What is the difference between a server motherboard and a regular one? Differences between server equipment and PC components. Some of the server tasks

For enthusiasts and overclocking fans, this Skylake processor has become not only a way to bypass Intel's overclocking limitations, but also a real opportunity to get a powerful quad-core Core i7 at the price of a Core i3. In this article we will look at another profitable option for purchasing iron in China.

On AliExpress, Taobao and Ebay you will find many Xeon server chips of the Sandy Bridge-EP and Ivy Bridge-EP generations, which went on sale in 2012 and 2013. At one time, these were one of the most high-tech and productive solutions, which is why such processors were quite expensive. For example, the retail price of the 8-core Xeon E5-2670, about which we'll talk further, when ordering a batch of 1000 pieces the cost was 1,550 US dollars. Now this model can be purchased for just a few thousand rubles, which is not surprising considering its age.

The presence of six and eight cores in outdated Xeon models attracts computer enthusiasts, and simply all those who want to save as much as possible when assembling system unit. Some Sandy Bridge-EP and Ivy Bridge-EP chips are comparable in price to modern ones dual-core processors Pentium and Core i3 generations of Skylake and Kaby Lake, and in terms of performance in multi-threaded applications - with quad-core Core i7. Sounds tempting, right? But, as with any adventure related to the purchase of iron in China, there are pitfalls. Therefore, we decided to study this issue in more detail by assembling a productive and inexpensive gaming computer based on the Xeon E5-2670.

⇡ Selection and purchase

The most attractively priced Sandy Bridge-EP models for the LGA2011 platform and Chinese motherboards on the X79 Express chipset are the Xeon E5-1620, E5-1650, E5-2650, E5-2660, E5-2665, E5-2670, E5- 2680 and E5-2690. Finding these chips for sale on sites such as Taobao and AliExpress is not difficult. The most popular options from this list are the 6-core Xeon E5-1650 and 8-core Xeon E5-2670. The first model has an unlocked multiplier. Statistics show that this processor, using high-quality air cooling, can easily be overclocked to 4.3-4.5 GHz. The Xeon E5-2670 does not have such an advantage, but under load all eight cores of the chip operate at a frequency of 3 GHz, which is also not bad. Other server processors no longer look so attractive in the price-performance coordinate system. The Xeon E5-1620 has only four cores, the E5-2650 (~2,000 RUR), E5-2660 (~3,000 RUR) and E5-2665 (~3,500 RUR) chips run on low frequencies, and E5-2680 (~7,500 rubles) will have to be paid extra.

The same trading platforms are full of Xeon processors of the Ivy Bridge-EP generation. The 10-core Xeon E5-2660 V2, Xeon E5-2670 V2 and Xeon E-2680 V2 seem to be interesting options - they are also compatible with the LGA2011 platform. An engineer Xeon E5-2680 V2, operating with all ten cores loaded at a frequency of 3100 MHz, costs an average of 8,000 rubles. A serial modification of this processor will cost significantly more.

As a result, our choice fell on the Xeon E5-2670 model (~4,000 rubles) with C1 stepping, since this chip costs less than the Xeon E5-1650 (~6,000 rubles), but without overclocking it turns out to be faster in all resource-intensive tasks, using multithreading. Specifications of these processors are shown in the table. For clarity, let's add to the comparison the Core i7-3970X Extreme Edition - the flagship chip among desktop solutions for the LGA2011 platform - and the Core i7-7700, since many sellers on TaoBao and AliExpress like to compare 8-core Xeons with 8-thread Core chips of the sixth and seventh generations.

It is important to understand that Sandy Bridge in 2017 is already considered an obsolete architecture. After its release in 2012, Intel introduced Ivy Bridge, Haswell, Broadwell and Skylake solutions, and when comparing a Sandy Bridge processor with a Skylake or Kaby Lake chip at the same frequency, a modern CPU will be faster by an average of 30%, and this is a noticeable increase. Plus, don’t forget that platforms evolve along with central processors.

	Xeon E5-2670	Xeon E5-1650	Core i7-3970X Extreme Edition	Core i7-7700
Platform	LGA2011	LGA2011	LGA2011	LGA1151
date exit	Q1 2012	Q1 2012	Q4 2012	Q1 2017
Codename	Sandy Bridge-E	Sandy Bridge-E	Sandy Bridge-E	Kaby Lake
Technical process, nm	32	32	32	14
Number of cores/threads	8/16	6/12	6/12	4/8
Clock frequency (Turbo Boost), GHz	2,6 (3,3)	3,2 (3,8)	3,5 (4,0)	3,6 (4,2)
Unlocked multiplier	No	Eat	Eat	No
Level 3 cache, MB	20	12	15	8
Built-in memory controller		DDR3-800/1066/1333/1600, 4-channel	DDR3-1066/1333/1600, 4-channel	DDR4-2133/2400, DDR3L-1333/1600, 2-channel
Number of PCI Express lanes, version	40, 3.0	40, 3.0	40, 3.0	16, 3.0
Integrated Graphics	No	No	No	HD 630
TDP level, W	115	130	150	65
Price at the time of sale, $	1 552	583	1 059	303

As you can see, it is very similar to the server Xeon E5-1650. But Intel desktop six-core processors (don’t forget about the existence of the Core i7-3960X), also designed for the LGA2011 platform, have higher clock speeds and more third-level cache: 15 versus 12 MB. In addition, it is worth noting that processors based on the Sandy Bridge architecture do not support the AVX2 vector instruction set.

On sale you can find Xeon with both C1 and C2 stepping - a new version corrects errors in the implementation of VT-d virtualization technology. C2 stepping does not provide any other advantages, but such versions of Xeon cost 800-1,000 rubles more. We have the C1 model at our disposal. This is not an engineering sample, there are full markings on the cover. The FPO (Finished Process Order) number 3135C087 indicates that the processor was manufactured in week 35 of 2011.

