entranceDDR2 speed standards. High-speed memory standards DDR2 Real memory bandwidth
Greetings - My dear readers and visitors! 🙂
Today I would like to consider such a question that, judging by the letters coming to me, worries a considerable number of users. Namely - the question of high-speed statuses random access memory DDR2: Are there any differences between them? What will be optimal? and etc…
So, today the market reigns DDR2, there are several basic high-speed standards for this memory. DDR2-533, DDR2-667, DDR2-800, DDR2-1066. There are others, but these are already so-called overclocker standards, the meaning of the use of which is available only with an extreme acceleration of the system.
So, DDR2-533. The youngest high-speed standard. Why 533? 533 is the frequency of the module in MHz. Today, the memory of this standard (in free sale for desktop PCs) is a bit - he leaves the market. Why? Because today we can observe, probably the lowest prices for DDR2 RAM. Many manufacturers, many models, a lot of memory, etc. In general, there are a lot of reasons, but now it's not about it ... DDR2-533 Earlier, it was relevant in the light of his greatest low cost, but today it's a little bit wrong - all the memory of Chief. DDR2-533 was used mainly in office and inexpensive home PCs. Today, the memory of this standard is used in some laptops.
DDR2-667 - middling. In good sense. In 2007, the memory of this standard was the most common in domestic universal desk computers. It works already at a frequency of 667 MHz. This memory today is sold very well and is as a rule as a rule in universal desktop PCs. It is also used in laptops.
DDR2-800. - "Older brother". The memory of this standard is very popular in this, 2008 for installation in universal, home and simply high-performance PC. This was fundamentally influenced by a significant decline in the price of memory of this standard. This memory operates at a frequency of 800 MHz. Also installed in laptops.
DDR2-1066. - high-performance memory. Not much different from 800-ki. The only difference is the best overclocking potential. Accordingly, this memory is very popular with overclockers, which set it in their powerful PCs. When overclocking, it gives even greater performance compared to DDR2-800.
So, brief description Dali. What are the differences? And the differences, to be honest, not so strong ...
If the memory works in normal mode, the PC is not accelerated by the user (or accelerates, but a bit), then as a rule there is a significant difference between the DDR2-533 and DDR2-1066 modules. And the difference between DDR2-667, for example, and DDR2-800 - can be said in everyday work at all noticeable. Why? This is associated with the organization of the memory subsystem and the principles of its work (I will not "ship you", just mention 😉). The same can be said about DDR2-533 against DDR2-667.
As I mentioned, my memory is very cheap today. Previously (in my opinion, even in the early 2007), prices for memory of a different standard differed quite significantly, and it was possible to stand before the choice - which will be optimal. Today the Council is such - since the price of the memory is low, it is better to take, as they say, "Maximum". Excessive memory performance will not hurt (Hmm, if it can be called superfluous at all ...). Therefore, in most cases today, the optimal solution for installation in desktop PCs (home, medium / highly productive) is DDR2-800. On average, installation of 2 GB memory kit (800-ki) will cost today on average, 1500 rubles.
Single amendment: If you are an overclocker, and plan to not slam you to overclock your computer, then it is better to buy a DDR2-1066 memory, it will provide the necessary level of overclocking potential.
P.S. Of course, it would be foolish to choose the RAM only focusing on the high-speed standard! There are still many nuances in the choice of memory, such as timings, etc. ... Therefore, choosing memory, you must definitely take into account all this (all moments and nuances) and seek help from a specialist.
P.P.S. Gradually, the market includes DDR3. What is it? Is it profitable to use this memory? ... I will try to answer all these questions in the pages of this site.
If the overclocking is not from an aesthetic point of view, when you feel a sense of moral satisfaction from the results achieved, and with the practical one, the final result of this process is to obtain greater productivity for less money. As a rule, overclocking lovers do not chase the top components, whether video cards, processors, or RAM. It is enough to competently approach the selection and overclocking of this or that component. system Block And with a certain fraction of luck, the component you purchased will be on the step, or even a little more productive than in the nominal mode. Of course, with sufficient amounts of financial resources, no one bothers to acquire the fastest or already guaranteed overclocked components, but for the fact for most of us it is still a non-dissigant luxury.
