Chipsets that support the Intel Xeon processor. G31 chipset: overclocking, specifications and supported processors GMA3100 graphics core

The diversity of Intel's chipset portfolio is difficult to understand, as some features overlap from model to model. Choosing a higher model number doesn't necessarily give you a more advanced chipset. So, for example, the G35 provides more advanced graphics, but does not support DDR3 memory and does not work with the newer ICH9 southbridge, as the G33 does. However, the Intel G31 chipset is actually an entry-level chipset and uses a minimum of power.

The G31 belongs to the category of mainstream desktop chipsets designed for "basic computing". This means that this set of system logic is completely unsuitable for high-end systems and does not support any advanced features. The G31 chipset has been designed as a cost-effective option for the average user. Therefore, it is focused on processors such as Core 2, Pentium Dual Core or similar Celeron based on the Core 2 microarchitecture.

The G31 chipset is limited to 4GB of memory, while the G33 and G35 support up to 8GB. The low-end chipset only supports dual-channel DDR2-800 memory (although this is not a disadvantage compared to DDR3) and works with the ICH7 southbridge instead of ICH8, ICH9 or ICH10. As a result, the G31 supports only four SATA / 300 ports, but provides two more UltraATA / 100 channels, while newer chipsets support either one legacy ATA channel or none at all. The G31 with ICH7 southbridge provides eight USB 2.0 ports, HD audio, traditional PCI slots, and just a 100Mbps network controller. If you need a faster Ethernet connection, then look for a motherboard that comes with a PCIe NIC to provide gigabit Ethernet. Both motherboards presented in our review are exactly like that. Finally, although the G31 chipset has one x16 PCI Express upgrade slot, it is not PCI Express 2.0 compliant.

Although FSB1066 was originally listed, all current G31 motherboards we reviewed support FSB1333 processors as well. Both of our motherboards support all Core 2 processors including Core 2 Quad processors. Incredibly, even a three-phase voltage regulator on the Foxconn G31 motherboard is enough for the Core 2 Quad Q9550 to work, in case you decide to abandon the dual-core processor in favor of the quad-core one.

Graphics core GMA3100

Integrated graphics won't deliver the performance and features that decent 3D graphics cards like Radeon HD4850 ​​/ 4870 and Nvidia GeForce 260/280... Even "mainstream" video cards on GPUs GeForce 9600GT greatly outperform integrated solutions. However, we decided to stick with the integrated graphics to keep power consumption as low as possible.

The G31 chipset contains the Intel GMA3100 integrated graphics core based on the GMA3000 core. This is the same graphics core that Intel first used in the 965 chipset line (called GMA X3000), and although the G965 supports Pixel Shader 3.0, the G31, G33, Q35 chipsets are limited to the SM 2.0 model, which means support for DirectX 9.0c. However, this is sufficient for all the functions of the Aero interface in the Windows Vista operating system. Motherboards with the G31 chipset typically have one analog display output, D-SUB15, and sometimes a digital DVI output. Since the GMA3100 is not suitable for HTPC (home theater), the motherboards are not equipped with HDMI outputs; you should also not expect two digital outputs from such cards.

CPU Intel Core 2 Duo was released in late summer 2006 and made a splash. Suddenly, AMD Athlon 64 X2 processors outperformed the product that provided better performance while using less power. Released at 2.66 GHz (2.93 GHz for the expensive Extreme version), the Core 2 Duo retained its superior status for over two years.

The first update consisted of increasing the bus frequency of the processors from FSB1066 to FSB1333 in the summer of 2007 with the release of the P35 platform and slightly modified processors. The second modification is the frequency of FSB1600 in high-end models and a change in the technical process from 65 nm to 45 nm. All Core 2 Duo E7000, E8000, and Core 2 Quad Q8000 and Q9000 processors are based on 45nm Wolfdale cores... Quad-core Yorkfield processors use two Wolfdale dies.

Over time, Wolfdale processors have improved; the most recent modification was the introduction of M0 stepping, which reduced the power consumption of the processor in idle mode compared to L. ... Of course, the reduced L2 cache capacity is another opportunity to reduce power consumption, which we did not fail to take advantage of.

Currently, the E7000 line consists of only two models: the E7300 at 2.66 GHz and the E7200 at 2.53 GHz. Both models are based on the Wolfdale core, but have only 3 MB of L2 cache and a reduced FSB1066 (for comparison: all Core 2 Duo E8000 processors have FSB1333 and 6 MB of L2 cache). The E7000 line supports neither Virtualization Technology (VT) nor Trusted Execution Technology (TXT); for our purposes, this is not a problem, since both technologies are irrelevant to users of "mainstream" desktops. The reduced L2 cache size, coupled with reduced clock speeds, makes this processor an excellent choice for a highly energy efficient computer, as you will see in our testing. Despite the claimed 65W thermal package, the Core 2 Duo E7200 consumes much less energy than this ceiling.

The 775th socket is far from new. Over the entire period of its existence, a huge variety of motherboards have been released, it is simply impossible to list all of them,. It will probably be much easier to specify which motherboard chipsets support Intel Xeon server processors. In simple terms, you should find out which chipset is installed on your motherboard in order to understand whether Intel Xeon wants to work on it or not.

Purchase

All the necessary hardware was purchased from our "narrow-eyed friends" on the Internet site https://ru.aliexpress.com at "ridiculous" prices (). Also, used THIS CASHBACK SERVICE , which made it possible to additionally save up to 15%.

If you plan to buy in domestic stores, then pay attention to CASHBACK SERVICE LETISHOPS ... It is not so profitable for Aliexpress, but there are many stores there, returns from 1 to 30% with every purchase.

Compatibility table

Below is a small but rather capacious table on the compatibility of chipsets and Xeon LGA771 processors.

Well, one more table. If you are sure that the motherboard is fully compatible with the chipsets indicated in the left half of the table, you can safely select the processors listed on the right.

During the installation process, you need to pay attention to the fact that in the overwhelming majority of cases you have to update the BIOS, perform a flashing, taking into account the following:

the 5xxx series are all Intel Xeons whose model numbers end in 5xxx. They can be combined with motherboards that support one or two physical central chips.

