Hello Giktimes! Modernization random access memory- the most basic type of upgrade in a PC, but only as long as you are lucky and do not stumble upon one of the many hardware incompatibilities. We tell you in what cases a set of cool RAM will not “start” on an old PC, why on some platforms you can increase RAM only with the help of “selected” modules, and we warn about other characteristic quirks of hardware.
We know about RAM that there is never too much of it, and that, depending on the age of the computer, you have to choose from very old DDR, old DDR2, mature DDR3 and modern DDR4. At this point, the guide at the level of “well, the main thing is to buy it, and then it will somehow work, or exchange it, if anything” could be completed - it’s time to consider the pleasant and not so specific in the selection of hardware. That is, cases when:
- It should work, but for some reason it doesn't
- the upgrade is not cost-effective or is it better to do it in a multi-step manner
- I want to carry out the modernization with “little blood” in accordance with the potential of the PC
Check where the controller is located
If you're upgrading an outdated computer not just for the "love of the art" but also for practical reasons, it makes sense to first evaluate how viable the hardware platform is before investing in it. The most archaic of the current ones are chipsets for Socket 478 (Pentium IV, Celeron), which extend from platforms with support for SDRAM PC133 (Intel 845 chipset, for example), through mainstream DDR-based options, up to later, strikingly more modern chipsets with DDR2 support PC2-5300 (Intel 945GC, etc.).
Previously, controllers were located outside the processor, but now, as it happens, they work from inside
Against this background, alternatives from the AMD camp of the same time look less colorful: all chipsets for Socket 754, which housed the Athlon 64, representatives of the K8 microarchitecture, support DDR memory, the same type of memory was supported by processors for Socket 939 (Athlon 64 and the first dual-core Athlon 64 X2). Moreover, in the case of AMD chips, the memory controller was built into the processor - now this approach would not surprise anyone, but Intel purposefully kept the controller in the chipset, precisely in order to combine processors for the same socket with new types of RAM.
For this reason, subsequent AMD chips for socket AM2/AM2+ with a RAM controller under the processor cover worked only with DDR2, and Intel with its “long-lived” Socket 775 extended the pleasure with DDR to the very tomatoes of DDR3! In more modern platforms, both processor manufacturers have switched to an on-chip CPU controller, and such tricks with supporting assorted RAM are a thing of the past.
When is it cheaper to change a chipset than to shell out for old memory?
This cumbersome list is not needed to impress readers with the breadth and abundance of chipsets in outdated PCs, but to provide a little unexpected upgrade maneuver. The essence of this simple maneuver is that sometimes it will be more rational to purchase a motherboard with support for cheaper and more modern memory, rather than fork out for the already rare RAM of the previous generation.Because the same amount of DDR2 memory on the secondary market will be at least 50% more expensive than DDR3 memory of comparable capacity. Not to mention that DDR3 has not yet been removed from the assembly line, so it can be purchased in new condition, in an inexpensive kit.
And with new chipsets, it becomes possible to expand RAM to values that are relevant today. For example, if you compare prices in Russian retail, then 8 gigabytes (2x 4 Gb) of DDR2 memory with a frequency of 800 MHz will cost you about 10 thousand rubles, and the same amount of DDR3 memory with a frequency of 1600 MHz (Kingston Value RAM KVR16N11/8, for example) - 3800-4000 rubles. Taking into account sales and purchases motherboard for an old PC the idea looks reasonable.
The realities of upgrading computers with native DDR and DDR2 support have long been known to everyone:
- memory modules with different timings and frequencies most often they manage to work together, and “alignment” occurs either according to the SPD profile in a less powerful module, or (what’s worse) the motherboard chooses a standard profile for working with RAM. As a rule, with the minimum allowable clock frequency.
- the number of modules, ideally, should be equal to the number of channels. Two memory sticks with a capacity of 1 GB each in an old PC will work faster than four modules with a capacity of 512 MB. Fewer modules means lower load on the controller, higher efficiency.
Two channels in the controller - two memory modules for maximum performance. The rest is a compromise between capacity and speed.
- modules of equal volume work more efficiently in dual-channel mode. In other words, 1 GB + 1 GB will be better than 1 GB + 512 MB + 512 MB.
- evaluate platform performance before purchasing memory. Because some chipsets do not reveal the potential of even their “antediluvian” type of RAM. For example, the Intel 945 Express platform is equipped with a dual-channel DDR2 controller supporting frequencies up to 667 MHz. This means that the platform will recognize the DDR2 PC6400 modules you purchased, but the modules will be limited in performance and will work only as PC2-5300, “identical to natural ones.”
The Intel LGA775 socket is one of the options when buying a motherboard with DDR3 support is easier and cheaper than upgrading memory with a platform within old version DDR
And, it seems, this list of nuances is enough to make you want to “drag” an LGA775-based computer to a chipset with DDR3 support. However, you will still laugh, but upgrading an old platform with new RAM also has its own nuances.
In debut platforms with DDR3 support (Intel x4x and x5x chipsets and AMD analogues of the same time), controllers can only work with old-style modules. An absurd situation? Yes, but the fact remains a fact.
The fact is that old systems do not speak the “language of communication” with modules that are equipped with high-density memory chips. At the everyday level, this means that this module, whose 4 gigabytes are “spread” across eight chips on the front side of the printed circuit board, will not be able to work in an old PC. And the old module, in which the same volume is implemented on 16 chips (8 on each side) with a similar volume and frequency, will be operational.
Such compatibility problems are typical, for example, for the desktop Intel G41 Express (the same one that carries a considerable share of the surviving Core 2 Duo or Core 2 Quad) or the mobile Intel HM55 (laptops based on the first generation Intel Core based on the Nehalem microarchitecture).
