New Celeron and Sempron processors: fast and cheap. AMD Sempron Socket754 processors Amd athlon 64 sempron series

Athlon 64 X2 is outdated, both physically and mentally. Such devices
were presented back in 2006. These were the first multi-core solutions
AMD company. Assessing their importance today is not particularly difficult. Their release was the first evolutionary step of this manufacturer in the field of high-tech solutions. It was he who significantly influenced the development of the computer industry. Nowadays you won’t surprise anyone with an 8-core CPU. This has already become the norm. But then such a decision produced a kind of revolution, the fruits of which we still enjoy to this day.

Story

The first 2-core CPU in the home PC niche was the product of AMD's eternal competitor, Intel. It was a Pentium processor with the index XE 840. It was installed in which was the main one for this manufacturer at that time. The increase in the number of cores has caused the need to reduce this, resulting in decreased performance in single-threaded applications. A similar result was obtained by the product of its constant competitor - the AMD Athlon 64 X2 processor. But due to the fact that such solutions were initially oriented towards multi-threading, the effect was not as strong as that of the main competitor. With the emergence of software that is capable of fully loading two physical cores, the balance of power gradually changed. And such solutions gradually replaced CPUs with 1 core from use. Yes, similar devices are still sold now, but they are mostly used for office PCs, where work in the foreground comes to the fore. office applications and low cost of the finished system. And for gaming systems it is recommended to take 4, 6 or 8 cores. As a last resort, you can opt for 2 cores, but this will significantly affect the quality of the game not in better side. This arrangement was laid out more than 5 years ago, and one of its founders was the AMD Athlon 64 X2 processor.

Modifications

Initially, such CPUs were installed in which was the most progressive from this manufacturer at that time. 4 processor models were immediately presented. The youngest of them was the AMD Athlon 64 X2 4200. The rest had a similar name, but differed in index. Modifications 4400, 4600 appeared, and the flagship of this line had the index 4800. Also, a mandatory attribute of the designations of these CPUs was “+”, which was added at the end of the name. The frequency of the base model was 2200 MHz. Also among the architectural features it is worth noting the cache, the size of which in the younger model was 1MB. Moreover, each of the cores accounted for only half of it. The remaining modifications could boast more high frequency and increased cache size.

Later decisions

A little later, more productive products appeared on the market. A logical development in this direction was the emergence of such CPUs for the AM2 platform. Their cache size was similar to that of their predecessor. But the frequencies increased significantly and amounted, for example, for the CPU of the AMD Athlon 64 X2 5000 model to 2700 MHz. Also another innovation was support for new memory, which was called DDR2. But, in principle, these processors, the period between the appearance of which is slightly less than 2 years, have a lot in common.

Conclusion

The AMD Athlon 64 X2 processor is one of the founders of the era parallel computing on one chip. If you look closely at it, you can easily find a lot in common with new AMD solutions. And this is not surprising, because they are built according to a similar architecture, which has undergone certain changes over the past 5 years, but also retained common features.

AMD is already quite for a long time divides its desktop processors into high-performance (for high-end systems) - the FX and X2 series, regular (middle-end) - Athlon 64, and budget (for systems entry level). The latter were called Sempron. Moreover, during the heyday of the SocketA platform, such a division also took place. True, then AMD had two lines - Athlon and Duron. It was the latest line that allowed users to get fairly high performance at a very affordable price (see review of Desktop Processors: Results of 2003).

When moving to 64-bit processors that were released for socket 754, there was no clear division between regular and budget processors. However, in AMD's policy one could notice signs of a gradual abandonment of the obsolete SocketA platform. And as soon as the production of Athlon64 Socket939 processors with a dual-channel memory controller was launched, AMD made it clear to users that SocketA was already dead and would not be supported in the future. And for budget systems All Socket754 processors will be designed. At the same time, the first processors under the Sempron brand appeared.

It is interesting to note that SocketA processors were (and are) also sold under the Sempron brand. Moreover, if we compare the performance of 64-bit processors with the speed best processors SocketA, the difference was quite impressive. Therefore, to create the junior models of the Sempron SocketA budget line, AMD engineers did not use traditional methods of creating cheap processors. We are talking about cutting down features and functionality, such as lowering the system bus frequency and reducing the size of the cache memory. As a result, the low-end Sempron SocketA processors were almost exact copies of the Athlon SocketA processors.