Problems can arise with Intel motherboards. Very rarely there are troubles with motherboards from MSI, Gigabyte, ASUS. This may be due to the fact that Intel motherboards have their own BIOS, which practically does not lend itself to manual flashing.

Nforce 680i and 650i chipsets from Nvidia according to the official version do not work with 45nm processors. It all depends on luck. Some motherboards with these chipsets combined and functioned normally with 45nm Xeon 4 cores, and some did not. To find out how you will be, see the list of boards that have successfully passed the test.

Zeon's power and system bus frequency must be supported by your computer's motherboard.

After the release of the new processors of the Intel Core 2 family on the market of budget motherboards, focused on working with these processors, confusion began. The fact is that the release of new chipsets intended for building inexpensive computer systems based on the presented processors is scheduled for a much later date and motherboard manufacturers have tried to fill the gap with solutions based on the Intel 945 Express family chipsets. As a result, a number of inexpensive motherboards with support for Intel Core 2 Duo processors appeared on the market, but ... only budget models of these processors operating with 800 MHz FSB frequency 1066 MHz). Such motherboards have one more drawback: since these solutions are focused on building inexpensive PCs, most of them are based on chipsets with integrated graphics (Intel 945G Express, Intel 945GC Express, Intel 945GZ Express), but their graphics core is Intel Graphics Media The Accelerator 950 (Intel GMA950) does not meet the requirements for working with the Windows Vista Aero interface (recently, this criterion has become one of the main ones when evaluating the performance of integrated graphics solutions). But now, with the release of the new Intel P31 / G31 Express chipset family, which replaced the Intel 945 Express chipsets, everything is falling into place. GIGABYTE UNITED has expanded its line of budget motherboards with new solutions based on the Intel P31 / G31 Express chipsets. In this article, we will tell you about the capabilities of two of them - motherboards Gigabyte GA-G31MX-S2 and Gigabyte GA-P31-DS3L.

Gigabyte GA-G31MX-S2 and Gigabyte GA-P31-DS3L motherboards, joining GIGABYTE UNITED's S Series portfolio, are designed to work with the full range of Intel LGA775 processors running at 800, 1066 or 1333 MHz, including with the new quad-core processors Intel Core 2 Quad and Intel Core 2 Extreme, as well as the upcoming Penryn-based processors, which will be released using a 45-nanometer process technology.

The motherboard Gigabyte GA-G31MX-S2 is made in the microATX form factor (dimensions - 24.4x21.5 cm) on a dark blue textolite, traditional for GIGABYTE UNITED products. It is based on the Intel G31 Express system logic set (Intel G31 Express + ICH7), the capabilities of which are complemented by the ITE IT8718F I / O controller.

To install system memory modules (it is assumed that unbuffered memory modules of DDR2 SDRAM 533/667 or 800 standard will be used), the board has two DIMM slots. The memory subsystem can operate in both one- and two-channel modes. The maximum system memory supported by the board is 2 GB.

The motherboard Gigabyte GA-G31MX-S2 has an integrated graphics core Intel Graphics Media Accelerator 3100 (Intel GMA 3100), fully supports API DirectX 9c and OpenGL 1.4 and meets all the requirements of the new Windows Vista operating system to work with the Aero interface. If the capabilities of this integrated graphics solution do not seem enough to the user, then a PCI Express x16 slot is at his service, where you can always install a discrete graphics card that meets his needs. By the way, in addition to this slot for installing additional cards that expand the capabilities of the motherboard, there are two more PCI slots and one PCI Express x4 slot.

To organize a network connection, Gigabyte GA-G31MX-S2 provides the user with a Realtek RTL8110SC gigabit Ethernet controller. "Sound" here is implemented by a combination of the HDA (High Definition Audio) controller integrated into ICH7 and the Realtek ALC888 audio codec, which provides the ability to reproduce sound in 7.1 format. There are six audio connectors on the output panel of the motherboard for connecting external audio devices.

In addition to the aforementioned audio connectors, the output panel of the board contains two PS / 2 connectors (for connecting a mouse and keyboard), parallel (LPT) and serial (COM) ports, a D-Sub video connector, an RJ-45 connector and four USB 2.0 ports.

In total, Gigabyte GA-G31MX-S2 supports eight USB ports: four more, in addition to the four on the output panel, can be implemented using additional expansion brackets, for which the board provides two connectors (two ports per connector).

To build a disk subsystem, the board has four SATA II ports, one PATA connector supporting up to two ATA 100 or ATAPI devices, and an FDD connector.

The system components are powered by a three-channel VRM unit.

To cool the microcircuits of the north and south bridges, plate light-alloy radiators are used. Note also that this motherboard has two 4-pin connectors for connecting the processor (CPU FAN) and system (SYS FAN) fans. These fans are speed controlled using Gigabyte's proprietary Smart Fan technology, which is part of the Smart feature set. By the way, the set of functions implemented in this motherboard should be discussed separately. Like all S-series motherboards from GIGABYTE UNITED, the Gigabyte GA-G31MX-S2 motherboard has a set of proprietary innovations - a set of S-functions. In this case, this kit includes a set of Smart and Safe functions (this is what the S2 combination in the model name stands for). The Smart feature set includes a number of smart utilities to simplify the work with the system and automate some actions. Among them:

• Download Center is an intelligent utility that allows you to find and download the necessary drivers and utilities from the official website of GIGABYTE UNITED;
• @BIOS is a utility that allows you to automatically find and download BIOS firmware from the official GIGABYTE UNITED website and update the BIOS code in a Windows environment;
• Q-Flash - a low-level utility (accessible from the CMOS Setup Utility menu) that allows you to flash the BIOS without loading any OC;
• Xpress Install - a utility that provides automatic installation of drivers and utilities necessary for the operation of the motherboard;
• Boot Menu - allows you to select the device from which to boot without changing the settings in the CMOS Setup Utility;
• Smart Fan is a technology for intelligent control of the rotation speed of cooling fans.

