Characteristics and labeling of RAM. How to find out which RAM: DDR, DDR2, DDR3 or DDR4 What frequencies do ddr2 have?

I received a question from Alexander Shilin:

Folks, I have this question, but if my mother’s ceiling says 600+, then 667 strips will do? I didn’t see anything with a frequency of 600 at all, I only saw 667 and higher.

To be honest, it was not possible to find a motherboard that supports memory with an operating frequency of no higher than 600 MHz, and RAM with a frequency of 667 MHz has almost disappeared from sale.

But we were able to find motherboards whose specifications stated support for DDR2 667/533/400, but not a word about DDR2 800. One of these boards is ASUS P5LD2 based on the Intel 945P chipset.

The chipset is old, and, most likely, when a computer with such a motherboard was assembled, no more than 1GB of memory was installed in it, or even only 512MB. However, no one has canceled the desire to increase computer performance by increasing the amount of RAM.

Only memory stores with the required characteristics of DDR2 667/533/400 are not available in stores, but only DDR2 800. Is it possible to install it? Will it work?

Can.

To verify this, let’s run the CPU-Z program, which I already praised when I wrote about it. Only this time we’ll open the SPD tab.

Here is an example for DDR2 PC2-5300, 667MHz:

DDR2 PC6400, 800MHz:

And here is the memory, officially labeled as DDR2 PC6400, 800 MHz, but supporting operation at 1066 MHz:

The most interesting line for us in this case is the Frequency line in the Timings Table section. Only the frequency value must be multiplied by 2 to obtain the values indicated in the price lists and manuals for the mat. boards

In general, SPD is a system of profiles hardwired into RAM, which tells the motherboard through the BIOS at what frequency a given stick is capable of operating.

And then it is clear that DDR2 PC2-5300, 667 MHz can operate not only at 667 MHz, but also at 533 MHz, and even 400 MHz.

The same can be said about DDR2 PC6400, 800MHz. The lack of mention in the plate about the possibility of operating at a frequency of 667 MHz is caused, I believe, to save space.

I think that the latest bar will work even at a frequency of 400 MHz. But from an economic point of view, buying in this case is very strange.

So buy DDR2 PC6400, 800MHz and feel free to install it on a motherboard that only supports DDR2 667/533/400. Everything will work great and even more reliably, because... such a bar will have a noticeable margin of safety, instead of working at the limit. 🙂

28 comments

Ilya(29 July 2009, 15:56)

On boards that support only slow memory, you can install fast memory - it will simply work on the maximum supported motherboard. speed board (i.e. low).

(29 July 2009, 16:01)

Ilya, in fact, wrote about this, only in order not to be unfounded, he added several images. 🙂

Anton Molodoy(30 July 2009, 11:30)

>ASUS P5LD2 on Intel 945P chipset.
I have exactly such a mother :)

>when a computer with such a motherboard was assembled, no more than 1GB of memory was installed in it, or even only 512MB.
I'm probably ebony. but I have 3GB. I love it when there is a LOT of memory.

(30 July 2009, 13:40)

Anton, geeks don't count. 🙂
I meant the standard configs that are sold to people.

Igor(27 August 2009, 00:56)

In general, I'm one radish confused in this memory. The laptop supports 533MHz, there was a double bank of 512MB PC4200 running at 266MHz. I installed the PC6400 (800) and thought it would work at 533 MHz. But it is not at all like that - 399 MHz. In short, I “clicked” the screenshots and pasted them here: http://komp-kompyuterov.narod.ru/index.html What’s what? Or is everything right 400x2=800.=)…I guess my enlightenment is later. Then why are they fooling people with eight hundred megahertz?

(27 August 2009, 07:01)

Igor, 800 is, obviously, when the two-channel mode is activated: 2 channels of 400 MHz in total give 800.

In the case of laptops, it's even trickier. This screenshot clearly shows that the maximum frequency (RAM Max support) is 533 MHz. Those. in the case of one bracket - 266 MHz.

But there is no need to be upset. 🙂 2GB is in any case much better than 512MB, and 800MHz is now no more expensive than 533.

Igor(28 August 2009, 09:51)

Well, at least the issue with “grabbing” from the swap has now been resolved. And sometimes it slowed down like a child. :)
Well, in short, I didn’t manage to indulge myself enough in the innovation. A terrible thing happened to the laptop. (I didn’t shed my mother’s blood, but..) By the way, as a result of what happened, when I tried to open an MP3 with Windows, MP writes that “the operation could not be completed due to lack of memory.” Well, isn’t it a mockery? :) And the classic player opens Fine. And there are still a lot of bad things present. Well, this already applies to Windows or security problems. Maybe there is a relevant topic here somewhere? Or are we going off-topic here? Then I’ll write about the problem globally.

(28 August 2009, 09:55)
Igor(30 August 2009, 04:06)

Well, as they say, once such a p... uh... sorting out began. 🙂 First, the total of the 1st episode; objects (folders, shortcuts, etc.) seemed to be nailed down and were not moved by any click, the “insert” context menu stopped working (always inactive), these same errors were not clicked in the error logs to see the description, when logging into accounts there was an empty window without choosing something, in the task manager, the absence of a loved one in the users tab and generally loss of administrator rights, partial or complete xs (message when trying to launch an application on drive D), processes in the task manager instead of +50 remained 30+, periodic reboots with a blue screen (quickly scrolling, you don’t have time to look at what’s written there), later we managed to figure out the error code
Error code 10000050, parameter1 8f640cec, parameter2 00000001, parameter3 805b641a, parameter4 00000000.
Error code 10000050, parameter1 c399ff20, parameter2 00000000, parameter3 bf80dd9b, parameter4 00000000.
something like this, when I try to scan for viruses, there are also reboots (in fact, I tried to fight them for 3 days), messages about a damaged file system in C, and so on and so forth. The main problem was to remove texts with passwords/logins. I was already mentally ready to rewrite it manually, but remembering about the Windows disk, I successfully used the file transfer wizard. (fine-soft ones are not as bad as they really are =))) I don’t remember how it all started, but definitely after that As soon as I started manipulating the memory, I still remember there was something freezing, scandisk and off we go. I tried to restore the system - again an error and a reboot. (now in the pad I write Ctrl+S after each sentence, because the bastard regularly reboots:(). Everything described was running with home editing, the second XP (cut off which was game editing) also almost didn’t start at all, complaining about the broken C. With safe The mode also didn’t work out well. Having pushed around, I brought up the heavy artillery and restored Acronis True Image Home 11.0 sector-by-sector logical C. Everything seemed to work normally (although right now there’s such confusion in my head that I can’t guarantee anything :)) And the second axis started working. I exchanged the memory (goodram) I think maybe the bracket was buggy. I inserted it, everything seemed to be fine in PC Wizard 2008, I even tested it, it showed something like my old 4200. Oh well, I connected to DSL and let’s download new things. The image of Acronis was already in October 2008, albeit with almost all the necessary programs. Well, here I am sitting here, stuffing an iron friend... and bam. Again the old song. There hasn't been a reboot... mother... for a long time. Similar codes, application error log is already corrupted. Something was freezing (again, beyond my memory:), Scandisk was checking something there. True, this time there was no folder on the disk where there is 000 at the end.
So I'm back again after the reboot. :) Some crap wanted to go to the Internet (it's disabled), I banned it in Komodo. Then I went into it to see in more detail what it was, clicked in the log... an error window and a reboot. After the error message savedump.exe and now there is no record of this event. Somehow I don’t even know what to think. Maybe it really is some kind of virus. Maybe some idiot (I can’t hold back anymore) registered in the MBR? Well, Acronis is registered there (recovery at boot). True, launching it with the choice F11 (recovery) 2-3 times yesterday, and even now displays MBR error 2. Maybe there’s something wrong here? In short, I have no strength. I lay it out and go to bed. Tomorrow (today) I’ll restore it again with aronise and see how it develops with the old memory. PS By the way, the day before I fitted the mouse with a double-click button... Maybe there’s something here? =)))))) ZYY I’m stuck, I can’t tear myself away. Overloaded again. And again I got into some kind of small-soft synchronizer. Something like this. ZYYY I couldn’t go into reboot with firelis, I spat and installed my RAM. It seems to last for some minutes. :) That memory was so hot...even though it’s a laptop.

