Pentium 4 2.40GHz processor

Number of nuclei - 1.

The basic frequency of the nuclei of Pentium 4 2.40GHz is 2.4 GHz.

Price in Russia

Family

Intel Pentium 4 2.40GHz Test

The data is obtained from user tests that tested their systems both in acceleration and without. Thus, you see the averaged values \u200b\u200bcorresponding to the processor.

Numerical operations

For different tasks are required different strengths CPU. The system with a small amount of fast nuclei is perfect for games, but will give way to a system with a large number of slow nuclei in the rendering scenario.

We believe that for budget gaming computer Suitable processor with minimum 4 nuclei / 4 threads. At the same time, individual games can load it 100% and slow down, and performing any tasks in the background will lead to FPS drawdown.

Ideally, the buyer should strive to minimize 6/6 or 6/12, but to take into account that systems with more than 16 threads are now applicable only in professional tasks.

The data is obtained from user tests that tested their systems as in acceleration ( maximum value In the table), and without (minimum). A typical result is indicated in the middle, a position among all tested systems is indicated in the color strip.

Accessories

We collected a list of components that users most often choose by collecting a computer based on Pentium 4 2.40GHz. Also with these components are achieved the best results in tests and stable operation.

The most popular config: Motherboard for Intel Pentium 4 2.40GHz - ASUS P8Z68-V, video card - GeForce GT 525m.

Alexey Shobanov

Continuing the Spring Prime Minister, Intel introduced another model in the processor line for high-performance systems for home and office - Intel Pentium 4 processor with a clock frequency of 2.4 GHz. The transition to the 0.13 micron technological process has significantly expanded the "frequency horizons", opening up before the flagship of the processor market from Intel, and now we seem to be quite ordinary to the quarterly presentations of new, increasingly rapid processors. Like his predecessors - Pentium 4 2 GHz and 2.2 GHz, also built on the basis of Northwood nucleus at 0.13 micron technology, new processor It has a second-level cache size of 512 KB, which is twice the size of the L2 cache in the younger models of this line, created on the basis of the Willamette kernel (0.18 micron technical process). Pentium 4 2.4 GHz is made in the MPGA-478 form factor using the FC-PGA2 enclosure (Flip-Chip Pin Grid Array), which today has the most advanced heat dissipation scheme. Speaking O. thermal mode The Pentium 4 processor on the new Northwood core cannot be noted not to note the fact that the transition to a new 0.13 micron technology allowed not only to increase the number of transistors to 55 million on a crystal, while reducing its size, but also reduce the power supply of the kernel to 1.5 V, thereby reducing heat dissipation. Thus, in the first processors on this core, which operate on a clock frequency of 2 GHz and 2.2 GHz, it constitutes 52 W and 55 W, respectively, and the new Intel Pentium 4 2.4 GHz does not exceed 58 W. For temperature control, the processor uses the so-called "Thermal Monitor" technology, the essence of which is reduced to the use of the thermal sensor and the TCC block (Thermal Control Circuit) controlling the supply of clock pulses per processor. At the same time, two modes of operation are provided: Automatic (Automatic Mode) and on-Demand Mode. Auto mode It can be activated via the BIOS system board. In this mode, when the processor temperature increases to a certain value, the TCC unit is activated and generates pulses blocking the supply of clock pulses, which actually causes a decrease in the processor clock frequency by 30-50% (in accordance with factory settings), increasing its downtime, which, in turn, reduces the temperature. The operation of the TCC block in the "on demand" mode is determined by the contents of the temperature control register (ACPI Thermal Monitor Control Register). In accordance with its state, the TCC block can be activated regardless of the processor temperature, while the idling duration of the processor can be varied more flexibly in the range between 12.5% \u200b\u200band 87.5%. And, of course, the ability to disconnect the computer during the catastrophic heating of the processor crystal to 135 ° C; In this case, the system tire is issued to thermTrip #, initiating power off. Like all its predecessors, the new processor is built in accordance with the Intel NetBurst microarchitecture, which implies the following innovations:

• 400-megahertse system tire;
• Hyper-Pipelined Technology;
• Advanced Dynamic Execution;
• EXECTION TRACE CACHE;
• Rapid Execution Engine;
• ADVANCED TRANSFER CACHE;
• Streaming SIMD EXTENSIONS 2 (SSE2).