Currently, AMD also produces Sempron SocketA processors. In particular, the company's product range includes models 2500+ and 2400+, operating respectively at frequencies of 1.75 GHz and 1.667 GHz (system bus 166 MHz; L2 cache memory volume - 256 KB). In addition, AMD has another SocketA model - Sempron 3000+ with a clock frequency of 2.0 GHz and L2 capacity = 512 KB. It is clear that from the point of view of the average user, these processors do not represent any value. But on the other hand, AMD works with large system integrators who have long-term commitments to corporate clients. And in this area it is not so easy to upgrade the SocketA platform to Socket754 (due to the very large quantities installed systems).

But in any case, the SocketA platform is already dead, and at present when we say Sempron we will mean Socket754 (and in the future Socket939). When forming the Sempron line, AMD engineers had to rack their brains. The fact is that the frequency ceiling of the 0.13 micron technical process is still limited by the 2.2 GHz bar, and the new 0.09 micron technical process does not provide a significant increase in operating frequencies. Of course, AMD submitted to the 2.8 GHz bar - this is the frequency at which the Athlon 64 FX-57 processors operate. But to produce mass-produced and cheap processors such as Sempron, it is necessary to have an order of magnitude better percentage of usable chips per wafer. Therefore, if you look at things realistically, then the frequency potential of the 0.09 micron technical process for Sempron production is around 2.4 GHz (although in the future, when changing cores and debugging the technical process, an increase to 2.6 GHz is possible).

Therefore, faced with a limitation in operating frequencies, AMD engineers were forced to solve two problems. Firstly, Sempron processors should not be inferior (or better yet, superior) in speed to Intel Celeron processors, and secondly, Sempron processors should not interfere with sales of AMD Athlon64 processors. While solving the first problem was not particularly difficult, solving the second problem arose. In my opinion, AMD marketers failed to come up with a way to divide the Socket754 Athlon64 and Socket754 Sempron processors into different market niches. As a result, they simply turned a blind eye to this problem, declaring that the entire Socket754 platform is intended for budget systems.

However, there are still a couple of differences between Athlon64 and Sempron. Firstly, Sempron processors do not support the execution of 64-bit instructions. But for most users this does not matter: the 64-bit Windows system is not yet fully debugged, and the number of optimized applications is small (I think most users are waiting for the release of a fundamentally new OS Windows Vista, which will take place in about a year). The second difference is more significant - the size of the second level cache in Sempron processors is reduced to 256 KB, and in some models - to 128 KB (this is quite significant compared to 1024512 KB for Athlon64 processors). At the same time, the performance ratings of Sempron processors were calculated with an eye to direct competitors (Intel Celeron). As a result, very funny situations arose more than once when in stores the Sempron 3100+ (S754) processor was more expensive than the Athlon64 2800+ (S754), and at the same time worked slower :).

If we talk about the size of the cache memory, then many users may be confused by the variety of combinations of clock frequency and cache memory size. In addition, overclockers have another problem when choosing a processor: for the same models, AMD, at different times, used different cores and steppings that have different overclocking potential. Bringing all the information about Sempron processors into one table is a rather difficult task: because some processors with the same performance rating had serious differences from each other (very often this applies to processors released for OEM). Therefore, we will limit ourselves to only listing the technical characteristics of the latest processors.

When looking at the table, it is not difficult to grasp the logic behind the formation of the performance rating. In particular, an increase in cache memory from 128Kb to 256Kb, with an equal clock frequency, leads to an increase in the rating by 200 points. And increasing the clock frequency by 200 MHz with the same L2 cache volume increases the rating by 400 points. True, the 3100+ processor, which differs from the 3000+ model in L2 capacity (256Kb versus 128Kb), falls out of this clear picture. But there is an explanation for this: AMD marketers estimate processors released at 0.09 microns to be somewhat “more expensive”. This is partially justified by the following. Although the change in technical process does not increase operating speed, processors released at 0.09 microns are slightly faster due to minor changes kernels. By the way, AMD quite often redesigns the core, and the main changes concern the built-in memory controller. Because no matter how good the Athlon64 is, there will always be an area where one or another characteristic of the processor can be improved, modified or corrected.