Igor(30 August 2009, 04:09)

How do I feel about banning unique content? :) True, I didn’t make any paragraphs...
The peephole test turned out well. :))

(30 August 2009, 08:33)

Igor, this doesn’t look like a memory anymore, especially considering its replacement.
It looks like:

1. Virus. It would be nice to boot from some Live CD and check “Dr.Web CureIt!”, since it doesn’t need installation.

2. But it looks even more like the death of the hard drive. Again, it’s better to run the check from a Live CD, but as a last resort, you can just try a Windows one. And look for a utility from the HDD manufacturer.

Igor(30 August 2009, 15:49)

3. And it also looks like a poltergeist. :)
In short, it’s memory, Goodram’s RAM. Probably some kind of incompatibility. Now on its native Hyundai Electronics, a real branded Korean, with civil stamping everything works without failures already in the morning. Even from the night - as established. And the other system started without problems - I ran through Perfect World. True, the remaining damage will have to be repaired. For the first time, I returned my memory to a much more dead system, so there was apparently no result.
Tested the system - no failures. The event log remains damaged
day. In Comodo Firewall, everything is also normal in its log. Dawes-
I installed some updates on my computer and after that the following appeared:
fight. msfeedssync.exe is breaking into the network. Firefox using IE
not running at all. Why the hell does it bother checking news feeds?
or whatever. Well, as for the HDD, my health is 88%, but before the crisis it worked fine in my opinion. Maybe he felt bad
when did you install the new memory? In general, I will restore it somehow
OS, I will update all the other hardware and the disk image in Acronis. Then maybe I’ll stick it in with goodram if I don’t give it back before then. And I need to think about what kind of memory to look for, or rather, find at least something that works for my machine. At that point, this is the only one in stock for laptops. And we know and use CureIt, since literally half a month ago I picked up “something” (Neshta) and tried to treat it on two computers. Now I checked with CureIt - everything is clean.
True, he always swears at Giljabi.exe from my lg_swupdate directory. But I think everything is fine here. :)

PS I wonder if there might be a virus in my memory that was already stored before me? (type from the manufacturer) :))

Igor(1 September 2009, 19:30)

Heh, in the chaos, such a detail as the amount of memory was not noticed
in one slot. Now I installed 1GB Kingston and so far everything is ok. And thinking
that it will continue to be ok. Now it’s M1 and M2 and not like in the “PC Wizard 2008 physical memory_2Gb” screenshot. With another M1 Yes
and I remember that I support 2GB, 1GBx2. Those. in two slots.
All that remains is, if necessary, to put another one in the “bottom” and voila - two-channel
Naya. Well, those who came here according to the subject will now know what horrors
may follow after a seemingly routine operation.

Sergey(November 18, 2009, 20:50)

Hello, Vladimir! I'd be glad to hear your advice.
The memory stick is DDR1 3200, 512 MB. What is better, installing another stick with the same characteristics (DDR1 3200, 512 MB) or a 1 GB stick (to get 1.5 GB)? By the way, the motherboard (Foxconn P4M800P7MA-RS2) has 2 slots for DDR1 and two slots for DDR2. Does it make sense to install DDR2?

(18 November 2009, 20:57)

Sergey, it’s better to install another 1GB and get 1.5GB in total.
You most likely will not notice the difference between DDR1 and DDR2, and in most cases it is impossible to install both types of memory at the same time.

Sergey(19 November 2009, 21:14)

Thank you. What is the probability that the new 1GB stick will work with the old 512MB? I heard that strips with the same parameters, plus a double channel, work better with each other.

Igor(24 November 2009, 16:58)

Absolutely nothing prevents them from working together if the mother supports such a number and in such slots. Memory needs to be increased for running applications. There will be no significant difference between 1.5 and 2 GB if, for example, 1 GB is consumed during operation of the most capacious one. The difference will be if it costs 1 GB and when the program is running, 1.5 GB is taken, i.e. “grabbed” from the swap and, accordingly, slows down due to access to the HDD. See: task manager->performance->peak. How much, when your favorite heavy machine is working, is how much RAM is needed. =) A two-channel system gives an increase of less than 10%, if I’m not mistaken, which is not the point, as important as described above. Well, as they say, this is my opinion, although it rules at noob-level users. =)

Sergey(February 25, 2010, 00:57)

Artyom(September 15, 2010, 12:51)

mat. My board is exactly from the series that is mentioned in this article, Asus soket 775 P5LD2 SE. Thank you, Vladimir) I’ll try.

Anton(31 January 2013, 14:05)

Hello, the following question:
The ASUS P5LD2 motherboard in its description says that the maximum RAM memory can be installed with a frequency of 667 MHz, but I bought 2 sticks of 2 GB and a frequency of 800 MHz, installed the computer and I really liked it because Previously there was 1GB OP.
But after that, space on the hard drive began to disappear, namely on drive “C” (Windows XP is installed on it)
Could this be due to a Motherboard limitation?
Or did I catch some kind of virus? because At the moment, Kaspersky without a license is not paid for = does not work.

(January 31, 2013, 2:09 pm)

Anton, is there a lot of space missing?
Windows has a page file, it can sometimes depend on the size of the RAM.
There is a sleep mode, when the entire contents of the RAM are saved to the hard drive - and the system always reserves a volume equal to the amount of memory. You can turn it off and the place will return.

Or maybe it's just some kind of coincidence.

Vas!(19 May 2013, 19:34)

Hello! Can you tell me: the Asus motherboard supports memory up to 800 MHz, now it costs 2 x 512 at 533 speed (pc-4300). Is it possible to expand by adding 1 or 2 GB but 800th memory? Zs-4300 is nowhere to buy. Will this combination of 2x512 MB on 533 and 1 or 2 GB on 800 work??? Thank you.

Opana(23 October 2015, 15:43)

Hello, I have slots for DDR3 and DDR4 on my motherboard, is it possible to add another 8Gb*2 DDR3@2133MHz to the 8Gb*2 DDR4@3200MHz?

Tony(27 March 2017, 16:29)

My question is, will this work in the Toshiba Satelit A 215 laptop? There the frequency will definitely be 667 hertz at the 800 hertz bar, and is there a risk that it won’t start at all? And in general, can you cram more than 4 gigs of RAM there? Or is there 4, the maximum?

Guest(2 July 2018, 10:22)

Ha, P5RD2-VM does not start with 800 memory (officially the ceiling is 667). But she found a crutch - if you stick one 667 and the other 800 together, then everything works.

Vadim(10 October 2018, 12:08)

asrock 945gcm-s does not support 800 MHz memory

Testing high-speed DDR2 modules: is there any point?

When most users hear the word "overclocking" or overclocking, they immediately imagine increasing the processor clock speed. But an equally important factor is the FSB frequency, which can be easily increased without any problems, providing a performance gain equal to several additional MHz on the CPU. However, the benefits of overclocking components are not always obvious, especially in Pentium 4 systems, where the benefits, for example, of high-speed memory are not noticeable every time.

In principle, there is nothing fundamentally bad about using the fastest memory itself. The maximum possible frequencies and associated delays are what distinguish elite modules. In the case of the Athlon 64, this means using DDR400 DIMMs, which support ideal CL2-2-2-5 latencies.