In a few words we describe these features of the Intel Pentium 4 processor architecture 4. 400-megahertic tire (as it is also called - Quad Pumped Bus) allows for a special organization on physical level Pass 4 data packets for one clock over the system bus with a frequency of FSB 100 MHz. Thus, this 64-bit bus has a peak bandwidth of 3.2 GB / s, providing high-speed processor exchange with other devices. Soon the 533-megahertsev QUAD PUMPED-tire is expected, which corresponds to the operation of the system bus on the physical frequency of FSB 133 MHz, while it is easy to assume, the rate of data exchange according to it will exceed the seemingly unattainable 4 GB / s. Hyper-Pipelined Technology involves the use of an unprecedented long 20-speed hyperconveyor (we recall that the R6 family processors have doubled the conveyor). This approach can significantly increase the processor clock frequency, although it leads to such a negative consequence as an increase in the rebooting time of the conveyor in the event of a transition prediction error. In order to reduce the likelihood of such a situation, the Advanced Dynamic Execution technology is applied in the Pentium 4 processors, which implies an increase in the command pool to 126 (in Pentium III, the pool of commands contained 42 commands) and an increase in up to 4 KB of branch buffers, which stores the address of the transitions already performed. This, coupled with an improved prediction algorithm, makes it possible to increase the likelihood of transition prediction by 33% compared with the P6 family processors and bring it to 90-95%. IN pentium processors 4 A multi-non-prior approach is implemented to the organization of the first level L1 cache. Although L1, as in most modern processors, consists of two parts: data cache (8 KB) and cache instructions, the latter feature is that now it stores up to 12 thousand already decoded micro operations, and those located in the order of their execution, Certain based on predictions of branching transitions. Intel Pentium 4 processor cache with such an organization was called Execution Trace Cache. Rapid Execution Engine is two arithmetic logic blocks (ALU) that operate on a double processor frequency. In the case of the processor described by us, the clock frequency of which is 2.4 GHz, this means that ALU blocks operate at 4.8 GHz, and given that they function in parallel mode, it is easy to calculate that the processor can perform four integer operations for Tact (just over 0.4 μs). The Cache of the second level L2 processors of the Pentium 4 family got the name Advanced Transfer Cache. Having a 256-bit tire operating at the core frequency, and an improved data transmission scheme, this cache provides the highest bandwidth, so important for streaming processing processes. As noted above, the initial processors on the Willamette kernel had a L2 cache of 256 MB, the transition to 0.13 micron technology allowed to increase the second-level cache to 512 MB. Such an increase in L2 cache has affected the processor productivity, allowing us to reduce the likelihood of misses when contacting. In the Pentium 4 processors, support for an enlarged set of streaming SIMD extensions (Streaming SIMD Extensions), which received the name SSE 2. In this set to the already existing 70 SIMD instructions, 144 new instructions were added. These instructions allow you to perform 128-bit operations with both integers and floating point numbers, giving a significant productivity gains on a number of tasks using streaming data processing. Here there is only one "but" - the code of the task is performed must be appropriately optimized and compiled.

With all the above improvements, the Pentium 4 model line processors are based on the same 32-bit Intel architecture (IA-32), and the new processor is no exception. As a result, Pentium 4 2.4 GHz is optimized for working with 32-bit software and shows traditionally stable and high-performance work with such operating systems such as Windows 98, Windows Me, Windows 2000, Windows XP and OS UNIX. We had the opportunity to test the work of the new processor from Intel, while the following test stand configuration was used:

• intel processor Pentium 4 2.4 GHz;
• motherboard MSI MS-6547 (on the SIS 645 chipset);
• hDD Fujitsu MPG3409AH-E 30 GB with file system NTFS;
• 256 MB random access memory DDR SDRAM PC2700 (Cl 2.5);
• gigabyte GF3200TF Video Card (GeForce 3 Ti 200, 64 MB) with NVIDIA Detonator V video driver. 27.42 (resolution 1024 × 768, color depth 32 bits, VSYNC - Off).

For testing we used operating system Microsoft Windows. XP. Test results are shown in the table.

Perhaps someone will ask the question: how many processor performance can be increased and generally needed for modern personal computer So powerful central processors? We want to answer this that the work of the central processor will always be found. Its computing power can be used by shifting on it the logic of other computer subsystems, thereby reducing the cost of the latter. Some experts raise the question that with further speed growth central processor It would be possible to shift on it and the computational load of the graphics card processor (which was already done in the past, but with completely different motivations).

In conclusion, I would like to note that the new processor from Intel - Pentium 4 2.4 GHz demonstrates stable operation and excellent performance on applications working with sound, video, 3D graphics, on office applications and games, as well as when performing complex computing tasks . In a word, on the basis of this processor, high-performance houses for home and office can be created, capable of satisfying the most demanding user requests and solve problems that prevent the highest possible requirements for the computing power of your personal computer.

ComputerPress 5 "2002

Comparison of IPC

For those who do not know: IPC (Instructions Per Cycle, the number of executable instructions for the clock) is a good indicator of how fast the processor works, and the simultaneous combination of high IPC and clock frequency values \u200b\u200bprovides maximum performance. That is what we see in the Intel Coffee Lake 8 generation processors, and although AMD is clearly lagging behind when it comes to frequencies, this company is actually approaching Intel in part of the IPC. For this reason, many of you are interested in this aspect of CPU testing.

To understand how far AMD has advanced in this direction, we decided to minimize the number of testing parameters and at the same time, to bring the situation to the actual working conditions as much as possible. The first and most obvious step here is to bring the core frequencies to a single constant value that we have done by fixing all CPU kernels at 4 GHz. All options bOOST technology And, thus, the core frequencies could not go beyond 4 GHz.

