will need to increase the BCLK frequency from 133 (2660 / 20) MHz to 200 (4000 / 20) MHz, but in this case it is theoretically possible to use memory with a frequency of 800 MHz (200 * 2 * 2 ) up to 2400 MHz (200*6*2).

The processor that came to us for testing was overclocked to 4209 MHz (BCLK - 210 MHz) with a supply voltage of 1,440 V, which in percentage terms is 58% of the “additive” relative to the standard mode. Further overclocking was limited by the stability of the system, i.e. The operating system could also start with a processor frequency of 4.5 GHz, but it and the applications worked with errors. If this were a Socket LGA 775 platform, then this result would become a record, but for now this is just an isolated fact, many of which make up statistics. For comparison, the previously tested Intel Core i7-860 was able to overclock to 4074 MHz (BCLK - 194 MHz) with a supply voltage of 1.296 V; The Intel Core i7-920 conquered the frequency of 3990 MHz (BCLK - 190 MHz) with a supply voltage of 1,360 V, and the Intel Core i7-940 was able to show stable operation at a frequency of 3910 MHz (BCLK - 170 MHz) with a supply voltage of 1,296 V.

Test package		Result		Productivity gain, %
		Rated frequency	Overclocked processor
	Rendering CB-CPU
	Shading, CB-GFX
Fritz Chess Benchmark v.4.2, knodes/s
Tom Clancy's H.A.W.X. Demo, High, 1280x1024, AA2x
	DirectX 9, High, fps
	DirectX 10, Very High, fps

The average increase in test programs was 37,9 %. Comparing again with the Intel Core i7-860, Intel Core i7-920 and Intel Core i7-940, which showed an increase in performance when overclocked 28,7% , 18,8% And 13,8% , the acceleration result of the Intel Core i5-750 can be described as extremely high. Judging by the capabilities of processors targeted at the Socket LGA 775 and AM3 platforms, the Intel Core 2 Quad Q9550 and AMD Phenom II X4 955 “accelerated” due to overclocking 18% And 13% respectively. Therefore, we can say that the Intel Core i5-750 processor has a very high overclocking potential, which provides the opportunity to get a lot of “free performance”.

Features of the memory controller built into the processor

Updating the location of the memory controller could not but affect its properties. That is why we will test all possible memory operating modes and evaluate changes in performance.

The first thing that came to mind was to fill all the motherboard slots for memory. Four memory sticks were installed in four slots, the same type as was used in testing.

It’s worth noting right away that neither the frequency nor the timings of the modules have changed their values, but the Command Rate parameter, which characterizes the delay of the controller when executing commands, has changed its value from 1T to 2T.

The following testing will show how much such a “change” will affect performance:

Test package		Result		Change in productivity, %
	Rendering CB-CPU
	Shading, CB-GFX
Fritz Chess Benchmark v.4.2, knodes/s
Tom Clancy's H.A.W.X. Demo, High, 1280x1024, AA2x
	DirectX 9 High, fps
	DirectX 10 Very High, fps

The performance drop is noticeable in all test programs. The average is 0.90%. Of course, this is not a lot, but, nevertheless, the conclusion is clear: due to the needs of modern games, the required amount of memory is at least 3 GB. And since two identical modules are needed to activate the Dual Channel mode, the best option would be to purchase two two-gigabyte memory sticks at once. The option “two one-gigabyte ones now and two more over time,” as you can see, is not entirely rational.

Actually, about Dual Channel and Single Channel... It is not uncommon that, due to financial difficulties, one stick of RAM is purchased, and later another one is purchased, sometimes with a capacity different from the first. We forcibly disabled Dual Channel mode by installing modules in only one channel to evaluate the performance drop in this case and obtained the following results:

Test package		Result		Decrease in productivity, %
	Rendering CB-CPU
	Shading, CB-GFX
Fritz Chess Benchmark v.4.2, knodes/s
Tom Clancy's H.A.W.X. Demo, High, 1280x1024, AA2x
	DirectX 9 High, fps
	DirectX 10 Very High, fps

The average performance drop was only 4.49%, although in some tasks it was more noticeable. The conclusion is as simple as in the previous experience: you should not save on buying memory when switching (purchasing) to the Socket LGA 1156 platform.

The next experience was nothing more than a forced memory slowdown. This experiment was carried out in order to determine the dependence of system performance on the frequency of RAM. What if you decide to save money and buy stale DDR3-800

Thanks to the connection between BCLK and memory frequency through x2, x4 and x6 multipliers, implemented in processors of the Intel Core i5-7*0 and Intel Core i7-8*0 lines, changing the memory frequency was not difficult. The results speak for themselves:

Test package		Result		Decrease in productivity, %
	Rendering CB-CPU
	Shading, CB-GFX
Fritz Chess Benchmark v.4.2, knodes/s
Tom Clancy's H.A.W.X. Demo, High, 1280x1024, AA2x
	DirectX 9 High, fps
	DirectX 10 Very High, fps

The average performance drop in test programs was 4.06%. This is even less than the loss of the Dual Channel mode. Of course, when performing tasks closely related to memory performance, the increase will be about 25%, but in all other applications this factor is not so significant. Thus, it is precisely on the memory frequency when purchasing a system that some savings are possible, although with dubious prospects.

Sufficient QPI bus bandwidth

And finally, I would like to check the feasibility of using the fast QPI bus, which directly connects the processor cores themselves and the memory controller with a PCI-E controller. The QPI bus was forcibly slowed down from 2400 MHz to 2133 MHz, which is a percentage reduction of -12.5%. The results of the performance changes are as follows:

Test package		Result		Decrease in productivity, %
	Rendering CB-CPU
	Shading, CB-GFX
Fritz Chess Benchmark v.4.2, knodes/s
Tom Clancy's H.A.W.X. Demo, High, 1280x1024, AA2x
	DirectX 9 High, fps
	DirectX 10 Very High, fps

So, with the QPI bus slowing down by 12.5%, the average performance drop was only 1.3%, which is a mere trifle. Obviously, the processors of the Intel Core i5-7*0 and Intel Core i7-8*0 lines received the high-performance QPI bus more as an “inheritance” from the processors of the Core i7-9*0 line than out of necessity. Considering that there are only three “consumers” of traffic on it (memory controller, PCI-E x16 v2.0 controller and DMI bus connecting the processor to the chipset), its bandwidth turned out to be somewhat unnecessary than necessary.

Conclusion

Intel is finally able to provide an Intel Core i5-750 processor that is affordable and worth the money spent. Firstly, the full implementation of Intel Turbo Boost technology makes the processor more flexible. Where else can you find a processor that independently increases the frequency of two cores at once by 540 (!) MHz? Secondly, its price, even taking into account some speculation about the new product, is more pleasant than that of other processors based on the Nehalem architecture, and it is even cheaper than the Intel Core 2 Quad Q9550 or AMD Phenom II X4 955. Thirdly, I would like to remember that even an entry-level motherboard based on the Intel P55 chipset, for example GIGABYTE GA-P55M-UD2, fully implements all the capabilities of the processor and at the same time costs only a little over $100. Thus, such a combination will be even cheaper than the average motherboard for the Socket LGA 775 platform with a processor of corresponding performance.

Subscribe to our channels