Modern P4 systems use DDR2 RAM. It is capable of operating at higher frequencies than conventional DDR, and latency is gradually improving. Today, the most common memory is DDR2-533 (266 MHz), which is gradually being replaced by 333-MHz modules (DDR2-667). Higher frequencies today are only available through overclocking, although chipset manufacturers are completely immersed in improving their products.

One might assume that the higher potential for “overclocking” DDR2 RAM will translate into a corresponding increase in performance, but, unfortunately, in reality the situation is different. A P4 system with DDR2-533 memory will be only slightly faster than with DDR400. And the transition to DDR2-667 gives less effect than one might expect.

At the same time, a growing number of manufacturers, including A-Data and Corsair, are releasing DDR2-667 modules that can handle low latency and high frequencies. We received modules from both manufacturers and installed them in an "overclocked" P4 system to see what happens at DDR2-1066 frequencies.

Memory "overclocking" is always relative

On an Intel system, the RAM bus always operates at some ratio relative to the FSB frequency. Most modern motherboards provide some flexibility in this regard, allowing you to select more than one ratio. The northbridge of the 945 and 955x chipsets offers four frequency ratios: 1:1, 3:4, 3:5 and 2:1. If we take the base FSB frequency of 200 MHz (FSB800) as a basis, we can get DDR2-400, DDR2-533, DDR2-667 and DDR2-800. The latter option has been possible for quite some time, but unofficially.

If you want to “overclock” the system without increasing the memory frequency, then increase the FSB frequency while simultaneously switching to a lower factor. Of course, you should also make sure that the CPU frequency does not exceed the permissible parameters, since it depends on the FSB frequency. For example, a 3.2 GHz Pentium 4 640 gets its specified frequency on a 200 MHz FSB through a multiplier of 16. If the FSB reaches 240 MHz, the CPU will have to run at 3.84 GHz. Very few processors are capable of handling this frequency.

To get DDR2-1066 memory without overclocking the system, we used a 1:1 ratio (memory bus to FSB), and increased the FSB frequency to 266 MHz. We used the 3.73 GHz Pentium 4 Extreme Edition as the processor.

We chose the 3.73 GHz Pentium 4 Extreme Edition because it runs at a 266 MHz FSB (FSB1066). If the memory bus/FSB frequency ratio is 1:1, the memory will operate in DDR2-1066 mode.

High frequency or low latency?

AData labels its DIMMs as DDR2-800, while Corsair is limited to 675 MHz. In any case, the delays CL3-2-2-8 work.

We decided to test both low and high memory latencies. Our experience with DDR1 memory shows that the choice should often be made in the direction of low latencies. This is precisely the reason why AMD delayed the introduction of the M2 socket and DDR2 memory until CeBIT 2006 - the company's engineers consider the advantages of DDR2 at 800 MHz to be too insignificant to change the system today.

At the same time, memory manufacturers are moving in different directions. AData indicates that its DDR2 DIMMs are capable of operating at 800 MHz. And it must be said that this statement is confirmed in practice. But for such frequencies it is necessary to increase memory delays. Corsair took a different route: the top DDR2 memory modules have a maximum frequency of 675 MHz, but at the same time the optimal latencies CL3-2-2-8 are given. This allows Corsair to achieve higher performance compared to DDR2-800 modules.

More power, less life

Since process limitations do not allow the production of commercially viable 400 MHz chips, the supply voltage must be increased to increase clock speeds. DDR1 modules require a nominal 2.5 V, so overclockers “overclock” them to 3.0 V and higher. But for DDR2 the base frequency is 1.8 V. In principle, 2.0 V for modules is not too high a load, and higher voltage levels are also sometimes set. This topic is hotly discussed on forums today.

Increasing the input voltage increases the memory's tolerance, resulting in higher clock speeds and aggressive latencies. But you have to pay for everything: increasing the voltage reduces the lifespan of memory modules.

Although AData has a strong position in the US market, it originated from Taiwan. AData's product portfolio is similar to that of other manufacturers and includes many types of SDRAM and flash memory.

On the company's website you can find various types of DDR2 modules, up to DDR2-1066, which AData powers from 1.95 V. However, the DIMMs sent to our laboratory were able to achieve DDR2-1066 mode only when the voltage was raised to 2.4 Q. Unlike many other manufacturers, AData products are aimed at extremely high frequencies, which is why the modules are certified for a CAS latency of 5 cycles. While lower latencies may also work, AData does not guarantee them.

We tested the AData modules, and each time we set the delays manually. In the DDR2-1066 class, 1 GB modules turned out to be the fastest, since they supported CL4-5-5-10 latencies. DDR2-800 mode worked with CL4-4-4-8, DDR2-709 with CL4-3-3-8 and DDR2-533 with CL3-3-3-8.

Corsair guarantees the operating frequency of the modules is 675 MHz. We launched the modules in DDR2-1066 mode, but it cannot be called completely stable. Unlike AData, Corsair has chosen the lowest latencies: CL3-2-2-8 for DDR2-667 - the best latencies we've seen. In addition, our tests show that performance at low latencies is often better than at higher clock speeds (and higher latencies). To ensure better compatibility, the SPD-ROM values are set to CL4-4-4-12. That is, the modules will work on all motherboards. If you wish to set larger delays, you must enter them into CMOS yourself.

Corsair modules also work in DDR2-800 mode. Although the manufacturer recommends a voltage of 2.1 V for DDR2-667, which provides CL3-2-2-8 delays, for DDR2-800 we had to raise the voltage to 2.2 V. By raising the voltage to 2.3 V, we were able to get 533 MHz (DDR2-1066), but the resulting level of stability no longer improved with increasing voltage. It should be emphasized that at 333 MHz (DDR2-667), these DIMMs are capable of competing with higher frequency competitors.

We chose Corsair DIMMs for our project mainly due to their low latency. Corsair's results in our charts are labeled with the manufacturer's name, and all other results are AData DIMM specific.

Impressive Corsair memory latencies.

CPU
Single-core CPUs	Intel Pentium 4 Processor 660 (3.6 GHz, 2 MB L2 cache)
Memory
Intel platform (DDR2-667)	2x 512 MB - DDR2-667 (333 MHz) Corsair CM2X512A-5400UL (XMS5400 V1.2) (CL3-2-2-8-1T @ 333 MHz) 2x 256 MB - DDR2-800 (400 MHz) A-DATA M2OEL6F3G3160A1D0Z (CL4-5-5-10 @ 533 MHz)
Motherboard
Intel platform	Gigabyte 8I955X Royal Intel 955X chipset
System hardware
Graphics card (PCIe)	nVidia Geforce 6800 GT (reference board) GPU: nVidia GeForce 6800 GT (350 MHz) Memory: 256 MB DDR-SDRAM (500 MHz)
HDD	Western Digital WD740 Raptor 74 GB, 8 MB cache, 10,000 rpm
Net	3Com 3C905B
DVD-ROM	Gigabyte GO-D1600C (16x)
power unit	Tagan TG480-U01, ATX 2.0, 480 W
Software
Chipset drivers	Intel Inf 7.0.0.1019
Graphics driver	nVidia Forceware 71.84
DirectX	Version: 9.0c (4.09.0000.0904)
OS	Windows XP Professional 5.10.2600, Service Pack 2

Tests and settings

Tests and settings OpenGL
Doom III	Version: 1.0.1262 1280x1024, 32 Bit Video Quality = High Quality demo1 Graphics detail = High Quality
Wolfenstein Enemy Territory	Version: 2.56 (Patch V 1.02) 1280x1024, 32 Bit timedemo 1 / demo demo4 Geometric detail = high Texture detail = high
DirectX 9
FarCry	Version 1.1 Build 1378 1280x1024 - 32 Bit quality options = High
Video
Pinnacle Studio 9 Plus	Version: 9.4.1 from: 352x288 MPEG-2 41 MB to: 720x576 MPEG-2 95 MB Encoding and Transition Rendering to MPEG-2/DVD no Audio
Auto Gordian Knot DivX 5.2.1 XviD 1.0.3	Version: 1.95 Audio = AC3 6ch Custom size = 100 MB Resolution settings = Fixed width Codec = XviD and DivX 5 Audio = CBR MP3, kbps 192 182 MB VOB MPEG2-source
Audio
Lame MP3	Version 3.97.1 Multi-threaded Alpha Wave 17:14 minutes (182 MB) to mp3 32 - 320 kbit VBR = level 3
Applications
WinRAR	Version 3.40 283 MB, 246 Files Compression = Best Dictionary = 4096 kB
3DS Max 7	Characters "Dragon_Charater_rig" 1600x1200 Rendering Single
Synthetic
PCMark 2004 Pro	Version: 1.3.0 CPU and Memory Tests
SiSoftware Sandra Pro	Version 2005, SR1 CPU Test = Multimedia Benchmark Memory Test = Bandwidth Benchmark

Conclusion: The advantage of high memory frequencies is small

Synthetic tests give a good difference between different DDR2 frequencies.