Sometimes motherboard/laptop manufacturers release new BIOS version in order to teach old platforms to work with new RAM revisions, but most often there is no talk of any long-term support for old equipment. And, unfortunately, there is no talk of any special series of memory for owners of “outdated, but not quite” PCs - memory production has moved forward and turning it back is very expensive.
In order not to bother with such concepts as “memory chip density,” at the household level, owners of old PCs are advised to look for Double-sided DIMM, dual-sided memory modules that are more likely to be compatible with debut DDR3-based platforms. In the Kingston model line, a suitable option would be HyperX Blu KHX1333C9D3B1K2/4G - 4 GB DDR3 module for desktops with sixteen memory modules on board. It's not so easy to find on sale, but if you want 16 GB on an old PC, know how to spin.
And yes, the “best of the archaic” chipsets, such as the Intel P35 Express, for example, are also content with DDR3 support at 1333 instead of the 1600 MHz typical for modern budget platforms.
HyperX Blu KHX1333C9D3B1K2 is one of the few ways to get 16 GB of RAM in older PCs
No diversity - no problem
After a long-term “stronghold of resistance” with the memory controller in the northbridge of Intel platforms, experiments stopped. All new Intel platforms and AMD provided a controller under the cover of the CPU itself. This, of course, is bad from the point of view of the longevity of the platform (you cannot do the trick and “transfer” to new type memory with an old processor), but RAM manufacturers adjusted and, as you can see, DDR3 memory has not lost its popularity even in 2017. Its carriers today are the following platforms:| AMD | Intel |
| am3 | lga1366 |
| am3+ | lga1156 |
| fm1 | lga1155 |
| fm2 | lga1150 |
| fm2+ | lga2011 |
The list of processor architectures based on these platforms is much more extensive! But there is less variety in the choice of memory, or rather almost none. The only exception is AMD processors for socket AM3, which, to the delight of budget-conscious buyers, are compatible with socket AM2, AM2+. Accordingly, the “reds” equipped such processors with a universal controller that supports both DDR2 memory (for AM2+) and DDR3. True, in order to “boost” DDR3 on Socket AM3 to frequencies of 1333 and 1600 MHz, you will have to additionally tinker with the settings.
This is roughly how new computers based on DDR3 and competing memory types compared in the recent past
The principles for selecting memory in the case of DDR3-based platforms are as follows:
- for FM1, FM2 and FM2+, if we are talking about an APU with powerful integrated graphics, you can and should choose the most powerful RAM. Even old chips based on FM1 are able to cope with DDR3 at a frequency of 1866 MHz, and chips based on the Kaveri microarchitecture and its “restyling” Godavari in some cases squeeze out all the juice even from extremely overclocked DDR3 at a frequency of 2544 MHz! And these are not “corn” megahertz, but truly useful in real work scenarios. Therefore, overclocking memory is simply necessary for such computers.
Performance gains in AMD APUs depending on RAM frequency (source: ferra.ru)
It’s worth starting, for example, with modules HyperX HX318C10F - they already work “in the base” at 1866 MHz and CL10, and when overclocked they will come in handy for clock-sensitive AMD hybrid processors.
AMD APUs desperately need high-frequency memory
- "antique" Intel processors on LGA1156 and its server brother LGA1366 platforms capable of riding high-frequency DDR3 only if the multiplier is correctly selected. Intel itself guarantees stable operation exclusively within the “up to 1333 MHz” range. By the way, do not forget that in addition to supporting ECC registered memory, the LGA1366 and LGA2011 server platforms offer three- and four-channel DDR3 controllers. And they remain, perhaps, the only candidates for upgrading RAM to 64 GB, because non-registered memory modules with a capacity of 16 GB are almost never found in nature. But in LGA2011, memory overclocking has become easily possible up to 2400 MHz.
- Almost all processors based on microarchitectures Sandy Bridge and Ivy Bridge(LGA1155) support RAM with frequencies up to 1333 MHz. Raise the clock generator frequency and thus get “easy” overclocking in this generation Intel Core is no longer possible. But models with an unlocked multiplier and the “correct” motherboard capable of going far beyond the notorious 1333 MHz, so for Z-chipsets and processors with the K suffix it makes sense to spend money on modules HyperX Fury HX318C10F - the standard 1866 MHz is “driveable” almost to the maximum values for Bridge processors. It won't seem enough!
- LGA1150, a carrier of chips based on Haswell and Broadwell microarchitectures, became the last of Intel’s “civilian” platforms with support for DDR3, but the methods of interaction with RAM have not changed much since the days of Sandy Bridge and Ivy Bridge. Unless support for mass DDR3 models with a frequency of 1600 MHz has finally come to life. If we talk about overclocking, then the theoretical maximum for processors with unlocked multipliers on overclocking motherboards is 2933 MHz! The maximum is the maximum, but with support for XMP profiles in modern DDR3 modules, achieving high frequencies on aging memory types is no longer difficult.
DDR4 is the fastest, most basic memory to upgrade and purchase
It’s hard to call DDR4 SDRAM memory a new product - after all, Intel processors Skylake, the first mass-produced CPUs with DDR4 on board, came out back in 2015 and managed to get a “restyling” in the form of slightly more optimized and efficient overclocking ones Kaby Lake. And in 2016, AMD demonstrated a platform with DDR4 support. True, it was just a demonstration, because the AM4 socket is intended for AMD “finally serious competition” RyZEN processors, which have just been declassified.
DDR4 is still very young, but in order to unlock the potential of four-channel controllers on the Intel LGA 2011-v3 platform, overclocker memory is already needed
With the choice of memory for supernova platforms, everything is extremely simple - the frequency of mass-produced DDR4 modules starts at 2133 MHz (they are also achievable on DDR3, but “in a jump”), and the volume starts at 4 GB. But buying a “starter” DDR4 configuration today is as short-sighted as being content with DDR3 with a frequency of 800 MHz at the dawn of its appearance.