The Safe feature set includes a number of utilities that improve the reliability of your computer system. Among them:

• Virtual DualBIOS - BIOS backup utility (a backup copy of the BIOS code is created in a closed disk partition with the ability to automatically recover in the event of a BIOS crash);
• BIOS Setting Recovery - a utility that allows you to restore custom CMOS Setup settings;
• C.O.M. (Corporate Online Manager) is a utility that allows the system administrator to remotely receive comprehensive information (versions of installed drivers and software of the client machine) about the configuration of user PCs on which the client software is installed, as well as to monitor in real time a number of subsystems of these computers (temperatures processor, fan speed, supply voltage);
• Xpress Recovery2 is a utility (loaded before loading OC and does not depend on it) that allows you to create a backup copy of the system with the possibility of its subsequent restoration from the created image;
• PC Health Monitor - Provides monitoring of the main operating parameters of the computer system through the CMOS Setup or EasyTune utility.

If the Gigabyte GA-G31MX-S2 is an excellent option for building compact computer systems with an integrated graphics core, for example, HTPC (Home Theater PC) systems or office PCs, then the second of the considered motherboards - Gigabyte GA-P31-DS3L - is already a solution more focused on the creation of productive home systems, in particular PCs for fans of computer games, although this may seem strange when it comes to a budget-level motherboard. Unlike the first model, the Gigabyte GA-P31-DS3L motherboard is made in the ATX form factor (dimensions - 32.4x21.5 cm). It is based on the Intel P31 Express system logic set (Intel P31 Express + ICH7), the capabilities of which are complemented by the ITE IT8718F I / O controller.

To install system memory modules (it is assumed that unbuffered memory modules of the DDR2 SDRAM 667 or 800 standard will be used), in this case, four DIMM slots are equipped - thus, the maximum system memory supported by the board is 4 GB.

Motherboard output panel
Gigabyte GA-P31-DS3L

Since this model does not have an integrated graphics core, there is only one way to build a video subsystem - to use a discrete graphics card, for the installation of which Gigabyte GA-P31-DS3L has a PCI Express x16 slot. In addition to this interface, there are three more PCI slots and three PCI Express x1 slots to accommodate additional expansion cards.

Like the Gigabyte GA-G31MX-S2, the Gigabyte GA-P31-DS3L motherboard has a Realtek RTL8110SC gigabit Ethernet controller and a sound controller, which is implemented by a bundle of the HDA (High Definition Audio) controller integrated into ICH7 and the Realtek ALC888 audio codec.

The possibilities of building the disk subsystem of this motherboard are similar to those described for the Gigabyte GA-G31MX-S2 model.

The output panel of the Gigabyte GA-P31-DS3L motherboard contains two PS / 2 connectors (for connecting a mouse and keyboard), parallel (LPT) and serial (COM) ports, a D-Sub video connector, an RJ-45 connector and four USB ports 2.0, a set of six audio jacks, and optical and coaxial SPDIF outputs.

The system components are powered by a four-channel VRM unit. By the way, when creating the Gigabyte GA-P31-DS3L motherboard, the manufacturer used the proprietary Ultra Durable technology, within which only high-quality solid-state capacitors are used in the electrical circuits of this solution, which have better characteristics and a much longer service life than electrolytic ones.

To cool the microcircuits of the north and south bridges, plate light-alloy radiators are used. To connect the cooling fans, the motherboard has three connectors: two four-pin connectors for the processor (CPU FAN) and SYS FAN2 system fan, and one three-pin connector for the SYS FAN1 system fan.

Like the motherboard described above, Gigabyte GA-P31-DS3L has a set of proprietary S-functions, but this time, in addition to the Smart and Safe sets, it also includes the Speed ​​set, which includes utilities so familiar to fans of GIGABYTE UNITED products. overclocking like EasyTune and MIT (Motherboard Intelligent Tweaker).

Summarizing the above, we note that although Gigabyte GA-G31MX-S2 and Gigabyte GA-P31-DS3L motherboards are built on related chipsets, they are aimed at different categories of users. At the same time, both of these models are undoubtedly united by the fact that they combine the promise of solutions based on chipsets of the new Intel 3x Express family and the availability of motherboards based on the Intel 945 Express chipset (the price of these motherboards, according to price.ru, is about $ 100 .) and can be an excellent solution for building low-cost computer systems with an eye to the future.

The Intel G31 chipset is designed to create low-cost office computers. Motherboards built on its basis have, as a rule, the MicroATX form factor and basic expansion options. The integrated graphics core Intel GMA 3100 is designed exclusively for work with office applications, as well as with simple 3D programs. We also note that budget motherboards do not have overclocking functions, and a few exceptions to this rule are not popular with computer enthusiasts.

In our today's review, we'll take a look at two boards based on the Intel G31 chipset. One of them is made by Foxconn, which has a fairly strong position in the budget sector. The second board was developed by abit, whose products have almost completely disappeared from store shelves. And, sad as it may seem, but, most likely, abit I-G31 is the last motherboard of the once famous company.

Specifications (edit)

Boxes

Foxconn packaging design:

We received the Foxconn product in a fully serial version, but the abit board came to our laboratory in a test version - only the board itself is included in the package, so we cannot evaluate its package bundle.

Foxconn G31MG-S complete set

The package bundle of Foxconn G31MG-S is not rich, but for a $ 45 product it includes all the necessary components.

Boards

The development of a budget product limits engineers to a fairly rigid framework, so motherboards of this type have exactly the same characteristics as those of competing products. This is clearly seen in the tested motherboards: both development teams were given the task "to create a motherboard based on the G31 chipset that costs less than $ 50". As a result, the boards are almost identical:

For cooling the "north bridge" massive heatsinks are installed on the boards:

They are necessary, among other things, because the GMA 3100 graphics core is built into the "north bridge", which increases the overall level of heat dissipation. It should be noted that the Foxconn model also has a radiator installed on the "south bridge", while abit has saved on this part.

The boards have two 240-pin DIMM sockets for DDR2 memory modules, and the maximum amount of memory supported is 4 GB.

Each of the cards has one PCI Express x16 slot, one PCI Express x1 slot and a couple of PCI slots.

Now let's talk about the expansion options. Both boards have four SerialATA II links, but the possibility of creating RAID arrays is not provided. The fact is that the boards use one of the cheapest versions of the ICH7 south bridge.