But even if AData and Corsair DIMM frequencies are impressive, the performance results are not so much.

In our opinion, moving from DDR-533 to DDR2-667 only makes sense if you maintain low latency (Corsair). Switching to DDR2-800 gives minimal performance gains, and DDR2-1066, with even higher latencies, is also not impressive. Moreover, the price of high-speed modules does not at all justify the performance increase that they provide.

For business applications, installing high-speed DDR2 DIMMs is not worth it for price reasons, and even for gamers, we recommend that the money be better spent on a high-end graphics card. In any case, we recommend buying branded memory modules, since reputable manufacturers pay more attention to testing and certification of their products.

Theoretical foundations and first results of low-level testing

DDR2 is a new memory standard approved by the Joint Electronic Device Engineering Council, which includes many manufacturers of chips, memory modules, and chipsets. Early versions of the standard were published already in March 2003, it was finally approved only in January 2004 and received the name DDR2 SDRAM SPECIFICATION, JESD79-2, revision A (). DDR2 is based on the well-known and proven DDR (Double Data Rate) technology. You could even say this: “DDR2 begins where DDR ends.” In other words, the first DDR2 will operate at frequencies that are the limit for the current generation of DDR-400 memory (PC3200 standard, clock frequency 200 MHz), and its further variants will significantly exceed it. The first generation of DDR2 memory, already currently produced by such vendors as , and , are its varieties DDR2-400 and DDR2-533, operating at frequencies of 200 MHz and 266 MHz, respectively. Next, a new generation of DDR2-667 and DDR2-800 modules is expected to appear, although it is noted that they are unlikely to appear at all and, moreover, will become widespread even by the end of this year.

To be fair, it is worth noting that DDR2 memory, as such, has appeared quite a long time ago - of course, this refers to memory on video cards. However, this variant of DDR2 (called GDDR2) is actually a special type of memory designed specifically for the video card market and is slightly different from the “desktop” version of DDR2, which is the focus of this review. general information

So, “desktop” DDR2-SDRAM is considered as an evolutionary replacement for the current generation of DDR memory. The principle of its operation is absolutely the same - data transfer (at the memory module level) is carried out via a 64-bit bus on both parts of the clock signal (ascending "edge" and descending "cut"), which provides twice the effective data transfer rate in relation to its frequency. Of course, at the same time, DDR2 implements a number of innovations that make it possible to make a leap to much higher frequencies (and, therefore, greater bandwidth) and larger capacities of chip arrays, on the one hand, and reduced power consumption of modules, on the other. How this is achieved, we will see later, but for now let’s turn to the “macroscopic” facts. DDR2 memory modules are manufactured in a new form factor, in the form of 240-pin DIMM modules, which are electrically incompatible with slots for DDR memory modules (in terms of number of pins, pin spacing and module pinout). Thus, the DDR2 standard is not backward compatible with DDR.

The table below shows the approved naming conventions and specifications for the first three DDR2 standards. It is easy to see that DDR2-400 has the same bandwidth as the current DDR-400 memory type.

The first DDR2 memory modules will be available in 256 MB, 512 MB and 1 GB variants. However, the standard provides for the possibility of building modules of significantly higher capacity, up to 4 GB, which, however, are specialized modules (not compatible with desktop options, at least for the moment). In the future, modules with even greater capacity are expected to appear.

DDR2 chips will be manufactured using FBGA (Fine Ball Grid Array) packaging, which is more compact than the traditional TSOP-II variant, allowing for higher chip capacities in a smaller size and improved electrical and thermal characteristics. This packaging method is already used by some DDR manufacturers as an option, but is recommended for use in terms of the JEDEC standard.

The voltage consumed by DDR2 modules, according to the standard, is 1.8 V, which is significantly less compared to the supply voltage of DDR devices (2.5 V). A quite expected (although not so obvious) consequence of this fact is a reduction in power consumption, which is important for manufacturers of both laptops and large workstations and servers, where the problem of power dissipated by memory modules is far from the least important. DDR2 from the inside

The DDR2 standard includes several important data-related changes to the DDR specification that allow higher frequencies to be achieved at lower power consumption. We will look at how exactly the reduction in power dissipation is achieved while simultaneously increasing the speed of the modules.

Data sampling

The main change in DDR2 is the ability to fetch 4 bits of data per clock cycle (4n-prefetch), as opposed to 2-bit fetch (2n-prefetch) implemented in DDR. Essentially, this means that at each clock cycle of the memory bus, DDR2 transfers 4 bits of information from the logical (internal) banks of the memory chip to the I/O buffers along a single data interface line, while conventional DDR is only able to transfer 2 bits per clock per line . The question quite naturally arises: if this is so, then why is the effective bandwidth of DDR2-400 the same as that of regular DDR-400 (3.2 GB/s), and not double?

To answer this question, first let's look at how regular DDR-400 memory works. In this case, both the memory core and I/O buffers operate at a frequency of 200 MHz, and the “effective” frequency of the external data bus, thanks to DDR technology, is 400 MHz. According to the 2n-prefetch rule, at each memory clock (200 MHz), 2 bits of information enter the I/O buffer along each data interface line. The task of this buffer is multiplexing/demultiplexing (MUX/DEMUX) of the data stream - in simple terms, “distilling” a narrow high-speed stream into a wide low-speed one, and vice versa. Since in a DDR SDRAM memory chip the logic banks have a data bus width between them and the level amplifier that is twice as wide as from the read latches to the external interface, the data buffer includes a 2-1 type multiplexer. In general, since memory chips, unlike modules, can have different data bus widths - usually x4/x8/x16/x32, the use of such a MUX/DEMUX (2-1) scheme implemented in DDR means that the internal the data stream of width X and frequency Y from the array is converted into an external stream of width X/2 and frequency 2Y. This is called peak throughput balancing.

Let us now consider the operating diagram of a DDR2 SDRAM type memory chip device, equal frequency and “equal width” (i.e. the same data bus width) relative to the DDR chip of the DDR-400 memory module. First of all, we note that the width of the external data bus remains exactly the same 1 bit/line, as does its effective frequency (in this example 400 MHz). Actually, this is already enough to answer the question posed above: why the theoretical bandwidth of equal-frequency memory modules such as DDR2 and DDR are equal to each other. Further, it is obvious that the use of a 2-1 multiplexer used in DDR SDRAM is no longer suitable in the case of DDR2 SDRAM, which fetches data according to the 4n-prefetch rule. Instead, a more complex circuit with an additional conversion stage of a 4-1 multiplexer is required. This means that the core output has become four times wider than the external interface of the microcircuit and the same number of times lower in operating frequency. That is, by analogy with the example discussed above, in the general case, the MUX/DEMUX 4-1 circuit converts the internal data stream of width X and transmission frequency Y from the array into an external stream of width X/4 and frequency 4Y.