The memory controller built into processors based on the LGA1151 platform is dual-channel, which means that you need to fit into a couple of modules, the capacity of which is enough for modern games. Today this volume is 16 GB (no, we’re not kidding - with 8 GB of RAM in 2017 you won’t be able to “deny yourself anything”), and as for the clock frequency, DDR4-2400 memory has become the right mainstream.
In server/extreme processors for the LGA 2011-v3 platform, the memory controller is already four-channel, and of all types of RAM, only DDR4-2133 is de jure supported, but overclocking memory based on the Intel X99 chipset with Intel Core i7 Extreme is not easy, but very easy . Well, a computer for maximalists needs memory for maximalists - for example, “the toughest” HyperX Predator DDR4 HX432C16PB3K2 with a clock frequency of 3200 MHz. According to the “go for a walk” principle, the LGA 2011-v3 platform must be equipped with all four modules - only in this case will the four-channel controller be able to realize the full speed potential of the memory subsystem.
In order not to cram the rules and exceptions
What can be added to the nuances of choice described above? A lot of things: specific all-in-one nettops with non-reference design of components, laptops of the same model with completely different potential for upgrades, individual capricious models of motherboards and other “rake” that are easy to stumble upon if you have not followed hardware trends on the forums enthusiasts.In this case, Kingston offers online configurator. With its help, you can select guaranteed compatible and efficient RAM for desktops, workstations, nettops, ultrabooks, servers, tablets and other devices.
There is a reason to check the compatibility of the PC hardware with the memory you are considering buying, so as not to return to the store and explain to consultants that “the memory is functional, but my computer needs DDR3-1600, which is not quite the usual DDR3-1600.”
Don't leave old people to their fate!
Don't you think - upgrading memory is really more troublesome than older computer. This article does not cover all possible difficulties and particulars in choosing memory (it is almost physically impossible, and you would be tired of going through the entire summary of such trifles). But this is not a reason to send still working hardware to the dustbin of history.
You can light up at any age
Because outdated PCs from our overclocking-enthusiast bell towers can still do a good job for less ambitious users or retrain as a home server/media center, and we won’t be performing yet another song to the “immortal” Sandy Bridge, which celebrated its sixth anniversary and is still good. I wish you high performance and fair winds in upgrading your PC!
Fast RAM is good, but fast RAM at a discount is even better! Therefore, do not miss the opportunity to purchase any of the HyperX Savage DDR4 and HyperX Predator DDR4 memory kits with a 10% discount using a promotional code before March 8 DDR4FEB in Yulmart. There is no such thing as too much memory, and even more so with powerful and cool memory for new PC platforms!
For getting additional information about products Kingston And HyperX please visit the company's official website. HyperX will help you choose your kit
High system memory bandwidth and low memory latency have always been relevant. Since the inception of AnandTech - since 1997 - memory has been evolving: the transition from EDO to SDRAM, from PC66 to PC133, from SDR to DDR, and even from VC to DRDRAM. Just using DDR SDRAM increases Athlon performance by 20-30 percent. In addition, we know how important latency is when memory bandwidth is high. The question arises: if processor manufacturers can produce so much powerful processors why can't anyone come up with something for them? effective method retrieving data from memory?
Let's consider the path that data takes before it gets from memory to the processor. When the processor reads from system memory, the first command is sent across the system bus to the chipset's northbridge, which then passes it on to the onboard memory controller. It is in these first steps that pitfalls are hidden. Sometimes (although rarely - after all, the system bus and memory buses are usually synchronized) there is not enough system bus bandwidth. As a result, the speed of reading from memory decreases. Much more often, large delays occur due to inefficient operation of the north bridge and memory controller.
Next, when the memory controller receives a read command, the request is sent to the memory via the memory bus, and after several operations the found data is sent back to the memory controller. The memory controller then receives this data and transmits it to the system bus interface in the northbridge, and then this data goes back to the processor.
As for the second half of this process, it all depends entirely on the type of memory used and the frequency of the memory bus. However, using the chipset and the system bus, you can influence the speed of the first and last few operations.
An intermediate L3 cache could have been used as a way to reduce latency and as a way to increase the channel utilization between the northbridge and the processor, but AMD chose to integrate the memory controller directly into the processor.
Rice. 6. Hammer processor circuit
This not only reduces delays in working with memory (now write/read requests bypass the external north bridge), but also significantly reduces the chances that the chipset will slow down the overall performance of the platform. We've seen plenty of examples of Athlons not achieving peak performance simply because of platforms that don't work as intended. Therefore, nothing better than getting rid of the source of the problems and integrating the memory controller into the processor was invented.
The Hammer architecture addresses the on-chip memory controller (MCT) and the on-chip DRAM controller (DCT). The memory controller is a generic interface between the Hammer core and the DCT controller. This controller understands what memory is in general, but it is in no way tied to the specific type of memory being used. The memory controller is connected to the DCT, which is a more specific device that only works with certain types of memory. Theoretically, AMD could create a Hammer with DDR SDRAM support, and a Hammer with RDRAM support simply by changing the DTC controller, but note that there is very little benefit from using RDRAM for the Hammer. One of the disadvantages of RDRAM is too long delays, which occur quite often. One way to solve this problem is to use RDRAM in conjunction with processors with long pipelines, as in the Pentium 4. It is clear that the Hammer pipeline is not so long, and its clock speed will not be able to compensate for the RDRAM delays, as is done in the Pentium 4. Therefore, AMD's solution staying with DDR SDRAM is quite reasonable.
