Informática
Comparison AMD vs INTEL
Introduction: AMD K6 vs Intel Pentium As a introduction we can say that there are a lot of different concepts between these kind of processors. In order to find out the main differences we will see basic concepts about processors technology. The evolution in the field of microprocessor, in the last 2 years, has been characterised by an even more increasing change; starting from spring 1997, when sixth generation CPUs have been introduced (AMD K6 and Intel Pentium II), clock frequency of processor has increased a lot coming over 500 MHz. Many think that the clock frequency of a processor is the most important element of general performance, while it is only one of the aspects to consider as it is as important as the architecture used on one side and the software optimisation on the other. For instance, two different CPUs both at 400 MHz of clock achieve different performance with some applications and the almost the same performance with others; this shows how it is wrong to consider only the clock frequency as choice parameter to measure the performance of a processo. Apart from the clock frequency, there are other important parameters such as the type and the amount of cache memory installed; this kind of memory allows to speed the data transfer from the processor to the system memory and affects the general performance; there are two types of cache:
K6-III CPU led to the definition of a new kind of cache, L3 or third level; this is the L2 cache on Socket 7 motherboard renamed as L3 as K6-III integrates it within the CPU core. Note how when you disable the cache external to the processor there is a decrease in general performance (the benchmark used is Ziff-Davis Business Winstone 99). K6-III behaves different than K6-2 and Pentium II: as its L2 cache is integrated on the CPU core it has been possible to disable only the L3 cache, that is the one installed directly on the motherboard; note how the L3 cache doen's affect the performance (a decrease of 3,5%). Turning to applications which use the floating point unit, such as 3D Studio Max 2, you can note how disabling the L2 cache doesn't affect a lot the performance; this is due to the fact that the floating point calculations doesn't use a lot the L2 cache. Note how the performance of K6-III aren't affected quite at all. A second parameter to consider is the internal architecture of the CPU Core. The instructions topologies carried by a processor are, in general, two:
The architecture chosen allows to obtain higher performance than the competitors in one, or both, topologie; Without describing the technical details, is important to know that CPU Core of AMD K6-2 and K6-III achieve, with the same clock frequency, better performance than that of Intel Celeron, Pentium II and Pentium III CPUs with integer calculations, while the contrary isn't true, where Intel CPU are better with floating point calculations. In detail L2 cache affect a lot the performance with integer calculations, while it doesn't affect quite at all the performance with floating point calculations. The special architecture of K6-2 CPU, with L2 cache working always at the bus frequency (100 MHz) and not bound to the clock frequency, allows, above a specific clock frequency, Intel CPUs to obtain better performance even with integer calculations due to the fact that with Intel CPUs the working frequency of the L2 cache increases as the clock frequency increases and consequently the performance too. This isn't true for K6-2 CPUs which have an L2 cache working at clock frequency. About 3D video performance, it deals a lot with processors; video and CPU are strictly connected. In the generation process of a 3D scene (let's think to a game) many factors are involved, both hardware and software; many think that most work is carried by the video card, and it is true, but the processor too plays an important role as it processes many calculations necessary to define the image. For instance, if we want to display a cube, the calculations for the corner position are carried by the processor and not by the video card; more complex is the scene, more calculations are to be carried and more will be the operations executed by the processor. These are floating point calculations, so the FPU (Floating Poin Unit) has to work hard. Considering the fact that 3D scenes aim at achieving real-life scene and showing many details leads to the creation of video processors more complex and with high performance, there will be an even increasing demand for floating point performance of the CPU to create scene which are even more complex (so with many polygons) and realistic. Intel and AMD, with their proprietary instruction sets within the Pentium III and K6-2 and K6-III CPUs, aim at increasing the performances of the processor while carrying the operations needed to create a 3D scene. These instructions, called respectively SSE and 3DNow!, are compatible among them and must be turned on by the software to work; practically, they are a good tool to overcome the performance limit of the FPU of the processor to allow the video chip to manage a high number of polygons and to create even more complex scenes. | |||||||||||||||||
AMD K6-III AMD K6 CPU has been launched in April 1997 and it has been the first CPU on the market of the sixth generation followed, a month later, by Intel Pentium II. This processor soon showed high performance with integer calculations (at that time the competitors were Intel Pentium MMX and Pentium PRO CPUs as well as the new Pentium II), with performance with floating point calculations as choke point. In June 1998 the model K6-2 has been introduced with some difference in comparison to K6.
In February 1999 there has been the third evolution of the K6 series, with the introduction of K6-III model; this CPU has been created integrating 256 Kbytes of L2 cache within the Core of K6-2 CPU. Following are listed the technical details of the two CPUs.