Since in this case the core of the memory chips is synchronized at a frequency half as high as the external one (100 MHz), while in DDR the synchronization of the internal and external data stream occurs at the same frequency (200 MHz), among the advantages of this approach there is an increase in the percentage of usable data. chips and reduction in energy consumption modules. By the way, this also helps explain why the DDR2 standard assumes the existence of memory modules with an “effective” frequency of 800 MHz, which is twice as high as the current generation of DDR memory. After all, it is precisely this “effective” DDR2 frequency that can be achieved now with DDR-400 memory chips operating at a native frequency of 200 MHz, if data is sampled according to the 4n-prefetch rule according to the scheme discussed above.

Thus, DDR2 means a rejection of the extensive development path of memory chips in the sense of a simple further increase in their frequency, which significantly complicates the production of stably operating memory modules in large quantities. It is being replaced by an intensive development path associated with the expansion of the internal data bus (which is a mandatory and inevitable solution when using more complex multiplexing). We would venture to assume that in the future we can expect the appearance of DDR4 type memory, which fetches not 4, but 8 bits of data from memory chips at once (according to the 8n-prefetch rule, using an 8-1 type multiplexer), and working at a frequency no longer 2, but 4 times lower in relation to the frequency of the I/O buffer :). Actually, there is nothing new in this approach; something similar has already been seen in memory chips like Rambus DRAM. However, it is not difficult to guess that the downside of this development path is the complication of the MUX/DEMUX I/O buffer device, which in the case of DDR2 must serialize four bits of data read in parallel. First of all, this should affect such an important characteristic of memory as its latency, which we will consider below.

On-chip termination

The DDR2 standard also includes a number of other improvements that improve various characteristics of the new type of memory, including electrical ones. One of these innovations is on-chip signal termination. Its essence lies in the fact that to eliminate excess electrical noise (due to signal reflection from the end of the line), resistors are used on the memory bus to load the line not on the motherboard (as was the case with previous generations of memory), but inside the chips themselves. These resistors are deactivated when the chip is in operation and, conversely, are activated as soon as the chip enters the standby state. Because the signal is now attenuated much closer to its source, it eliminates electrical noise within the memory chip when transmitting data.

By the way, in connection with on-chip termination technology, one cannot help but dwell on such a point as... the heat dissipation of the module, which, in general, is primarily designed to actively reduce the new DDR2 standard. Indeed, such a signal termination scheme leads to the emergence of significant static currents inside the memory chips, which leads to their heating. Well, this is true, although we note that the power consumed by the memory subsystem generally, this shouldn’t increase at all (it’s just that the heat is now dissipated elsewhere). The problem here is slightly different, namely, the possibility of increasing the frequency of operation of such devices. It is very likely that this is why the first generation of DDR2 memory is not DDR2-800 modules at all, but only DDR2-400 and DDR2-533, for which the heat dissipation inside the chips still remains at an acceptable level.

Additional delay

Incremental Latency (also known as Lazy CAS) is another enhancement introduced in the DDR2 standard that is designed to minimize instruction scheduler downtime during data transfers from/to memory. To illustrate this (using a reading example), let's first consider reading Bank Interleave data from a DDR2 device with an added latency of zero, which is equivalent to reading from regular DDR memory.

The first step is to open a bank using the ACTIVATE command along with the first address component (the row address), which selects and activates the required bank and row in its array. During the next cycle, information is transferred to the internal data bus and sent to the level amplifier. When the amplified signal level reaches the required value (after a time called the delay between determining the row and column addresses, t RCD (RAS-to-CAS Delay), a read command with auto-precharge (RD_AP) can be issued for execution together with column address to select the exact address of the data to be read from the level amplifier. After issuing the read command, the column selection strobe delay t CL (CAS signal delay, CAS Latency) is executed, during which the data selected from the level amplifier is synchronized and transmitted. to the external pins of the microcircuit. In this case, a situation may arise when the next command (ACTIVATE) cannot be sent for execution, since the execution of other commands has not yet finished. So, in the example under consideration, the activation of the 2nd bank must be delayed. by one clock cycle, since at this moment the read with auto-recharge command (RD_AP) from bank 0 is already executed. Ultimately, this leads to a break in the sequence of data output on the external bus, which reduces the actual memory bandwidth.

To eliminate this situation and increase the efficiency of the command scheduler, DDR2 introduces the concept of an additional (additional) delay, t AL. When t AL is non-zero, the memory device monitors the READ (RD_AP) and WRITE (WR_AP) commands, but delays their execution for a time equal to the value of the additional delay. The differences in the behavior of a DDR2 memory chip with two different t AL values are shown in the figure.

The upper figure describes the operating mode of the DDR2 chip at t AL = 0, which is equivalent to the functioning of the DDR memory chip device; the lower one corresponds to the case t AL = t RCD - 1, standard for DDR2. With this configuration, as can be seen from the figure, the ACTIVATE and READ commands can be executed one after the other. The actual implementation of the READ command will be delayed by the amount of additional delay, i.e. in reality it will be executed at the same moment as in the diagram above.

The following figure shows an example of reading data from a DDR2 chip assuming t RCD = 4 clock cycles, which corresponds to t AL = 3 clock cycles. In this case, by introducing additional latency, ACTIVATE/RD_AP commands can be executed consecutively, in turn allowing data to be issued continuously and maximizing actual memory throughput.

CAS issuance delay

As we saw above, DDR2, in terms of external bus frequency, operates at higher speeds than DDR SDRAM. At the same time, since the new standard does not involve any significant changes in the production technology of the chips themselves, static delays at the DRAM device level should remain more or less constant. The typical intrinsic latency of DDR DRAM devices is 15 ns. For DDR-266 (with a cycle time of 7.5 ns.) this is equivalent to two clock cycles, and for DDR2-533 (with a cycle time of 3.75 ns.) four.

As memory frequencies further increase, it is necessary to multiply the number of supported delay values for the CAS signal output (towards b O higher values). The CAS delay values defined by the DDR2 standard are presented in the table. They are in the range of integers from 3 to 5 measures; the use of fractional delays (multiples of 0.5) is not allowed in the new standard.

DRAM device latencies are expressed by cycle dimension (t CK), i.e. are equal to the product of the cycle time by the selected CAS delay value (t CL). Typical latency values for DDR2 devices fall in the range of 12-20 ns, based on which the CAS latency value used is selected. Use b O Larger delay values are impractical for reasons of performance of the memory subsystem, and smaller ones due to the need for stable operation of the memory device.

Recording delay

The DDR2 standard also makes changes to the write latency specification (WRITE commands). The differences in the behavior of the write command in DDR and DDR2 devices are shown in the figure.

DDR SDRAM has a write latency of 1 clock cycle. This means that the DRAM device begins to “capture” information on the data bus on average one clock cycle after the WRITE command is received. However, given the increased speed of DDR2 devices, this period of time is too short for the DRAM device (namely, its I/O buffer) to successfully prepare to “capture” the data. In this regard, the DDR2 standard defines write latency as the CAS issuance delay minus 1 clock cycle (t WL = t CL - 1). It is noted that linking the WRITE delay to the CAS delay not only allows you to achieve higher frequencies, but also simplifies the synchronization of read and write commands (setting up Read-to-Write timings).

Recovery after recording

The procedure for writing to SDRAM memory is similar to the reading operation with the difference in the additional interval t WR, which characterizes the recovery period of the interface after the operation (usually a two-cycle delay between the end of data output to the bus and the initiation of a new cycle). This time interval, measured from the moment the write operation ends until the moment it enters the regeneration stage (Auto Precharge), ensures the restoration of the interface after the write operation and guarantees the correctness of its execution. Note that the DDR2 standard does not change the write recovery period specification.