We also note here that the "south bridge" provides operation of one ParallelATA link. Further, both boards have eight USB 2.0 ports. The port configuration is the same for both cards: four ports on the rear panel and four additional ports. Now - a few words about the High Definition Audio sound subsystem. The abit board has the Realtek ALC662 (5.1) codec, Foxconn has the ALC888 (7.1) codec. Both boards have high-speed network interfaces - the abit board has a Realtek RTL8111C controller, and Foxconn has an RTL8111B controller.

Foxconn G31MG-S motherboard is equipped with one COM and one LPT port, but abit I-G31 does not support these interfaces.

BIOS

The BIOS of the boards is based on the Award BIOS Phoenix version.

The main memory settings of the Foxconn board are in the "Fox Central Control Unit" section, and the abit memory settings are in the "Advanced Chipset Features" section:

An important parameter that affects performance is setting the memory frequency.

Now let's look at the sections of system monitoring.

Both boards display the current values ​​of the processor and system temperatures (abit also determines the PWM temperature), as well as operating voltages. The boards determine the rotational speed of the fans (abit - three, Foxconn - two) and have the function of adjusting the rotational speed of the processor fan depending on the CPU temperature. On the Foxconn board, this technology is traditionally called Smart Fan, and on abit - FanEQ:

The boards provide the user with access to all the latest Intel technologies.

Moreover, on the Foxconn board, the power saving settings are placed in a separate section ("Green System Mode")

The boards also allow you to adjust the amount of memory allocated for the needs of the integrated graphics core.

Overclocking and stability

Before moving on to overclocking, let's take a look at the power converters. The PWM of the abit I-G31 board has a three-phase circuit, in which there are two 820 uF capacitors, four 680 uF capacitors and four more 1000 uF capacitors. Foxconn's power supply module also has a three-phase circuit, in which six 820 uF and three 330 uF capacitors are installed.

On the Foxconn board, all overclocking functions are concentrated in the "Fox Central Control Unit" section:

Moreover, there are relatively few functions, and they are scattered across numerous subsections.

As regards the abit board, its BIOS lacks any overclocking functions at all.

However, the overclocking capabilities of the Foxconn board have little practical effect. In particular, the maximum stable FSB frequency is 350 MHz.

So, Intel has been paused for almost three years since the release of the revolutionary i9xx series of chipsets. Recall that then the desktop systems were added at once: a new type of socket and a new power connector, DDR2 memory, PCI Express bus (including the option for connecting video accelerators) and High Definition Audio. Then, over the course of two generations of chipsets (i945 / 955/975 and i965), only increases in FSB and memory frequencies followed, as well as support for new processor families (first dual-core, and then - Core 2).

Now we are meeting a new generation of chipsets, which, along with a radical change in numbering, offers an update to such important architectural characteristics of the system as the general-purpose bus and the type of memory.

Intel X38 Express

It is logical to start reviewing the new family of chipsets with its top representative, which, however, has not yet been released on the market and will appear only in the third quarter, like the entire second wave of new chipsets. Note that earlier the model number of the top product was set with an increased numerical index (i915 - i925), but now the top product can be easily distinguished by the X prefix, which Intel is responsible for any general improvements (not only for chipsets, but also for processors, video accelerators) ... The following block diagram lists the key features of the X38:

• support for "new" processors of the Celeron and Pentium families, as well as all processors of the Core 2 (Duo / Quad / Extreme) family with a system bus frequency of 800/1066 MHz, including future models with a system bus frequency of 1333 MHz;
• dual-channel memory controller DDR2-533 / 667/800 or DDR3-800 / 1066/1333 with support for up to 4 DIMM modules with a total capacity of up to 8 GB (with ECC) and Fast Memory Access and Flex Memory technologies;
• 2 graphic interfaces PCI Express 2.0 x16;
• DMI bus (~ 2 GB / s bandwidth) to the new ICH9 / R / DH / DO south bridge.

It is very noticeable that all the key characteristics of the chipset have undergone changes. Let's take a look at the innovations point by point.

Processor support... Here it should be noted right away that officially all 3x series chipsets do not support Celeron D, Pentium 4 and Pentium D processors (as well as their Extreme Edition versions). The lack of support is not due to the changed characteristics of the processor bus, but to the new FMB standard for creating motherboards (specifically, the VRM processor power supply module), which provides support for future processors created according to the norms of the 45-nanometer process, instead of the old ones produced on the basis of 90 (and more) -nanometer technology. Of course, there is no direct connection between the applied chipset and the power subsystem on the motherboard, but manufacturers, in the vast majority of cases, follow Intel development standards, so it seems extremely unlikely that we will see a significant number of models based on Intel 3x supporting processors of the pre-Core era. 2 ". Not to mention motherboards with simultaneous support for Prescott and Penryn.

As for Core 2 support, the X38 is doing just as well as possible: all current and future Core 2 Duo, Core 2 Quad and Core 2 Extreme models (including quad-core versions) will officially work with this chipset, and all of them will have 1333 MHz bus supported. From the younger families of the new processors (Celeron 400 and Pentium E2000), everyone will be able to work on the X38, although for marketing reasons Celeron 400 support for the top chipset has not been announced.

Memory support... The capabilities of the DDR2 controller have not changed in all the new chipsets (in fact, no development in this area is expected, everything that is in the specification has already been implemented), but Intel 3x-based motherboards will also be able to work with DDR3 memory. The features and theoretical performance of the new type of memory have already been analyzed in a separate article on our website, but here we will restrict ourselves to considering the practical aspects. The first question that usually arises is whether it is possible to simultaneously support DDR2 and DDR3? The situation here is no different from the transition from DDR to DDR2: Intel does not officially test such combinations and does not check them for compatibility, but no one bothers motherboard manufacturers to do it on their own. Our readers, who regularly check the news, are undoubtedly already familiar with several models of combo boards, and even today we tested one of them (however, we will hardly see the combo models on the X38). Note that the simultaneous Work DDR2 and DDR3 memory is, of course, impossible: at startup, the board will initiate work with memory of one type or another.