This article discusses the option of purchasing not only central processor, but also motherboard with kit random access memory. The fact is that it is not easy to find a cheap motherboard based on the X79 Express chipset from a well-known manufacturer on Avito and other flea markets. There are options, but for a “worn” board of the GIGABYTE GA-X79-UD3 level they are asking 13-15 thousand rubles. People are not shy about offering crazier options. Fortunately, enterprising Chinese quickly figured out how they could organize sales of old Xeons, and at the same time earn extra money. Therefore, simultaneously with server CPUs of the Sandy Bridge-EP and Ivy Bridge-EP generations, inexpensive motherboards from little-known or completely unknown manufacturers are being traded. It is this combination - Xeon, “some kind of X79 motherboard”, DDR3 registered ECC memory - that is the most profitable offer today. Chinese boards work easily with regular RAM modules, but register memory kits on AliExpress and Taobao are cheaper and allow you to save additional money. In addition, sellers sell packages that immediately include one or another Xeon Sandy Bridge-EP, a motherboard and RAM (almost always register memory).

As with any product purchased abroad, there are pitfalls. Buying computer equipment abroad is always a risk, and even more so in the case of such platforms as AliExpress and Taobao. Firstly, we contact unverified sellers, and the presence of a rating system only helps partly. Although sellers promise 100% performance of the hardware, no one is immune from deception, defects, or substitution. Secondly, there is no official warranty for such components, and the seller’s warranty, even if it is provided, will not be easy to use. Thirdly, after sending, there comes a tedious wait for the order, during which you can only hope that the Russian and Chinese post offices will not lose or break the parcel. In most cases, everything works out, but, as Forrest Gump said, shit happens.

It is more profitable to buy a processor, board and memory on the Taobao site - this is a kind of AliExpress for our own people (for the Chinese). Nobody bothers you to use the services of third-party “purchasing” services. The cost of Chinese motherboards for the LGA2011 platform, which we will talk about later, on AliExpress is noticeably inflated for unknown reasons. But the prices for the processor and RAM are quite tolerable.

Many sellers sell goods both individually and as a set. In the second case, it’s simply more convenient - the package will contain everything you need and upon receipt you can immediately begin assembling the PC.

	LGA2011	LGA1151	AM3+	LGA1151	LGA1151	LGA1151
CPU	Intel Pentium G4560	AMD FX-8320E	Intel Core i3-6100	Intel Core i5-7400	Intel Core i7-6400T (QHQG, L501C679)
Motherboard	H110 Express	AMD 760G	H110 Express	H110 Express	Z170 Express
Memory	2 × 8 GB DDR3-1600	2 × 8 GB DDR4-2400	2 × 8 GB DDR3-1866	2 × 8 GB DDR4-2133	2 × 8 GB DDR4-2400	2 × 8 GB DDR4-2133
Price	14,000 rub. (including delivery)	15,500 rub.	17,000 rub.	18,000 rub.	22,500 rub.	25,000 rub.

As a result, we purchased a kit that included a Xeon E5-2670, 16 GB of DDR3-1600 registered RAM and motherboard mATX form factor based on X79 Express. This set cost, taking into account all additional expenses, 14,000 rubles. And this is noticeably cheaper than options with modern Pentium G4560 or Core i3-6100. All components arrived safe and sound - the seller did not spare any packaging material.

⇡ Motherboard and RAM

In fact, decide motherboard for Xeon Sandy Bridge-EP is more difficult than with the processor itself. As mentioned above, getting a model with an X79 chipset from a well-known manufacturer is very problematic and expensive. Among the Chinese offers there are several inexpensive models, but all of them are not without drawbacks and require optimization. Formally, Chinese motherboards for the LGA2011 platform are divided into two types. Although the device name uses the “X79” marking, in most cases Intel C602/C604 server chipsets are used.

The first and, perhaps, the most popular option is motherboards of the ATX form factor produced by HUANAN, as well as its clones. Original boards are marked X79; the manufacturer's name is printed on the PCB or heatsink. Clones are designated X79Z - on such motherboards you can find a PCI slot (the original HUANAN does not have it). It is better to take boards of new versions: 2.46 and 2.47 for X79 and 2.4b for X79Z. We recommend avoiding old modifications, such as v150, 2.43 and rev 2.01. Firstly, there are no improved BIOS versions for them - this means that you will not be able to adjust the rotation speed of even PWM fans, and you will not run the RAM at an effective frequency of 1866 MHz. Secondly, these boards are not suitable for overclocking the Xeon E5-1650.

HUANAN and clones are the most expensive options. Some models have an M.2 slot, but it operates in SATA 3 Gb/s mode, so it is practically useless. But they are all equipped with two SATA 6 Gb/s connectors, so they are quite suitable for building a PC using an SSD. Additionally, most have an internal USB 3.0 connector.

The second type includes motherboards for the LGA2011 platform, made in the mATX form factor. They are cheaper, but have fewer ports on the I/O panel, as well as denser component wiring. The most popular models belonging to the second type are marked G218a. Among them there are versions v1.0, v1.1a (this is the board that was purchased in the end) and v 1.1b. In addition, you can find boards marked MS-7777. They do not support registered ECC memory, are not equipped with an internal USB 3.0 connector, and even internal USB 2.0 ones are not all working. And these motherboards are equipped with only three expansion slots. Another option is boards labeled YW-X79-E. They have support for DDR3 ECC REG memory. These devices are less common than others, and therefore their capabilities have not been fully studied by enthusiasts.

		YW-X79-E

As a result, you need to choose among motherboards whose power converter is equipped with a metal radiator. For assembly in a case, it is better that the board has an internal USB 3.0 connector. Plus, register memory is now relatively inexpensive. In general, we are looking towards models 2.46/2.47, 2.4b and v1.1a. We chose the G218a, preferring its more compact form factor. This motherboard also costs 1,000-2,000 rubles less than its ATX counterparts.

The chipset (after all, we are dealing with a microcircuit produced using a 65-nm process technology) in all Chinese motherboards is cooled by a small aluminum radiator. In a case with mediocre air circulation, the chip most likely will not last long.

The markings of the connectors on the PCB are similar to those that MSI uses in its solutions.

The seller included a simple cooler along with the processor, motherboard and memory. Also in the box we found instructions in English, a disk with drivers, a pair of SATA cables, a plug for the I/O panel on the case and an adapter for mounting the cooler.

Please note that the LGA2011 processor socket in the G218A-V1.1a and other “Chinese” products is designed slightly differently than in serial motherboards from well-known brands. It does not have mounting threaded holes at the edges, but four holes are drilled in the PCB, the distances between which correspond to the LGA1366 standard. This fastening system was used. It turns out that any coolers whose mounts are designed for use on LGA1366 platforms are compatible with Chinese motherboards. If such support is not provided, then the cooling system must be selected from among models that are attached using a crossbar with two eyes at the ends. For example, Deepcool Gammaxx S40 is suitable. Just use the adapter included in the kit. First we install it, then we hook the cooler to the lugs, as if we were installing a cooling system for an AMD processor.