Thus, the latencies of DDR2 devices in general can be considered one of the few characteristics in which the new standard is inferior to the DDR specifications. In this regard, it is quite obvious that using equal-frequency DDR2 is unlikely to have any advantages in terms of speed compared to DDR. How this really is, as always, will be shown by the results of the relevant tests. Test results in RightMark Memory Analyzer

Well, now is the time to move on to the test results obtained in the test package version 3.1. Let us recall that the main advantages of this test in relation to other available memory tests are its wide functionality, openness of the methodology (the test is available to everyone for review in the form) and carefully developed documentation.

Test bench and software configurations

Test bench No. 1

Processor: Intel Pentium 4 3.4 GHz (Prescott core, Socket 478, FSB 800/HT, 1 MB L2) at 2.8 GHz
Motherboard: ASUS P4C800 Deluxe on Intel 875P chipset
Memory: 2x512 MB PC3200 DDR SDRAM DIMM TwinMOS (timings 2.5-3-3-6)

Test bench No. 2

Processor: Intel Pentium 4 3.4 GHz (Prescott core, Socket 775, FSB 800/HT, 1 MB L2) at 2.8 GHz
Motherboard: Intel D915PCY based on Intel 915 chipset
Memory: 2x512 MB PC2-4300 DDR2 SDRAM DIMM Samsung (timings 4-4-4-8)

Software

Windows XP Professional SP1
Intel Chipset Installation Utility 5.0.2.1003

Maximum Real Memory Bandwidth

The maximum real memory bandwidth was measured using the subtest Memory Bandwidth, presets Maximal RAM Bandwidth, Software Prefetch, MMX/SSE/SSE2. As the name of the selected presets itself suggests, this series of measurements uses a standard method for optimizing read operations from memory Software Prefetch, the essence of which is to prefetch data that will be later required from RAM in the L2 cache of the processor. To optimize writing to memory, the method of direct data storage (Non-Temporal Store) is used, which avoids cache clogging. The results using the MMX, SSE and SSE2 registers turned out to be almost identical; for example, below is the picture obtained on the Prescott/DDR2 platform using SSE2.

Prescott/DDR2, maximum real bandwidth

Note that there are no significant qualitative differences between DDR and DDR2 on equal-frequency Prescotts in this test. But what’s more interesting is that the quantitative characteristics of DDR-400 and DDR2-533 memory bandwidth turn out to be very close! (see table). And this is despite the fact that DDR2-533 memory has a maximum theoretical memory bandwidth of 8.6 GB/s (in dual-channel mode). Actually, we don’t see anything surprising in the result obtained - after all, the processor bus is still an 800 MHz Quad-Pumped Bus, and its bandwidth is 6.4 GB/s, so it is the limiting factor.

As for the efficiency of write operations in relation to reading, it is easy to see that it remains the same. However, this again looks quite natural, since in this case the write bandwidth limit (2/3 of the read bandwidth) is clearly set by the microarchitectural features of the Prescott processor.

Memory latency

First of all, let's take a little more detail on how and why we measured the “true” memory latency, since measuring it on Pentium 4 platforms is, in fact, a far from trivial task. This is due to the fact that processors of this family, in particular the new Prescott core, are characterized by the presence of a rather “advanced” asynchronous hardware data prefetcher, which makes it very difficult to objectively measure this characteristic of the memory subsystem. Obviously, using sequential memory traversal methods (direct or reverse) to measure its latency is completely unsuitable in this case; the Hardware Prefetch algorithm in this case works with maximum efficiency, “masking” latencies. The use of random memory bypass modes is much more justified, however, truly random memory bypass has another significant drawback. The fact is that such a measurement is performed under conditions of almost 100% D-TLB miss, and this introduces significant additional delays, as we have already written about. Therefore, the only possible option (among the methods implemented in RMMA) is pseudorandom a memory traversal mode in which each subsequent page is loaded linearly (negating D-TLB misses), while traversal within the memory page itself is truly random.

However, the results of our previous measurements showed that even this measurement technique significantly underestimates the latency values. We believe that this is due to another feature of Pentium 4 processors, namely, the ability to “capture” two 64-byte lines from memory into the L2 cache each time it is accessed. To demonstrate this phenomenon, the figure below shows the dependence of the latency of two consecutive accesses to the same memory line on the offset of the second element of the line relative to the first, obtained on the Prescott/DDR2 platform using the test D-Cache Arrival, preset L2 D-Cache Line Size Determination.

Prescott/DDR2, data arrival via L2-RAM bus

It is clear from them (the random bypass curve is the most indicative) that access to the second element of the line is not accompanied by any additional delays up to 60 bytes inclusive (which corresponds to the true size of the L2 cache line, 64 bytes). The 64-124 byte area corresponds to reading data from the next memory line. Since latency values in this area increase only slightly, this means that the subsequent memory line is actually “pumped” into the processor’s L2 cache immediately after the requested one. What can be made of all this? practical conclusion? The most direct: in order to “deceive” this feature of the Hardware Prefetch algorithm, which works in all cases of memory bypass, it is enough to simply bypass the chain with a step equal to the so-called “effective” length of the L2 cache line, which in our case is 128 bytes.

So, let's move directly to the results of latency measurements. For clarity, here are the L2-RAM bus unloading graphs obtained on the Prescott/DDR2 platform.

Prescott/DDR2, memory latency, line length 64 bytes

Prescott/DDR2, memory latency, line length 128 bytes

As in the case of real bandwidth tests, the latency curves on another platform Prescott/DDR look absolutely the same at the qualitative level. Only the quantitative characteristics differ somewhat. It's time to contact them.

* latency in the absence of L2-RAM bus offload

It is easy to see that the latency of DDR2-533 was higher than that of DDR-400. However, there is nothing supernatural here - according to the theoretical foundations of the new DDR2 memory standard presented above, this is exactly how it should be.

The difference in latency between DDR and DDR2 is almost imperceptible with a standard 64-byte memory bypass (3 ns in favor of DDR), when the hardware prefetcher is actively running, however, with a “two-line” (128-byte) chain bypass it becomes much more noticeable. Namely, the minimum latency of DDR2 (55.0 ns) is equal to the maximum latency of DDR; if we compare the minimum and maximum latencies with each other, the difference is approximately 7-9 ns (15-16%) in favor of DDR. At the same time, it must be said that the almost equal values of “average” latency obtained in the absence of L2-RAM bus offload are somewhat surprising, both in the case of a 64-byte bypass (with data prefetch) and a 128-byte bypass (without it). ). Conclusion

The main conclusion that arises based on the results of our first comparative testing of DDR and DDR2 memory can be generally formulated as follows: “DDR2’s time has not yet come.” The main reason is that it is still pointless to fight to increase the theoretical memory bandwidth by increasing the frequency of the external memory bus. After all, the bus of the current generation of processors still operates at a frequency of 800 MHz, which limits the real bandwidth of the memory subsystem at 6.4 GB/s. This means that at present there is no point in installing memory modules with a higher theoretical memory bandwidth, since the currently existing and widely used DDR-400 memory in dual-channel mode fully justifies itself, and in addition has lower latency. By the way, about the latter - an increase in the frequency of the external memory bus is inevitably associated with the need to introduce additional delays, which, in fact, is confirmed by the results of our tests. Thus, we can assume that the use of DDR2 will be justified, at least, not earlier than the moment when the first processors with a bus frequency of 1066 MHz and higher appear, which will overcome the limitation imposed by the processor bus speed on the real bandwidth of the memory subsystem as a whole.

Now, having learned what it is and what and how it serves, many of you are probably thinking about purchasing a more powerful and productive RAM for your computer. After all, increasing computer performance with additional memory RAM is the simplest and cheapest (unlike, for example, a video card) method of upgrading your pet.