In the context of assembling systems, DDR3 is good for everyone: less heat dissipation (the supply voltage is lowered, so even DDR3-1066 will emit less than DDR2-800), a different location of the key in the socket will not allow confusing DDR2 and DDR3 slots on combo boards. As you already know, DDR3 is supposed to operate at frequencies up to 800 (1600) MHz, and the X38 will immediately allow you to use almost the fastest version - DDR3-1333. The situation with the availability and timings of available memory at the time of launching Intel 3x is awful. DDR3 modules are not yet massively presented on the market, and in such conditions even "elite" manufacturers (like Corsair) allow themselves to sell modules with frankly mediocre characteristics at an insane price. We recommend all our reasonable readers to wait, as over time, of course, prices will fall and characteristics will grow. In the meantime, analysts predict that DDR3 will reach 50% of its market presence only in 2009, and by the end of 2007 this type of memory will hardly gain even 10%. And of course, in the practical part of the article we will see why we are offered to overpay.

PCI Express 2.0... Here Intel strikes a preemptive blow, not only having finally created a chipset with support for two full-speed PCI Express x16 interfaces, which has long been boasted by top competitors' products (in the vast majority of cases, there is no need to talk about a really tangible gain from such a configuration, but the principles are more expensive), but also by implementing the host controller of the second version of the standard. In practical terms, the use of PCI Express 2.0 does not prevent the use of old video cards, since the connectors are used the same, and compatibility in both directions is observed. As regards the graphical interface, the PCI Express 2.0 innovations are likely to be not very interesting, with the exception of two. First, the performance of each PCI Express lane has been doubled, so that a connection with one lane (PCIEx1) now has a bandwidth of 500 MB / s in each direction at the same time, and for a 16-lane PCIEx16 interface, the total bandwidth will be 16 GB /with. We emphasize that in the foreseeable future, the systems will not get any practical benefit from this.

Secondly, the power supplied via the bus was increased by the same 2 times: the PCIEx16 slot of the first version of the standard provided up to 75 W, but now the video card can get 150 W. (The question immediately arises, how will these "extra" watts get to the bus - will motherboards on X38 have a special additional power connector?) However, how immediately after the start of i915 / 925 video cards for PCI Express began to appear, but with their own power connector onboard (bus 75 W was not enough), and now top-end video accelerators only condescendingly look towards the power supply from the PCIEx16 slot, suggesting, at best, to abandon one of the two on-board power connectors. However, here, of course, the "merit" of SLI / CrossFire is great: top-end video cards are primarily designed for pairing, and if one can theoretically still have enough bus power, then the second video accelerator, thoughtlessly deprived of its own power connector, simply cannot in such conditions to start. As for the ability to combine a pair of Intel X38-based video cards, everything is the same here: CrossFire is officially supported, SLI is not officially supported and will not be available in the foreseeable future.

Also paired with the X38 will be a new south bridge from the ICH9 family, the functionality of this family will be discussed in detail below.

Intel P35 Express

Let's briefly list the main functional characteristics of the north bridge of this chipset:

There are already fewer innovations here, of the most significant only DDR3. Processor support is limited to the same models based on 65nm and the future 45nm process technology, but due to the reasons described above (for motherboards based on P35, a simplified FMB design is assumed), Core 2 Extreme models (especially quad-core ones) will not work in motherboards based on P35. Also, the chipset lacks support for DDR3-1333 memory (in fact, it lacks a divider for setting such a memory frequency). Instead of PCI Express 2.0, the standard PCI Express x16 (first version) graphics interface is used, and like the P965 and earlier chipsets, the P35 does not allow flexible configuration of this interface to support CrossFire. However, as before, this fact does not stop the motherboard manufacturers - they create solutions for CrossFire based on the P35, connecting the second slot to the south bridge (where PCIEx1 peripheral interfaces go to it). The south bridge for this chipset is also one of the ICH9 family.

Intel G33 Express

The main integrated chipset of the new family has a somewhat illogical name G33, while in functionality it is on a par with the P35. The reason is that in the third quarter Intel will release another integrated chipset (now G35), with an improved graphics core, and it was necessary for the newcomer not to catch up with the top-end X38. So, the G33, which is a variant of the P35 with an integrated graphics core, has the following architecture:

Let's briefly list the main functional characteristics of the north bridge of this chipset:

• support for "new" processors of the Celeron and Pentium families, as well as Core 2 Duo / Quad processors with a system bus frequency of 800/1066 MHz, including future models with a system bus frequency of 1333 MHz;
• dual-channel DDR2-533 / 667/800 or DDR3-800 / 1067 memory controller with support for up to 4 DIMMs with a total capacity of up to 8 GB (without ECC) and Fast Memory Access and Flex Memory technologies;
• PCI Express x16 graphic interface;
• integrated graphics core GMA X3100 with support for Clear Video technology;
• DMI bus (~ 2 GB / s bandwidth) to the new ICH9 / R / DH south bridge.

Again, this chipset differs from P35 only in the presence of integrated graphics.

Integrated GMA X3100 graphics... Let's hope that the video drivers for the X3100 will be ready quickly, and we will finally see everything that we have been promised since the days of the X3000 (G965). In fact, the new video core has not undergone major changes compared to the GMA 950 (i945G), so it is significantly inferior in architecture to the X3000; we will analyze the differences when / if we can fully carry out all the tests. In the meantime, we recall that the Clear Video technology is designed to hardware accelerate and improve the quality (deinterlacing + color correction) video playback (including HD), as well as provide digital video interfaces (including HDMI) for image output. Of course, full support for the Aero interface is promised in Windows Vista. The GMA X3100 also supports playback of HD DVD and Blu-ray discs, we will look at the details after testing the boards on the G33.

Intel G31, G35, Q35 Express

Let's say a few words about the rest of the new line of chipsets. All of them will be released to the market in the third quarter of 2007.

The G31 is an entry-level integrated chipset, which can hardly be called new. In fact, its functionality is at the level of the 945G chipsets, which it is intended to replace. Even the south bridge of this chipset is the same old ICH7 / R - thus at the same time the leapfrog with PATA support, which is not at all desirable in the corporate sector, which has been going on since the days of ICH8. G31 supports Core 2 Duo (but with FSB not higher than 1066 MHz) and memory up to DDR2-800.