The first processors based on the Hammer architecture had either a 64-bit or 128-bit DDR SDRAM controller. The DCT controller can support clock frequencies of 100, 133, or 166 MHz for DDR200, DDR266 or DDR333 SDRAM. AMD has made it clear that later versions of the Hammer DCT will replace the DDR controller with a DDR-II controller.
Memory Bandwidth Comparison
| Memory type | 64-bit DCT | 128-bit DCT |
| DDR200 | 1.6GB/s | 3.2GB/s |
| DDR266 | 2.1 GB/s | 4.2GB/s |
| DDR333 | 2.7 GB/s | 5.4GB/s |
The location of the memory controller directly on the chip also means that the speed of memory access directly depends on the clock frequency, since the data already reaches the processor, bypassing the system bus. As an example at the Microprocessor Forum, AMD gave a theoretical 2GHz Hammer with a memory latency of only 12 ns (you can see the Hammer pipeline on the right). Obviously, this does not include the time it takes to read data from memory, but in any case, it turns out to be much faster work via the outer north bridge. So, AMD is going to increase the number of instructions executed per clock by increasing the speed of reading data from memory. As a result, Hammer actuators will be better equipped with data than Athlon actuators.
Rice. 8 Read time
data from memory
So, the built-in memory controller takes over one of the main functions of the external north bridge. AMD went further and practically built the northbridge into the processor die. The only thing left to the traditional external north bridge is the AGP controller. This will virtually eliminate all the performance issues that would have arisen when using Hammer with chipsets of its time, and it will also make motherboard manufacturers happy because it will greatly simplify the layout of the tracks between the memory and the processor.
Below is an example of a single-processor Hammer system.
Rice. 9. Typical “architecture” of AMD Hammer
As you can see, the only chip available on the motherboard (except for the south bridge) is the AGP 8X controller. It communicates with the processor via HyperTransport bus. Probably, in search of a cheap solution, chipset manufacturers will simply create one single chip that will perform all the traditional functions of the south bridge plus the functions of the AGP 8X controller.
In addition, only two memory banks are visible in the image. AMD stated that Hammer-based single-processor systems will support a maximum of 2 unbuffered DIMMs.
Memory controller
Memory controller- digital circuitry that controls the flow of data to and from main memory. May be a separate chip or integrated into a more complex chip such as a northbridge, microprocessor, or system-on-a-chip.
Computers using Intel microprocessors have traditionally had a memory controller built into the chipset (northbridge), but many modern processors such as the DEC/Compaq Alpha 21364, AMD Athlon 64 and Opteron, IBM POWER5, Sun Microsystems UltraSPARC T1, and Intel Core i7 processors have integrated memory controller located on the same chip to reduce memory access latency. Although integration increases system performance, the microprocessor is tied to one type of memory, which does not allow the combination of processors and memory of different generations. To use new types of memory, it is necessary to release new processors and change their socket (for example, after the advent of DDR2 SDRAM, AMD released Athlon 64 processors that used the new Socket AM2 socket).
Integrating a memory controller with a processor is not a new technology; back in the 1990s, the DEC Alpha 21066 and HP PA-7300LC used integrated controllers to reduce system costs.
Tasks
The memory controller contains the logic circuits necessary to perform read and write operations in DRAM, as well as to update data stored in DRAM. Without periodic updates, DRAM memory chips lose information as leakage currents drain the capacitors that store the bits. Typical reliable storage time is a fraction of a second, but not less than 64 milliseconds according to JEDEC standards. Over longer periods of time, information is only partially retained.
Multi-channel memory
Fully buffered FB-DIMM memory
Notes
Wikimedia Foundation. 2010.
- Counter-offensive of the Eastern Front
- Control (values)
See what a “memory controller” is in other dictionaries:
Interrupt controller- (English: Programmable Interrupt Controller, PIC) a chip or built-in processor unit responsible for the ability to sequentially process interrupt requests from different devices. Contents 1 PIC 2 APIC ... Wikipedia
memory access controller- - [E.S. Alekseev, A.A. Myachev. English Russian Dictionary in computer systems engineering. Moscow 1993] Topics information technology in general EN memory access controllerMAC ...
Computer memory cell- The "RAM" request is redirected here. See also other meanings. The simplest scheme for the interaction of RAM with the CPU RAM (also random access memory, RAM) in computer science, memory, part of the computer memory system, into which ... Wikipedia
Programmable Interrupt Controller- Interrupt controller is a microcircuit or built-in processor unit responsible for the ability to sequentially process interrupt requests from different devices. English name Programmable Interrupt Controller (PIC). As a rule... ... Wikipedia
Direct Memory Access- (English Direct Memory Access, DMA) mode of data exchange between devices or between the device and main memory (RAM) without the participation of the Central Processor (CPU). As a result, the transfer speed increases, since the data is not... ... Wikipedia
programmable logic controller- PLC [Intent] controller A control device that performs automatic control through software implementation of control algorithms. [Collection of recommended terms. Issue 107. Management Theory. Academy of Sciences of the USSR. Committee scientifically... ... Technical Translator's Guide
Function controller- Schematic location of the south bridge on system board South Bridge(from the English Southbridge) (functional controller), also known as an I/O hub controller from the English. I/O Controller Hub (ICH). This is a chip that implements ... Wikipedia
USB controller- as part of a personal computer platform, it provides communication with peripheral devices connected to the universal serial bus. A USB controller is an intelligent device capable of interacting with... ... Wikipedia
Programmable logic controller- Massively used programmable logic controller of the SIMATIC S7 300 family Programmable Logic Controller (PLC) or programmable electronic controller ... Wikipedia
professional graphics controller- The controller has 320 KB of memory. Resolution - 640x480 image elements. Ability to display 256 colors from a palette containing more than 16 million shades. Topics information technology in general EN... ... Technical Translator's Guide
Memory
Memory is a device for storing information. It consists of random access and permanent storage devices. The random access memory device is called RAM, read only memory - ROM.