Last month, we introduced you to a wide range of memory, manufactured by OCZ, including inexpensive Modules of PC2-4200 and PC2-5300 standards. And before the material, and, especially after its publication, our address received numerous requests to test exactly the budget DDR2 memory available to many. Given this fact, in today's material it will be about checking the overclocker potential of budget modules operational dDR memory2 PC2-4200 standard 1 GB. We all managed to get twelve modules for tests or six pairs DDR2-533 from manufacturers as Patriot, Kingston, Transcend, Twinmos, Hynix and Samsung:
advertising
All DDR2 SDRAM modules 240-pin, without supporting the error correction mode, with a bilateral location of chips in the FBGA package. The rated voltage of the entire memory provided for tests is 1.8 V, the effective frequency is 533 MHz with a theoretical bandwidth of 4.2 GB / s. Two-channel budget kits (and there are also available on sale) in today's testing participation did not take.Before switching to the section with an overview and testing of memory modules, I suggest you familiarize yourself with the test configuration and testing technique.
1. Test configuration, toolkit and testing technique
Memory was tested in two-channel mode On the following system unit configuration:
- Motherboard: ASUSTEK P5B DELUXE / WIFI-AP (Intel P965), LGA 775, BIOS 1004;
- the chipset has cooler Master Blue Ice Pro cooler (~ 4500 RPM, 22 ~ 26 DBA);
- CPU: Intel Core 2 DUO E6400 2133 MHz, 1.325 V, L2 2 x 1024 KB, FSB: 266 MHz X 4, SL9S9 MALAY (CONROE, B2);
- CPU cooling system: Zalman CNPS9700 LED, ~ 1700 RPM;
- Thermal interface: Zalman ZM-STG1;
- Video card: NVIDIA GeForce. 7950 GX2 2 x 512 MB (500/1200 [Email Protected]/ 1580 MHz);
- Disk subsystem: SATA-II 320 GB, Hitachi (HDT725032VLA360), 7200 RPM, 16 MB, NCQ;
- Case: ATX ASUS Ascot 6ar2-B Black & Silver (on blowing, blowing out and on the side wall of the housing are installed 120 mm COOLER MASTER MASTER, ~ 1200 RPM, ~ 21 DBA);
- Power supply: MGE Magnum 500 (500 W) + 80 mm Fan Glacialtech SilentBlade (~ 1700 RPM, 19 DBA).
For cooling of the RAM modules, 80 mm fan of the Austrian company NOCTUA operating on ~ 1800 RPM was installed. Modules were installed pairwise in the first and third board slots.
Well, as they say it went such a thing ... uh ... Timelessness. 🙂 at first the total episode; Objects (folders, labels, etc.) as if nails are nailed and not moved by any click, stopped working "insert" context menu (always inactive), in error logs did not have these mistakes to see the description, when entering accounts Empty window without choosing anything, in the Task Manager, the lack of yourself is a favorite in the tab. With a blue screen (quickly pelled, you do not have time to smoke what is happening there), later it was possible to find 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.
approximately in such a spirit, when trying to scan to viruses, too, reboot (actually with them and tried to fight for 3 days), reports of the Pokhatnaya file System on C, well, etc. and TP. The main problem was texts with passwords / login to remove. I was already morally ready for rewriting into manual, yes, remembering the disk with Windows safely used the file transfer wizard. (Not so badly smallest as they are in fact \u003d))) why it all started already and not mention, but just after As the manipulation of memory started, there was still a freezing of something, Skandisk, and went to go. I tried and restoring the system to make a mistake and reboot. (now I write in Pad and after each Ctrl + S proposal, because I regularly rebuild the Gadan :(). Everything described was held with Hoham Edishin, the second XP (cropped by Geim Edishin) was also almost never launched at all, complaining of C. with safe regime Also, the intelligent nothing went out. Purchased I pulled out heavy artillery and Acronis True. Image Home 11.0 Liked the Logic C. Logic. Everything seemed to work normally (although it shouldn't guarantee anything in my head right now :)) And the second axis earned. Exchanged Memory (GoodRam) I think there was a gossible plank. Inserted, it seems fine Everything in PC Wizard 2008 even tested, something showed how my old 4200. Well, okay, I connected to DSL Yes, let's download new clothes. Robust then Akronis was as much as October 2008, albeit with almost all the necessary programs. Well, here I am sitting on the iron friend here ... and Baz. Again the old song. Reboot ... Mother ... Long type there was no. Similar codes, application error log is already damaged. It was hanging something (again by my memory :), Scandisk picked something there. True, this time on the folder disk it was not the same where such 000 at the end.