The G35 is an interesting integrated chipset with a revised graphics engine, which Intel promises will be the first [integrated] solution with DirectX 10 support. In due time, of course, we'll talk about the G35 (and its GMA X3500) in more detail. The rest of the G35 promises to be very similar to the G965 (note that this also applies to the integrated video), and architecturally it will be related to Intel 3x chipsets only by support for 45nm Wolfdale and Yorkfield and the new Core 2 Duo with 1333MHz FSB (memory DDR3 is also not supported). The old ICH8 / R / DH will be used as the south bridge for the G35.

The Q35 (and its simplified Q33) is the backbone for Intel vPro business systems, an integrated chipset with disabled gaming capabilities. Most interesting will be the combination of the Q35 with the ICH9DO (Digital Office) southbridge, which will support technologies such as AMT (Active Management Technology) 3.0, Trusted Execution Technology and Virtualization Technology. The Q35 also does not support DDR3 memory.

Intel ICH9 Southbridge

New chipsets - updated south bridges. Quite match its northern counterparts, ICH9 has a number of evolutionary improvements over ICH8, and also supports (ICH9R only) one technology that can be considered revolutionary. Let's briefly list the main functional characteristics of the new family of south bridges:

• up to 6 PCIEx1 ports;
• up to 4 PCI slots;
• 4/6 (4 for ICH9, 6 for ICH9R) Serial ATA II ports for 4/6 SATA300 devices (SATA-II, second generation of the standard), with support for AHCI mode and functions like NCQ (for ICH9, this mode is guaranteed to work only under Windows Vista), individually mute, eSATA and port splitters support;
• the ability to organize a RAID array (only for ICH9R) levels 0, 1, 0 + 1 (10) and 5 with the Matrix RAID function (one set of disks can be used in several RAID modes at once - for example, on two disks, you can organize RAID 0 and RAID 1, a separate part of the disk will be allocated for each array);
• 12 USB 2.0 devices (on two EHCI host controllers) individually disabled;
• Gigabit Ethernet MAC controller and a special interface (LCI / GLCI) for connecting a PHY controller (i82566 for Gigabit Ethernet, i82562 for Fast Ethernet);
• Intel Turbo Memory support;
• High Definition Audio (7.1);
• strapping for low-speed and outdated peripherals, etc.

ICH9R traditionally differs from ICH9 in the presence of support for RAID arrays, as well as two extra SATA ports. Special versions of the south bridge ICH9DO (Digital Office) and ICH9DH (Digital Home) are based on ICH9R, but the first of them additionally offers Active Management Technology 3.0, Trusted Execution Technology and Virtualization Technology, and the second - Viiv Technology (the positioning of both of these variations is obvious) ...

Of the minor evolutionary changes, the number of USB 2.0 ports increased to 12, the implementation of the eSATA function and port splitters (which is important for external eSATA connectors) for SATA chipset ports, as well as SATA connectors (like USB, starting with ICH8) are now subject to individual disconnection. An alternative to creating RAID arrays for data safety can be the new Intel Rapid Recover Technology, which allows you to create a disk image on another hard drive, quickly update it without touching unchanged files, and quickly recover data if the first hard drive is damaged. The south bridge still integrates a Gigabit Ethernet MAC controller, but we have not seen it used in any i965-based board - apparently, for ordinary desktop systems, a network controller from Marvell, Broadcom, Realtek and others like them, connected via the PCI Express bus turns out to be cheaper. At the same time, users of corporate vPro systems will certainly appreciate the features of the proprietary Intel controller. It would be strange to expect the return of PATA support after it was abandoned in ICH8, and it really did not happen - Intel considers this issue closed despite the abundance of problems with "substitutes" for chipset PATA.

The most intriguing feature of the new series of south bridges is the support for Intel Turbo Memory technology (known as Robson Technology in the process of development). Its essence consists in installing a module with a certain amount of NAND flash memory on the board (for a start, it is supposed to produce versions with 512 MB and 1 GB). Basically, apparently, the module will be installed in the PCIEx1 slot, although, in principle, other connection options are possible (for example, to the contacts for the external USB port). Users of Windows Vista will benefit from Turbo Memory, and unlike, say, USB key fobs with flash memory, the module integrated on the board can be used by the new Microsoft OS for both ReadyDrive and ReadyBoost.

Briefly, in the first case, we get the opportunity to use a flash drive as a cache memory for a hard drive - for linear read-write operations, there can be no big gain here (flash memory is slower than a hard drive), so the use of ReadyDrive will be observed with regular operations of exchange of small portions of data, which are typical for reading-updating the paging file (the access time for flash memory is noticeably shorter than that of a hard disk). An additional advantage is the reduction in the number of calls to the hard disk (data is merged to the disk in batches, during idle moments, and reading is not performed at all if there is the necessary data in the Turbo Memory cache), which saves energy - of course, this is a real gain only for mobile phones devices.

ReadyBoost, on the other hand, expands the available memory for pre-reading and caching data (from the hard disk), and although flash drives cannot compete with RAM in speed, reading not from the hard drive, but from flash memory with its low random access time allows you to significantly speed up the loading of applications and opening files (numbers are called up to 2 times). The downside of Turbo Memory is the potential fragility of flash drives, the best of which are characterized by the number of rewrite cycles of the order of a million (possibly several million), which, even with some reserve capacity, can lead to a loss of drive capacity long before the end of the PC's life. which one is installed.

Heat dissipation... The heat dissipation of the new chipsets deserves a separate mention. Despite being manufactured using the same 90nm process technology and more complex logic, the 3x series chipsets consume significantly less than their predecessors. So, the TDP for the P35 is 16 W (for the P965 - 19 W), and this despite the fact that the TDP of the new chipset is calculated based on the increased FSB (1333 MHz) and memory (1066 MHz DDR3) frequencies, that is, under equal conditions, the difference is much more than 3 watts in favor of the P35. Likewise, the new chipsets have noticeably lower maximum idle heat dissipation (5.9 W for the P35 and 10 W for the P965), although here a small concession is allowed for newcomers: measurements in idle are carried out for the case of 2 DIMMs, and not 4, as before ... In principle, the G33 is characterized by the same consumption values, but since this chipset can be used without an external video card, we will give its heat dissipation for this case for reference: in idle - 5.75 W (versus 13 W for the G965), and the TDP is 14.5 Watts (the G965 has a record 28 watts).

compare the reference heatsink for the new chipsets and the applied MSI

As a result, the difference is so noticeable that it can be easily detected even by touch when touching the chipset heatsinks. By the way, the reduced heat dissipation, of course, entailed a rework of the standard cooling system, and Intel's documentation provides a recommended version of a chipset cooler, with a significantly lower weight and surface area. Fortunately, those P35-based motherboards that we saw (including models from Intel itself) retained the same type of heatsinks (used for i945 / 965 chipsets), but top-end products from all manufacturers, of course, will continue to be equipped with powerful designs using heat pipes - the position obliges, although now it will become relevant only in case of serious overclocking. As a result, we have a turning point in an extremely unpleasant trend, when, after the hot i965 and the burning nForce 600i, it seemed that soon it would be necessary to develop new standards for chipset cooling devices.