The device has four DIMM slots, which means that Xeon Sandy Bridge-EP processors with a 4-channel DDR3 controller will be able to fully realize its capabilities. The v1.1a model supports ECC register memory.

The poor layout of some components immediately catches your eye. The internal USB 3.0 connector is inconveniently routed. And installing a long video card equipped with a dual-slot cooling system will lead to overlap of the PCI Express x1 closest to it and two SATA 3 Gb/s connectors.

By the way, the G218a has a total of six connectors for connecting drives. SATA 6 Gb/s ports are indicated in red, and 3 Gb/s ports are indicated in blue. Also on the PCB there are connectors for connecting case mini-jacks, COM and three USB 2.0 bundles at once.

The board received only three connectors for connecting fans. Two of them are four-pin. We will tell you more about setting the rotation speed of the turntables below.

The power subsystem has seven phases. Its elements are cooled by a small asymmetrical aluminum radiator. A screw mount is used, and a small layer of thermal paste is applied between the power converter and the heatsink. In the circuit, each channel has one inductor and a pair of M3004D transistors manufactured by UBIQ Semiconductor. The power subsystem is controlled by the ISL6366 PWM controller. I note that the G218a design uses exclusively solid-state capacitors.

Under load, the converter heats up noticeably, but the temperature does not go beyond 100 degrees Celsius. Ineffective cooling is... Therefore, when choosing options without additional cooling of the power subsystem, we were eliminated.

Please note that the RAM also gets noticeably warm. This problem affects all register modules. The forums report that kits without radiators have lower temperatures. In any case, to cooling the components future system based on Xeon and X79 boards must be approached with all responsibility.

The I/O panel of the G218a is frankly scanty. There are a pair of USB 3.0 Type A ports. There are four more of the same connectors, but USB 2.0. From the markings we see that there should have been optical S/PDIF, but the space for soldering this connector is empty.

A simple Realtek RTL8111E controller is responsible for the network. A similar version of this gigabit chip is RTL8111H - . The audio path is based on a six-channel Realtek ALC662. Modern budget motherboards for Skylake and Kaby Lake processors use better chips - Realtek ALC887 or Realtek ALC892.

Controllers G218A-V1.1a- gallery

BIOS version G218a does not have many settings. All Chinese motherboards use similar firmware. Yes, you have to deal with the good old blue interface without mouse support and the inability to update from a flash drive without creating a bootable DOS partition.

Apito Setup Utility- gallery

There is no way to regulate the voltage of the processor and RAM. At all. There is no access to manual timing control for RAM modules. Among the useful functions: setting the RAM frequency divider, CPU multiplier and Smart Fan - controlling the speed of fans connected to four-pin connectors depending on changes in processor temperature.

There are plenty of Samsung, Micron and SK Hynix RAM kits on sale. Kits with registered ECC memory for the LGA2011 platform are still cheap and cost significantly less than conventional RAM. But they are also noticeably slower. If you already have a set of regular DDR3 memory, then use it. If you buy register strips in China, then you can actually buy a two-channel DDR3-1600L kit with a capacity of 16 GB for 3,000 rubles. Considering that we are assembling a computer based on a fairly old platform, it makes sense to buy to capacity, namely, purchase four modules of 8 GB each. In my case, the seller sold the “board - processor - memory” kit only together with the Samsung M393B1K70DH0-YK0 set, consisting of two sticks with a total capacity of 16 GB. The letter “L” in the name of the RAM standard means that the kit operates at a voltage reduced to 1.35 V.

Practice shows that most of the DDR3-1333 and DDR3-1600 registered RAM purchased in China “starts up” at an effective frequency of 1866 MHz. Since we cannot change the timings in the BIOS, when the operating speed increases, most kits start with delays of 12-12-12-32.

⇡ Overclocking and tuning

Together with system board, processor and RAM, the seller supplied a simple cooler - just with fasteners for the LGA1366 socket. The fan does not have PWM, so it rotates at a frequency of about 2400 rpm all the time. In this mode, the cooler’s operation is audible: from 30 centimeters, the measuring device recorded an acoustic pressure of 42 dB.

At the same time, the efficiency of this cooling system is quite sufficient for the Xeon E5-2670. The processor frequency in LinX 0.7.0 drops to the standard 2.6 GHz. The BIOS provides settings to change the CPU power limit, but they do not work. In games and benchmarks, the operating frequency of all eight cores is 3 GHz. afudos newbios.rom /gan .

I think there is no point in reminding once again that an unsuccessful BIOS flashing process can lead to a “bricked” system. In this case, you will have to take up the programmer. Owners note that problems may arise with flashing HUANAN X79 rev2.47 motherboards. If you are unlucky, you will have to download this BIOS version using the programmer. But in the case of the G218a, the BIOS update went quickly and without any hiccups.

As already noted, the Xeon E5-2670 does not overclock. The multiplier can only be changed in the Xeon E5-1620 and E5-1650 models. You cannot manually adjust the voltage, but the processor changes it automatically in the range from 0.6 to 1.35 V. This is enough for the E5-1650 to accelerate to 4.3-4.5 GHz. Please note that high-quality cooling is necessary, since the TDP of the processor is 130 W. The multiplier is set in the BIOS Advanced/CPU Power Management Configuration menu.

Eight-core Xeon E5-2665, E5-2670 and E5-2680 are still overclockable, but only using the SetFSB utility (we used version 2.3.178.134). Clock generator operating frequency realistically increase by 3-7 MHz, which will give a total of an additional 90-210 MHz ( what luck with the iron?) for all eight cores.

The utility does not support generators such as ICS932SQ420DGLF and ICS932SQ420DKL. Therefore, go to the Diagnosis tab, in the Clock Generator menu, select the very last item - PLL Diagnosis. Click on the Get FSB button. A table will appear at the bottom of the SetFSB screen. Select the parameter that intersects at row “00” and column “06”. In the Bin field, change the default 00011000 to the following set of zeros and ones:

• 00011001 - for setting 101.30 MHz on the bus.
• 00011010 - for setting 102.47 MHz on the bus.
• 00011011 - for setting 103.78 MHz on the bus.
• 00011100 - for setting 105.08 MHz on the bus.
• 00011101 - for setting 106.25 MHz on the bus.
• 00011110 - for setting 107.55 MHz on the bus.
• 00011111 - for setting 110.03 MHz on the bus.