2 generation processors Ryzen were tested on motherboard ASRock X470 Taichi Ultimate, and COFFEE Lake Processors - on the ASRock Z370 Taichi board. In both configurations in all tests, the same G.Skill Flarex DDR4-3200 memory was used with the "Xtreme" memory profile and the same MSI GTX 1080 TI Gaming X TRIO video card.

We can immediately say that this article does not contain recommendations for potential buyers - we conducted testing in purely research purposes.

Coffee Lake processors originally have a clear advantage in the clock frequency.

In this review, we included the test results of Intel Core i7-8700K processors, Core i5-8600K and AMD Ryzen 7 2700x, Ryzen 5 2600x and Ryzen 7 1800x, Ryzen 5 1600x.

So, now processors 1600x, 2600x and 8700K have the same resource: 6 nuclei and 12 streams.

The 1800x and 2700x processors have an advantage of 8 cores and 16 streams, while 8600K with 6 cores and 6 threads, on the contrary, is at a disadvantage.

All this should be borne in mind when we go further. Let's start the results.

Benchmark

Let's start with the dough on the continuous bandwidth of memory. Here we see that processors 1 and 2 generations Ryzen have almost the same bandwidth - about 39 GB / s. Meanwhile, COFFEE LAKE processors, working with the same memory, are limited to the value. bandwidth About 33 GB / s, which is 15% less compared to Ryzen processors.

Go to Cinebench R15 test. Here we see that the 2600x processor shows higher results compared to 1600x - 4% more in multithreaded mode and 3% more in one-threaded. And if we look at 8700K, we will see that it is 4% faster than 2600x in one-threaded mode and 4% slower in multi-threaded.

As you could expect, with the same clock frequency, the Ryzen processors with 8 cores and 16 threads in multi-threaded mode are easily bypass 8700K. I brought these results here simply because they had me. With the appropriate request, I could spend this test, for example, with Core i7-7820x.

The next point is to edit the video in PCMark 10, and this test gives more clear results, although before that we observed a noticeable difference between the 1600x and 1800x processors. And here we see confident 10% progress when moving from 1600x to 2600x, and this puts AMD to one level with Intel in terms of IPC performance (at least in this test).

As the results of Cinebench R15, used to the maximum AMD SMT technology (Simultaneous Multi-Threading) looks more efficient than intel technology HT (Hyper-Threading). Here, the 1600x processor was faster than 8700K by 3.5%, and 2600x - by as much as 8%, and for this example, this is a significant difference.

Productivity / Performance in applications

For the next test, we took Excel, and here the 8700K processor was about 3% faster than 1600x, on the same clock frequency. However, 2600x is able to compete with 8700K: it showed the same completion time when performing a test task - 2.85 ° C is an impressive result.

In the HandBrake test, the results of AMD Ryzen processors were not so shiny: here we see that 2600x can only compete with 8600K, and in comparison with 8700K it turns out 15% slower.

Go to Corona benchmark. Here we see that the 2600x processor can reduce the rendering time by 8% compared with 1600x, and at the same time it turns out only 3% slower than 8700K. Thus, in this test, Intel still keeps the advantage in IPC, but it is minimal.

The next test is Blender, and here 2600x was only 2.5% faster than 1600x, and 4% slower than 8700K. Not too big difference, and again Intel keeps the advantage in IPC - in this test it is less than 5%.

In Benchmark V-Ray, we see that the 2600x processor has exceeded the result of 1600x by 4% and was only one percent more slower than 8700K, i.e. Essentially, it turned out to be with him at the same level.

Game benchmarks

It is time to consider a number of game results, and here AMD processors fall off. As I have repeatedly spoke earlier, the Intel Ring Bus ring bus is just better suitable for games, and we see it even when comparing this Intel solution with their own architecture based on Mesh Interconnect, developed for processors with a large number of nuclei. The internal bus AMD Infinity Fabric is experiencing a number of problems, and these problems will remain as long as the game processors do not require more nuclei.

Thus, although the processor is 2600x and exceeds 1600x by 8% in the game Ashes of the Singularity, it is at the same time noticeably losing 8700K - for as many as 11% slower. The fact that Intel processors work with a significantly higher clock frequency, will increase this difference to 20% or even more.

In Game Assassin "S Creed: Origins We see a small 2% superiority of the processor 2600x over 1600x, while the 8700K processor for as much as 14% faster.

This difference slightly decreased during the installation of high graphics settings, but still when we compare the average frame rate values, 8700K is 12% faster than the processor 2600x.

IN Battlefield 1. With the settings ultra, we see that the 2600x processor is 9% faster than the 1600x processor, but still 7% slower than the 8700K processor.

This difference becomes even more at average settings, since the influence of the GTX 1080 Ti video card decreases. Here, the 2600x processor again demonstrates 9% performance growth relative to 1600x, but now it is 10% slower than 8700K, which even with these settings looks like a GPU performance limit.