RAM - volatile memory
RAM is designed for recording, reading and storing programs (system and application), initial data, intermediate and final results. Direct access to memory elements. Other name - RAM(Random Access Memory) random access memory. All memory cells are combined into groups of 8 bits (1 byte) and each such group has an address at which it can be accessed. RAM is used for temporary storage of data and programs. When you turn off the computer, the information in RAM is erased. RAM is volatile memory. Modern computers typically have between 512 MB and 4 GB of memory. Modern application programs often require 128–256, or even 512 MB of memory for their execution, otherwise the program simply will not be able to work.
RAM can be built on dynamic chips (Dinamic Random Access Memory - DRAM) or static (Static Random Access Memory - SRAM) type. Static memory has significantly higher performance, but is much more expensive than dynamic memory. For register memory (MPC and cache memory) SRAM is used, and the main memory RAM is built on the basis of DRAM chips.
ROM is non-volatile memory.
In English-language literature, ROM is called Read Only Memory, ROM(read-only memory). Information in ROM is written at the factory of the memory chip manufacturer, and its value cannot be changed in the future. ROM stores information that is independent of the operating system.
The ROM contains:
Program for controlling the operation of the processor itself
Programs for controlling the display, keyboard, printer, external memory
Programs for starting and stopping the computer (BIOS – Base Input / Outout Sysytem)
Device testing programs that check the correct operation of its units every time you turn on the computer (POST -Power On SelfTest)
Information about where on the disk it is located operating system.
CMOS - non-volatile memory
CMOS RAM is non-volatile computer memory. This write-multiple write chip has a high cell density (each cell is 1 byte in size) and low power consumption - it has plenty of power batteries computer. Received its name from the technology of creation based on complementary metal-oxide semiconductors ( complementary metal-oxide semiconductor- CMOS). CMOS RAM is a database for storing PC configuration information. Computer startup program Setup BIOS used to set and store configuration parameters in CMOS RAM. Each time the system boots, the parameters stored in the CMOS RAM chip are read to determine its configuration. Moreover, since some computer startup parameters can be changed, all these variations are stored in CMOS. The BIOS SETUP installation program, when writing, saves its system information in it, which it subsequently reads (when the PC boots). Despite the obvious connection between the BIOS and CMOS RAM, they are completely different components.
Keywords of this lecture
controllers, chipset, ports, USB, COM, LPT, BIOS POST, CMOS, Boot, I/O devices,
(controller- regulator, control device) - a device for controlling various computer devices.
Chipset(chipset)
A set of chips designed to work together to perform a set of functions. Thus, in computers, the chipset located on the motherboard acts as a connecting component that ensures the joint functioning of memory subsystems, central processor(CPU), I/O and others. Motherboard (motherboard, MB, also used name mainboard- main board; slang. Mother, mother, motherboard) is a complex multilayer printed circuit board, on which the main components of a personal computer are installed (central processor, RAM controller and RAM itself, boot ROM, controllers of basic input-output interfaces), chipset, connectors (slots) for connecting additional controllers using USB, PCI and PCI-Express buses.
North Bridge(Northbridge; in selected Intel chipsets, memory controller hub, Memory Controller Hub, MCH) - chipset system controller on the motherboard x86 platform, to which the following are connected as part of the organization of interaction:
via Front Side Bus - microprocessor,
via the memory controller bus - RAM,
via the graphics controller bus - video adapter,
connected via internal bus south bridge.
South Bridge(Southbridge; functional controller; I/O Controller Hub, ICH). Usually this one chip on the motherboard, which through the Northbridge connects “slow” (compared to the CPU-RAM connection) interactions with the central processor (for example, bus connectors for connecting peripheral devices).
AGP(from the English Accelerated Graphics Port, accelerated graphics port) - developed in 1997 by Intel, a specialized 32-bit system bus for a video card.
PCI(English: Peripheral component interconnect, literally - interconnection of peripheral components) - an input/output bus for connecting peripheral devices to the computer motherboard.
Ultra DMA(Direct memory access, Direct memory access). Different versions of ATA are known under the synonyms IDE, EIDE, UDMA, ATAPI; ATA (English: Advanced Technology Attachment) - parallel interface for connecting storage devices (hard drives and optical drives) to the computer. In the 1990s it was standard on the IBM PC platform; is currently being replaced by its successor - SATA and with its advent it received the name PATA (Parallel ATA).
USB(English Universal Serial Bus - “universal serial bus”, pronounced “yu-es-bee” or “oo-es-be”) - a serial data transfer interface for medium- and low-speed peripheral devices in computing. To connect peripheral devices to the USB bus, a four-wire cable is used, with two wires (twisted pair) in a differential connection used to receive and transmit data, and two wires to power the peripheral device. Thanks to built-in lines USB power supply allows you to connect peripherals without its own power supply (the maximum current consumed by the device via the USB bus power lines should not exceed 500 mA).
LPT-port ( standard device printer "LPT1" Line Printer Terminal or Line PrinTer) in operating systems of the MS-DOS family. IEEE 1284 (printer port, parallel port)
COM-port (“com port” Communication port, Serial port, serial port, serial port) is a bidirectional serial interface designed for exchanging bit information. This port is called serial because information is transmitted through it one bit at a time, bit by bit (unlike a parallel port).