For example, to set the bus frequency to 106.25 MHz, change the value 00011000 to 00011101 and press the Update key. When you change the FSB for the first time, the computer may freeze. We turn it off by holding down the Power key on the case, and then start the system again. If you've loaded, that means the board has overclocked via bus up to 106.25 MHz. If you don’t boot, turn off the power to the system unit, wait ten seconds, turn on the system again - the settings will be reset. We try to set a lower clock generator frequency in SetFSB: we do all the same operations, but in the Bin field we specify the set 00011100 (105.08 MHz). And so on until we find a stable combination for Xeon, motherboard and RAM.

Enthusiasts report that the bus frequency of Chinese motherboards is overclocked using SetFSB to 106.25 or even 107.55 MHz. We were less fortunate, as the system only worked stably at 103.77 MHz. As a result, the processor frequency increased to 3113 MHz, and the RAM worked in DDR3-1937 mode with a set of timings of 12-12-12-32.

This acceleration will last exactly as long as power is supplied. Power outages or disconnecting the system unit from the outlet will lead to resetting the clock generator settings. Then the overclocking procedure using SetFSB will have to be repeated.

The built-in PCI Express controller of Sandy Bridge-EP processors supports the 3.0 standard. But sometimes the PEG ports of the motherboards into which the video card is installed operate in PCI Express 2.0 mode. In the case of NVIDIA video cards, this situation is corrected by a special patch. Just run it with administrator rights and reboot the system.

Some users have noted that overclocking an 8-core Xeon bus above 105 MHz causes PCI Express 3.0 mode to switch back to PCI Express 2.0 mode. But in general there is nothing wrong with PCI Express x16 2.0: the interface bandwidth is quite enough for video cards of the GeForce GTX 1070 and GeForce GTX 1080 level.

Marketers are actively pouring information into consumers' ears about the number of cores in processors. It seems that if you install it, you will get a very powerful gaming system. As far as you should know, prices for components for the server segment of the market are very different from solutions for home or office equipment. What happens if you install 12-core CPU Xeon E5 into a “regular” computer for everyday tasks? Today we will answer your question.

As they say, everything is learned by comparison. We will test speed of two computers in various tasks: games, benchmarks and resource-intensive “combat/real-life” applications. Server CPUs are usually aimed at use in multiprocessor systems, but also in desktops on ASUS X99-Deluxe and ASUS Rampage V Extreme motherboards. E5-4650 started up without problems. And I didn’t have to spend money on memory - it works great with a simple unbuffered DDR4 SDRAM.

Let's not delay any longer and introduce the nominees for today's battle: Intel Xeon E5-4650 v3(Haswell-EP, 12 cores + Hyper-Threading, 2.0 - 2.6 GHz, L3 cache 30 MB) and Intel Core i7-5960X Extreme Edition(Haswell-E, 8 cores + Hyper-Threading, 3.0 - 3.5 GHz, L3 cache 20 MB). The nominees were installed on an ASUS X99-Deluxe motherboard with 4 DDR4-2133 4 GB modules. Storage system presented SSD drive Crucial M550 512 GB. It was a discrete video card. Performance comparisons were made under operating room control Microsoft systems Windows 8.1 Pro x64.

First of all, we ran the “synthetic” utilities performance testing. They were " PCMark 8 Pro" And " 3DMark Pro" by Futuremark. The first program was launched in the “Home”, “Work” and “Creative” modes. These three scenarios simulate the load of a typical home computer, office use, entertainment and work with multimedia content, respectively. 3DMark Pro is known for its optimization for multi-core systems, but the results of both utilities show that server processor for gaming computer– the choice is irrational.

For comparison of speed in applications We have selected several “heavy” packages with optimization for multiple threads, which should load processors to the maximum and show the real state of affairs. 3DS Max 2015 was working on the final rendering of one frame of Space_flyby. Finereader 12.0 translated pre-scanned text with a large number of formulas and graphs into text form. "Photoshop CC 2014" processed several 24-megapixel frames with digital camera a pre-prepared script. “WinRAR 5.1” loved archiving so much that this time it packed a folder with data of various formats with a volume of 1.7 GB. “x264” was encoding an AVC file with a source bitrate of about 30 Mbit/s (1080p and 50fps).

Based on the results of measurements server Intel Xeon outperformed Core i7 in only 2 applications. The final position of the E5-4650 cannot be called enviable - a large lag at a significantly higher cost. However, do not forget that Xeon is entrusted with tasks unusual for him. To fulfill the “maximum plan”, we will check the performance of the systems in real games, although the test is in 3DMark already told us a lot.

Note that usually the maximum load in modern games falls on graphics card. In view of this, expect a large separation of one configuration from the arc in high resolutions(1920*1080) is not necessary. Therefore, we performed the comparison in the same gaming applications at a slightly lower resolution - 1280*800 (in fact, this is the same as installing a more powerful card without reducing the resolution). The results are amazing.

Bottom line quite predictable - use server processor in gaming computer illogical. For a much higher cost, the owner buys a system with lower performance. So, Intel Xeon E5-4650 goes to its rightful place - on a multiprocessor server motherboard and continues to “serve the homeland” in its inherent tasks. E5 won only one thing today – in energy consumption under load (utility). Of course, consumption was measured test bench entirely (without monitor). The effect of power supply efficiency (80 Plus Platinum) should be minimal. It is much more logical to spend a slightly smaller amount on a CPU than the cost of a server processor, but buy a TOP-segment video card, for example, a new product for 2017.

Server equipment, unlike home systems, has high reliability and performance. The servers can host tens of thousands of sites, it must work non-stop for years, while the load on home system many times less. To ensure high reliability and computing power of server equipment, developers use various thermal control systems, and cooling is installed on almost every important component (not only on the video card and processor, as on home PCs).

In addition, it is worth knowing that server devices allow you to replace important components without turning off the power, that is, during operation. The system is equipped with a huge number of sensors and testing programs, which, if one or another part fails, will issue a message to the administrator asking for replacement. The replacement itself will occur without data loss or power off, that is, all sites on the server will be accessible all the time.