K6-2 K6-III Socket Socket 7 Socket 7 L1 Cache 64 Kbytes (32K data; 32K istructions) 64 Kbytes (32K data; 32K istructions) L2 Cache from 512Kbytes to 2 Mbytes, depending on 256 Kbytes, inside the cpu Core; L3 Cache - from 512Kbytes to 2 Mbytes, depending on Core voltage 2.2V (2.4V for 450 and 475 Mhz versions) 2.4V Clock frequency 300-333-350-366 400-450 Mhz Bus frequency 95Mhz; 100 Mhz 100 Mhz Core type CXT CXT K6-III CPUs must be installed on Super 7 motherboards, that is those including following requirements:
Note how all the Super 7 motherboards for K6-2 CPUs, if including power voltage of 2.4 V, can use K6-III CPU too after updating the BIOS. There are some problems connected to the electrical features of the CPU; due to the integration of the L2 cache within the CPU Core, the requirements for a processor have been increased in comparison to K6-2, as shown here:
K6-2 K6-III Current 7.35A (266 Mhz) 12.40A (400 Mhz) Maximum heat dissipation 14.70W (266 Mhz) 26.80W (400Mhz) Medium heat dissipation 8.85W (266 Mhz) 16.10W (400Mhz) Note the increased maximum heat dissipation (the maximum values remains however the same as K6, that is 233 MHz working at 3.3V with a value of 30.2W), which can be explained both with power voltage increased at 2.4V and high clock frequency as well as the power maximum request which achieves 13.5 Amperes with version at 450 MHz of clock. It is important to check the Super 7 motherboards, above all if one of the first manufactured, to see if they can be used with K6-III CPUs; the readme file of the newest version of the BIOS must be read in orther to do this. The certification for K6-III CPU ensures that the electrical part of the motherboard matches the power requirements of the K6-III so to avoid stability problems and misbehaviours. K6-III CPU can also been installed on some Socket 7 motherboards including Intel 430HX and 430TX chipset with power voltage at 2.4V; these motherboards don't include bus frequency of 100 MHz, but this doesn't affect the performance as we can see in the following illustration: Setting the 2x multiplier of the motherboard (multiplier included in all Socket 7 motherboards), the CPU multiplies at 6x so to obtain the clock frequency of 400 MHz (66x6) similar to what happens with K6 CPU and Pentium MMX which turn on the 3.5x multiplier when the 1.5x motherboard is set. Is necessary to visit the Web site of the manufacturer and check the certification for K6-III (always with the problem above said of the maximum power), as well as the presence of a BIOS update which detects the CPU.
The main limit of the 3DNow! instructions is a wide machine range (that is a number of PC including 3DNow! CPU) but a limited range of software. It is possible to see with games software. Microsoft API DirectX support 3DNow! instructions starting from the version number 6, but often game programmers don't use them or only partially; it refers to a more efficient support of the 3DNow! instructions starting from the version number 7 of DirectX. Practically, with 3DNow! instructions there is an increase of 10-15% of fps with DirectX 6 programs even if the patch for Quake 2 (a very popular game developed with 3D technology) specifically designed by AMD shows how the gain in performance can be higher if the instructions are used in the correct way: Using a 3dfx VooDoo 2 SLI accelerator is possible to note an increase in performance, about 68% more at a resolution of 800x600 and 65% at a resolution of 1024x768 using K6-III CPUs; using a Riva TNT accelerator, the gain in performance is lower, between 15% and 20% depending on the resolution. Using 3DNow! instructions with the rendering software leads to an increase in general performance even if limited to 12-14%; anyone use nowadays rendering through software, but this kind of test can show the average increase that we can expect turning on the 3DNow! instructions. 3D Mark Max benchmark allows to check the impact of 3DNow! instructions in rendering the polygons of the 3D scenes; turning on 3DNow! instructions lead to an increase in performance changing from 150% and 180%, a very high value. These data are only theoretical values as it can't be reproduced with real applications, but they clearly show how the appropriate 3DNow! instructions can speed up the processing of the geometry of the scene. Using 3DNow! instructions doesn't affect only the increase of displayed fps; in fact if there are games requiring many fps to be played (such as Quake 2), there are some other which need less frames per second (think to graphical adventures); in these cases, 3DNow! instructions enrich a 3D scene, more details are displayed to increase realism without changing the number of frames displayed (more processing power available, more polygons inserted in the scene and processed by the CPU, more detailed scene).