So I returned again after the rebut. :) Some kind of crap wanted on the Internet (it is disabled), forbade in the Domato. Then he went into him to see more about what it was, clicked in the journal ... The window of errors and reboot. After the message Savedump.exe error and now there is no record of this event. Something I really do not know what to think. We can really do any virus. Maybe what a stzuka (no longer holding back) in MRR prescribed? Well, there is an Acronis (restoration recovery). True launching it with the choice of F11 (recovery) 2-3 times yesterday was, and even now highlights MBR ERROR 2. Can there be what lies here? In short, everything is no strength. I lay out and go to bed. Tomorrow (today) again I will restore Aroneis and look at how to develop with old memory. Shl to the way and the mouse has adjusted the on the day before with the double click button ... Maybe what is there? \u003d)))))) Zyy shudded, I will not break away. Overloaded again. And again I climbed into there is no synchronizer small-scale. Something like this. I could not go to the flap to go in rebuilding, spit put his RAM. It seems that some minutes holds. :) That memory was such a hot ... although the laptop after all.
So, both chipset (i925x and i915p) support the memory of the DDR2-400 and DDR2-533 standard. This means that with the frequency of the system bus (FSB) equal to 200-mertz, in the first case, the memory operates at 100mgez, in the second - 133mgerz (and due to the features of the DDR2 architecture, this value is consistent and we get 400 and 533 minutes, respectively).
Also note that ASUS has announced the support of the DDR2-600 memory in their motherboardsoh. In particular, P5AD2 Premium (i925x) and P5GD2 (i915p) have the appropriate parameter in the memory frequency setting section. In other words, engineers of the company opened access to the informal frequency divider \u003d 34 (MemcPU).
But in practice, we could not launch the system with this divider.
The first set of memory modules, which appeared in our laboratory, became KingMax DDR2-533. 
To enlarge - click (~ 170kb).
Please note that for the production of these modules, Kingmax uses chips with its own marking.
According to SPD, these modules have the following timings: 3-3-9-3 for frequency 100mgers, and 4-4-12-4 for the frequency of 133mger. At regular frequencies, the modules worked all the tests immaculately. And now we will try to find out what the potential for this memory during acceleration is accelerated. To do this, we collected a stand with ASUS P5GD2 on the I915P chipset and the Pentium4 processor 3.2Ggerz. The maximum possible frequency of FSB, in which the system worked stably appeared 240mger. This result is slightly lower than on the ASUS P5AD2 board (i925x; Max. FSB \u003d 250mgerz), but unfortunately the last BIOS for this board (V 1004) does not make it possible to manually change the memory timings.
As a result, when overclocked with a multiplier 12 (menu item "DDR2-400") The maximum memory frequency is 120mgerz (DDR2-480), and when using a multiplier 23 (menu item "DDR2-533") The frequency is 160mgerz (DDR2-640) .
So, results:
As we can see, Kingmax DDR2-533 memory perfectly worked on minimal timings, up to the maximum possible frequency (specifically for this system). Naturally, with raising timings, the results have not changed. In addition, a significant amount of frequency extension is felt. For example, on the ASUS P5AD2 board, this memory worked at a frequency of 166 (DDR2-667) of the Merz (FSB \u003d 250mger; SPD timings).
How good is this result? Perhaps all other DDR2 modules will demonstrate the same indicators. To clarify this issue, we tested a couple more DDR2-533 modules of an unknown manufacturer (collected on Samsung chips).
samsung chips 
As it turned out, these noname modules showed significantly worse results. In particular, at minimum timings, the maximum possible frequency of memory was equal to 120mgerz (maximum using divider 12). And when using a multiplier 23 and frequency FSB \u003d 200 gger (i.e., the memory frequency \u003d 133mgerz DDR2-533) could only start, but could not load Windows. The situation has somewhat improved the increase in VMEM voltage to 2.0V - Windows booted, and the tests have passed successfully. But increasing the frequency of memory literally to 1mgers led to malfunctions.
The actual potential of both memory sets we defined. But how to use this information in practice? Immediately ask questions: "What mode is most productive: DDR2-400 or DDR2-533?", "How much does performance change when changing timings?".