We observe a similar picture in the game Far Cry.where the 2600x processor is 10% faster than 1600x is very big progress, but even here it turns out to be 8% slower than 8700K.

Comparison of power consumption

This test for energy consumption was not carried out in the most realistic conditions, since when installing a single clock frequency of 4 GHz, many energy-saving options were disabled. From a scientific point of view, this is also not quite a clean experiment, because I had to increase the voltage on Ryzen processors in excessive value - to stabilize all nuclei at an increased frequency of 4 GHz.

Taking into account all the above, we see that systems with processors 1600x and 2600x consume exactly the same amount of energy, while the system with 8700K consumes 3% less, i.e. Under these conditions, this processor is slightly more efficient.

In testing S. Far Cry. The power consumed everywhere was almost the same - all processors bring the overall energy consumption of the system to about 380 W.

In the Blender test, we see a reduction in power consumption by 10% when moving from the processor 1600x to 2600x. For a 2600X processor, this is an impressive achievement, but it still consumes more power to 21% than the 8700K processor.

This time in the Handbrake test, the system with a processor 2600x demonstrated 7% more power consumption than the system from 1600x, and on a terrifying 32% greater than the system from 8700K.

Conclusion

Despite the fairly large deficit of the clock frequency (compared to analogues from Intel), processors of 2 generations of Ryzen in test applications are not so often far behind their competitors, and now we can understand why - comparing them on the same clock frequency 4 GHz. For example, in the Cinebench R15 application, we see that in a single-core mode, their performance is below only 3%, but in multi-core mode SMT helps AMD processors work up to 4% faster than Intel.

In our study, AMD processors were 3% slower than Intel in Corona test, but in benchmarks such as V-Ray, Excel and video editing have shown almost the same result with them. In Handbrake, they were 15% slower, but in PCMark 10 (test on the image in physical phenomena) - 8% faster. Of course, this is a game of gemina, and I am ready to argue - some AMD fans hoped that we would need a gaming performance deficit mainly on the clock frequency. Unfortunately, this is not so.

The main problem here is in the method of connecting the AMD processor nuclei, or rather, the CCX modules. Intel Ring Bus has a very low delay and the distribution of resources always chooses the shortest path. However, as soon as we add additional kernels, the ring bus is incremented in size - it takes more rings to connect all nuclei - and its efficiency is reduced. Thus, Intel processors with a large number of cores (for example, 28) need a more optimal method of connection of the nuclei. And in these cases, the architecture with Mesh InterConnect works perfectly.

However, we already know that for 6-, 8- and 10-core processors it is not the most the best decision, and that is why core processors i7-7800x, 7820x and 7900x in games are noticeably inferior to 8700K. The 8700K processor has an average delay time between nuclei of about 40 ns, and 7800x this time is from 70 to 80 ns.

Ryzen processors are a bit more complicated: inside the CCX module, the delay between the cores is close to what we see from the 8700K processor, and does not depend on the speed of the DDR4 memory. However, as soon as we go beyond the limits of CCX, the delay between the nuclei increases to 110 ns, and this is already associated with the memory of DDR4-3200. With a faster memory, the delay between the ccx module kernels is reduced, since the AMD Infinity Fabric bus is tied to the clock frequency of memory, and the low delay dram here also helps a lot.

Another problem is the games themselves, since almost all popular games are developed based on the CPU only with several cores, and we only begin to observe some steps taken in the direction of partitioning tasks on the parallel processing by the CPU kernels. Before the emergence of processors, Ryzen games were developed and optimized almost exclusively under Intel processors. Now the situation is gradually changing, as the game characteristics of Ryzen processors are improved, but we are unlikely to see them in the near future with the Intel processors with the Ring Bus.

However, in terms of performance IPC AMD definitely reduced the gap. The cache with a reduced delay also actually helps, and, thus, buying a 2 generation processor Ryzen carries some benefits before buying a Coffee Lake processor. It will be interesting to observe the battle between these processors, which will unfold in 2018 and further.

Found the unpleasant problem of the limit of the clock frequency. Having reached the threshold of 3 GHz, the developers were encountered with a significant increase in energy consumption and heat dissipation of their products. The 2004 technology level did not allow substantially to reduce the size of the transistors in a silicon crystal and the output from the current situation was an attempt to not increase frequencies, but to increase the number of operations performed in one beat. Given the experience of server platforms, where multiprocessor layout has already been tested, it was decided to combine two processors on one crystal.

Since then, a lot of time has passed, a CPU with two, three, four, six and even eight cores appeared widely access. But the main market share is still occupied by 2 and 4-nuclear models. Amend the situation is trying to AMD, but their architecture Bulldozer did not meet hope and budget eight cores are still not very popular in the world. Therefore, the questionwhat is better: 2 or 4-core processorstill remains relevant.