PS/2- connector used to connect a keyboard and mouse. It first appeared in 1987 on IBM PS/2 computers and subsequently gained recognition from other manufacturers and became widespread in personal computers and workgroup servers. series personal computers IBM on processors of the Intel 80286 and Intel 80386 series, produced since April 1987. /2 – computer version.
Since the appearance of processors based on the Nehalem core, everyone considered an integrated three-channel memory controller to be one of their advantages. Not just an integrated memory controller (ICM for short), namely three-channel. It’s clear why this is “cool” - after all, AMD had single-channel and dual-channel memory controllers five years earlier, so an additional channel, and even the fastest memory at the moment, such as DDR3, looked like a very serious advantage. According to many users, this is also one of the main factors to which the Core i7 line processors owe their high performance. It is worth noting that she Intel company did not refute this opinion in any way, for which it paid a little - the truly mass-produced processors of the Nehalem architecture, which will be released in early autumn, are designed for the LGA1156 design, which involves the use of only two memory channels. It would seem that this is a serious drawback of the new models, which will not allow them to compete with their older brothers. But is it?
In reviews of motherboards, we have already tried to evaluate the usefulness multi-channel mode memory in LGA1366 processors, and the results turned out to be, to put it mildly, disappointing. For modes, of course, and not for users. However, the tests were carried out on a very limited number of applications, so they did not give a final answer to the question of whether the three-channel mode is needed in practice. Now we have decided to fill this gap. More precisely, at first there was simply a desire to try not a three-channel, but a two-channel mode, for a subsequent more correct comparison of the performance of the Core i7 900 and 800 series: so as not to build hypotheses about what most influenced the results (if they, indeed, will turn out to be significantly different). However, simply “running” the tests from the latest version of our method in yet another configuration is too boring, and such a confrontation of only two versions cannot result in a good article, so we have complicated the task a little.
Test bench configuration
All testing was carried out using Core processor i7 920, Intel DX58SO (“Smackover”) motherboard and a reference video card based on NVIDIA GeForce GTX 275 - in short, everything is as it should be, according to version 4.0 of our testing methodology. Only the memory was different. In addition to the Kingston kit we usually use, we also took a kit from Apacer, which has half the volume. All modules support operation over high frequencies, than the official ones for the Core i7 920 1066 MHz, but we tested them exactly at this frequency according to the 8-8-8-19 scheme.
The result was four configurations presented in the table:
Why them? We need two three-channel ones in order to clearly understand what is important in some application: three-channel or total volume? This will be clearly visible in the results: if both 3×2 and 3×1 are winners, then there is a benefit from three channels, if only the first, then the application simply needs a lot of memory (more precisely, it is able to use it). Without 3x1 it would be difficult to come to a definite answer. The usefulness of participating in 2x2 tests is obvious - this is how they are equipped modern systems on Core 2 and AMD processors, and this is what will become quite popular for some time for systems based on LGA1156 (of course, it would be possible to test the memory in a 2x1 configuration, but this is from the point of view of systems not related to the budget sector, not at all not interested). 1x4 looks extremely synthetic, since it is unlikely that anyone, having two 2 GB memory modules, will install them in one channel, “disregarding” the others, however... We need it to improve general education. Yes, and DDR3 modules with a capacity of 4 GB have already appeared. Unfortunately, this is still an exotic product that has not even reached our hands (otherwise the 2x4 variant would definitely be on the list of those being tested), but the mass distribution on the market of both such modules and kits based on them is only a matter of time.
Detailed results of all subtests, as usual, are presented in the table, in Excel format. Note that in today's testing they will sometimes be even more interesting than the overall average indicators for the groups, so those who are interested in detailed information should not deny themselves the pleasure of getting to know them.
Shooting
But first, we decided to test the performance of each of the options in a synthetic application, which today was Everest 4.6 (yes, this is far from latest version popular test package, however, the “real” software is not updated instantly, so these results are very interesting to us even if we assume that 4.6 is poorly optimized for Nehalem).
And the very first results are somewhat discouraging - as we see, there is no visible increase from the use of the third ICP channel. Moreover, three modules from Apacer cope with this task more slowly than two from Kingston. At the same time, single-channel mode is a clear outsider. The theoretical memory bandwidth of DDR3 1066 is 8528 MB/s, which is what we are stuck with - this is understandable. But adding another channel increases the reading speed not by two, but by less than one and a half times, and the third does not give anything at all.
With the recording speed it's even more fun - the single-channel mode honestly came up against the theoretical bandwidth, and increasing the number of channels gave only less than 20% in all cases.
And finally, access delays. The obvious leader here is the two-channel mode (remember that in this diagram, the smaller the numbers, the better), although single-channel access does not worsen the situation much, but in the three-channel mode the delays increase relatively significantly: by a quarter.
It is already possible to draw certain conclusions. As we remember from the behavior of other architectures with ICP (AMD K8/K10), they are most susceptible to delays when accessing memory, which is very noticeable in real applications. It is unlikely that Nehalem will behave exactly the opposite. Moreover, all this is against the backdrop of identical read and write speeds, that is, the dual-channel mode should become the leader. Single-channel is no longer a fact that it will be too fast: the delays are lower, but the bandwidth is also much lower, and this cannot but have an effect. We'll check how strong it is. And along the way, let’s see how different applications treat different total amounts of memory: synthetic benchmarks cannot provide any information on this matter.
3D visualization
Both three-channel configurations were outsiders, from which we can conclude that the main thing for this group of applications is access delays. But these two options behave differently, and studying the detailed test results shows a rather mixed picture, from which we can conclude that for some applications not only three, but also four gigabytes of memory are no longer enough.
Rendering 3D scenes
Rendering is generally not very sensitive to the characteristics of the memory system, which could have been expected initially - the main thing here is the “number crushing” abilities of the computing cores and their number (and “virtual” computation threads are also perceived positively). Moreover, there are no special requirements for the amount of memory - as long as it is enough for the calculated scene and overhead. For our tests, 3 GB is quite enough, as the diagram above shows us.