Performance research

Test stand:

• Processor: Intel Core 2 Duo E6600 (2.4 GHz)
• Motherboards:
  • MSI P35 Neo Combo (BIOS V1.0B16 dated 04/20/2007) based on Intel P35 chipset
  • Gigabyte 965P-DQ6 (BIOS D25) based on Intel P965 chipset
  • EVGA nForce 680i LT SLI (BIOS P03) based on NVIDIA nForce 680i LT SLI
• Memory:
  • 2 x 1 GB Corsair XMS3-1066C7 modules (DDR3-1066)
  • 2 x 1 GB modules Corsair CM2X1024-9136C5D (DDR2-1142)
• Video card: ATI Radeon X1900 XTX, 512 MB
• Hard Drive: Seagate Barracuda 7200.7 (SATA) 7200 RPM

Software:

• OS and drivers:
  • Windows XP Professional SP2
  • DirectX 9.0c
  • Intel Chipset Drivers 8.2.0.1014
  • NVIDIA Chipset Drivers 9.53
  • ATI Catalyst 6.8
• Test applications:
  • RightMark Memory Analyzer 3.72
  • 7-Zip 4.10b
  • WinRAR 3.41
  • XviD 1.0.2 codec (29.08.2004)
  • SPECviewperf 8.01
  • Doom 3 (v1.0.1282)
  • FarCry (v1.1.3.1337)

Test platform

Due to the fact that we received several MSI motherboards based on the P35 chipset for testing at once, including one with simultaneous support for DDR2 and DDR3, as well as a set of DDR3 memory modules from Corsair, today's testing will help answer two questions at once. Firstly, we will find out how the DDR2 and DDR3 speeds on the same platform (P35) compare, and secondly, we will compare both versions of this platform with other chipsets on the market today. As the latter, it is logical to take the P965 (which is being replaced by the P35) and the top-end chipset of the latest NVIDIA series - nForce 680i LT SLI (we have already found out that there is no difference between nForce 680i LT SLI and nForce 680i SLI in terms of speed and functionality, and we had a board based on nForce 680i LT SLI).

Comparing the two types of memory turned out to be more difficult, since the pre-release BIOS versions of MSI boards were practically not ready for DDR3: the BIOS of the P35 Neo Combo did not provide an opportunity to set the normal (for DDR3) supply voltage (1.5 V) and timings (they were limited by the standard the DDR2 scheme, so it was impossible to set values ​​higher than 6 for the main timings). At the same time, the Corsair modules we had in DDR3-1066 mode did not agree to work with timings lower than 7-7-7, so the board had to be started in the mode of setting timings using SPD. Additional problems were created by the novelty of the platform, which did not allow checking the accuracy of timing (and other parameters of memory operation) by numerous utilities for Windows. Fortunately, the latest version of CPU-Z already understands both the P35 and DDR3 chipsets, so, with discounts on everything said above, we managed to bring some clarity.

In DDR3-1066 mode (SPD timings), according to CPU-Z, the timings were as follows: 7-7-7-20. Since the board did not allow setting the values ​​of the main timings higher than 6, we ran DDR2 memory at 1066 MHz with timings of 6-6-6-18 to bring the results as close as possible. At the same time, at 800 MHz, our DDR3 modules unexpectedly easily agreed to work even at 4-4-4-12 timings, which made it possible to compare this configuration with the P965 and nForce 680i LT SLI in their standard mode with [email protected] Since we did not have a P965-based motherboard at hand that would allow us to run our memory in DDR2-1066 mode, the previous generations of chipsets in this mode are represented only by an NVIDIA product (recall, according to tests it is extremely close to i965).

Now, before proceeding to the presentation of the test results, let's consider the issue theoretically. Under equal conditions (at the same frequency with the same timings), DDR3 cannot be noticeably faster than DDR2, and the main hopes for acceleration from the use of a new type of memory can only relate to modes with reduced timings at high frequencies. Indeed, according to the absolute values ​​of timings, the modes [email protected] and [email protected] are equal, so if memory manufacturers manage to launch low-latency modules, DDR3 may be more efficient even under “normal” conditions.

The second possible advantage of DDR3 is the increased bandwidth, since this memory can run on b O higher frequencies. Unfortunately, this gain may only appear on future processors, since at 1066 MHz FSB the bandwidth of this bus is only ~ 8.5 GB / s, which corresponds to the bandwidth of dual-channel DDR2-533! As practice shows, in such cases, usually increasing the memory frequency "by one notch" can still bring a small gain, but in reality even DDR2-800 is more than enough even for future processors with a 1333 MHz bus, while current processors are neither DDR3-1066 nor Moreover, DDR3-1600 is not needed.

Test results

Traditionally, we will start with a low-level study of the memory potential using a test developed by our programmers.

Let us remind you that despite the similar speed indicators in real applications, the NVIDIA and i965 chipset look very different in the synthetic RMMA test, so we will not focus on this difference.

The P35, which is noticeably inferior to both competitors in read speed, demonstrates a very interesting effect: when the memory (both DDR2 and DDR3) operates at 1066 MHz, its performance is higher than in the DDR2-800 mode, although the nForce 680i LT SLI has decrease slightly. Let us leave this fact, which is weakly consistent with our theoretical reflections, for the time being, and turn to other correlations. Actually, it remains for us to note that DDR3 looks noticeably worse than DDR2 even with equal timings. We deliberately do not indicate the exact size of the differences here, since it would be hastily to estimate the percentage differences before moving on to real tests.