And... Here you are standing at the display case with packages of RAMs. There are many of them and they are all different. Questions arise: Which RAM should I choose?How to choose the right RAM and not make a mistake?What if I buy a RAM and then it doesn’t work? These are completely reasonable questions. In this article I will try to answer all these questions. As you already understand, this article will take its rightful place in the series of articles in which I wrote about how to choose the right individual computer components, i.e. iron. If you haven't forgotten, it included the following articles:
—
—
—
This cycle will continue, and in the end you will be able to assemble for yourself a super computer that is perfect in every sense 🙂 (if finances allow, of course :))
In the meantime learning how to choose the right RAM for your computer.
Go!

RAM and its main characteristics.

When choosing RAM for your computer, you must take into account your motherboard and processor because RAM modules are installed on the motherboard and it also supports certain types of RAM. This creates a relationship between the motherboard, processor and RAM.

Find out about what RAM does your motherboard and processor support? You can go to the manufacturer’s website, where you need to find the model of your motherboard, as well as find out which processors and RAM it supports. If you don’t do this, it will turn out that you bought a super modern RAM, but it is not compatible with your motherboard and will gather dust somewhere in your closet. Now let's move directly to the main technical characteristics of RAM, which will serve as unique criteria when choosing RAM. These include:

Here I have listed the main characteristics of RAM that you should pay attention to first when purchasing it. Now we will reveal each of them in turn.

Type of RAM.

Today, the most preferred type of memory in the world is memory modules DDR(double data rate). They differ in release time and, of course, technical parameters.

DDR or DDR SDRAM(translated from English: Double Data Rate Synchronous Dynamic Random Access Memory - synchronous dynamic memory with random access and double data transfer rate). Modules of this type have 184 contacts on the strip, are powered by a voltage of 2.5 V and have a clock frequency of up to 400 megahertz. This type of RAM is already obsolete and is used only in old motherboards.
DDR2- a type of memory that is widespread at this time. It has 240 contacts on the printed circuit board (120 on each side). Consumption, unlike DDR1, is reduced to 1.8 V. The clock frequency ranges from 400 MHz to 800 MHz.
DDR3- the leader in performance at the time of writing this article. It is no less common than DDR2 and consumes 30-40% less voltage compared to its predecessor (1.5 V). Has a clock frequency of up to 1800 MHz.
DDR4- a new, super modern type of RAM, ahead of its counterparts both in performance (clock frequency) and voltage consumption (and therefore characterized by lower heat generation). Support for frequencies from 2133 to 4266 MHz is announced. At the moment, these modules have not yet entered mass production (they promise to release them into mass production in mid-2012). Officially, fourth generation modules operating in DDR4-2133 at a voltage of 1.2 V were presented at CES by Samsung on January 4, 2011.

Amount of RAM.

I won’t write much about memory capacity. Let me just say that it is in this case that size matters :)
Just a few years ago, RAM of 256-512 MB satisfied all the needs of even cool gaming computers. Currently, for normal functioning of the Windows 7 operating system alone, 1 GB of memory is required, not to mention applications and games. There will never be too much RAM, but I’ll tell you a secret that 32-bit Windows uses only 3.25 GB of RAM, even if you install all 8 GB of RAM. You can read more about this.

Dimensions of the planks or the so-called Form factor.

Form - factor- these are the standard sizes of RAM modules, the type of design of the RAM strips themselves.
DIMM(Dual InLine Memory Module - a double-sided type of module with contacts on both sides) - mainly intended for desktop desktop computers, and SO-DIMM used in laptops.

Clock frequency.

This is a fairly important technical parameter of RAM. But the motherboard also has a clock frequency, and it is important to know the operating bus frequency of this board, since if you bought, for example, a RAM module DDR3-1800, and the motherboard slot (connector) supports the maximum clock frequency DDR3-1600, then the RAM module as a result will operate at a clock frequency of 1600 MHz. In this case, all sorts of failures, errors in the operation of the system, etc. are possible.

Note: Memory bus frequency and processor frequency are completely different concepts.

From the tables above, you can understand that the bus frequency, multiplied by 2, gives the effective memory frequency (indicated in the “chip” column), i.e. gives us the data transfer speed. The name tells us the same thing. DDR(Double Data Rate) - which means double the data transfer rate.
For clarity, I will give an example of decoding in the name of the RAM module - Kingston/PC2-9600/DDR3(DIMM)/2Gb/1200MHz, Where:
—Kingston- manufacturer;
- PC2-9600— name of the module and its capacity;
- DDR3(DIMM)— memory type (form factor in which the module is made);
— 2Gb— module volume;
- 1200MHz— effective frequency, 1200 MHz.

Bandwidth.

Bandwidth- a memory characteristic on which system performance depends. It is expressed as the product of the system bus frequency and the amount of data transferred per clock cycle. Throughput (peak data rate) is a comprehensive measure of capability RAM, it takes into account transmission frequency, bus width and the number of memory channels. The frequency indicates the potential of the memory bus per clock cycle - at a higher frequency, more data can be transferred.
The peak indicator is calculated using the formula: B=f*c, Where:
B is the bandwidth, f is the transmission frequency, c is the bus width. If you use two channels to transmit data, we multiply everything received by 2. To get a figure in bytes/s, you need to divide the result by 8 (since there are 8 bits in 1 byte).
For better performance RAM bus bandwidth And processor bus bandwidth must match. For example, for an Intel core 2 duo E6850 processor with a system bus of 1333 MHz and a bandwidth of 10600 Mb/s, you can install two modules with a bandwidth of 5300 Mb/s each (PC2-5300), in total they will have the system bus bandwidth (FSB) equal to 10600 Mb/s.
Bus frequency and bandwidth are denoted as follows: “ DDR2-XXXX" And " PC2-YYYY". Here "XXXX" denotes the effective memory frequency, and "YYYY" the peak bandwidth.

Timings (latency).

Timings (or latency)- these are time delays of the signal, which, in the technical characteristics of the RAM, are written in the form “ 2-2-2 " or " 3-3-3 " etc. Each number here expresses a parameter. In order it is always " CAS Latency"(working cycle time), " RAS to CAS Delay"(full access time) and " RAS Precharge Time» (pre-charge time).

Note

So that you can better understand the concept of timings, imagine a book, it will be our RAM that we access. Information (data) in a book (RAM) is distributed among chapters, and chapters consist of pages, which in turn contain tables with cells (like in Excel tables, for example). Each cell with data on the page has its own vertical (columns) and horizontal (rows) coordinates. To select a row, the RAS (Raw Address Strobe) signal is used, and to read a word (data) from the selected row (i.e., to select a column), the CAS (Column Address Strobe) signal is used. The full reading cycle begins with the opening of the “page” and ends with its closing and recharging, because otherwise the cells will be discharged and the data will be lost. This is what the algorithm for reading data from memory looks like:

the selected "page" is activated by applying the RAS signal;
data from the selected line on the page is transmitted to the amplifier, and a delay is required for data transmission (it is called RAS-to-CAS);
a CAS signal is given to select a (column) word from that row;
data is transferred to the bus (from where it goes to the memory controller), and a delay also occurs (CAS Latency);
the next word comes without delay, since it is contained in the prepared line;
after access to the row is completed, the page is closed, the data is returned to the cells and the page is recharged (the delay is called RAS Precharge).

Each number in the designation indicates how many bus cycles the signal will be delayed. Timings are measured in nanoseconds. The numbers can have values from 2 to 9. But sometimes a fourth one is added to these three parameters (for example: 2-3-3-8), called “ DRAM Cycle Time Tras/Trc” (characterizes the performance of the entire memory chip as a whole).
It happens that sometimes a cunning manufacturer indicates only one value in the RAM characteristics, for example “ CL2"(CAS Latency), the first timing is equal to two clock cycles. But the first parameter does not have to be equal to all timings, and may be less than others, so keep this in mind and do not fall for the manufacturer’s marketing ploy.
An example to illustrate the impact of timings on performance: a system with memory at 100 MHz with 2-2-2 timings has approximately the same performance as the same system at 112 MHz, but with 3-3-3 timings. In other words, depending on latency, the performance difference can be as much as 10%.
So, when choosing, it is better to buy memory with the lowest timings, and if you want to add a module to an already installed one, then the timings of the purchased memory must match the timings of the installed memory.