For answers to these questions, we have spent another series of tests. Results - on the next page.
| Content: | ||
| P.1 - Kingmax DDR2-533 memory P.2 - DDR2 memory optimization on i925xi915P chipsets |
DDR2 memory optimization on i925xi915P chipsets
So, the first question is: "What mode is most productive: DDR2-400 or DDR2-533?". To do this, we have installed a full-time FSB \u003d 200 guts frequency, SPD timings.
| PCMark 2002 (Cpumem) | QUAKE3 (FASTEST; FPS) | WinRAR (Kbatiuti) | ScienceEMark (MBits) | |
| ASUS P5AD2; DDR2-400 (SPD) | 681610558 | 456.1 | 398 | 4089.81 |
| ASUS P5AD2; DDR2-533 (SPD) | 680410472 | 456 | 398 | 4086.32 |
| ASUS P5GD2; DDR2-400 (SPD) | 682510204 | 468.8 | 398 | 4052.54 |
| ASUS P5GD2; DDR2-533 (SPD) | 680810264 | 469.8 | 388 | 4027.34 |
The first two lines of the results are obtained on a system with ASUS P5AD2 board on the I925x chipset and biosomes 1004.
The second two lines - on the P5GD2 board on the I915P chipset with the BIOS version 1001.
As can be seen from the table, the performance of both modes is approximately the same. It turns out that it is not necessary to force the memory to work in DDR2-533. Surely we will get higher performance in DDR2-400 mode and lower timings.
The fact is that the bandwidth of the DDR2-400 fully satisfies the needs of the Pentium4 processor operating on 200 (800QPB) Merce Tire. The bandwidth of the processor tire is 6.4gb, and prop. The ability of the DDR2-400 memory is 6.4gb. And when using DDR2-533, the memory bandwidth increases to 8.5gb, but the pumping speed of the processor tire does not change. As a result, the user does not receive any speed gain.
For clarity, I will give the following example: if we need to fill the capacity of 2 liters from the crane in the kitchen (naturally fully open, we will spend the same time if we use a plastic bottle with a neck with a corresponding crane diameter (DDR2-400), or a mug with large Sorry (DDR2-533). That is, the performance rests on the diameter of the water tap (or processor tire.
Now let's see how the performance depends on the memory timings (FSB \u003d 200HGER; memory frequency divider \u003d 12 or "DDR2-400"):
| PCMark 2002 (Cpumem) | QUAKE3 (FASTEST; FPS) | WinRAR (Kbatiuti) | ScienceEMark (MBits) | |
| 3-3-8-3 | 680010242 | 449,4 | 390 | 4025,2 |
| 3-3-9-3 | 681810225 | 470,3 | 389 | 4106,77 |
| 3-4-8-4 | 680010060 | 459 | 360 | 4081 |
| 3-4-12-4 | 680910067 | 454,4 | 351 | 4079,48 |
| 4-3-8-3 | 680010079 | 446 | 378 | 4023,68 |
| 4-4-8-4 | 68059893 | 452,9 | 350 | 3987,71 |
| 4-4-12-4 | 67829896 | 430 | 343 | 3970 |
| 5-5-15-5 | 67769502 | 413 | 306 | 3829,31 |
Standard timings are highlighted in bold.
In general, the picture is clear: the lower the memory timings, the higher the performance of the system. But there are small subtleties. For example, if we reduce the timings from 3-3-9-3 to 3-3-8-3, then the performance will decrease (and quite noticeable). This suggests, with non-standard timings, there are some violations in synchronizing the transmission of alarm, which leads to a slowdown in work. As a result, setting timings at random, it can seriously slow down the computer.
What we get in the end.
1) For a regular user, you can completely calmly set the memory modules into the DDR2-400 mode (i.e., the memory frequency divider \u003d 12), as well as timings on SPD. As a result, the system will work at least not worse than with DDR2-533 memory. And in most cases, the system will work faster, due to lower timings.
2) for experienced usersDisconnecting your system is also recommended DDR2-400 mode. Due to lower timings - we get higher performance. And at the expense of a higher memory frequency divider, we at a high frequency FSB has a relatively small memory frequency that runs on low timings.
From this point of view, Kingmax DDR-533 memory looks very attractive. The most important thing is that it maintains performance at the frequencies of 160mgers (DDR2-640) and higher, with the minimum possible timings. In this case, the memory modules operate on standard voltage, which means an increase in the potential of acceleration, with an increase in VMEM voltage.
According to the use of DDR i modules, we know that the compatibility of memory is played. various models motherboards. It happens that the same memory shows completely different results on the boards of various manufacturers. There are also cases, the dependence of the memory of memory from different versions BIOS, for the same motherboard.
It is clear that ordinary users This situation relates little - they receive a ready-made computer from the company, and all the headache for the selection of memory falls on the collectors, this company itself. But the overclockers will have to sweat choosing the desired bundle "Motherboard + RAM".
All questions, comments and wishes can be asked in.