Pentium III Pentium II has been launched at the end of April 1997; at the very beginning of that month AMD had launched the first x86 6th generation cpu, the K6, but Pentium II has soon become the fastest CPU available, both for clock frequency and general performance. There are many improvements from Pentium MMX (launched in January 1997):
Pentium II has been launched with clock frequencies of 233 and 266 MHz, followed by the 300 MHz version in September 1997; in January 1998 the version at 333 MHz was launched and it was the first CPU using a production process at 0.25 microns against the 0.35 microns of previous versions; this leads to distinguish two types of Core used for Intel Pentium II CPUs: Klamath at 0.35 microns and Deshutes at 0.25 microns. In April 1998 the version at 350 and 400 MHz of clock frequency have been launched, with bus at 100 Mhz, while in September 1998 the version at 450 MHz of clock frequency has been introduced. In February 1999 the Intel Pentium III CPU has been officially presented, available at 450 and 500 MHz of clock frequency: following are listed the main features of these CPUs:
Pentium II Pentium III Slot type Slot 1 Slot 1 Cache L1 32 Kbytes (16K data; 16K instructions) 32 Kbytes (16K data; 16K instructions) Cache L2 512 Kbytes 512 Kbytes Core voltage 2.0V - 2.8V 2.0V Clock frequency 233-266-300-333-350-400-450 Mhz 450-500 Mhz Bus frequency 66Mhz; 100 Mhz 100 Mhz Core type Klamath (233-266-300 Mhz) Deshutes Following are listed the electrical features of the processor:
Pentium III Current 14.5A (450 Mhz) Maximum heat dissipation 25.3W (266 Mhz) It can seem that Pentium II and Pentium III are, except for the clock frequency, the same, but it isn't true: in Pentium III 70 new instructions have been introduced, called SSE (Streaming SIMD Extensions), which speed up 3D, videoconferencing and multimedia programs. In detail, they can be divided in three groups:
From this comes that the software has to use all the 70 instructions, as it happens with 3DNow! instructions of AMD K6-2 and K6-III CPUs. This can be made in different ways: integrating the necessary code within the software (3D application or game), support through the driver of the video card (for the instructions which speed up the running of 3D applications) or through API Microsoft DirectX 6.1. Nowadays there are only few applications which use SSE instructions, so we can't forecast the spreading of these in houses' software; it is true that when they are used the performance increases as shown in these examples: The CPU 3D Mark test included in the 3D Mark 99 Max benchmark allows to check the impact of SSE instructions on the calculation process of the geometry of a 3D scene; note how using these instructions the performance increases of 62%. Another benchmark which shows that the performance increases when using the SSE instructions is Lightning and Transformation included within the 3D Winbench 99 version 1.1 of Ziff-Davis; the performance obtained by Celeron at 500 MHz is the same of an ideal Pentium III at 500 MHz without SSE instructions. Note how the performance increases of 72%, a very high value; the limit of these data is the fact that they refer to ideal situations which can't be reproduced in real life. One of the main features of Pentium III is a serial code which allows to identify the processor used; following there is a screenshot of the utility which allows to display the serial code. There have been many arguments on this feature of Pentium III: it is technically possible, in fact, that someone reads the serial code while you are online (such as when you visit a site) without noting it and this led Intel to disable the default serial code. It is difficult to understand why Intel introduced such a feature, except for the fact that it allows to identify the processor thanks to the serial code. | |||||||||||||||||
Overlockability AMD cpu, like K6-2 or K6-III, haven't any kind of frequency multiplier lock, so is possible to operate with these changing it, or increasing bus frequency. It is de definition of OVERLOCKING. In general is better, with Socket 7 cpu, to increase the bus frequency, because the L2 cache works at that frequency, and the L2 cache speed is very mportant for the general performance. With the K6-III the situation changes a lot, because the L2 cache is integrated in cpu Core and so the conversion into L3 cache of the motherboard cache, allows to understand how an increase of bus frequency isn't important, if not useless, for general performance. So is necessary to obtain the highest clock frequency supported by the cpu, even if, to do this, we have to use a not very high bus frequency. Anyway is important to work at about 100 Mhz bus frequency, to obtain high performance also from L3 cache and system memory in the event of cache miss, that is data that isn't possible to load on cpu L1 and L2 cache. For example, with a K6-2 cpu we obtain higher