Memory operating modes.

RAM can operate in several modes, if of course such modes are supported by the motherboard. This single channel, two-channel, three-channel and even four-channel modes. Therefore, when choosing RAM, you should pay attention to this module parameter.
Theoretically, the speed of operation of the memory subsystem in dual-channel mode increases by 2 times, in three-channel mode – by 3 times, respectively, etc., but in practice, in dual-channel mode, the performance increase, unlike single-channel mode, is 10-70%.
Let's take a closer look at the types of modes:

Single channel mode(single-channel or asymmetric) – this mode is activated when only one memory module is installed in the system or all modules differ from each other in memory capacity, operating frequency or manufacturer. It doesn’t matter what slots or memory you install into. All memory will run at the speed of the slowest memory installed.
Dual Mode(dual-channel or symmetrical) - the same amount of RAM is installed in each channel (and theoretically the maximum data transfer rate is doubled). In dual-channel mode, memory modules work in pairs: 1st with 3rd and 2nd with 4th.
Triple Mode(three-channel) – the same amount of RAM is installed in each of the three channels. Modules are selected according to speed and volume. To enable this mode, modules must be installed in slots 1, 3 and 5/or 2, 4 and 6. In practice, by the way, this mode is not always more productive than the two-channel one, and sometimes even loses to it in data transfer speed.
Flex Mode(flexible) – allows you to increase the performance of RAM when installing two modules of different sizes, but the same operating frequency. As in dual-channel mode, memory cards are installed in the same connectors of different channels.

Generally, the most common option is dual-channel memory mode.
To operate in multi-channel modes, there are special sets of memory modules - the so-called Kit memory(Kit set) - this set includes two (three) modules, from the same manufacturer, with the same frequency, timings and memory type.
Appearance of KIT kits:
for dual channel mode

for three-channel mode

But the most important thing is that such modules are carefully selected and tested by the manufacturer itself to work in pairs (triples) in two- (three-) channel modes and do not imply any surprises in operation and configuration.

Manufacturer of modules.

Now on the market RAM Such manufacturers as have proven themselves well: Hynix, amsung, Corsair, Kingmax, Transcend, Kingston, OCZ…
Each company has its own for each product marking number, from which, if deciphered correctly, you can find out a lot of useful information about the product. Let's try to decipher the module marking as an example Kingston families ValueRAM(see image):

Explanation:

KVR– Kingston ValueRAM i.e. manufacturer
1066/1333 – operating/effective frequency (Mhz)
D3- memory type (DDR3)
D (Dual) – rank/rank. A dual-rank module is two logical modules wired onto one physical channel and alternately using the same physical channel (needed to achieve the maximum amount of RAM with a limited number of slots)
4 – 4 DRAM memory chips
R – Registered, indicates stable operation without failures or errors for as long a continuous period of time as possible
7 – signal delay (CAS=7)
S– temperature sensor on the module
K2– set (kit) of two modules
4G– the total volume of the kit (both slats) is 4 GB.

Let me give you another example of marking CM2X1024-6400C5:
From the labeling it is clear that this DDR2 module volume 1024 MB standard PC2-6400 and delays CL=5.
Stamps OCZ, Kingston And Corsair recommended for overclocking, i.e. have the potential for overclocking. They will have small timings and a clock frequency reserve, plus they are equipped with radiators, and some even coolers for heat removal, because When overclocking, the amount of heat increases significantly. The price for them will naturally be much higher.
I advise you not to forget about fakes (there are a lot of them on the shelves) and buy RAM modules only in serious stores that will give you a guarantee.

Finally:
That's all. With the help of this article, I think you will no longer be mistaken when choosing RAM for your computer. Now you can choose the right RAM for the system and increase its performance without any problems. Well, for those who will buy RAM (or have already bought it), I will dedicate the following article, in which I will describe in detail how to install RAM correctly into the system. Do not miss…

Best RAM 2019

Corsair Dominator Platinum

The best memory among classmates with high performance and innovation in RGB technology. DDR4 standard, speed 3200MHz, default timings 16.18.18.36, two 16 GB modules. The strips have bright Capellix RGB LED backlights, an advanced iCUE program, and Dominator DHX heat sinks. The only problem is that the height of the module may not be suitable.

Corsair, as always, outdoes itself with each new model, and the Dominator Platinum is no exception. Today it is the favorite DDR4 memory kit for gamers and owners of powerful workstations. The appearance of the modules is sleek and stylish to appeal to gaming enthusiasts, DHX cooling works efficiently, and the performance of the slats is ready to become a legend. In any case, it will provide the user with flagship parameters for many years. Now the memory has a new design, a new, brighter Corsair Capellix backlight with 12 LEDs. iCUE proprietary software provides flexible memory tuning for maximum performance. If you have changed the motherboard or processor, or maybe the graphics accelerator, the memory can be configured as native for any new component.

The price tag of the memory is slightly higher than that of other manufacturers, but this is compensated by the highest quality and amazing performance.

Description

In addition to dividing by throughput and capacity, modules are divided by:

the presence of an additional memory chip for error correction code. They are designated by ECC symbols, for example: PC2-6400 ECC;
the presence of a specialized addressing chip - register.
"Regular" modules are designated as "non-registered" or "unbuffered". The register in buffered - "registered" - modules improves the signal quality of command-address lines (at the cost of an additional clock latency during access), which allows you to raise frequencies and use up to 36 memory chips per module, creating high-capacity modules that are typically used in servers and desktops. stations. Almost all DDR2 Reg modules currently produced are also equipped with ECC.
the presence of an AMB (Advanced Memory Buffer) chip. Such modules are called fully buffered, designated by the letters F or FB, and have a different key location on the module. This is a further development of the idea of registered modules - Advanced Memory Buffer buffers not only address signals, but also data, and uses a serial bus to the memory controller instead of a parallel one. These modules cannot be installed in motherboards designed for other types of memory, and the key position prevents this.

As a rule, even if the motherboard supports registered and unbuffered (regular memory) modules, modules of different types (registered and unbuffered) cannot work together on the same channel. Despite the mechanical compatibility of the connectors, Registered memory simply will not run in a motherboard designed to use regular (unbuffered) memory and vice versa. The presence/absence of ECC does not affect the situation in any way. All this applies to both regular DDR and DDR-II.

It is absolutely impossible to use Registered memory instead of regular memory and vice versa. Without any exceptions. The only exception at present are dual-processor LGA1366 boards, which work with both regular and Registered DDR-III, but you cannot mix two types of memory in one system.

Advantages over DDR

Higher Bandwidth
Generally lower power consumption
Improved design to promote cooling

Disadvantages compared to DDR

Typically higher CAS latency (3 to 6)
The resulting delays at the same (or even higher) frequencies are higher

DDR2 is gradually being replaced by DDR3.

Literature

V. Solomenchuk, P. Solomenchuk PC hardware. - 2008. - ISBN 978-5-94157-711-8

Notes

Links

Types of Dynamic Random Access Memory (DRAM)
Asynchronous	FPM RAM · EDO RAM
Synchronous	SDRAM · DDR SDRAM · Mobile DDR (LPDDR) · DDR2 SDRAM · DDR3 SDRAM · DDR4 SDRAM
Graphic	VRAM · WRAM · MDRAM · SGRAM · GDDR · GDDR2 · GDDR3 · GDDR4 · GDDR5
Rambus	RDRAM · XDR DRAM · XDR2 DRAM

Wikimedia Foundation.