The difference between 2 and 4 core processor

At the hardware levelthe main difference between the 2-nuclear processor from 4-nuclear - Number of functional blocks. Each core is essentially a separate CPU, equipped with its computing nodes. 2 or 4 such CPUs are combined with each other internal speed and overall memory controller to interact with RAM. Others functional nodes Also may also be common: most of the modern CPU individual is the first (L1) and second (L2) level, blocks of integer computing and floating semicolons. Cache L3, characterized by a relatively large volume, one and is available to all nuclei. Separately, you can mark the already mentioned AMD FX (as well as the Athlon and APU APU CPU a): they are not only the cache memory and the controller, but also blocks of floating semicolons: each such module simultaneously belongs to two nuclei.

AMD Athlon quad-core processor circuit

From user point of viewthe difference between 2 and 4 core processor It is the number of tasks that the CPU can handle over one clock. With the same architecture, the theoretical difference will be 2 times for 2 and 4 cores or 4 times for 2 and 8 cores, respectively. Thus, with the simultaneous operation of several processes, the increase in the amount should entail the growth in the speed of the system. After all, instead of 2 operations, quad-core CPU in one point in time can be performed at once.

What caused the popularity of dual-core CPU

It would seem that if an increase in the number of cores entails an increase in performance, then on the background of models with four, six or eight cores in two-cores there are no chances. Nevertheless, the global leader in the CPU market, Intel, annually updates the range of its products and produces new models of everything with a pair of cores (Core i3, Celeron, Pentium). And this is against the background of what even in smartphones and tablets on such CPU users look with distrust or contempt. To understand why the most popular models are precisely processors with two cores, several main factors should be taken into account.

Intel Core I3 - the most popular 2-core processors for home PC

Problem compatibility. While creating software Developers strive to make it so that it can function both on new computers and already existing CPU and GP models. Considering the range on the market, it is important to ensure that the game is working fine and on two nuclei, and on eight. Most of all existing home PCs are equipped with a dual-core processor, so the support of such computers is given the most attention.

The complexity of parallelization of tasks. To ensure the effective involvement of all cores, the calculations produced during the program of the program should be divided into equal streams. For example, a task that can optimally use all the kernels, having allocated one or two processes each of them - the simultaneous compression of several videos. With games - more difficult, since all operations performed in them are interconnected. Despite the fact that the main work performs graphic processor Video cards, information for the formation of a 3D picture prepares the CPU. It is so that each kernel has processed its portion of data, and then supplied its GP synchronously with others is quite difficult. The more simultaneous calculation streams need to be processed, the hardest implementation of the task.

Continuity of technology. Software developers use for their new projects already existing developments exposed to repeated upgrades. IN some cases It comes to the point that such technologies are rooted in the past for 10-15 years. Development based on a draft decade ago, cardinal processing for perfect optimization is very reluctant, if not at all. As a result, there is an inability of software rational use of the PC hardware capabilities. S.T.A.L.K.E.R game game Call of Pripyat, published in 2009 (in the heyday of multi-core CPU) built on the 2001 engine, so it does not know how to load more than one core.

S.T.A.L.K.E.R. Only one 4-nuclear CPU is fully engaged.

The same situation with the popular online RPG World Of Tanks: the Big World engine on which it is based, was created in 2005, when multi-core CPUs were not yet perceived as the only possible way development.

World of Tanks also does not know how to distribute the load on the kernel uniformly

Financial difficulties. The consequence of this problem is the previous paragraph. If you create each application from scratch without using existing technologies, its implementation will cost the fabrical sums. For example, the cost of developing GTA V amounted to more than 200 million dollars. At the same time, some technologies were still not created "from pure sheet", and borrowed from previous projects, as the game was written under 5 platforms at once (Sony PS3, PS4, Xbox 360 and One, as well as PCs).

GTA V is optimized for multi-core and knows how to uniformly load the processor

All these nuances do not allow to fully use the potential of multi-core processors in practice. Interdependence of manufacturers hardware And software developers generates a closed circle.

Which processor is better: 2 or 4-nuclear

Obviously, with all the advantages, the potential of multi-core processors still remains unrealized to the end. Some tasks do not know how to evenly distribute the load and work in one stream, others do it with mediocre efficiency, and only a small share of fully interact with all nuclei. Therefore, the questionwhat better processor, 2 or 4 kernelsTo buy, requires a careful study of the current situation.

The market contains products of two manufacturers: Intel and AMD, characterized by the implementation features. Advanced Micro Devices traditionally make focus on multi-core, while Intel is reluctant to take such a step and increase the number of nuclei only if it does not lead to a decrease in specific performance in the calculation of the kernel (to avoid which is very difficult).

An increase in the number of cores reduces the final performance of each of them.

As a rule, the general theoretical and practical performance of the multi-core CPU is lower than that (built on the same microarchitecture, with the same technical processorrum) with one nucleus. Caused by the fact that the kernels use common resources, and this is not the best way affects performance. Thus, it is impossible to simply purchase a powerful four or sixyuclear processor with the calculation that it will definitely not be the weaker dual-core from the same series. In some situations, it will be noticeable. As an example, it is possible to launch old games on a computer with an octalized AMD FX processor: FPS at the same time lower than on a similar PC, but with quad-core CPU.