Cost of components for servers

It is worth considering that servers must be very reliable; commercial damage due to server malfunction and downtime during repairs can reach millions of rubles, so manufacturers make server equipment super-efficient and reliable. Any server component is an order of magnitude more expensive than its counterpart from the world of consumer computer electronics.

For example, they will allow you to easily assemble a high-performance professional station.

Model MBD-X10DRI-LN4+-OE-ATX is equipped with two sockets 2011, with support for Intel Xeon E5-2600 v3 processors (installed on the latest generation supercomputers from the USA and China).

The board supports up to 24 memory slots of the new DDR4 standard and works with a total amount of RAM of 1.5 TB. As you can see, the difference with a top-end home PC, on which you can install a maximum of 4 RAM slots with a capacity of no more than 64 GB, is striking.

In addition, it is worth remembering that it is extremely important for the server fast work with the network, which is why gigabit ones are usually installed.

Unlike workstations, servers do not have many USB connectors installed, usually limiting themselves to one or two ports.

Difference between server motherboards and custom ones

Server motherboards must provide high-speed data reception and transmission; there is no need to have audio connectors, even the monitor output is usually one and test (VGA standard), for the initial installation of the OS. By the way, the servers use either Microsoft systems or various UNIX variants.

Processors from Intel Corporation are usually used for server stations. On this moment the following families are relevant:

• Intel Xeon E7 (socket LGA1567);
• Intel Xeon E5 (socket FCLGA2011-3);
• Intel Xeon E3 (socket FCBGA1440);
• Intel Xeon D (socket FCBGA1667);
• Intel Itanium (socket LGA1248);
• Intel Atom (socket UTFCBGA1380).

Each of these processors has its own parameters, for example, the E7 series has up to 45 cores and a 45 MB cache memory, in addition, it supports up to 2 TB of RAM.

Requirements for servicing server stations

All devices operating as a server are head and shoulders above user equipment in terms of performance and power. However, they also cost a decent amount, because they have to work for a long time without turning off and resting. It is worth knowing that server equipment must be well maintained. There should be a minimum of dust in the room and constant air conditioning should be provided. Thus, internal cooling systems receive already conditioned air and cope more easily with heating the server elements. The power supply of server stations must have several backup lines and in the event of an accident the server should not be turned off, the automation should switch it to uninterruptible source power supply, fed through a backup power supply line.

Server stations require round-the-clock monitoring of the status of components (each critical node has a temperature sensor and its operation) so that the administrator or technical service upon receiving a message about the accident, they immediately replaced the failed component “hot”, without turning off the power to the equipment.

Very few publications write about servers and server hardware. And the main reason is technical complexity - there are many differences here from ordinary consumer hardware, and a limited readership. Such articles are of interest only to administrators and those who make purchasing decisions, and to some enthusiastic readers who are interested in professional-grade hardware. However, server hardware is closer to desktop hardware than you think, and additional knowledge never hurts.

When people think of servers, they think of large computers, heavy boards and outrageous performance, but the reality is often different. Today there are many form factors and a huge amount of hardware and software, so it is difficult to come up with a universal definition of the word “server”.

Although professional and consumer hardware have many similarities, we believe that it is the emphasis on certain functions and qualities that allows us to classify Hardware to a professional level. For example, your home PC should be fast, quiet, upgradable and, of course, reasonably priced. It will work for several years, and will often remain idle for several hours, and the user will have the opportunity to replace the failed hardware or simply remove accumulated dust. There are different requirements for servers: reliability, 24/7 availability, Maintenance without stopping work.

First and most importantly, the server must be reliable. Whether it is a database server, a file server, a web server, or another type of server, it must be very reliable because your business depends on it. Secondly, the server must be always available, that is, hardware and software should be selected in such a way that downtime is minimal. Finally, prompt technical service is very critical in a professional environment. That is, if an administrator needs to perform a task, it must be performed as efficiently as possible without conflicting with the criteria mentioned above. That is why server performance is often a consequence of accounting necessary requirements and long-term strategies, and not the result of some emotional step, as is often the case with gaming PCs.

In our article we will talk about server components and describe the technologies common to servers and consumer PCs, as well as talk about the differences and advantages. Since all professional-grade components are much more expensive than regular ones, we will begin our excursion with this question.

Professional means expensive

If you buy professional components or servers and workstations, you will quickly find that they cost more than regular consumer hardware. And the reason often lies not in some complex technology, but in the specifications of professional components, in their testing and validation. For example, the Core 2 Duo Conroe processor is very close to the Xeon Woodcrest in performance. But the differences lie in the sockets used, the specifications and the systems in which these processors are installed. Server hard disks are specifically designed for continuous operation 24/7, while desktop hard drives are not.

We usually assume that any consumer product is compatible with all others, which is not always the case, but most often. Therefore, you can replace one compatible component with another, and most likely there will be no problems. But this approach is no longer acceptable if you plan to upgrade the server or perform maintenance.

New products for the professional market are being developed with a predictable upgrade path in mind, as manufacturers want these products to work with existing systems, current and future generations of components. AMD and Intel customers regularly receive company roadmaps for their products, which provide a glimpse into the future. Consumers can buy a product with confidence that they will receive support and upgrade capabilities over time.

Warranty and replacement of components is also very important. If a broken desktop HDD Under warranty, any new model, then professional solutions often require exactly the same components. Therefore, the administrator needs to search for the exact same product, while regular users, on the contrary, will be unhappy if they do not receive the latest generation components (which, by the way, is cheaper for most manufacturers).

The magic word for the professional market is validation. When a game-changing product is about to be released, it will be reviewed and tested on popular hardware systems. The validation process ensures that companies can deliver highly complex systems to the enterprise market. Indeed, a business can only be built if the IT platform works flawlessly.

AMD Opteron (Socket 940), Intel Xeon Dempsey and Xeon Woodcrest (Socket 771): popular dual-core server processors.

Of course, you're probably familiar with the Athlon, Celeron, Core 2, and Sempron processor lines, which are desktop processors for home and office computers. But AMD and Intel have products aimed at professional customers: AMD Opteron, Intel Xeon and Itanium. Opteron is built on the AMD64 architecture, like Athlon processors 64 and Sempron, and Xeon is based on Core 2 or Pentium NetBurst architecture, depending on the model.

Professional processors typically have more interfaces—multiple HyperTransports in the Opteron, two independent FSBs (one per processor) in the Intel world—and a richer set of features, which are often required for server applications and workstation software.