performance with 124x3.5 (434 Mhz) instead of 100x4.5 (450 Mhz), because the L2 cache works faster. With the K6-III, instead, is the contrary, because the L2 cache works at 450 Mhz instead of 434 Mhz. The overclockability range of K6-III is reduced, because the cpu Core voltage (2,4V) doesn't allow to overvolting the cpu more than 2,5V, to avoid cpu damage, even if it is impossible to generalize: not always two cpu are identical. Intel CPUs manufactured after August 1998 has a locked frequency multiplier so it isn't possible to set a different one in comparison to the default one independent from the fact that this is higher or lower; the frequency multipliers follow this kind of order: Processor multiplier Bus frequency Intel Pentium II 266 Mhz 4x 66 Mhz Intel Pentium II 300 Mhz 4.5x 66 Mhz Intel Pentium II 333 Mhz 5x 66 Mhz Intel Pentium II 350 Mhz 3.5x 100 Mhz Intel Pentium II 400 Mhz 4x 100 Mhz Intel Pentium II 450 Mhz 4.5x 100 Mhz Intel Pentium III 450 Mhz 4.5x 100 Mhz Intel Pentium III 500 Mhz 5x 100 Mhz Intel Celeron 266 Mhz 4x 66 Mhz Intel Celeron 300 Mhz 4.5x 66 Mhz Intel Celeron 300 A 4.5x 66 Mhz Intel Celeron 333 Mhz 5x 66 Mhz Intel Celeron 366 Mhz 5.5x 66 Mhz Intel Celeron 400 Mhz 6x 66 Mhz Intel Celeron 433 Mhz 6.5x 66 Mhz Intel Celeron 466 Mhz 7x 66 Mhz Generally Intel CPUs have a high margin of overclockability; in this case, Pentium III CPU at 500 MHz achieves with no problem 560 MHz using a bus frequency of 112 MHz while the version at 450 MHz achieves 504 MHz always with bus at 112 MHz. We can also achieve higher values acting on the cooling system, also because a value of 550 MHz of clock is near to the maximum limits of clock frequency of Deshutes core with production process at 0.25 microns. | |||||||||||||||||
Test configuration The test configuration is the following:
Socket 7 Slot 1 cpu AMD K6-2 400 Intel Pentium III 500 motherboard Asus P5A Abit BH6 memory 2x64 Mbyte Hunday PC100 2x64 Mbyte Hunday PC100 video cards Diamond Viper V550 (chip NVidia Riva TNT) Diamond Viper V550 (chip NVidia Riva TNT) 3D video cards 2 Creative Labs 3D Blaster VooDoo 2 8 Mbytes 2 Creative Labs 3D Blaster VooDoo 2 8 Mbytes hard disk Quantum Fireball EX 6.4 Gbyte Quantum Fireball EX 6.4 Gbyte cd-rom Pioneer 36x EIDE Pioneer 36x EIDE operative systems Windows 98 Windows 98 driver Windows 98: NVidia Detonator Windows 98: NVidia Detonator The tests have been carried out on following criteria:
| |||||||||||||||||
High-End applications Performances of the floating point unit of Intel cpus is quite superior to K6-2 and K6-III ones, as can easily be seen in High-End Winstone 99 results. K6-III has higher performances, at the same clock speed, than K6-2, thanks to the 256 full speed L2 cache memory, with an average gap of 9% compared with Intel cpus at the same clock speed (k6-2 obtains an average 23% gap when compared with Intel cpus). Measuring the time needed to apply some filters with Adobe Photoshop 5.02 we can clearely see that the best performances are achieved by Intel cpus, but K6-2 and K6-III are quite near; K6-III 450 Mhz has the same performance as Pentium II 400 Mhz. Best performances are obtained using Pentium III at 500 Mhz and celeron at 500 Mhz; we can clearely see that, even if Photoshop 5.02 is SSE optimized, this instructions are not used in this benchmark. Also with 3D Studio Max 2 we can clearely see that the performances of Intel's CPU are quite superior to AMD's one; the 256 Kbytes L2 speed working at clock frequency lets the K6-III to obtain performances 10% faster than K6-2 at the same clock speed, but still remains a certain gap with Intel cpus at the same clock speed.
| |||||||||||||||||
Conclusions The comparison between Intel Pentium III and AMD K6-III clearly shows the differences, both in terms of performance and in terms of concepts, between the two CPUs:
This analysis shows that the CPU representing the best compromise between performance and price is Intel Celeron: this processor has, at the same clock frequency, similar performance as Pentium III with applications which don't use SSE instructions. K6-III is a big step forward in comparison to K6-II: the integration of 256 Kbytes of L2 cache working at the clock frequency allowed to obtain a big gain in performance both with Business applications and High-End software even if the performance of floating point unit (the same as in K6-II) is lower than that of Intel CPUs. Pentium III is practically a Pentium II working at a higher clock frequency and including the SSE instructions set; even if these CPUs, if correctly used, allow to obtain performance gains, the software support isn't still clear. The version at 500 MHz of Pentium III is, in general, the best performing processor available on the market, but its price is too high if compared to the performance. Finally, we can distinguish between three kind of processor needs:
|
1
Descargar
Enviado por: | Chechiman |
Idioma: | inglés |
País: | España |