Do I need multi-core today

Does this mean that many nuclei do not need? Despite the fact that the conclusion seems natural - no. Easy everyday tasks (such as web surfing or work with several programs at the same time) react positively to an increase in the number of processor cores. It is for this reason that the manufacturers of smartphones make focus on the quantity, lowering the specific performance in the second plan. Opera (and other browsers on chromium engine), Firefox run each open tab In the form of a separate process, respectively, the more nuclei, the faster the transition between tabs. File managers, Office programs, players are not resource-intensive. But with the need to frequently switch between them, the multi-core processor will increase the performance of the system.

Opera browser each tab assigns a separate process

Intel is aware of this, because Huperthreading technology, which allows the kernel to process the second stream by the forces of unused resources, appeared during Pentium 4. But it does not allow to fully compensate for the lack of performance.

In the Task Manager, a 2-core processor with HUPER Threading is displayed as 4-nuclear

The creators of the games, meanwhile, gradually catch up with the missed. The emergence of new generations of Sony Play Station and Microsoft Xbox consoles stimulated developers to pay more attention to multi-core. Both consoles are created on the basis of AMD eight-year chips, so now the programmers do not need to spend a lot of strength to optimize when porting the PC game. With the growing popularity of these consoles - with relief were able to sigh and those who were disappointed in the acquisition of AMD FX 8xxx. Multi-core rehabilitate the market positions, which can be verified by the example of reviews.

"Top" at that time of desktop processors who oversail a 2-gigahertz border. Today's day in the rules at both companies appeared on a new model, and therefore, there is a reason to hold another comparison or correct the shortcomings of the old one. The study of new models is always wondering if they differ architecturally, but today is not the case. Old kernels, the next stage of multiplication coefficients - that's new processors. The "reverse" fact deserves attention: Athlon XP 2100+ is the latest model on the Palomino kernel, not even in the production plan and covering the place before the release of the new core Thoroughbred.

Intel processors are also killed. Very soon there will be a transition to a bus 533 MHz, so that we have an instance, too, in some way "farewell".

Well, we will try to extract the maximum benefit from this testing. First, you can compare the new model with the previously, and by difference in tests in tests to evaluate scalability. Secondly, you can put a fresh version of the tests used and add new - good, such articles are usually not used for intermediate comparison. Finally, thirdly, always remain relevant completely useless and completely win-win attempts to identify the absolute leader in speed.

To solve the first task, add a 2.2-gighertz model in a pair of Intel Pentium 4, and ATHLON XP 2000+ AMD ATHLON XP 2100+, and test every couple on the same chipset. Based on the experience of the already mentioned large comparison, to solve the third task, we choose the three most interesting platforms for the Intel processor, and for the AMD processor, we will limit ourselves to one - the most quickly almost everywhere Via KT333 + DDR333. Well, before the update of the test set - please chandelice with the results.

Test conditions

Test stand:

• Processors:
  • Intel Pentium 4 2.2 GHz, Socket 478
  • Intel Pentium 4 2.4 GHz, Socket 478
  • AMD Athlon XP 2000+ (1667 MHz), Socket 462
  • AMD Athlon XP 2100+ (1733 MHz), Socket 462
• Motherboards:
  • EPOX 4BDA2 + (BIOS from 05/02/2002) based on i845D
  • ASUS P4T-E (BIOS 1005E version) based on i850
  • ABIT SD7-533 (BIOS 7R version) based on SIS 645
  • SOLTEK 75DRV5 (BIOS T1.1 version) based on Via KT333
• 256 MB PC2700 DDR SDRAM DIMM SAMSUNG, CL 2 (used as DDR266 on i845D)
• 2x256 MB PC800 RDRAM RIMM SAMSUNG
• ASUS 8200 T5 Deluxe GeForce3 Ti500
• IBM IC35L040AVER07-0, 7200 rpm, 40 GB
• CD-ROM ASUS 50X

Software:

• Windows 2000 Professional SP2
• DirectX 8.1
• Intel Chipset Software Installation Utility 3.20.1008
• Intel Application ACCELERATOR 2.0
• SIS AGP DRIVER 1.09
• Via 4-in-1 Driver 4.38
• Nvidia Detonator V22.50 (VSYNC \u003d OFF)
• CPU Rightmark RC0.99.
• RAZORLAME 1.1.4 + Lame Codec 3.89
• Razorlame 1.1.4 + Lame Codec 3.91
• VirtualDub 1.4.7 + DivX Codec 4.12
• VirtualDub 1.4.7 + DivX Codec 5.0 Pro
• WinACE 2.11
• WinZip 8.1.
• eTestingLabs Business Winstone 2001
• eTestingLabs Content Creation WinStone 2002
• BAPCO & MADONION SYSMARK 2001 OFFICE PRODUCTIVITY
• BAPCO & MADONION SYSMARK 2001 INTERNET CONENT CREATION
• BAPCO & MADONION SYSMARK 2002 OFFICE PRODUCTIVITY
• BAPCO & MADONION SYSMARK 2002 INTERNET CREAT CREATION
• 3DStudio Max 4.26
• SpecViewPerf 6.1.2
• Madonion 3DMark 2001 SE
• iDSOFTWARE QUAKE III ARENA V1.30
• Gray Matter Studios & Nerve Software Return to Castle Wolfenstein V1.1
• Expendable Demo.
• DRONEZMARK.