There are two different versions of Opteron processors on the market: one uses Socket 940 with DDR memory, the second uses Socket 1207 (Socket F) and DDR2 RAM. As with all AMD64 processors, the memory controller is part of the processor, which is a significant advantage as the number of processors grows: not only do you get more memory controllers to accommodate more memory, but each processor runs its own block of memory. Of course, this raises coherence problems and increases the complexity of multiprocessor systems, but it also throughput the total also turns out to be higher. Opterons for Socket 940 Opterons use PGA packaging, that is, the legs are on the processor. Opteron for Socket 1207 switched to LGA packaging, where the legs are on the socket and flat contacts are on the processor.

These days, dual-core processors should be your choice. Dual-core processors, even with lower clock speeds, are superior to single-core models in the server market. Dual-core Opterons for Socket 940 are built on Egypt and Italy cores; the latter version is more advanced. But today we recommend choosing models for Socket 1207 (Socket F), thanks to support for DDR2 memory and the ability to upgrade to quad-core processors, which will appear sometime this year.

AMD's current 1207-pin Socket F is suitable for current dual-core and future quad-core Opteron processors.

Intel Xeon processors are available in different types, and previous versions used Socket 604. Modern platforms are based on Socket 771, which is an LGA socket. There are different Intel Xeon processors, but we recommend sticking only with dual-core models. In the table http://www.intel.com/products/processor_number/chart/xeon.htm there is full list processors.

Models from 5030 to 5080 are manufactured using a 90nm process technology and are based on the now outdated NetBurst architecture. We recommend Woodcrest-based Xeon processors, with model numbers ranging from 5110 (1.6 GHz) to 5160 (3.0 GHz). They are produced using 65nm technology, require less energy, but provide high performance. The E53xx line is built on quad-core processors Clovertown with frequencies from 1.6 to 2.66 GHz.

Xeon processors do not have an integrated memory controller. Instead, they rely on the motherboard's quad-channel DDR2-667 memory controller. To provide sufficient throughput for dual or quad core processors, modern platform Socket 771 (Blackford) provides two independent FSBs (DIBs), one for each processor.

Intel is the first manufacturer to introduce quad-core processors. Clovertown is assembled from two Woodcrest dual-core dies placed in one package.

Intel Xeon Dempsey (65nm NetBurst), Woodcrest (65nm dual-core Core 2) and Clovertown (65nm quad-core Core 2).

Server memory works on the same principle as normal memory for consumer PCs. The modern standard is DDR2 memory (Double Data Rate SDRAM of the second generation). DDR2 works with more prefetch buffers (4 instead of 2), so the interface frequency can be doubled compared to DDR1.

Compared to consumer memory, professional memory has two different mechanisms designed to preserve data integrity. The register memory contains a small chip, the so-called "register", which is responsible for updating the signal. If the memory of a regular PC cannot consist of more than four (or sometimes six) DIMMs - the signals pass through all memory modules and attenuate, then register memory easily allows the installation of eight modules. In addition to the register, DDR2 memory contains on-chip termination, which prevents signal reflection.

The second mechanism is the ECC error correction code. Instead of storing the standard 64 bits, an ECC DIMM channel adds another memory chip that can store another 8 bits, allowing data to be recovered. Therefore, single-bit errors can be corrected on the fly.

All AMD processors Opterons for Socket 940 require DDR333/DDR400 registered memory, while the Socket F generation (Socket 1207) requires DDR2-667 registered memory.

Fully-Buffered DIMMs (FB-DIMMs) use a so-called buffer component, a high-power chip that converts parallel signals into serial interface. Its main purpose is to connect more than eight memory modules to the controller. With Intel's quad-channel DDR2 memory controller, you can install eight 2GB DIMMs on each of the four channels, if motherboard manufacturers want to support that configuration.

FB-DIMMs are more expensive, run hotter, and are no faster than regular registered memory. Yes, they are most likely the future of servers with large amounts of memory; the same technology is used for current Intel platforms Xeon.

Click on the picture to enlarge.

As an example, we took the server motherboard Asus board P5MT (it is used in servers entry level, since it allows you to use regular processors rather than more expensive server processors). Server motherboards do not support overclocking and are usually equipped with a large number of interfaces, as well as expansion slots with high bandwidth.

The 133 MHz PCI-X bus continues to be the dominant interface for expansion cards. It is built on the parallel PCI bus, which is found in almost every PC today. PCI-X is 64 bits wide, while the PCI bus in your computer is 32 bits. PCI-X 133 supports bandwidth up to 533 MB/s. However, it should be remembered that the PCI-X controller bandwidth is distributed among all connected devices.

The PCI Express (PCIe) interface is more modern. PCI Express is a serial interface that uses multiple lines to connect a device to a controller. Professional expansion cards use PCIe x4 slots (four lanes), but there are also x1, x8 and x16 PCIe cards/slots. PCIe x16 is typically used for high-end graphics cards; graphics workstations carry two full PCIe x16 slots for two graphics cards.

Motherboards for servers and workstations usually contain a built-in network controller. It can be built on the same components found in consumer-grade motherboards, but usually includes more powerful chips that provide, for example, hardware support for TCP/IP computing or other functions to increase performance.

This board is equipped with four DDR2 memory slots, one Socket 775 connector for installing a Pentium 4 or Core 2 processor, one 32-bit PCI slot, one PCI Express x16 slot for a video card or a powerful storage controller, as well as two PCI-X 133 slots. Two Broadcom gigabit Ethernet controllers are responsible for networking capabilities. The motherboard has an ATi graphics processor. It is of course outdated, but it is sufficient to display the desktop or command line, which is what is required for server operating systems.

All other interfaces and components are also found on consumer hardware: south bridge, UltraATA/100 or Serial ATA controllers, voltage regulators, etc. The significant difference, again, lies in the validation process, during which manufacturers test how their products work with others and publish compatibility lists.

The ATi RageXL chip is many years old and doesn't support 3D graphics, but it's good enough for servers. Moreover, most of the time no one looks at the screen there.

A little higher we already mentioned a motherboard with an integrated video card. All server motherboards are equipped with a very simple graphics processor with a small amount of dedicated memory - solutions that take memory from the RAM are not popular here. The successor to RageXL today can be considered the ATi ES1000 graphics processor, which initially worked in the consumer market, but then appeared in servers due to improvements in hardware and drivers. Administrators don't even need to think about installing a special or updated version of the driver: the driver comes with the OS and is certified.