So, let's look at the processors that will take part in today's testing.

From left to right: Sempron 3100+, 3300+ and 3400+. In addition, our testing will involve one of the slowest and cheapest processors for Socket 754. This is Sempron 2600+, operating at 1.6 GHz and having 128 KB of L2 cache.

The processor is based on a 0.09 micron Palermo core with DH8-D0 stepping.

The next Sempron processor differs from all others in that it is manufactured using a 0.13 micron process technology.

It is based on the Paris stepping DH7-CG core.

If you don’t look at the markings, then visually all processors designed for Socket754 are completely identical, both on the front and back sides.

Try to guess what processor this is? Yes, I don’t remember myself :).

Just like the 2600+ model, it is based on the Palermo core and has an L2 cache of the same size. But the main difference between the 3300+ model and the 2600+, in addition to the clock frequency, is the new core stepping (DH8-E3). In addition to the next improvements to the memory controller, this stepping has expanded functionality. In particular, the processor supports the execution of SSE3 instructions.

And finally the Sempron 3400+ model. This processor operates at a clock frequency of 2.0 GHz, and the L2 cache size is 256 KB.

It is also based on the Palermo core, but the core stepping is the most recent - E6. Thanks to it, the processor, in addition to executing SSE3 instructions, is capable of executing 64-bit instructions (i.e. x86-64).

Thus, AMD is transferring support for AMD64 technology to budget processors. It is worth noting that Intel was the first to release budget processors with x86-64 support (Celeron D model with EM64T technology), and AMD acted as a catching up party. In addition, the starting price of $150 for the Sempron3400+ model does not allow us to classify it as a budget processor (in my opinion, in this regard, the Athlon64 Socket754 is much more attractive).

Now let's list specifications Sempron 3400+ processor:

• Processor core - Palermo
• Cool"n"Quiet technology support
• AMD64 technology support
• NX-bit technology support
• Stepping - E6
• Technological process - 0.09 microns
• Clock frequency 2.0 GHz (multiplier = 10)
• HTT bus frequency = 200 MHz
• Core area 84 sq. mm.
• Number of transistors 63.5 million.
• L1 cache size: 128 KB
• L2 cache size: 256 KB
• Standard voltage: 1.4V
• Typical Heat Dissipation: ~62W
• Maximum case temperature: 69C (note - the case, not the core itself)

The processor has a single-channel memory controller (like other Socket754 processors), and supports the following memory standards: DDR200, DDR266, DDR333 and DDR400.

It seems most likely that, following the 3400+ model, AMD company will release junior models of processors based on the Palermo core with E6 stepping. Thus, even owners of the most budget systems will be able to get x86-64 support.

In order for a user to be able to distinguish a processor with x86-64 support when purchasing a processor, it is necessary to know the marking features. In particular, processors based on the E6 stepping have the letters BX at the end of the marking. And, for example, processors based on the Palermo core stepping E3 have the last marking letters - B0.

To test Socket 754 processors, we chose the Epox EP-8NPA-SLI motherboard.

This motherboard was one of the first to offer the nForce 4 chipset for this socket.

Board: Epox EP-8NPA-SLI (nVidia nForce 4 SLI)
Memory: single channel DDR400 (PC3200)

When the Athlon 64 was announced, three versions of processors for Socket 754 were announced: the original Clawhammer core with 1 MB of L2 cache; the same core with the cache halved; as well as the Newcastle core with 512 KB L2 cache and a smaller die surface area. In early 2005, Sempron joined them with an L2 cache of only 256 KB.

All processors for Socket 754 work with a single-channel DDR400 memory interface, resulting in maximum speed transfer rate is 3 GB/s. At the same time, memory capacity greater than 1 GB is often problematic to install. The processors operate with a 200 MHz Hyper Transport channel, a maximum thermal dissipation of 89 W, and a clock frequency that varies from 1.8 GHz to 2.4 GHz. There are no plans to further increase the frequency for this socket. Available chipsets include nVidia nForce 3 150/250, VIA K8T800 and SIS's 755FX.