Test results

We have repeatedly tried to formulate the criteria of the optimal processor test. Of course, the ideal is unattainable, but today we make our first step in his direction - launch the project CPU Rightmark (). For details and news of the project, we send you to his site, here we also give brief explanations that should help you understand the essence of the test experiment and its toolkit.

So, CPU Rightmark is a processor test and memory subsystem that performs numerical simulation of physical processes and solving problems from the three-dimensional graphics area. Speaking very briefly, one program block is numerically solves the system. differential equationscorresponding to modeling in real-time behavior of the system of many bodies, the other block visualizes the solutions found also in real time. Each unit is implemented in several variants optimized for various processor command systems. It is important to note that the test is not purely synthetic, but written using techniques and means of programming, typical of the tasks of its area (three-dimensional graphic applications).

The solutions block of the differential equations are written using a set of commands of the X87 co-processor commands, and also has an option optimized for a SSE2 set (C cycle vectorization: two cycle iterations are replaced by one, but all operations are made with two-element vectors). The speed of this unit indicates the performance of the bundle processor + memory when performing mathematical calculations using valid double accuracy numbers (characteristic of modern scientific tasks: geometric, statistical, modeling tasks).

The results of this subtest show that the speed of working with the X87 FPU instructions at Athlon XP is higher, however, due to the support of the SSE2 set (naturally absent at Athlon XP), Pentium 4 turns out to be much faster. We emphasize that the SSE commands are not used in this block, so the results of the test run in SSE use modes are omitted (they simply coincide with the corresponding MMX / FPU and MMX / SSE2). We note almost the ideal scalability of the test on the CPU frequency - here the effect of memory is almost reduced to zero due to the effective caching and character of the operation of the block with intensive calculations with a relatively small amount of data exchange.

The visualization unit in turn consists of two parts: the pre-processing unit of the scene and the ray trace block and drawing. The first is written in C ++ and is compiled using a set of commands of the X87 coprocessor. The second is written in the assembler and has several options optimized for different sets of instructions: FPU + GeneralMMX, FPU + EnhanceDMMX and SSE + EnhanceDMMX (similar separation to blocks is typical for real-time visual realization tasks). The total speed of the visualization block indicates the performance of the processor + memory when performing geometric calculations using valid single accuracy numbers (typically for three-dimensional graphic programsOptimized by SSE and Enhanced MMX).

Again, the speed of working with the X87 FPU instructions at Athlon XP is significantly higher, however, use when calculating SSE again displays forward Pentium 4, despite the support of this Athlon XP processor set. At the same time, in terms of performance on megahertz, both processors go practically in the total - Pentium 4 receives a separation corresponding to its higher frequency. We emphasize that the SSE2 commands are not used in this block, so the results of the test run in the SSE2 activation modes are omitted (they simply coincide with the corresponding MMX / FPU and SSE / FPU). Note the excellent pentium 4 + SIS 645 bundle, caused, obviously, the greatest speed of access to memory at low latency. In general, the rendering process is accompanied by a rather active shipment of data, which makes the contribution of the chipset and the type of memory used to the total system performance.

The total performance of the system is calculated by the formula: overall \u003d 1 / (1 / mathsolving + 1 / rendering), so that a very significant winnings of Pentium 4 when using SSE2 in a block of calculating the physical model almost does not give an increase in performance without using SSE in the visualization unit. But when performing calculations using SSE, an additive from the inclusion of SSE2 is quite an impressive value. (Note that this characteristic is valid for specific selected testing conditions, the possibility of testing the test allows you to set almost any ratio of the deficiency time of the physical model and visualization (by changing the screen resolution or accuracy of calculations).) Since Athlon XP does not support the SSE2 set, its performance It sufficiently depends on the speed of drawing the scenes, where it is inferior to Pentium 4 when using the SSE set, although it remains an absolute champion of a "clean" speed of operations with only MMX and FPU. Note that from tested chipsets under Pentium 4 i845D looks slightly better i850 (probably due to greater latency in the latter), and the champion is SIS 645 due to the above.

A new version of the popular LAME encoder is already available for a long time, but we all had no case to apply it. As part of the preparation of this article, testing and the old, we used until now, version 3.89, and the last officially available version 3.91 were carried out. The results coincided completely (within the error), which is quite consistent with the lack of reference to the speed optimization of the code in the list of innovations of the program. (By the way, the encoder has already correctly supported working with all the available advanced multimedia sets of commands and registers.) The test, as you see, is excellent scaled by the processor frequency, since it is performed here to make effective preliminary data caching, but there are a number of questions on fairly low productivity. Pentium 4 on i850 and SIS 645. It seems the most reasonable suggestion that such an effect on performance has BIOS BACK: Product from Abit We have not seen yet in the case, but the fee from ASUS on the i850 is familiar to us, and when using previous version Firmware (once again send you to the past) there was no such recession. Athlon XP is still a leader in this test, and the version 2000+ is completely enough for victory.