Workstations, on the other hand, require more powerful hardware. ATi is positioning FireGL graphics accelerators based on the Radeon X1000 line for this market. Nvidia offers the Quadro FX line, which is very close to the GeForce 7000 family. The differences between consumer and professional chips can be small, for example in driver optimization. Professional graphics cards provide excellent performance in specialized applications, but they also cost much more.

Hard drives are another interesting aspect regarding servers and workstations. A few years ago, server hard drives used Small Computer System Interface (SCSI) and spindle speeds of 10,000 or 15,000 rpm, which significantly outperformed desktop drives with a speed of 7,200 rpm. Server hard drives are still faster, although the difference is not so great.

Market professional hard drives divided into three segments. The first high-capacity segment uses conventional 3.5" Serial ATA hard drives, validated for 24/7 operation. The performance segment is trying to maximize storage density, which is why we are seeing more and more 2.5" high-performance hard drives appearing on 10 000 rpm with Serial Attached SCSI (SAS) interface. The high-performance segment relies on SCSI or SAS hard drives with a rotation speed of 15,000 rpm.

Server and workstation hard drives typically require active cooling because they are optimized for maximum reliability and performance. All professional hard drives come with a five-year warranty.

Power supplies for the professional sector are specially designed with maximum reliability in mind. Any decent power supply can correct the consequences of one missing phase, but professional solutions can handle more serious failures. Some also provide surge protection, although here we get overlap with the area that lies under the responsibility of uninterruptible power systems (UPS).

Professional power supplies are modular and provide redundancy in the form of two modules, each of which is capable of providing sufficient power to the system. If one power supply fails, the system will continue to operate from the second unit.

Hello Giktimes! Popular belief says that your neighbor’s grass is always greener, and the computers that meticulous entrepreneurs purchase for their needs are more reliable and productive than retail models flavored with marketing. A whole caste of enthusiasts hunts for server components and idolizes the performance of enterprise-class hardware. Let's figure out whether large organizations are really splashing around in the “IT paradise”, or have the geeks created an idol for themselves out of thin air?

There are no barriers to enthusiasts, especially if these barriers are erected by insidious marketers who have divided everything electronic devices for corporate and consumer! Because even in the media with advertisements about the mysterious “user experience,” software and hardware developers say, “The camera of this smartphone provides professional quality pictures!” And in other ways, the cliche about professionals who don’t use nonsense has been exploited for a long time. And if you’re looking for the notorious “professional equipment” and quality of services, then it’s better to ask for enterprise-class hardware and service methods, right?

The motives that guide restless enthusiasts lie on the surface - even though consumer technology is developing more vigorously due to the appetites of buyers, “battle-hardened” corporate-class components will clearly be more reliable, and even cheaper on the secondary market. Somehow geeks play on video cards for workstations and assemble powerful and “eternal” home PCs with server hardware! So, it makes sense to try your luck?

And, of course, there is a bit of this sense in such an undertaking, but with the acquisition of corporate “attributes” for home conditions, you can “get stuck” and, at best, overpay for unclaimed functionality, and at worst, go into the red in comparison with the options available to retail buyers. Let's figure out what the catch is in using hardware designed for corporations.

The server one is also a gaming one. Intel Xeon in home PCs

The first thing technology enthusiasts like to use from the enterprise segment is server processors. Not exotic ones, but the most “understandable” ones, that is, based on the x86 architecture. This pleasure is not cheap, so “Zeonovods”, relatively speaking, include two camps with slightly different guidelines in PC construction:

Xeon - initially not for games and “racing” in benchmarks, but sometimes they are useful

Enthusiasts focused on High-End components. This is a level where large-scale versions of Intel Core i7 are no longer enough, and when looking at the LGA-2011 platform (of any generation), thoughts come to mind that “supercharged” Core i7s offer “the same eggs,” only in smaller quantities and without acceleration.

Because, since we’re talking about price, there have been times in history when eight-core Xeons turned out to be a third cheaper and significantly “cooler” than 6-core Core i7 Extreme Editions. For example, this was the case after the debut of Intel Haswell-E chips in 2014 - firstly, the price difference between the six-core Core i7-5960X and the “civilian” quad-core i7-4790K was a measly 15%. And secondly, the junior eight-core server Xeon E5-2609 v4 cost about 30% less than the candidate from the Haswell-E camp. At the same time, unlike the “just” Core i7, the Xeon has a lower TDP level and does not have graphics integrated into the processor that are useless for enthusiasts.

At the same time, there is tons of L3 cache in all three models, and the frequency, although lower in the Xeon, does not allow the belief that “there is no such thing as a superfluous core” and “very soon games will be optimized so that they run quickly on 8 or more cores.” thrifty lovers of resting speed, after which the hot guys send junior versions of Xeon to Intel chipset X99 and... do not admit to anyone how things are in games.

Therefore, four cores, diluted with the help of Hyper-Threading, are almost always more effective in games than eight low-frequency “pots” in the Xeon, which cannot even be overclocked (locked multiplier, near-zero overclocking on the bus).

"Kulibins", who wanted to modernize the old platform with minimum costs . For example, purchase to replace an old one Core processor 2 Duo is not the old Quad, but a much cooler and high-frequency quad-core Xeon X5460, which, using a simple adapter, can be installed not in a server motherboard with Socket 771, but in a “civilian” one for Socket 775.

The main thing in this scenario is to take care of high-quality cooling (server “stones” sport a TDP of about 120 W instead of 95 W for standard quad-core processors), but in the end this option of upgrading from a very old platform to a “tolerably old” one justifies itself, especially since on some On motherboards the processor can be overclocked up to 4 GHz.

And after all, “Zions” have advantages with which they compensate for their multi-core sluggishness in games! For example, the ability to support multiprocessor configurations, with which video/music/photo encoding and CAD modeling are much faster than in the top-end Core i7 Extreme. Support for register memory with ECC, for example, allows you to correct errors on the fly, and this comes in handy when uptime is high (it’s a server!). Support for huge amounts of RAM and a huge number of cores will also come in handy when the server needs to process incoming connections as quickly as possible. But all this is almost useless on a home PC.