Scientific and engineering calculations
And in this group, another class of applications appears, in addition to those who need as much memory as possible and for whom the volume is not important - those who begin to work slower depending on the increase in RAM. At first glance, the situation is inexplicable - if the speed drops due to a lack of memory, it is easy to understand, but simply no one should “notice” the excess. On the other hand, why shouldn’t it? The effectiveness of caching may well depend on the amount of RAM, and even should depend on it. If a particular application uses only a small amount of memory, and a constant one, it will “get” a different amount of processor cache. For example, with six installed gigabytes, only half of the 8 MB L3 cache will be allocated for data from the “foreground” program (do not forget that someone can also “live” in the remaining memory, albeit not very actively, but at the same time use the cache claim), and with three, 2/3 of 8 MB will be servicing them. A curious effect, of course, it’s a pity that it lies somewhat aside from the main topic of our research. With it, everything is as usual - on average, the two-channel mode is the fastest, and of the two three-channel options, despite the presence of the renegade applications mentioned above, the one with the higher total memory capacity is more productive.
Raster graphics
Basically, everything is clear, because among raster editors we encounter all three already defined “groups” of applications. Although with some variations - for example, both Corel products don’t care how much memory and what kind - 3 or 4 GB doesn’t matter, but as long as it’s not 6. But we discovered just a very “memorable” application - Adobe Photoshop. Moreover, what is very interesting here is not the overall result of the subtests, but some of them individually. More precisely, one - Convert. And it is so interesting that we will duplicate in the article the corresponding piece of the table with “raw” data.
| Core 2 Quad Q9300 2×2 | Core i7 920 3×2 | Core i7 920 2×2 | Core i7 920 1×4 | Core i7 920 3×1 |
| 0:09:07 | 0:04:45 | 0:08:05 | 0:08:12 | 0:17:42 |
Conclusion? Despite the fact that most reviews on the Internet that compare processors of different architectures in this application (a minority of reviews simply do not test Photoshop, so you can even say that all articles of this kind) claim that the Core i7 is simply an ideal processor for Photoshop, as we see, there is nothing particularly outstanding in it. What is ideal here is not the kernel architecture, but the amount of memory. At 6GB, the Core i7 920 is twice as fast as the Core 2 Quad Q9300, which comes with just 4GB. These are the comparisons that are found in most articles (including on our website, but other resources behave similarly): 3x2 for processors under LGA1366 and 2x2 for Core2, AMD Phenom and so on. But if we limit the first of the processors to the same 4 GB (and it doesn’t matter how it’s dialed), then it turns out... that the difference from the Core 2 Quad is well within the acceptable range in terms of the difference in clock frequency. And if we “take away” just one more gigabyte of memory from the Core i7 (it would seem 3 or 4: not much difference), then the result will worsen even more doubled! This is the most illustrative example, however, other subtests behave in a similar way, even microscopically, but a difference is always found. And there’s nothing to be done - Photoshop really “loves” memory, and the more “weigh” the files processed in it, the more it “loves” it, and all performance testing utilities in this application(and not just our self-written tests), naturally, they operate with large files.
However, it cannot be said that the high results are not at all due to the Core i7 itself, but only to the preferences of a large amount of memory. The three-channel ICP allows you to install more memory, all other things being equal. But we will talk about this in detail a little later.
Data compression
Archive programs do not know how to use too much memory, so it simply harms them - they are very sensitive to the available cache memory capacity. The main RAM is even more susceptible to delays, which is why we have this picture - the slowest configuration is 3x2, and latency prevents 3x1 from coming out on top.
Compilation (VC++)
The project we are compiling does not require a lot of memory, so latency is important, as is some read and write speed. Therefore, the dual-channel memory access mode turned out to be the best here, but the single-channel mode only slightly outperformed the three-channel ones - the latency is lower, but so are other parameters.
Java
The Java machine test turned out to be very sensitive to the speed of reading from memory, but its total volume is also quite important to it. This is exactly the picture that could be expected everywhere if the naive assumptions were true that three-channel memory access is the key to high performance, but there is never too much memory. The only pity is that among the tested applications these dreams were confirmed literally a couple of times. But just an example when they are confirmed.
Audio encoding
An excellent task - there are, one might say, no requirements for the memory system. During rendering they were also almost absent, but here they are completely absent. An ideal processor benchmark, however, is disgusting for testing the system as a whole.
Video encoding
But here everything is almost as it should be in the “naive theory”. The only thing that spoils the picture is the insufficiently noticeable loss of the two-channel mode. More precisely, it would be almost unnoticeable. And for the fact that it exists at all, we owe it to exactly one application - DivX. An example of good optimization for all the features of today's Core i7. We will check how he behaves in “tomorrows” in less than a month.
Gaming 3D
Very, very calm, slightly unclear overall picture. However, beneath the surface calm, there is a real storm lurking in the detailed results. The preferences of the games are very divided, and we will leave which ones as a task for independent study. The main conclusion is for games (precisely as a set, and not for one specific game) the memory configuration issue is not particularly important. In general, it is even less necessary to decide than the issue of choosing a central processor (of course, if we are not talking about a very budget sector, such as Core 2 Duo or even Pentium/Celeron). The main question facing the “hardcore” gamer today will be: “Will I be able to use a multi-GPU or will I have to somehow limit my desires?”
Why do we need a three-channel ICP at all?