Introduction
We continue to explore the speed of the computer's RAM and its effect on the PC performance as a whole. In our previous materials, the "influence of memory volume on computer performance" and "the effect of memory timings on computer performance" we found out that for average modern computer It will be enough to have 512 MB of memory, preferably with minimal timings. But today, many users change their computers to newer. And as a result of the upgrade, it is necessary to establish a new memory - DDR-II. IN ordinary computers This memory can work at 400 or 533 MHz frequencies (without acceleration). With acceleration, you can get 667 MHz. We decided to find out what the difference in memory capacity at 400 and 533 MHz frequencies.
Test platform
BIOS modern motherboards allows you to choose a frequency of 400 or 533 MHz for RAM DDR-2. In the first case, a fission coefficient 1 is used, in which the memory frequency is equal to the FSB frequency. In the second - 4: 3, due to which the frequency is 200: 4/3 \u003d 533 MHz.
This is a modern platform designed to use high performance in computers. It is built on the chipset Intel i925xthat has only DDR-2 memory support, and when using PAT optimization technologies. In this computer, ventilation is very well calculated, so for overheating we did not have to be afraid.
Test system
- Intel Pentium 4 2.8 GHz (800 MHz FSB, 1024 KB L2, LGA 775)
- 80 GB MAXTOR DIAMONDMAX 9 (7200 RPM, 8 MB) S-ATA
- Sapphire RX600 Pro 128 MB PCI Express
- Windows XP Professional (ENG.) SP2
- Catalyst 5.3.
Test memory different applicationsTo see the speed difference or vice versa to show that it is not. Here we will need the following tests:
Sisoft Sandra 2005.
Synthetics
Emulation of real tasks
PCMark 2004 Patch 120
Test RealWorld.
Well, let's turn to tests.
Synthetic tests
First test - Rightmark Memory Analyzer. Unlike many other tests, it shows not only the average speed, but also the maximum.
The synthetic test, as it should be, shows a tangible difference in performance, especially in maximum values. Naturally, here DDR2-533 turns out to be better. Let's look at another synthetic test - Sandra 2005.
Our next test is Sandra 2005. Good alternative RightMark Memory Analyzer will confirm or refute the results obtained.
This synthetic test also shows a tangible difference between DDR2-533 and DDR2-400. And this means that in synthetic tests without a doubt 533-megahertz memory is an absolute leader. But the computer is not alive alone. It's time to make real tasks.
This test is not a real-world benchmark, but still show us the dependence of the computer's overall performance from the memory rate. According to the experience of previous tests, you can build assumptions that ... and in general, we look at the results.
But here we see that in general the DDR2-400 wins in DDR2-533. Why? The answer is simple - this is the error of measurement. PCMark04 is present, but this means that the difference between DDR2-533 and DDR2-400 is within the measurement error. In other words, there are no differences in common tests, but memory testing shows that PCMark04 really did not confuse anything. See which picture it turns out - the test tells us that the difference in the speed of memory, of course, is, but it does not affect anything. I wonder if it is not true? After such a test, we can only run Real World Benchmark to check in real tasks.
Well, let's see how memory will behave in the test of real tasks. SYSMARK2004 is a kind of reference point. Often, its results put a point in testing and continue tests does not make sense. And sometimes its results contradict the previous one and have to continue tests in search of truth. I wonder how will it be this time?
The results in the overall rating differ only to anything: one unit of almost one hundred sixty. Internet Content Creation test tests are equally equally. And this means that we can already draw conclusions and summarize.
The tests have shown that for computers of the average price range, the DDR2-533 and DDR2-400 memory is equally well suited. In real applications, the difference in the performance of the computer is meager or absent and at all. If I had to buy a value up to $ 1000 when buying a computer, I was offered to save the twenty, installing DDR2-400 instead of DDR2-533, I would not think agreed - the speed difference is not justified by the difference in the cost of these memory modules.
But there is one point: today, the DDR2-400 memory modules are very difficult on sale. The DDR2-400 REGISTERED ECC modules are used for workstations and servers, and cost more expensive. Normal non-ECC DDR2-400 modules are a rarity that disappears from the sale. So if you buy a ready-made computer with DDR2 memory, most likely 533-MHz memory will be installed in it. Well, if you select the configuration yourself and try to save as much as possible, you can safely take DDR2-400. There will be no catastrophic difference in performance.
The memory "OCZ PC3200 Titanium" is provided by Svega Computer, the official OCZ distributor in Russia.
Mikhail Degtyarev (Aka Like Off)
24/07.2005