And it is useful for it - many cores at high frequencies. If these conditions are met, the processor itself is compatible with the LGA 2011 or LGA 2011-3 platforms and is cheaper than “just” Core i7 - there is a point in purchasing it. Otherwise, it is better to either get by with mass-produced quad-core processors with eight threads, or design a workstation for specific use cases (rendering, encoding).

High-frequency Intel Xeons (if they are cheaper than mainstream CPUs) can be a good help not only in work, but also in games (source: ferra.ru)

Mow down frags on a workstation with hacked NVIDIA drivers

If using a server processor can be played in spite of, rather than because of, the installed hardware, then graphics, which should be used for video modeling or design, have historically been cool in gaming disciplines. In the confrontation between AMD and NVIDIA, even the scenarios for “misuse” of video accelerators have always been different: “red” gaming video cards were in great demand among miners until recently, and NVIDIA Quadro, historically, was persuaded to retrain as a gaming video card.

Professional NVIDIA video cards Quadro is significantly more productive than its gaming relatives

Moreover, Quadro is quite suitable for these purposes - the fact is that gaming GeForce is most often a professional video card with partially disabled pipelines GPU(from marketing considerations to chip rejection) at a more affordable price. For example, the new professional video card Quadro P6000 contains the most “complete” version of the GP102 graphics chip and for this reason outperforms the cool gaming GeForce 1080 in performance by almost 20%, and the mighty Titan X based on the same Pascal architecture invariably leaves behind.

In general, among fans of NVIDIA video cards, a proprietary sport has long been formed - using hardware modifications to bring the GeForce closer to the Quadro (for example, the GTX 680 is similar to the Quadro K5000 in terms of performance), while game lovers, on the contrary, cross a hedgehog with a snake, “picking” drivers and make professional video cards work faster in post-shooting games/ride-alongs/adventure games. Such activity does not allow one to “play as intended,” but one can only envy the persistence of the enthusiasts.

In mobile workstations, almost every NVIDIA Quadro video card has a funny pattern: every NVIDIA Quadro mobile video accelerator is equal to gaming GeForce a class lower in gaming tasks and a couple of levels cooler than the gaming GeForce in CAD disciplines.

Performance of mobile NVIDIA Quadro in comparison with GeForce analogues (source: msi.com)

For example, the Quadro M2000M performs at the level of the GeForce GTX 960M in games, but as soon as it comes to simulation, it “jumps” in results to the GeForce GTX 980M. Roughly the same ratio holds true for other Quad models: the M5000M competes with the GTX 980M in games, and the M1000M competes with the 950M in games.

NVIDIA Quadro M6000 compared to the fastest gaming graphics cards
(source: techgage.com

Ice cream for the kids, flowers for the lady: priorities in corporate memory and storage

Server RAM is not compatible with motherboards in home PCs, not because someone decided so “to spite” end customers. It’s just that server RAM is designed a little differently - it contains a register between the chips and the system memory controller in order to reduce the electrical load on the controller and be able to install more modules in one memory channel.

In other words, additional chips and the ability to automatically recognize and correct errors greatly increases the fault tolerance of this type of memory, but also increases its cost. In a word, don’t be surprised if you find that even low-frequency (by the standards of the DDR4 standard) modules will be 50% or more more expensive than their “household” counterparts - the inhuman requirements for endurance in systems that are turned on 24/7 have noticeably modified server RAM. In everyday use, it will be neither faster nor more efficient than its “civilian” counterparts, so for high performance you should turn to gaming kits - for example, HyperX Savage, if you need easy-to-overclock memory for gamers, and HyperX Predator, if you want to get the most out of the subsystem RAM maximum. For standard frequencies, the budget Kingston ValueRAM is great - reliable, install it once and forget it.

A server processor in a home PC can be useful, but instead of register memory it is better to purchase a standard DDR3/DDR4 kit

Enterprise SSD class have also undergone “tuning” in the direction of reliability - they, for example, have the ability to flexibly manage the reserve volume to suit the needs of the controller. The larger the volume, the lower the wear of the cells and the higher the durability of the drive. And a huge number of algorithms that are effective in difficult operating conditions, especially in terms of data safety in case the drive turns off in emergency mode. The firmware has been reconfigured for minimal latency in multi-user access mode and strives for stable performance even with an abnormally large volume of write and read operations. A home computer does not survive such a load, even if you “torture” the SSD with torrents. On the other hand, industrial SSDs are also not record holders in typical operations - typical SATA drives will quickly become obsolete “morally” in terms of memory capacity, rather than completely exhaust the number of rewrite cycles available for the cells - verified by a long-term comparative test involving HyperX models. And speed records with the same level of reliability have long been passed to drives based on the NVMe interface, which are implemented in one of the newfangled form factors “on top” of PCI-Express. In the Kingston/HyperX model line, the “king of the hill” was and remains the Predator SSD PCI-E.

The longevity benefit of purchasing an enterprise-class SSD does not compare to the performance benefits of a PCI-e gaming drive.

If you can’t, but really want to, then you can

Enterprise-class hardware is not so different from its “civilian” counterparts that it is considered unsuitable for use as a home PC, you just always need to proceed from whether the game is worth the candle. Because the situation is as follows:

Buying a platform that uses error correcting error correction (ECC) register memory for your home is a bad idea. Excessive durability does not compensate for expensive components and the average (in comparison with gaming analogues) level of performance will not please, especially since the prices for server memory are noticeably higher than for the average DDR3/DDR4 module.

Enterprise-class drives in home computer needed if you are paranoid, extremely worried about the safety of data in the event of power outages, and are worried about the reliability of modern SSDs in general. Organization-oriented drives will allow you to “maximize” reliability indicators so that your soul can be at peace.

A server processor for games... an interesting and quite effective idea, but only when we are talking about a cheaper (compared to mainstream analogues) and, most importantly, high-frequency model. Or about upgrading an old computer to a server CPU at “little cost,” that is, for almost nothing. And yes, ideally the platform should be borrowed from the “regular” Extreme series of mass-produced processors.

Professional video cards do an excellent job not only with modeling, but also with games. But it should be remembered that in mobile workstations (with a “stifled” TDP) a professional middle-class video accelerator will be able to compete in gaming disciplines only with budget-class gaming video cards. And desktop professional video cards, in turn, although fast in all work scenarios, are exorbitantly expensive, and are certainly not suitable as an economical option for “work and play.”

Be that as it may, you cannot skimp on high-quality and fast RAM... But today you can! We remind you that from February 2 to February 20, all memory kits