A new version 5.0 DivX codec came out quite recently, but taking into account the great popularity of this product, it is not difficult to predict it active use Already in the near future, without waiting for new releases with error corrections. Well, we follow in the course of folk wishes and go to the application of the version of DivX 5.0 Pro. We also conducted similar testing with the version of DivX 4.12, and the results of the comparison of codecs are as follows: the coding operation accelerates quite significantly - more than a minute, and regardless of the processor, chipset and type of memory. Also note that DivX 5.0 Pro forms a little big output video file. To the comparison of the actual processors in this test, we need nothing to add - everything was already said in the last article, but for good scalability of the coding should be paid attention to.

In archiving WinACE, as when coding MPEG4, the effect of memory subsystem (due to the large amount of data sent) is about twice the effect of increasing the processor frequency. Athlon XP in this test is still better than His news.

In archiving WinZip, we note that some Pentium 4 lag on SIS 645 and full equality in other cases.

The results of Winstones look at a rarity logical and understandable, but having remembered about frequent inexplicable failures and bursts in these tests in the past, we, perhaps, refrain from commenting.

Let me remind you that you still have to say decisive "I do not believe!" The results of Athlon XP in the SYSMARK test, since, due to the curricularity of individual programmers, the WME 7.0 version, which is part of the Internet Content Creation group applications for this test, did not know how to identify the support of the SSE instruction set at Athlon XP. Fortunately, we finally begin testing in the updated version of the benchmark - Sysmark 2002, in which this problem is solved.

Briefly about the differences in test applications:

As you can see, no replacement, only updates of versions. The algorithm for calculating the final points of officially known changes has not undergone, although we would suggest the recalculation of some proportionality coefficients.

It is interesting to compare the results of the old and new packages in the office subtest: first, it was probably introduced a certain corrective coefficient, which led to a decrease in the indicators of both sides. Secondly, it is obvious, due to the converted Microsoft Office package, Pentium 4 began to win in this subtest, although in SYSMARK 2001 both processor platforms walked.

In the Creating Content, the situation is even more interesting: due to the normal SSE recognition at Athlon XP in MS WME 7.1, the AMD processor added, but it is part of the subtest of a new package that is rewritten to support SSE2 adobe version Photoshop 6.0.1, so Pentium 4 gets even greater increase.

As a result, from dubious leadership in Sysmark Pentium 4 proceeds to the leadership obvious. Pay attention to how great the performance of pentium systems is growing in this test with increasing processor frequency, and an almost missing similar effect for the Athlon system.

Rendering in the 3DStudio Max is perfectly scaled and usually does not demonstrate signs of dependence on the speed of working with memory, so that we can only guess what happened in last firmware BIOS for ASUS P4T-E engineers companies. The diagram is clearly seen that rendering on Athlon XP accelerates in proportion to an increase in the processor frequency, but just due to a much higher frequency of Pentium 4 2.4 GHz goes into this test in the gap, although the speed of another 2.2-gigahertz model was approximately equal to Athlon XP 2000+.

In SpecViewPerf, in general, nothing interesting: the results are almost everywhere equal, with a light advantage of Pentium 4, and only in the DX-06 noticeably ahead of Athlon XP. Note that the test speed is almost independent of the processor speed.

When switching to a new processor Intel game Benchmark makes a small jerk, but it does not help him reach even to the results of Athlon XP 2000+.

Adding to test games Return to Castle Wolfenstein, based on the QUAKE III engine, the situation naturally has not changed. Moreover, relative indicators in these two games are similar almost one in one. I also add a DRONEZ, distinguished by the engine, but not the character of the results, and only ancient Expendable is left for Athlon XP ... We note that all the games are approximately equally well scalable in the processor frequency, which also plays intel.

conclusions

Farewell to the Palomino kee was not too much: it is impossible to say that Athlon XP is so far behind his rival, and indeed everywhere this lag at all occurs, but there is a tendency. With a real frequency, with the PR-rating of Lee - AMD lags behind Intel in terms of magic numbers in the name of processors, and the increase in frequency increases (no matter what "Duta" is considered to be in Pentium 4) in most of our tests gives an advantage in absolute indicators it is the pentium line 4. Many applications "learned", finally, about support for SSE in Athlon XP, which gave some splash, but this is a dead end, but the optimization under SSE2 is still not completed, and the further - the more apps Will go to the AMD Camp in the Intel Camp.

However, the post of Palomino leaves in a decent state. Standing last model from the existing competitors is not a catastrophic, the price is attractive, and we are with great andm I wonder for AMD attempts to return leadership with the new core.