As we can see, there is no great benefit from using the third channel of the memory controller in the Core i7 LGA1366. The channel is there, it can be used, but the results do not always improve. More often than not, they even get worse. So why did Intel make the ICP three-channel? Out of a desire to flex our muscles (a competitor has two, but we’ll do all three)? Perhaps there was such a temptation, but it’s unlikely - after all, three channels come at a fairly high price. And in the literal sense: the layout of boards becomes very complicated, and complicated means expensive. Processors can be made inexpensive (and the Core i7 920 we used today is a clear example of this - its retail price is the same as the Core 2 Quad Q9650), but the platform itself turns out to be a bit expensive. And without any particular benefit - for most “typical user” applications, you can now easily limit yourself to two 2 GB modules and not worry (especially considering the percentage of those still using 32-bit OS, where more RAM simply will not be used). As it was said in a good joke about a baby camel and his mother: “Why do we need these bells and whistles if we still live in a zoo?”
The fact of the matter is that the current Core i7s essentially live in a zoo. The best fit for it will be “real” desktop models designed for the LGA1156 version, the main (and indeed the only) difference from LGA1366 is its support “only” for dual-channel memory mode. And LGA1366 is initially a server platform. Servers require a lot of memory. Not 4, not 8 or even 12 GB, but really a lot. There, even fifty gigabytes can easily turn out to be in demand, or even insufficient. How can you install more memory in one system? The total volume is equal to the product of the number of modules and their volume. Therefore, it is necessary to increase either the number or the capacity of each module. The second is complicated and, generally speaking, does not depend on processor/chipset manufacturers. Moreover, the industry's adoption of denser memory chips has a beneficial effect on all server platform manufacturers at the same time, so it cannot become a competitive advantage.
This means that we need to increase the number of supported modules. And it is equal (in the general case) to the number of memory controllers multiplied by the number of modules supported by each. The latter is the product of the number of supported channels and the number of modules simultaneously working on each channel. Increasing the latter is a very difficult task, since at the same time it is necessary not to deteriorate the speed characteristics, at least. This problem even manifests itself in desktop systems, where more than two or three modules per channel are not used. For example, it could be like this: one module is DDR3 1333, two are DDR3 1066, three are DDR3 800. A lot of slow memory, of course, is sometimes better than a little fast memory, but it is still undesirable to incur such costs. And sometimes it’s impossible.
Intel has been working on the problem of increasing the number of memory modules supported by one controller channel for a long time and not without success. However, it turned out that the final result (FB-DIMM) satisfies the initial requirements, but its use causes a lot of undesirable side effects.
There is only one way left - firstly, move the memory controller to the processor, which in a multiprocessor system automatically provides us with support for several memory controllers. Secondly, increase the number of memory channels. Both were done. Result? A dual-Xeon system, as well as a dual-Opteron system, has two memory controllers. Only in the first both are three-channel, and in the second - two-channel, which gives us six and four memory channels, respectively. When installing two memory modules per channel (a very gentle mode), the first system will have 12 of them, and the second - 8. Let's say each module has a capacity of 4 GB, then the first system will have 48 GB, and the second - 32 GB. In a number of tasks, this will immediately provide the first system with a significant advantage. How can you use the same modules to add up to 48 GB of memory in an Opteron server? It’s easy - we install three modules per channel and... the entire memory system starts to work slower, since, for example, the delays will have to be greatly increased. And it turns out: with the same memory speed, system “i” has one and a half times its volume than system “a”, and with equal volume, system “i” works with memory faster than system “a”.
This is why the Xeon needs a three-channel memory controller. It is also needed in Opteron, but it was not possible to do it at the time. Just like now Intel failed to implement four channels. All the same, both manufacturers should go down this path, since one of them has already tried alternatives (namely FB-DIMM and increasing the number of modules on the channel) and was not very satisfied.
Why is all this in a zoo, on the desktop of an ordinary user? That's right - there's no need. Those who need it will buy a multiprocessor workstation and reduce the task to the previous one. The majority of people somehow didn’t have the desire to install 8 GB in their computers (although this has been available for a long time), so it makes no difference to them - you can install 12 or something. Moreover, now with two modules per channel of a dual-channel memory controller you can get 16 GB, and the question of how much worse/better this is than 24 GB for a normal computer user is akin to the question of how many angels will fit on the tip of a needle.
Total
When looking at the final diagram, a logical question arises - why did we do all this? It’s clear that almost everyone reached the finish line at the same time. The hypothetical single-channel mode showed its relative meaninglessness; the dual-channel mode, as could be expected from tests in synthetics, turned out to be the fastest. A 2% spread between the best and worst cases on such a representative number of applications is a very good result. It shows that, be that as it may, basically our current testing methodology continues to be a processor testing methodology, and other system characteristics have very little influence on the overall final score.
But! It’s too early to rest on this - as we see, the overall score turned out to be an idyll precisely because different applications balance each other out, but they behave completely differently. Some people need a lot of memory, for some, increasing it on the contrary is a hindrance, for some, the volume is not important, but low latency is vital, but DivX, in fact, “disdained” all objectively existing memory parameters and gave preference to a three-channel mode in any form. Therefore, when comparing systems with different configurations memory within the framework of one article (or independently), in specific tests you should not forget to ask how exactly this or that result was obtained. However, we don’t have much time left to tinker with different configurations - LGA1156, let us remind you, supports only two memory channels, so with these processors everything will be simple and logical. We will continue to test devices in the LGA1366 design in a 3x2 configuration, but sometimes we will also remove 2x2 from storage (when it is undesirable to make mental corrections for the features of the memory system). It would be possible to even completely switch to the latter, but there is no point - on average, they are, of course, somewhat faster, but support for three memory channels is an exclusive feature of LGA1366, so let them take the rap for it. We just need to remember that three-channel memory access on this platform does not increase performance at all, but quite the opposite.
Loading...