Socket 754 is equipped only with a single-channel memory interface. Therefore, problems arise when expanding memory.

Open AMD Athlon64 3400+ with core size 17.5 x 11.5 mm.

With the release of Athlon64 3700+, Socket 754 fell into oblivion after two years of existence.

Socket 754: AMD Athlon64, Sempron
Board: Asus K8N-E Deluxe (nForce 3 250 GB)
Memory: DDR400 (PC3200)

Question: What are the marking features of AMD processors?
Answer: The marking of AMD processors is called OPN (Ordering Part Number). At first glance, it is quite complex and looks more like some kind of cipher, although if you understand it, you can get quite detailed information about their main technical parameters:

1. The first two letters indicate the processor type:
   • AX - Athlon XP (0.18 µm);
   • AD - Athlon 64, Athlon 64 FX, Athlon 64 X2;
   • SD - Sempron.
2. The third letter indicates the TDP of the processor:
   • A - 89-125 W;
   • O - 65 W;
   • D - 35 W;
   • H - 45 W;
   • X - 125 W.
3. For Sempron processors, the third letter has a slightly different meaning:
   • A - Desktop;
   • D - Energy Efficient.
4. The next four numbers are the processor rating (the same one that is indicated in all price lists along with the processor type, for example, Athlon 64 4000+) or, in other words, the Model Number. It is a number that (from AMD's point of view) characterizes the performance of a given CPU in abstract units. Although there are some exceptions - in Athlon 64 FX processors, for example, instead of rating numbers, the letter index “FX (model index)” is indicated.
5. The first letter of the three-letter index indicates the type of processor case:
   • A - Socket 754;
   • D - Socket 939;
   • C - Socket 940;
   • I - Socket AM2;
   • G - Socket F.
6. The second letter of the three-letter index indicates the supply voltage of the processor core:
   • A - 1.35-1.4 V
   • C - 1.55 V;
   • E - 1.5 V;
   • I - 1.4 V;
   • K - 1.35 B;
   • M - 1.3 B;
   • Q - 1.2 V;
   • S - 1.15 V.
7. The third letter of the three-letter index indicates the maximum temperature of the processor core:
   • A - 71°C;
   • K - 65°C;
   • M - 67°C;
   • O - 69°C;
   • P - 70°C;
   • X - 95° C.
8. The next number indicates the size of the second level cache (total for dual-core processors):
   • 2 - 128 KB;
   • 3 - 256 KB;
   • 4 - 512 KB;
   • 5 - 1024 KB;
   • 6 - 2048 KB.
9. The two-letter index indicates the type of processor core:
   • AX, AW - Newcastle;
   • AP, AR, AS, AT - Clawhammer;
   • AK - Sledge Hammer;
   • BI - Winchester;
   • BN - San Diego;
   • BP, BW - Venice;
   • BV - Manchester;
   • CD - Toledo;
   • CS, CU - Windsor F2;
   • CZ - Windsor F3;
   • CN, CW - Orleans, Manila;
   • DE - Lima;
   • DD, DL - Brisbane;
   • DH - Orleans F3
   • AX - Paris (for Sempron);
   • BI - Manchester (for Sempron);
   • BA, BO, AW, BX, BP, BW - Palermo (for Sempron).

For example, the AMD Sempron 3000+ processor (Manila core) is labeled as SDA3000IAA3CN. But nothing lasts forever in our world, and AMD is soon going to rename its processor lines, introducing a new, much more descriptive alphanumeric scheme. New system assumes, along with the traditional brand and class designation, also an alphanumeric model code

1. The first character in the processor model name determines its class:
   • G - High-end;
   • B - Mainstream;
   • L - Low-End.
2. The second character determines the processor's power consumption:
   • P - more than 65 W;
   • S - 65 W;
   • E - less than 65 W (Energy Efficient class).
3. The first digit indicates that the processor belongs to a specific family:
   • 1 - single-core Sempron;
   • 2 - dual-core Athlon;
   • 6 - dual-core Phenom X2;
   • 7 - quad-core Phenom X4.
4. The second digit will indicate the performance level of a specific processor within the family.
5. The last two digits will determine the processor modification.