When testing the write speed, we are not interested in the limit values ​​reached when using the direct data storage method, since they will be the same on a processor of the same architecture. In terms of the actually achievable write speed into memory, the picture is roughly the same as when reading: the new chipset is noticeably slower than its competitors, DDR3 is slower than DDR2 (especially at 800 MHz), and the transition to the memory frequency of 1066 MHz is still accelerated by the P35 with both types of memory, but slows down the NVIDIA chipset.

Finally, the memory latency test, and here the first surprise is the implementation in the P35 memory controller of a technology similar to NVIDIA's DASP - when pseudo-random reading from memory (without going beyond one page) drastically reduces latency, several times. Obviously, we are dealing with the same prefetch caching buffer. Nevertheless, even in such a successful test of pseudo-random reading from memory, the P35 is significantly inferior to its competitors (in this case, nForce 680i LT SLI). In the comparison between DDR2 and DDR3, the old memory type wins again on the P35, this difference is especially noticeable in the DDR2 / 3-1066 mode, where DDR3 has higher timings.

Curiously, even here switching to 1066 MHz memory frequency leads to an acceleration, although the ratio of the absolute values ​​of timings should have led to the opposite: taking into account the clock time, CL4 for DDR2 / 3-800 corresponds to 10 ns, and CL6 for DDR2-1066 - 11.25 ns (not to mention CL7 for DDR3-1066 - 13.13 ns). Why is that? Two possible explanations come to mind. Firstly, the bus frequency of the Core 2 Duo E6600 and DDR2 / 3-1066 memory is noteworthy: perhaps such a synchronous mode of operation provides some advantage. However, the absence of such an effect in the NVIDIA chipset suggests that some internal optimizations of the memory controller also affect, as in the i965, which allow you to get a small gain from running the memory at any higher frequency.

Well, now let's move on from considering theoretical aspects to real tests, and here, with numbers in hand, we will evaluate the advantage of certain configurations.

So, according to the real results, it is already possible to draw the first conclusions. On the one hand, all the ratios we noticed earlier have been preserved: P35 is slightly (now we can say specifically - up to 7%) inferior to P965 and nForce 680i LT SLI, DDR2-800 on P35 is faster than DDR3-800 with equal timings (by 3%) , and DDR2 / 3-1066 on the P35 is faster than the same type of memory at 800 MHz (it is impossible to give an exact estimate here, since the timings of DDR2 and DDR3 are different), and without a discount for significantly O higher timings. On the other hand, it should be noted that the 7% difference is observed only in one test, and work with DDR2-800 is obviously not the P35's strong point. The differences are further obscured by the fact that [email protected] is a memory with almost extremely low latency, while [email protected]- the standard version, which the same Corsair and the company will probably very soon offer an alternative with significantly reduced timings.

But let's not rush to a conclusion, let's look at the results of other tests.

There are no surprises to be expected from testing the video encoding speed (measured according to our open method), here, as usual, all competitors look the same, since the limiting factor is the processor performance.

In the package of professional 3D applications SPECviewperf, only NVIDIA chipsets manage to show themselves, which is most likely caused by their optimized graphics bus controller, since different modes of memory operation (and even different types of memory) affect speed only purely nominally.

We also see nothing new in games, only the fact that in one of the Doom 3 modes (for the first and last time during today's testing) the P35 is the absolute winner (and of course, with memory operating at 1066 MHz) is worth noting. However, the difference between the chipsets in Doom 3 is generally small, no more than 3%, and the losses due to the use of DDR3 instead of DDR2 on the P35 are even less - about 2%. In FarCry, the spread of results is slightly more significant, up to 4%, but all three patterns we have marked today remain valid.

conclusions

It is difficult to evaluate chipsets that bring several revolutionary innovations at once. In this case, the announcement turned out to be smoothed, because PCI Express 2.0 will appear only in the third quarter, with the release of X38, and no compatibility problems are expected due to the transition to the new version of the standard. The second novelty, DDR3 memory, did not make much of an impression on us with its speed characteristics, but, fortunately, at least in the first generation of chipsets there will be a choice between DDR2 and DDR3, so we can safely wait for a decrease in prices and an increase in the characteristics of a new type of memory. Support for new processors is probably the main trump card of the Intel 3x series. However, by the time these new processors become available, it may become clear that they are supported by other chipsets, including competitors' products, of which FSB 1333 MHz support is declared for at least the nForce 600i series, and no one can say about the actual support for 45nm models. The new south bridge is moderately progressive, adding a little here and there, and its main intriguing feature, Intel Turbo Memory, needs practical testing before reaching a verdict.

Before moving on to evaluating the performance, I would like to note that we are waiting, firstly, for confirmation of the speed level shown by MSI boards. Indeed, all three motherboards that came to us demonstrated absolutely the same level of performance when working with DDR2 (two of them support only this type of memory), however, just before the end of the tests, we received a new firmware for the P35 Platinum, which slightly (by several percent) increased the speed this model. In addition, although we cannot say that combined solutions are inferior in speed to dedicated solutions, general concerns of this kind remain, so it's too early to put an end to the issue of DDR3 performance. If we take into account the output of [inexpensive, that is, mass] processors with 1333 MHz FSB, the picture may change additionally. Nevertheless, having performed a significant amount of tests, it would be foolish not to draw any conclusions from it. Our conclusions are as follows: taking into account all the above and implied reservations, the chipsets of the new series still look a little slower than the old ones (both i965 and NVIDIA nForce 600i), DDR3 memory under equal conditions can lead to a 2-3% performance loss, and for the P35 it is better a memory operating at a frequency of 1066 MHz is suitable, regardless of timings.

Speaking globally about the fate of new chipsets on the market, X38 will undoubtedly find its own, albeit few, fans of top-end solutions, being one of the best chipsets on the market in terms of functionality. P35, after leaving the early start strip, should show a decent level of performance, and its solid functionality, low heat dissipation, support for promising processors and memory type allow today to recommend buying a motherboard based on a new chipset instead of solutions of a similar class from competitors and old chipsets. O th company. Turbo Memory technology, with a clear demonstration of all the promised properties, can become another very weighty argument for Intel 3x. We promise to tell you about the integrated options separately later.