Thus, the latest dual- and quad-core processors will be designated as AMD Phenom X2 GS-6xxx and Phenom X4 GP-7xxx. Economical mid-class dual-core processors are Athlon X2 BE-2xxx, and budget AMD Athlon and Sempron will be called Athlon X2 LS-2xxx and Sempron LE-1xxx. And the notorious number 64, indicating support for 64-bit architecture, will disappear from the name of the Athlon processor.

Question: How do Sempron processors differ from Athlon 64 processors?
Answer: Modern processors of the Sempron series, intended for the budget segment of the market, differ from full-fledged prototypes - Athlon 64 processors reduced to 128 (or, in selected models, up to 256 KB) second level cache volume. In addition, the HyperTransport bus in Sempron processors only operates at 800 MHz, while in the Athlon 64 its frequency can reach 1000 MHz; Less significant is the lack of support for Pacifica virtualization technology. Everything else, including a dual-channel memory controller, support for the 64-bit AMD64 architecture and the SSE3 instruction system, is available in full.

At the same time, we should not forget that such sophisticated Sempron processors are produced mainly in versions for Socket AM2 and Socket 939. Older Sempron models for Socket 754, for example, have only a single-channel memory controller.

Question: What are the features of the Socket AM2 processor socket?
Answer: Today in the desktop segment, AMD is experiencing an “orgy”, when you can find processors on sale in at least four (!) variants: Socket 754, Socket 939, Socket 940 and Socket AM2 (and this is not to mention the rare Socket A, which are still occasionally found on store shelves). True, AMD came to its senses in time and with the release of the Socket AM2 platform, it again returned to the path of unifying the processor socket for desktops, for which it has always been respected by upgrade lovers.

Socket AM2, which will replace Socket 754 and Socket 939, has 940 pins (like the server Socket 940, but they are not compatible!), and is used in mass-produced single- and dual-core Athlon 64 processors, prestigious Athlon 64 FX and budget Sempron. Socket AM2 processors work with DDR2 memory with frequencies from 533 to 800 MHz (PC4200, PC5300 or PC6400) in dual-channel mode; Registered and ECC memory are not supported. Otherwise, AMD processors for Socket AM2 are completely identical to processors for Socket 939, the production of which is currently discontinued.

Question: Is AMD's future platform for Socket AM2+ and Socket AM3 compatible with existing solutions?
Answer: In the near future we expect another transition to new type memory - DDR3 (see FAQ on DDR3. In accordance with AMD's plans, at the beginning of 2008 the modern Socket AM2 will be replaced first by Socket AM2+, and then by Socket AM3. The only serious difference between Socket AM2 and Socket AM2+ will be the introduction of support for the new high-speed HyperTransport 3.0 buses. Its use will significantly increase throughput processor-chipset (as well as processor-processor in the case of multiprocessor solutions). Socket AM3 processors will also gain support for new DDR3 memory. The characteristic features of the new platforms compared to the modern Socket AM2 are given in the table:

In this regard, the question inevitably arises about the compatibility of promising AMD platforms with existing ones.

So, Socket AM2 and Socket AM2+ processors and motherboards will be fully compatible with each other. Of course, if you install a new CPU with HT 3.0 support in Socket AM2, it will communicate with the chipset at the speed of the old HT 1.0. Socket AM3 processors, thanks to their memory controller that works with both DDR2 and DDR3 memory, will be the most versatile and can be installed in Socket AM3, Socket AM2+ and Socket AM2 motherboards (providing the latter platform with a very decent service life). But they will not have backward compatibility - neither Socket AM2 nor Socket AM2+ processors can be installed in Socket AM3 boards.

Question: What is Cool"n"Quiet?
Answer: Energy-saving Cool"n"Quiet technology came to AMD desktop processors from the mobile sector and allows you to reduce heat generation and power consumption when they are not fully loaded. On this moment this technology is implemented in all processors of the AMD K8 family - Athlon 64, Athlon 64 X2, Athlon 64 FX, Sempron. Naturally, the motherboard must also support this technology (the corresponding item must be activated in the BIOS).

There is nothing radically new in Cool"n"Quiet technology. During operation, the operating system monitors the processor load, and if it is less than a certain threshold, the operating frequency and supply voltage of the processor are reduced. Reducing the operating frequency of the processor is carried out by reprogramming its registers (using special program- processor drivers). By lowering the frequency and voltage, the processor will consume much less power, heat up less and, if the cooler is equipped with a thermal control system, the system noise will decrease.

When the processor load increases, everything happens along the same chain (OC-driver-processor-cooler), but vice versa - the processor will return to the nominal frequency. There can be up to a hundred such switches between different modes per second; for user programs, all this happens completely unnoticed, and even if it affects the overall performance of the Cool"n"Quiet system, it is insignificant.

The user determines the degree of system response to changes in processor load by selecting one or another policy in the applet Windows Power Options- from the minimum level (switching to power saving mode only when idle) to severe energy saving (the processor will almost always be in a state of reduced power consumption).

Introduction64-bit systems with x86-64 architecture are gaining more and more popularity. Processors supporting the corresponding AMD64 and EM64T technologies are widespread on the market, the most popular user operating system Windows XP is officially available in a 64-bit version, and the number of software products using 64-bit modes is steadily growing. In fact, if you do not take into account the difficulties of the transition period (in particular, dampness of drivers, problems with some programs, etc.), for widespread migration to x86-64, only a small thing is missing: budget 64-bit systems. However, this problem is beginning to be resolved. Not long ago, Intel began offering LGA775 Celeron D processors with EM64T technology, and following Intel, AMD decided to take a similar step, releasing CPUs from the Sempron line with activated AMD64 technology.
Thus, owners of inexpensive platforms get a great chance to join 64-bit: now creating an x86-64 system does not require large financial expenditures. The cost of junior processor models with EM64T or AMD64 technologies has dropped to $60-$70.
Along with the launch of Sempron processors with AMD64 technology, AMD also introduced another, faster CPU model in this line. The new $134 processor received a rating of 3400+. This is exactly the processor that appeared in our laboratory. Since the main highlight of this CPU, which is based on the updated Palermo kernel revision E6, was compatibility with the x86-64 architecture, we decided to test this CPU in 64-bit mode, in operating system Windows XP Professional x64 Edition. That is, this article will be devoted to studying the operation of budget processors in a 64-bit environment.

More about AMD Sempron 3400+

Overclocking

Motherboard DFI LANPARTY UT nF3 250Gb (NVIDIA nForce3 250Gb);
Cooler AVC Z7U7414001;
Memory Corsair CMX512-3200XL (2 x 512MB, DDR400 SDRAM);
Graphics card PowerColor RADEON X800 XT (AGP 8x, 500MHz/500MHz);
HDD Western Digital Raptor WD740GD (SATA150).

How we tested

Processors:

AMD Athlon 64 3800+ (Socket 939, 2.4 GHz, 512KB L2, Venice);
AMD Athlon 64 3000+ (Socket 939, 1.8 GHz, 512KB L2, Venice);
AMD Athlon 64 3000+ (Socket 754, 2.0 GHz, 512KB L2, NewCastle);
Sempron 3400+ (Socket 754, 2.0 GHz, 256KB L2, Palermo revision E6);
Intel Celeron D 351 (LGA775, 3.2 GHz, 256KB L2, FSB=533MHz, Prescott-256);

Motherboards:

ASUS P5LD2 Deluxe (LGA775, i945P);
DFI DFI NF4 Ultra-D (Socket 939, NVIDIA nForce4 Ultra);
DFI LANPARTY UT nF3 250Gb (Socket 754, NVIDIA nForce3 250Gb).

Memory:

Corsair CMX512-3200XL, 2 x 512MB, DDR400 SDRAM, 2-2-2-5;
Corsair CM2X512A-5400UL, 2 x 512MB, DDR2-533 SDRAM, 3-2-2-8.

Graphic cards:

PowerColor RADEON X800 XT (PCI-E x16).
PowerColor RADEON X800 XT (AGP 8x).

Disk subsystem: Maxtor MaXLine III 250GB (SATA150).

Performance

conclusions